It is assumed that deserts have low species diversity. It is also assumed that biotic interactions in deserts are unimportant in the determination of community structure. The common explanation for the supposed low diversity of organisms in deserts is that the harsh environmental conditions limit the number of organisms that are able to live in this environment. This is because only few have the physiological capability to live under the stressful desert conditions, or because low primary productivity limits the amount of energy available to higher trophic levels. However, recent reviews demonstrated that almost all major animal taxa are represented in deserts and that their diversity may be comparable to that of many mesic habitats (Polis, 1991a). Polis (1991a) offered seven possible non-exclusive hypotheses on the processes that shape desert community structure, but data are still lacking in support of the relative importance of any of these processes.

It seems that adaptations for living in desert environments have evolved independently many times and in many animal groups. Therefore, physiological constraints do not prevent the colonization of the desert by most types of organisms (Dawson et al., 1989). Low primary productivity is a more plausible mechanism that may limit the number of organisms in deserts. The positive correlation between productivity level and the number of organisms at higher trophic levels has been suggested for a long time (Elton 1927, Lindeman 1942), and has some empirical (e.g. Wright 1983, Rosenzweig 1975) and theoretical (Oksanen et al.,1981; Oksanen, 1990) support: low primary productivity reduces the number of animals at higher trophic levels. This alone reduces diversity and also reduces the intensity of biotic interactions within desert communities (Noy-Meir 1974, 1985, but see Polis 1991ab). Biotic interactions are important in maintaining high diversity in natural communities.

The model developed by Fretwell 1977 and Oksanen et al. (1981) relates the number of ‘effective’ trophic levels to habitat productivity (an effective trophic level is one that, if not limited from above, can limit the level below it). The model also suggests the kind of biotic interactions that limit population densities at each trophic level present in the habitat. According to this model, (a) the number of effective trophic level increases with the productivity of the environment and (b) within a certain community, when trophic levels are counted from the top down, odd ones are controlled by competition and even ones by predation (the exploitation ecosystem hypothesis, EEH). Accordingly, low-productivity deserts include at most one trophic level (=trophic link), i.e. plants, that compete for resources (mainly water). Animals of all trophic levels are scarce and hence their diversity is expected to be low. Thus, deserts form a one trophic-link environment in which interactions among animal species are unimportant. The EEH was attacked from several directions arguing that its basic assumptions are too simplistic to suit the complexity of the natural environment and food web structure (see a review by Parsson et al., 1996; Polis and Strong, 1996). However, some of the factors mentioned by the model opponents are unrelated to habitat productivity (e.g. refuges, prey defense) or have only a weak connection to productivity (e.g. omnivory is possible only in ³ 3 links trophic chains) and may be marginally important in most food webs (Hairston and Hairston, 1993, 1997). However, objections to the EEH all suggest that the higher trophic levels have reduced effectiveness in controlling lower trophic levels. Hence, in these models, there will be fewer effective trophic links in a community. Thus, the EEH model can be regarded as suggesting the potential number of links energetically possible in an environment with a certain input of primary productivity.

A major habitat characteristic that changes with productivity in terrestrial ecosystems, and one not dealt with by the EEH, is plant structure. In terrestrial ecosystems, plant competition changes from soil resources to light as the productivity of the environment increases. As a result, first plant cover and then plant height change as we move from poor to richer environments. These changes affect, among other things, the exposure of active species to mobile, visually-orienting predators. The change in habitat structure with increasing productivity can (a) add feeding niches for herbivores and (b) reduce the per capita predation intensity on all trophic groups. Thus, the increase in plant structural complexity as habitat productivity increases counteracts the effect of increasing predator densities. We will refer to this effect of the change in plant cover with productivity as the productivity-structure hypothesis (PSH). This hypothesis does not replace the EEH but may be regarded as a complementary one.

Plant cover in desert habitats ranges from almost barren terrain to low densities of small shrubs and perennial grasses, and variable densities of annuals depending on seasonal precipitation. Deserts are distinguished by short spurts of primary productivity in the short rainy season(s) and long periods of minimal productivity in between. Also, there is high between-year variability in primary productivity resulting from the variability in precipitation levels. The short periods of high primary productivity enable the development of only one generation per season for most herbivorous insects. This environmentally-imposed univoltinism results in a one- year time lag in the interactions between herbivorous insect and their food resource, and limits the ability of these insects to closely follow rapid changes in primary productivity (Ayal 1994). As a result, most of the primary productivity in deserts ends up as seeds, or as dry plant material which accumulates in the habitat because the low moisture limits microbial decomposition. These two resources are the only long lasting rich resources in deserts and form the basis for the abundance of macrodetritivores and seed predators, which then transfer energy into the higher trophic levels (Crawford 1986). Hence, in deserts, macrodetritivores (e.g. termites, darkling beetles, isopods) and seed predators (e.g. ants) replace the herbivorous insect as the lower link of the arthropod-dependent food web. These two groups support a rich fauna of predatory arthropods (Polis and Yamashita 1991) and reptiles (Vitt 1991) as a second trophic link. The third link in this web are arthropodivorous birds (e.g. Ayal and Merkl, 1994 and see Polis 1991b) and mammals (e.g. Parmenter and MacMahon 1988, Shamir personal communication) that feed both on the first and the second trophic groups (i.e they are omnivores).

Low primary productivity in deserts results in low plant cover. Therefore, most desert habitats are transparent: animals that move in them are exposed and are easily located by visually orienting predators. This is especially true for surface dwelling arthropods, the majority of animals that live in deserts, which are exposed to high risks of predation by visually-orienting vertebrate predators. Of these predators, birds seem to be the most important as many of them are highly vagile and can move into the desert in high numbers in the appropriate season, and between habitats within the desert according to the local distribution of arthropods. Most of the arthropod-feeding birds take prey within a specific size-range and most are omnivores. Thus, the effect of bird predation on the distribution and abundance of macrodetritovores depends on the relative predation intensity by birds on intermediate-size predators (Ayal, personal observations).

The local productivity of a desert habitat depends mainly on the horizontal re-distribution of rainwater, which is a function of topography. The redistribution of water results in a mosaic of habitats with a wide range of plant cover. This, in turn, results in variation in predation intensities by different trophic guilds. In low productivity habitats plant cover is very low and predation intensity of arthropodivorous birds is high, especially on large arthropods and small reptiles. In low-productivity habitats the density of large predatory arthropods that are active on the surface (e.g. spiders, scorpions, solipugids) and of lizards is low and their foraging activity may be quite limited. As a result, predation intensity on small, burrow-dwelling macrodetritivores such as termites and isopods should be low, their densities may be relatively high and they may be food limited (Ayal, personal observations). On the other hand, in relatively rich habitats, plant cover is high enough to reduce the efficiency of predation by birds on both predatory arthropods and large macrodetritivores. Hence both groups are quite abundant in such habitats. The abundance of predatory arthropods in the productive habitat imposes a high predation level on small macrodetritivores and it seems that in such habitats, predatory arthropods and lizards may control the densities of macro-detritovores (Ayal 1977).

We suggest therefore that despite the low productivity in deserts, predation plays a major role in determining animal community structure. This may explain both the unexpectedly high diversity of arthropods in deserts and the abundance of arthropod predators as well.

Bibliography

The measurement of biodiversity is beset with problems of scope and definition; the term covers a hierarchy of conceptual issues ranging from genetic and functional diversity, through species diversity and up to community and biome level concerns. Even restricting our attention to species-level diversity measures, many indices present themselves, reflecting different (spatial or temporal) scales of analysis. Biodiversity is an intrinsically scale-dependent subject. Whittaker’s (1977) classic distinctions between a , b , g , d , and e diversity are explicitly scale-dependent, although the specific scales evoked remain rather ill-defined. Measurements are dependent not only on spatial scale, but on temporal scale as well; classical analyses of "point" (a ) diversity – e.g. insect collections from a single suction trap – can only be meaningfully analysed if we allow a wide range of time to pass at that "point" in space. At a single point in space and time, the diversity cannot generally exceed a single individual of a single species (unless we count parasites and symbionts…)! Coarser scale diversity measures (g and e diversity) refer to even less well-defined (spatial) scales of analysis; the diversity within a "landscape" or a "region." These measures of inventory diversity are linked by measures of transitional diversity; b diversity measures the difference between different "points" within a "landscape," whereas d diversity measures the diversity of "landscapes" within a "region."

A second classical ecological tool may provide an alternative approach. The species area curve ("SAC") can be used to define the relationship between diversity and scale, as it reflects the shift in species count over a range of scales. With such a curve, one can read off a , g and e diversities by noting the height of the SAC at different points (with area values corresponding to a single "site", a "landscape" and a "region", respectively). Similarly, b and d diversities can be assessed from the slope of the SAC at different scales, as the curve rises due to the differences between sites in species composition. There remains some debate as to whether the SAC is approximately linear in log-log space or in semi-log space. In either case, the implication is that some aspect of b diversity is scale - invariant: in the former case the multiple of species gained per multiple of area is approximately constant, in the latter the absolute number of species gained per multiple of area.

Such diversity patterns can be seen as the sum of the spatial patterns in individual species incidence. A fractal species distribution gives rise to a linear plot of (log) area occupied by a species as a function of (log) scale. The sum of many such linear "scale area" curves should itself be linear. Scale area curves do not appear to be precisely linear at relatively fine scales, however, and this curvature may help explain the semi-logarithmic nature of some observed species-area plots. These points will be examined using examples of species distributions taken from a multi-scale survey of annual plants performed in the northern Negev.

Although arid regions are commonly viewed as being impoverished environments for vertebrates, many deserts contain rich faunas with exceptionally high species diversities. Parts of arid southern Africa and North America, for example, contain local rodent faunas with over a dozen species; in arid Australia up to nine species of insectivorous marsupials and 40 species of lizards may co-occur. This contribution will review the factors that are known to influence species diversity and/or species richness in the vertebrate groups that have been studied most intensively in these respects. We will also point out topics that need further investigation.

Factors that Influence Reptile Diversity

Abiotic factors: Species richness of reptiles in North America is highly correlated with amount of sunshine received by an area. It is also correlated with evapotranspiration dfs, and weakly correlated with rainfall. Being poikilotherms, many reptiles benefit from the abundance of radiation in arid lands, and this is not restricted only to diurnal heliothermic species: nocturnal species may also benefit indirectly from sunfall in an area, through high ground temperatures.

The structure of the habitat has been found to be very important in determining reptile species diversity. Regions containing heterogeneous habitats support diverse reptile faunas. Most often this phenomenon has been exemplified with respect to vegetational types, for example in arid Australia, where several desert habitats co-occur in certain areas ("spinifex", mulga, etc) many reptile species can coexist. Certain vegetation types, such as the unique Australian "spinifex" habitat (Triodia spp.) form ameliorated conditions, which enable the survival of species, that otherwise would not be able to tolerate desert conditions. Other features of habitats, such as soil, rock cover and rock features, are likely to be involved in the determination of species diversity, but no studies have checked their contribution to reptilian diversity independent of the vegetation. Sandy substrates definitely increase species richness, not only through the diversity of plant species that they impose on different parts of sand dunes, but also by supporting several fossorial species, in addition to the above-ground lizards and snakes. Soil infertility is one of the explanations that have been suggested for the high reptile species-richness in arid Australia: the low productivity restricts the diversity of birds and mammals, while ectotherms such as reptiles can still be supported.

Biotic factors: Reptile species richness is correlated with the diversity of the resources utilized. The diversity of lizards and snakes, for example, was found to be partly dependent on prey diversity (arthropods and small mammals, respectively). The diverse termite fauna in Southern Africa allowed the evolution of termite specialists among desert lizards. In Australia, the infertile soils on which "spinifex" grows have a rather poor arthropod community, but a rich termite fauna, which in turn, support many species of lizards.

Australian desert lizards seem to have taken some of the ecological roles normally occupied by other taxa (snakes and mammals) in arid regions of other continents. In North America, reptile species richness is inversely correlated with the richness of birds. It has been suggested that ectothermic reptiles might enjoy some competitive superiority over endothermic birds in arid environments. Lizard biomass in North America also varies inversely with the biomass of small seed-eating mammals. Here a direct connection is less obvious, and it has been suggested that certain vegetation types invest most of their primary productivity into either leaves or seeds, supporting either insect-eating lizards or seed-eating mammals.

Niche overlap among lizard species decreases in areas rich in lizards, and this fact, combined with the direct relationship between food and lizard diversity, indicate that competition among reptiles is likely to influence species diversity. The direct and indirect effects of predation on reptile diversity have been studied less intensively than in other vertebrate groups (mammals, birds, amphibians and fish), and more effort need to be invested in these areas. Mutualism is another relationship that deserve more attention, but already now it is clear that where burrows are dug in hard soils by one species, co-habitation in the burrows often occurs. The extremely long and deep Uromastyx burrows in the Negev Desert, for example, support several species of arthropods, geckos, snakes and rodents.

Factors that Influence Mammal Diversity

Abiotic factors: Because of its major effect on primary productivity, rainfall is an important determinant of diversity in many deserts. In North American and Middle Eastern deserts, the local richness of rodents has been shown to follow Tilman-type curves, with high richness occurring in areas of poor productivity and lower richness occurring with increasing productivity. Such patterns are not evident for mammals in arid Australia or in the Namib of southern Africa. In the latter desert, much productivity is generated in inland areas of higher rainfall, and this "subsidises" local desert faunas in the Namib after prevailing winds. In parts of arid Australia, local richness of mammals is affected by regional rainfall patterns, with many species moving from drought-struck sites to localities that have received rain. A similar pattern is shown even more dramatically by birds.

Habitat is a second, critically important factor contributing to mammalian diversity. At the regional scale, landscape features such as riparian strips provide structurally complex and productive environments that allow ingress of mesic-adapted species into desert regions, as well as sites of refuge for desert species during drought.

Dramatically increased local diversities of mammals (and many other vertebrates) have been found along riparian corridors in many deserts, including the Namib, Thar and Simpson. At a similar scale, substrate has a large effect on species diversity and composition. There is some evidence that sandy substrates contain relatively more species than rocky, gravel or loess substrates in several world deserts, but there has been little attempt to determine whether sand is preferred (e.g. easier for burrowing) or simply provides a larger area than other substrates. Upland deserts are often relatively species rich, perhaps because they offer structurally complex environments with provision for shelter. Many mammals and other vertebrates, tend to be confined to particular substrates, and show morphological specialisations that permit exploitation of those substrates. At a local scale, microhabitat is a further important determinant of diversity. Numerous studies have shown correlation of small mammal, avian and reptilian richness with a microhabitat complexity. Often, strong associations are found, such as Ningai species and spinifex in arid Australia, or Wagners gerbil and rocky substrates in the Negev. In a few cases species richness has been shown to be causally dependent on microhabitat by experiments that have manipulated habitat complexity. Such experiments, in North America, Australia and the Kalahari, have shown that simplification of local microhabitat structure leads to predictable declines in small mammal richness. In the latter two areas, avian and reptilian diversity was also reduced on manipulated plots, with individuals in each taxonomic group avoiding simplified sites.

Biotic factors: Following from the important, broad-scale influence of rainfall and primary productivity on diversity, many studies have investigated the relationships between local species richness and the abundance and diversity of food. Classical early work on North American Heteromyids suggested that species packing in local communities depended on the partitioning of different-sized seeds. Similar relationships have been inferred in desert granivorous birds, and insectivorous mammals and lizards. However, more recent work suggests that some partitioning of food types or sizes may be a secondary consequence of species foraging in different microhabitats where food spectra also differs. Recent work suggests further that the Heteromyids paradigm may not be readily transferred to other desert mammalian systems; in contrast to the strict granivory exhibited by many heteromyids, rodents in other world deserts appear more omnivorous and opportunistic in dietary range.

Given that many desert small mammals can potentially take a broad range of prey types and sizes, some inquiry has been directed at identifying factors that constrain habitat and diet choice. In North America, and to a lesser extent, South America and the Middle East, competition has been seen as an important force organising community structure. The strongest evidence for competition has come from a series of elegant removal experiments which have demonstrated that dominant competitors can reduce the abundance and diversity of subordinate species at least in local areas. Competition has been demonstrated among rodents and between rodents and ants, and has also been inferred to occur between birds and other two predominantly granivorous groups.

Direct and indirect predation also shape mammalian community structure. Experimental removals of avian and mammalian predator species have shown that mammalian prey can increase subsequently in abundance and diversity. In contrast, manipulation of predation risk has suggested that prey species often shift to protected microhabitats when risk is high, and may thus allow more species to pack into local areas.

A final, general group of biotic interactions can be lumped under the heading of facilitation. Such interactions have been poorly studied, but may be pervasive in desert systems. Seminal studies in North America have shown that the foraging activities of rodents and ants facilitate the local presence of granivorous birds via their effects on habitat structure. In arid Australia, the presence of dasyurid marsupials is facilitated by the burrowing activities of rodents, dragons and other organisms. Dasyurids cannot dig, and thus depends on the engineering activities of other taxa for shelter. A similar situation occurs in the Eastern Negev Desert, where the penetration of a rock-dwelling rodent (Acomys) to densely vegetated sandy areas is possible because it inhabits burrows dug by gerbils.

Human Impacts on Vertebrate Diversity

Cattle and other herbivores divert primary productivity, destroy refuge areas of trees and waterholes, and trample the desert crust with their hooves. Arid Australia has certainly suffered from the pastoral industry, but may be unusual in that foxes, cats, rabbits and other species have been introduced recently too. In combination, these have greatly reduced the diversity of native desert mammals, especially medium-sized ones, and slightly reduced avian diversity. Decrease in reptile richness has been documented in North America, in regions subject to overgrazing or extensive use by off-road vehicles. Nevertheless, there are indications that a limited use by herds is beneficial in certain arid regions to the diversity of lizards. Habitat alteration by humans, such as the addition of trees and housing projects in arid lands, alter species composition of the lizard fauna, but it is not clear yet if they affects lizard diversity.

The geographic variability of biological diversity is known long ago and attracted attention of many ecologists and biogeographers for several decades. The study of geographical patterns of biodiversity called into being both theoretical models (e.g. productivity-diversity relationships) and empirical rules (e.g. Rappoport’s rule). On the other hand, one of the oldest concepts in biogeography, namely a concept of biogeographic regions, is based mainly on patterns of species and higher taxa distribution. This concept now incorporates other concepts including the consideration of a province as a response to both ecological and historical processes. It is clear that geographic patterns of local biodiversity are determined by different factors. Each of these factors acts on its own spatial scale. The entire continuum of spatial scales can be to divided into three groups according with relative roles of different factors determining biodiversity. (A) Local scale refers to faunistically homogenous areas. On this scale, local biodiversity is determined by ecological factors only (e.g. habitat diversity). (B) Regional scale refers to a region that is composed of some faunistically similar local areas. Areas within a region usually differ in their geological history and, as a result, in modern landscape (and, thus, habitat) structure. Biogeographic among-area differences within a region are usually the results of vicariance of species-congenerics or species of closely related genera. Local biodiversity on this scale is determined by both historical (geological history, speciation) and ecological factors. (C) Continental scale refers to a wide area consisting of several different biogeographic regions (provinces). Fauna of each province has its own history of origin. Local biodiversity on this scale is determined mainly by historical factors (evolution). Consequently, if we want to evaluate the relative importance of ecological and historical factors in biodiversity patterns, we should restrict our consideration to regional scale.

Different aspects of biogeographic regionalization have been severely criticized. One of the main points of criticism is the problem of choice of taxa for regionalization because regionalization results may differ depending on the taxonomic group selected. Rodents and lizards were selected for our analysis because both groups are well represented in deserts and geographic distribution of these taxa is known relatively well. Comparison of schemes of regionalization based on different taxa is beyond the scope of this presentation. Our sole aim has been to denote the regions with relatively homogeneous rodent and lizard faunas (provinces) for subsequent comparisons between them.

Two possible directions of biogeographic classification can be considered. The first is a classification of geographic regions based on their faunal composition (regionalization), while the second is a classification of species accordingly to their geographic distribution (areographic analysis). In mathematical terms this is equivalent to R/Q analyses of a presence/absence matrix.

The borders of the definitions of what is a "desert" and, consequently, what species should be called "desert animals" are somewhat obscure. So, we need to confine our consideration to particular areas and to particular species. We have chosen to restrict ourselves to those areas that are defined to be "arid" and "hyper-arid". In addition, we restricted ourselves to the analyses of the arid-adapted component of deserts and not the whole species assemblages found within them.

We selected species for the whole set analyzed and assigned each species to a particular type of geographic range. First, we selected species which we consider to belong to desert faunas. The next step was to make a classification of regions. First, we divide a map of a region into quadrates 2º by 2º, and compile a list of species for each quadrate. Second, we combine adjacent quadrates that have the same species set and calculate values of Jaccard index for each pair of the resulting units. Finally, we perform cluster-analysis of these units based on the matrix of Jaccard indices. According to the results of the above analysis, the entire Great Palaearctic Desert Belt can be divided into 19 and 15 faunistically-homogenous areas (provinces) for rodents and lizards, respectively.

The species diversity of a region is largely determined by a combination of current environmental forces and of historical processes. Species richness provides a useful measure of diversity if the area considered is delimited in space and the constituent species are enumerated and identified. This is exactly the case for the zoogeographic units defined.

Patterns of distribution of species within biogeographically-homogenous spatial units are reflected by the well-known species-area relationships. These relationships are caused by increasing environmental diversity with increases in the area considered. Our data demonstrated that rodent species richness increases with the area of a province. However, some provinces are outliers in both positive and negative directions. The Central Saharan, Kashgharian, Syrian and Nubian provinces contain fewer species than expected, while the Gobian and Central Kazakhstan provinces have faunas richer than expected. As stated above, the environmental component of rodent species diversity is reflected by a linear trend, whereas the historical component is reflected by deviations of the provincial points from this trend. Each case of far outliers has its specific causes. For example, the faunal poverty of the Central Saharan province may be a result of the recent increase in its desert area; the recent fauna was originally formed in a smaller area and then dispersed within its present boundaries. The extreme richness of the two Central Asian faunas can be explained by long history of desert evolution here. A plot of the overall species richness of the faunas against the areas of the provinces reflects both the environmental and historical determinants of the faunal composition. Taking into account only those species that originated in a given province rather than the entire species pool, we can expect that the relation between this subset and the provincial areas would mainly reflect the historical component. Indeed, there is a strong positive correlation between the number of species with local origin and the area of a province. It seems that the larger provinces support more intensive speciation.

A similar trend occurs for lizard faunas. There are two groups of outliers from the trend. Three Central Asian provinces (Sungorian, Kashgharian and Gobian) have less species than expected. This can be explained by climatic constraints (extremely cold winter). Iranian and Mesopotamian provinces have more species than expected. The cause of this extreme richness can be a very complicated relief that provide intensive speciation.

The environmental component of species-area relationships is determined by an increase in environmental (=habitat) diversity with an increase in area. We attempted to present this component by plotting species richness directly against environmental diversity. We used the number of major plant associations (=vegetation types) as a measure of the latter. There is clear positive relationship between rodent species richness and environmental diversity. In addition, lizard species diversity also demonstrate positive but non-significant correlation with plant association diversity. The number of plant associations can reflect not only habitat diversity but also food resource diversity, directly for rodents and indirectly (determining insect diversity) for lizards. The latter can explain stronger relationships between animal species diversity and the number of plant associations for rodents comparing with lizards.

Patterns of species diversity (=richness) considered at a lower spatial scale (within a biogeographically-homogenous area among environmentally-homogenous units) reflect current ecological relations and are free from the influence of historical processes. To understand spatial changes in species richness, areas of equal size should be compared. For this comparison we have chosen quadrats of 2° by 2° which follow lines of latitude and longitude, a method often used in similar studies. We assume that this scale is a trade-off between two conditions, namely relative homogeneity of spatial units and the availability of faunistic data. For each quadrat, species richness of desert rodents and lizards was calculated by overlaying distribution maps of individual species on the grid map and by summing the number of species occurring in each quadrat.

The distribution of rodent and lizard species richness across Great Palaearctic Desert demonstrates several "hotspots" of high species richness. A possible explanation of this pattern of distribution is that the area of the Saharo-Gobian realm is so large that environmental diversity is higher there than anywhere else. Furthermore, this realm takes the form of a long belt, which has repeatedly been divided into several parts and later unified. As a result, different parts of the belt evolved as independent biogeographic units during certain historical phases. These processes have led to the present multiple "hotspot" distribution throughout the entire realm.

Species richness, as a measure of the structural complexity of an ecosystem, is determined to a great extent by the level of productivity. Two patterns in the relationship between species richness and the productivity of an ecosystem can be demonstrated. Both patterns have been shown to occur in deserts, namely linear and hump-shaped relationships. In the former, species richness simply increases monotonically with increasing productivity as a result of increased resource availability. In latter model, species diversity plotted against a productivity gradient will increase from low productivity until some intermediate value is reached. Thereafter, competition between species occurs so that certain species outcompete others, thereby resulting in a reduction in species richness with a further increase in productivity, thus causing a hump-shaped curve.

Productivity in deserts is directly determined by the level of annual precipitation (e.g. Southgate et al. 1996), so the rainfall can be a measure of general productivity. Most provincial rodent and lizard faunas do not show any correlation between rodent species richness and the level of precipitation. This lack of pattern may be the result of a relatively narrow range in the variation of precipitation within the boundaries of these provinces. Other provinces demonstrate significant patterns of rodent species richness along rainfall gradients (Libyan, Iranian, Turanian, Central Kazakhstan and Gobian provinces for rodents, Northern and Southern Saharan, Hadramaut, Gobian and Kashgarian provinces for lizards). For lizards, in most cases (three of five) relationships between productivity and species richness were hump-shaped. Surprisingly, in most cases for rodents and in one case for lizards this relationships were negative; that is, the higher annual precipitation the poorer the desert rodent community. However, we have only taken into consideration species that are regarded as desert-adapted. High levels of precipitation cause a transition to non-desert environments and true desert species consequently drop out from the fauna as ecosystem productivity increases, and are replaced by non-desert species. The total number of both desert and non-desert species sometimes increases toward the borders of desert zones (e.g. Abramsky and Rosenzweig 1984), but the number of desert species decreases.

The two exceptions are provided by the rodent fauna of the Libyan province and by the lizard fauna of the Hadramaut province. These two faunas display an increase in the number of desert species with increasing productivity. There is well pronounced rainfall gradient in each of the provinces and both of them have no contact with any terrestrial non-desert ecosystem being partially bounded by open sea. The geographic position of the provinces provided specific conditions for the formation of its fauna when desert animals evolved to use free ecological space that in other cases is occupied by non-desert species. These cases emphasizes the role of historical processes in producing the spatial patterns of species diversity even on a small geographic scale.

Community ecology has been characterized as the "messy middle ground" (Lawton 1999), in which systems are so complicated as to be impossibly contingent but not so large that statistical order emerges. The arguments Lawton and others before him have made for the overwhelming impact of contingency in community ecology are compelling; however the solution of simply ignoring this intermediate level is highly unappealing to those of us who find the repeatability of local patterns of distribution and abundance of organisms fascinating. I suggest the solution to the lack of strong rules perceived by many in community ecology is two fold. First, we should take a much smaller leap in scale than to entire regions and continents—rather than look only at single communities in great detail, we must sacrifice some detail within communities and compare the processes of local community organization along small-scale gradients, such as in productivity or disturbance. At this scale of comparison, the important general principle of tradeoffs comes into major play and is critical for explaining patterns in the dominant traits of species that characterize different parts of gradients. This focus on tradeoffs across environments combined with comparative experiments has been used to great success in elucidating generalizations about the local processes that determine patterns along gradients such as secondary succession for plants (review in Tilman 1990; see also Tilman and Wedin 1991a,b, Wedin and Tilman 1993) or ephemeral to permanent freshwater ponds for several vertebrate and invertebrate groups (Werner and Anholt 1993, Werner and McPeek 1994, Skelly 1995). In other words, questions in community ecology that deal with "why more (or fewer) species coexist in habitat x than y?" or "why species of type a dominate in habitat x but of type b dominate in habitat y?" are much more likely to lead to general answers about local processes of species interactions than questions such as "how do a and b coexist in habitat x?" or "why are there 10 species in habitat y?".

A second important reason for the lack of large generalizations in community ecology and the apparent highly contingent nature of processes is that community ecologists have not (for good logistical reasons) tended to use large sample sizes of species or environments within their studies nor, until very recently, to synthesize their results or observations quantitatively across studies. The large sample sizes taken from across a broad range of systems are a major source of the ability of macroecology to generate provocative general patterns (Brown 1995). However, this advantage is not a unique property of a regional or continental scale or set of questions, but of a methodological approach and there is no intrinsic reason why questions about local-scale processes in community ecology cannot be addressed with large samples by similar synthesis among observations or experiments.

In this paper, I will discuss two relatively recent developments in community ecology that are beginning to allow quantitative development and testing of general rules: 1) synthesis among experiments (data mining from the literature) to generate large sample sizes and broad comparisons that are essential for deriving generalizations and rules (Gurevitch et al. 1992, Gurevitch and Hedges 1999, Osenberg et al. 1997, Osenberg et al. 1999). 2) Development of new experimental approaches that generate information on many species (or sites) in single experiments rather than from pairs of species (whether competitors or predator-prey) at a time. I will illustrate both these general approaches by focusing on some central questions about species interactions and plant biodiversity in deserts.

Species interactions in unproductive environments

Darwin (1859) placed a very strong emphasis on the importance of biotic interactions, especially competition, as a force of natural selection in all kinds of environments and, while less explicit, as a force organizing communities of organisms.

In plants, a number of experimental and observational studies have shown that competition (as well as facilitation) does indeed often occur in deserts (review in Fowler 1986, Kadmon and Shmida 1990, Guo et al. 1998) and some have therefore considered this argument settled However, this accumulation of case studies does not address whether competition is systematically any less important in desert conditions—given the prevalence of competitive interactions in plants in general (Goldberg and Barton 1992), the mere existence of examples of competition in water-limited systems does not necessarily imply that competition is equally important as in more productive systems. To address this question quantitatively, Goldberg et al. (1999) synthesized results from published field experiments on competition in plants. Overall, we found competition intensity to vary widely with standing crop (as a surrogate for productivity) but with a weak tendency for competition to decrease with productivity—a pattern not predicted by either of the main existing hypotheses. As often happens when large volumes of ecological data are available, the patterns exists as more of a boundary constraint than a deterministic relationship (Blackburn et al. 1992, Guo et al. 1998). Here, I break down this pattern in search of more complex rules that can account for more of the variation. Based on the data available so far, the general rules for plants seem to be that, at a given productivity level, competition intensity is greater: a) for water-limited systems rather than nutrient-limited systems, b) for size and size-related fitness components (growth rate, seed production) compared to survival, c) sites that have experimentally-enhanced productivity rather than naturally-determined productivity, d) for seedlings interacting with other seedlings rather than seedlings interacting with adults or adults with adults. In addition, positive interactions are common and e) are more frequent at low productivity for size and size-related fitness components, but f) are more frequent at high productivity for survival. Some of these results are quite surprising and, while needing to be confirmed with larger sample sizes, suggest interesting hypotheses about the mechanisms of interactions among plants.

An important point about this meta-analysis, and indeed, most of the data available on competitive interactions in terrestrial plants and animals, is that they are concerned with the consequences of competition for components of individual fitness. Thus, to directly address hypotheses about patterns in the strength of competition as a selective force and the role of different processes in shaping traits of individual organisms (Darwin 1859, Grime 1977, Parsons 1993), we need to know how these different components combine to determine individual fitness (McPeek and Peckarsky 1998). This is especially important given the evidence that different components of fitness exhibit different trends in competition intensity with productivity. For example, it can be argued that survival at low resources is more critical to population persistence in desert environments where droughts are common, while growth and reproduction are more important in environments in which resources are less pulsed (Goldberg and Novoplansky 1997). If correct, then Grime’s (1973) argument that competition is more important in high productivity is consistent with the data that effects on growth are greater than effects on survival regardless of productivity.

Bibliography

Understanding landscape components of biodiversity is an obstacle that may be overcome by using relatively simple landscapes such as those of arid lands. The difficulty of exploring biodiversity at a landscape level mainly results from the increase in complexity with the increase of the spatiotemporal scales. Such complexity may come from two directions: First, with the increase of scale, diversity becomes more diverse -- there are more resources, more species of different systematic and functional groups, more variability in productivity among patches and more habitat patches that may produce more complex structures. Second, the interaction between the different diversities may give rise to additional ecological structures, hence making it more difficult to distinguish between ecological patterns and processes. Furthermore, an ecological pattern may now be a product of the scale of which it has been observed, a product of a lower scale, a product of a higher scale, or a product of the interaction between the different scales.

The advantage of arid and semi-arid lands is that they provide relatively simple landscapes with a relatively easy recognition of their components and a relatively simple interaction between their different diversities. As a result, we can use currently developed spatially-explicit realistic models that are built upon major lessons from metapopulation dynamics and landscape ecology, to test predictions of landscape-scale processes and patterns of arid lands. In this paper we will describe a new spatially-explicit, multi-species, process-based, object-oriented landscape simulation model -- SHALOM -- that has been purposely designed to explore ecological complexity of large scales. Following the description of the model, we will present several simulation results (predictions) to demonstrate the strengths of using such models for understanding biodiversity processes and patterns.

Consistent with the current landscape ecology terminology, SHALOM has physical classes and biological classes. Each class has functions and characteristics that are strongly based on ecological realism. The physical classes include: landscape, habitat, cell and patch. A landscape is a matrix of cells that define the entire area under study. It is also an abstract class that serves as the system controller. Each cell of the landscape is characterized by a single habitat. A habitat is defined as a place relatively homogeneous for physical and biological characteristics. Its physical characteristics are temperature and precipitation. We assume that temperature and precipitation play an important role in characterizing the physical environment from the point of view of the organisms. In arid lands, strong evidence suggests that temperature and water are the major limiting factors. The biological characteristics of a habitat are the resources it offers and the proportion of each of these resources out of the total productivity of the habitat. The adjacent cells sharing a habitat type form a patch. The patch-specific characteristics include: productivity, energy supply and resource-proportion energy supply. Productivity linearly correlates with temperature and precipitation. The area of the patch is the sum of the areas of its cells. Energy supply is the amount of energy per unit time for the entire area of the patch. Resource-proportion energy supply is the amount of energy per unit time offered by each resource represented in the patch.

The biological classes include: population, species and community. A species is the set of individuals in the landscape that share biological and physical characteristics. Each species has a characteristic body size, habitat utilization, resource utilization, and a dispersal coefficient. Body size allows for using allometric relations with other characteristics such as birth rate, death rate and metabolic rate. Habitat utilization and resource utilization resemble the physical and biological characteristics of a habitat. Thus the model compares what is offered by a patch with what is required by a species in it. Habitat utilization is defined by the temperature and precipitation requirements which also determine the species' niche. Resources are distributed discretely. The resource-proportion use is the fraction of the use of each resource relative to the species total resource use. The dispersal coefficient of a species determines the intensity of dispersal when and if such is invoked. A population is the group of individuals of a species occurring in a particular patch. All breeding occurs within a patch's borders. Hence, a species is a metapopulation. Some characteristics specifically characterize the requirements and pressures a particular population faces in each patch. The community is the set of non-zero populations in a patch.

Processes of SHALOM are modeled on local and global scales. At the local scale, populations grow continuously with allometric birth and death rates, and are affected by:

We conducted two simulation studies to ask questions regarding 1) the effect of process interaction on body-size distribution of mammals in heterogeneous environment, and 2) the effect of habitat generality of larger species on large patterns of species diversity.

We simulated a simple 4-patch landscape, each patch having its unique habitat, with 26 species that differ only in body size, ranging between 5 to 1585 g. We used allometric values of Eutherian mammals for the simulated species. For precipitation and temperature we used values similar to those of the semi-arid lands in Israel.

The results show that interspecific competition alone reduces species diversity in each patch and in the entire landscape. Stochasticity depresses mean population sizes, and open opportunities for species to avoid competitive exclusion. Stochasticity also allows different patches to have different species composition, which is determined by which large species becomes locally extinct at random. With both demographic and catastrophic stochasticities, overall species diversity is similar to the overall species diversity of interspecific competition alone. However, demographic and catastrophic stochasticities differ in their effects on species diversity. We used the effects by each stochasticity as their "signatures" on community structure. Dispersing individuals move between patches and reestablish the local populations of their species. Thus, dispersal neutralizes the randomness of the assemblages produced by stochasticity.

We also simulated a 16-patch landscape, each patch having its unique habitat. In this simulation, not all habitats were suitable for all species. The results show that body size and species abundance have a log-normal relationship: larger species have lower abundance because their populations have lower sizes everywhere, while smaller species have a lower abundance because they occupy only a fraction of the available patches (habitats) in the landscape. Additionally, the results show that geographic range increases non-linearly with log body size: larger species can persist in a higher diversity of patches (habitats) and hence being able to occupy more habitats at the periphery. In general, the two patterns are highly consistent with those observed in natural systems. However, here they emerge as predictions of a theoretical simulation model.

In sum, the model represents a synthetic approach that provides ways to explore high-level ecological complexity and suggests predictions for future studies of macroecological biodiversity questions. Furthermore, SHALOM, with its functions and procedures, opens new opportunities to study combined ecosystem, community and population processes. Within the variety of existing ecosystems, arid and semi-arid lands should be the first candidates for applying the model’s approach because they provide relatively simple systems at a large ecological scale.

Hutchinson (1959) posed the problem best, "Why are there so many types of animals?" We can extend this to ask why there are so many types of species. What does this question mean? In its broadest sense, we want to understand the forces and processes that have led to the current total number of species on Earth in general and in arid zones in particular. But total species diversity is simply the sum of regional diversity, and regional diversity is largely the sum of local diversity. So we first want to understand what determines local and regional diversity. We focus on the role of environmental heterogeneity in the guise of various resources, the physical structure of the environment, and the spatial relationship of habitats in the landscape. To this we add the role of species interactions as determined by the various attributes of the coexisting species.

The problem outlined above is not unique to arid environments, but certain aspects are. In particular, the resource most likely to be limiting in deserts is water. Organisms can be directly limited by water, as is the case in many plant species. Or, the effects of water may be indirect through its effect on the types and amounts of resources produced by one organism that are exploited by another or in structural aspects of various habitats produced by plant cover. The seeds produced by annual plants and eaten by granivorous rodents is one such example. Bush/open microhabitat selection that promotes coexistence in desert rodents in the Great Basin Desert is another.

Many possible approaches can be used in the study of species diversity, ranging from marcoecological approaches to mechanistic ones. We advocate an evolutionary approach in which biodiversity is viewed as an outcome of mechanisms of species coexistence based on Darwinian and Malthusian laws. These processes operate at several scales (foraging: the area visited by a forager in a single foraging bout; intermediate: the area visited by a forager over several foraging bouts, i.e., its home range; landscape, the area over which individuals can move during dispersal; regional: the area over which a population can receive immigrants or their descendants), and optimal behavior (adaptive responses in foraging and dispersal to environmental heterogeneity) of individuals is the thread tying it all together.

We propose that biodiversity is largely the sum of local species coexistence summed over larger areas. Thus, to understand biodiversity requires an understanding of mechanisms of species coexistence. A mechanisms of species coexistence consists of two key ingredients. First, there must be some sort of environmental variability, a niche axis along which species can segregate. This variability occurs at a foraging scale and can vary from differences in predatory risk according to habitat structure to differences in food types according to encounter rates, handling times, or gut passage time. Second, there must be an evolutionary tradeoff among species such that each has a part of the niche axis at which it is superior to any of its competitors. This commonly occurs when each has a set of environmental conditions in which it is the most efficient species in its community. Efficiency can be measured in practice by applying foraging theory to measure giving-up densities (GUDs: the density of resource left in a resource patch at the optimal quitting harvest rate; Brown 1988) in resource patches. This is because an efficiency is the ratio of output to input, and the GUD occurs at the resource density where the output from foraging (i.e., the harvest rates) equals the input (i.e., the foraging costs); the species with the lowest GUD is the most efficient and can deplete resources under those conditions to the point where competitors cannot profit.

One can imagine any number of possible mechanisms of species coexistence, but they mostly fall into mechanisms involving habitat selection in time, habitat selection in space, or resource partitioning. We illustrate some of these mechanisms using desert granivores.

Mechanisms involving habitat selection in space include habitat selection in a mosaic of habitat types, microhabitat selection (typically bush/open), and spatial variation in resource abundance among patches. Gerbils and larks in the Negev Desert of Israel coexist via habitat selection across a substrate mosaic of sand, rock, and loess. Gerbils are more efficient and have lower GUDs on sand; larks are more efficient and have lower GUDs on loess. Escape substrate (substrate surrounding a foraging patch over which a forager must flee to reach refuge; can affect likelihood of escaping a predator attack and thus the cost of foraging due to predatory risk) and foraging substrate (substrate on which or in which food resources are to be found; can affect harvest rates and energetic costs of foraging) effects contribute to each species' success on its best substrate. The availability of water increases the efficiency of the diurnal larks, but not the nocturnal rodents, possibly because water loss in diurnal foraging causes larks to pay to be active where nocturnal animals do not. Desert rodents on Great Basin Desert sand dunes coexist via bush/open microhabitat partitioning. Large kangaroo rat species with their anti-predator morphology are relatively more efficient in the risky, open microhabitat, while smaller, more energetically efficient deermice and pocket mice are more efficient in the safety of the shrubs. Spatial variability in resource abundance underlies coexistence between a kangaroo rat and a ground squirrel in a Sonoran Desert creosote flat. The relatively small size and lower metabolic demands of the kangaroo rat allow it to exploit resource patches more efficiently, while the larger size and high travel speeds of the ground squirrel allow it to travel faster and more cheaply between rich patches.

Mechanisms involving habitat selection in time include coexistence on a resource whose abundance varies in time and coexistence based on seasonal variability in foraging costs. Two mechanisms involve temporal variability in resource abundance, one in which resource abundance varies on a seasonal time scale, and one on daily time scale. In a semi-arid Kalahari savannah, coexistence between a gerbil and a striped mouse may involve the foraging efficiency of the gerbil and the ability of the striped mouse to reproduce at a high rate following a pulse of productivity. At the Sonoran Desert creosote site, the kangaroo rat species coexists with a pocket mouse and a ground squirrel via seasonal variation in foraging costs brought about by seasonal changes in predatory risk. Changes in risk are due to seasonal activity of rattle snakes, population density changes in owls, and migratory behavior of hawks. Changes in abundance of each type of predator has disproportional effects on the risk experienced by different rodent species. This leads to a seasonal rotation of foraging efficiency where each species has a time of the year where it is most efficient. In the Negev Desert, coexistence between two sand dune gerbils involves daily pulses of seed renewal and temporal partitioning. The two gerbils differ in size, and the large species can monopolize rich patches early in the night using interference, while the small species can profitably exploit the leftovers later on.

Mechanisms involving resource partitioning include diet partitioning according to food type or food particle size and diet partitioning based on the complementarity of food and water. Desert animals that coexist by consuming seeds versus insects or seeds versus leaves are common place. Deermice and grasshopper mice in the deserts of North America may be one such example. Seed size partitioning among coexisting rodent species has often been hypothesized, but support for this mechanism has not been found. Finally, the foraging efficiency of granivorous birds have been shown to increase in the presence of water by as much as double. This complementarity of food and water may allow species that are not typically found in desert to invade and eliminate desert species when water is provided by human activities. This creates a serious conservation problem in deserts. However, we have no evidence that such a mechanism actually promotes coexistence.

These examples illustrate several points. First, several mechanisms can operate within each community. Which mechanisms operate depends on the range of environmental variability occurring at the site and the various characteristics of the species in the species pool. Second, a given species can be involved in more than one mechanism of species coexistence. This can be true both within a community with different competitors and in different communities. The case of Dipodomys merriami illustrates this point well. At the Sonoran desert site, it coexists with its competitors via two different mechanisms, and it coexists by yet a third mechanism at the Great Basin Desert dune site. Third, the role of habitat structure and resource diversity affects species diversity by providing the axes of environmental heterogeneity necessary for species coexistence. Habitat structure in particular can provide variability in predatory risk along which species can segregate.

At a more intermediate scale, habitat selection takes on added importance, and species diversity becomes strongly dependent on habitat diversity. At the landscape scale, optimal foraging decisions give way to decisions of optimal dispersal and settlement. Species that are competitors at a foraging scale within patches can be mutualists at a landscape scale providing knowledge of habitat quality can be obtained from the presence of competitors in a candidate site for settlement. Such effects can decrease alpha diversity while increasing beta diversity. In addition, source-sink relationships between habitat patches across a landscape can act to increase local species diversity by maintaining species in local communities largely by immigration. Thus, we can expect increased landscape diversity to lead to increased species diversity within communities. To the extent that complementary or essential resources are available in different habitat patches, landscape diversity can actually add to beta diversity, too. More regional effects such as the storage effect can also increase beta diversity.

Our discussion of species diversity to this point has been from an evolutionary point of view in that we have assumed that individuals make optimal foraging decisions. However, to the extent that evolution occurs in the context of a particular community, communities may represent coevolved assemblages. As such, a community may be the result not just of species sorting according to mechanisms of species coexistence as described above, but may actually be an evolutionary stable strategy or coalition. ESS limit to species diversity can be set by aspects of the environment such as temperature and productivity in conjunction with the fitness functions of the organisms.

Species interactions in the context of food webs can also yield insights into species diversity. Water limitations to productivity can limit the number of controlling trophic levels found in arid ecosystems. Consequently, we can expect arid ecosystems to be more similar to arctic systems than to ecosystems in more mesic areas. Also, optimal behavior on the part of both predators and prey can be expected to modify the strength both of predator-prey interactions and competitive interactions. This will affect the length and complexity of food webs.

Finally, approaches to understanding species diversity based on foraging theory can be applied to problems of conservation and management of biodiversity. Tools such as habitat suitability indices based on the theories of optimal patch use and optima habitat selection and management plans for ecological communities based on mechanisms of species coexistence can be used in the conservation of biodiversity.

We conclude that biodiversity can best be understood by applying Darwinian and Malthusian principals. Problems regarding biodiversity, then, can be addressed and solved by using tools based on these.

Substantial work over the last 5 decades has focused on the ecological and the genetical differences between populations of the same species across their distributional ranges. A major subject of investigation has been the difference in levels of genetic diversity among "core" and among "peripheral" populations, with special attention paid to the unique ecology and genetic structure of isolated peripheral populations . Two major hypotheses have been suggested regarding the levels of genetic diversity in peripheral and core populations . The rationale of each of these hypotheses is based mainly on the interaction between local selection and gene flow. According to the "Carson hypothesis," the large, continuous core populations will maintain higher genetic diversity as the result of balancing selection in contrast with peripheral populations that are usually smaller and more isolated (Carson 1959). Higher genetic diversity in core populations was also suggested in later models by and . Mayr (1963, 1970) proposed that only a relatively limited number of genotypes can thrive under the harsher environmental conditions found at the periphery. In addition, he argued that while gene flow is expected to be multidirectional among core populations it would be mostly unidirectional from core to peripheral populations (Mayr 1963). This would contribute to increased genetic drift and subsequently to a limited genetic variation at the periphery of the range. The dynamics of local selection and limited gene flow would both contribute to reduced genetic diversity at the range periphery. However, in some cases gene flow from the core may compensate for the effects of local selection and genetic drift at the periphery. In such cases genetic diversity may actually be homogenous throughout the species’ range .

According to the alternative hypothesis, termed the "Fisher hypothesis" , core populations undergo stabilizing selection and maintain relatively low genetic diversity relative to peripheral populations, which are subject to diversity-enhancing fluctuating selection (Fisher 1930). Fluctuation of selection forces is assumed to be stronger at the periphery, where the environment is more heterogeneous in both space and time.

Most natural ecosystems are subject to increasing anthropomorphic influences, affecting population structure, dynamics, abd directions of selection. While most biodiversity studies are focused on the community and the ecosystem levels, the within-species and within-population genetic structure are often neglected (e.g. . This research is crucial for enabling conservation and rehabilitation attempts of specific economically-important populations of wild game animal, fodder and medicinal plants or wild progenitors of crop cultivars. Indeed, the UN Rio de Janeiro conference specifically called for the conservation of genetic diversity within species as well as at the interspecific and ecosystem scale (United Nations Conference on Environment and Development 1992).

In this chapter we review terminology, methodology, and empirical findings related to the ecological and the genetic differences between core and peripheral populations. In different studies the genetic diversity is estimated by quantitative genetics of morphological traits and by allozymic diversity. We then discuss the findings in light of the geographical and ecological changes along the clines and present some consequences these findings have for conservation efforts of natural populations.

Definitions

A large number of geographical, ecological and evolutionary terms are employed to differentiate between peripheral and more central populations of a given species. Among the most popular are categorizations such as "periphery-core" and "margin-center" . Even such "simple" definitions may become tricky when the shape of the distribution range is asymmetrical and complex, as is often the case in natural populations. In this chapter we will follow the terminology of Brussard (1984) who distinguished between geographical and ecological definitions: he used the terms "core" and "periphery" to describe the position within a distribution area, and terms such as "central", "optimal", "favorable" and "marginal" when referring to environmental factors. According to these definitions, a habitat is classified as favorable or marginal according to the relative numbers of individuals it can maintain. The discrepancy between the geographical and the ecological definitions is exemplified in one of the earlier studies on Trimerotropis sparsa in which the geometrical center of the distribution range was found to be less favorable than peripheral parts of the range, where the species might not occur at all (White 1951). For the purpose of the current work we only relate to the geographical ranges of the species distributions.

Early studies

Early work done in the 50’s yielded contradictory results. One of the earliest studies was performed by da Cunha and Dobzhansky (1954), who compared chromosomal polymorphism in core vs. peripheral populations of Drosophila. According to their working hypothesis, the amount of adaptive polymorphism carried in a population is a function of the variety of the ecological niches it exploits. They found that core Drosophila willistoni populations were highly polymorphic relative to those at the periphery, where the species was less common and less ubiquitous than its competitors. They interpreted this result in a ‘Carsonian’ fashion, as in this case the center habitats were both more rich and more diverse . White (1951), however, found no diminution of chromosomal variability towards the distribution periphery , and other early studies found an increase in genetic diversity at the periphery of the range (recently summarized by . The topic continues to draw investigation, but no clear pattern has emerged . Reviewing the case of protein electrophoretic diversity, Parsons (1989, p.43) notes that: "...variability levels in central vs. marginal populations have revealed a rather confused situation. For an endangered fish, Poeciliopsis occidentalis, in Arizona, geographically peripheral populations show less electrophoretic variation than do central populations. In contrast, some Drosophila populations show higher electrophoretic variability at the margins...Hence, comparisons of electrophoretic variability under differing ecological circumstances must be approached with extreme caution."

Recent work on core-peripheral clines in Israel

The steep environmental gradients of Israel provide an excellent arena for testing hypotheses on genetic diversity across core-periphery clines. Located at the confluence of three continents and four biomes (Mediterranean, desert, steppe, and sub-tropical African), Israel encompasses steep climatic and ecological gradients within a small geographical range . A sharp gradient occurs from the Mediterranean areas of the north, where mean annual rainfall is over 500 mm (up to 1000 mm), to the arid regions of the southern Negev desert, where mean annual rainfall drops below 50 mm and where variability in rainfall on both temporal and spatial scales is high . A steep part of this north-south climatic gradient is a 60 km-wide transition belt at the northern margins of the Negev desert, over which rainfall decreases several-fold from over 450 to less than 150 mm. This zone represents an "ecotone" between the Mediterranean steppe and the desert where a relatively sharp rainfall and floristic gradient occurs and where various plant and animal species reach the margin of their global distribution . Other species continue their distributions further into the desert or the Mediterranean zones, with small isolated populations at the extreme periphery of their range. These sharp natural gradients provide a unique opportunity to examine diversity trends from core to periphery in populations which are geographically proximal, and thus potentially connected by gene flow, but which experience contrasting environments, selection pressures, and population dynamics.

In the chapter we review recent studies from different systematic groups that test hypotheses on diversity trends across this cline. We look at detailed studies of quantitative genetics of an annual legume (Trifolium purpureum) and a perennial clonal grass (Dactylis glomerata) (Danino et al., in preparation), morphological and allozyme diversity of Hordeum spontaneum, the wild progenitor of cultivated barley (Volis et al. 1998), and the morphological and electrophoretic diversity of a game bird, the chukar partridge (Alectoris chukar) . These species all have high densities in the Mediterranean region of Israel, and their populations become smaller and more isolated towards the arid periphery of the Negev.

The chukar partridge

Genetic diversity

Chukars were sampled from the Mediterranean core to the northern Negev ecotone (i.e. area of transition between Mediterranean and desert ecosystems), which comprises the edge of the species’ continuous distribution. Allozyme diversity of 32 allozyme loci, was determined for birds collected in five locations in 1990 and 1993. Genetic diversity, as estimated by the percentage of polymorphic loci, mean number of alleles, and observed and expected heterozygosity increased from the core to the ecotone. Single and multi-loci Hardy-Weinberg and linkage disequilibria increased significantly from populations located in the Mediterranean region to those at the ecotone, despite the close geographical proximity of these two sources. Populations exhibited some isolation by distance effects when subjected to substantial gene flow. Peak diversity was found in the intermediate ecotone area, located in-between the extreme periphery of the range and the core. This area supported the highest overall genetic diversity as well as highest levels of allelic endemism.

Morphological diversity

Morphological diversity was compared in five chukar populations in Israel and the Sinai Peninsula across the species range from the Mediterranean core, through the ecotone and towards the extreme periphery of the range. Rachel Nissani collected the birds in the 1970’s in order to test for Bergmann’s Rule. Each bird was measured for 35 morphological traits and 23 ratios between traits were calculated. We analyzed these data to discern differences in phenotypic diversity based on the coefficient of variation. Using a specially developed statistical estimator, termed "Estimator in Dependent Sample," diversity level in each of the study populations was estimated . Peak diversity was found in the population from the ecotone for both males and females in all traits tested. Diversity showed a hump-shaped pattern across the range, decreasing from the ecotone towards both the Mediterranean core and the extreme desert periphery of the chukar range.

Overall, both genetic and morphological diversity showed peak levels in chukar populations located at the Mediterranean-desert ecotone. We recommend that substantial research and conservation efforts be devoted to this area, located in-between the core and periphery of the range. Rapid urbanization of Israeli landscapes, and specifically of the ecotone region, threatens to disrupt unique genetic structures and essential genetic connections among chukar populations.

The purple clover (Trifolium purpureum) and the common dactyl (Dactylis Glomerata)

Plants were collected at 3 Mediterranean (core) and 4 semi-arid (periphery) populations in Israel. Phenotypic variability of naturally-grown mother plants was assessed on 8-10 morphological and life history traits. Plants were grown under favorable (H) and limiting (L) water availability for a full season in the ecological growth facility in Sede-Boqer, and were subsequently harvested and evaluated for response to water availability. The results were used to evaluate differences between core and peripheral populations in components of genetic variability, reactions to water availability, and heritability of growth behavior.

On-site parental plant traits were highly variable for both species at all sites, and they differed markedly between sites within both the core and the periphery. Regions differed significantly only for a limited number of traits. The effect of the site (i.e. population) was highly significant for most traits under both water treatments, whether calculated across regions or separately for each region (with one exception).

Generally, the differences in geographically-related factors between the core and the periphery sites were not larger than those between geographically proximal sites. Moreover, significant GxE interaction was observed in most cases. The large differences found between the sampled populations suggest that each and every one of the studied populations may represent a unique collection of genotypes. It is possible that the Israeli populations of the studied plants may all represent the "periphery" of their species distribution; in such a case more "genuine core" populations of these species may be found at the more temperate regions of central Europe (Dactylis) and the more mountainous regions of Asia (Trifolium). Alternatively, tracking down fine-scale core-peripheral differences in a small geographical region like Israel may necessitate sampling a much larger number of populations along their steep ecological transition region. Such a study might overcome intense background "noise" in the environment imposed by geology, disturbance and anthropogenic factors.

Wild barley (Hordeum spontaneum)

The geographical range of the wild barley (Hordeum spontaneum) follows the distribution of open park forests and herbaceous formations of the Mediterranean and the Irano-Turanian regions. The distribution of H. spontaneum is restricted to areas with altitude (< 1500), rainfall (>100 mm) and mean winter temperatures (> -5° C) and its primary habitats extend into the peripheral loess and sand deserts, where the plants are restricted to runoff enriched washes and ravines.

From populations sampled in the Mediterranean and the Irano-Turanian regions 6 peripheral and 12 core populations were compared for phenotypic variability (15 traits) and phenology (3 traits) after plants were grown under higher (H) and low (L) water availability in an experimental field at Sede Boqer. Amount of variation in phenotypic traits expressed through coefficient of variation was significantly higher in peripheral populations as compared with core populations under H, but no apparent difference could be detected under L. Using Principal Component Analysis peripheral populations were found to be more distinct from each other than core populations. Significant interactions (core/periphery) x (water) were detected for 8 morphological traits with significantly larger response to water stress in core populations. The results indicate that core populations of the wild barley are less genetically variable and more similar to each other than peripheral populations.

Implications for the conservation of biodiversity

While the study of genetic and phenotypic diversity in core and peripheral populations originally focused on speciation processes, its implications today are highly relevant in the context of human-induced environmental changes. These changes are affecting species distributions, persistence, and levels of diversity. In this chapter we consider the distribution of genetic diversity across species ranges, particularly across ecological gradients, and its implications for the conservation of biodiversity. We discuss the above mentioned evidence regarding trends in diversity across core-periphery clines. Based on the premise that populations with higher genetic diversity are better adapted or pre-adapted to environmental change we suggest in the chapter specific measures for the conservation of biodiversity across core-periphery clines, especially those clines most threatened by Global Climatic and Environmental Changes.

Bibliography
Bitan, A., and S. Rubin, 1991 Climatic atlas of Israel for physical and environmental planning and design. Ministry of Transport.

Brown, J.H., D. W. Mehlman, and G. C. Stevens, 1995 Spatial variation in abundance. Ecology 76: 2028-2043

Brussard, P. F., 1984 Geographic patterns and environmental gradients: the central-marginal model in Drosophila revisited. Annual Review of Ecology and Systematics 15: 25-64.

Carson, H. L., 1959 Genetic conditions that promote or retard the formation of species. Cold Spring Harbor Symposium Quantitative Biology 24: 87-103.

Da Cunha, A. B., and T. Dobzhansky, 1954 A further study of chromosomal polymorphism in Drosophila willistoni in its relation to the environment. Evolution 8: 119-134.

Fisher, R. A., 1930 The Genetical Theory of Natural Selection. Clarendon Press, Oxford.

Hoffmann, A. A., and P. A. Parsons, 1991 Evolutionary Genetics and Environmental Stress. Oxford University Press, Oxford.

Kadmon, R., and A. Danin, 1997 Floristic variation in Israel: a GIS analysis. Flora 192: 341-345.

Kark, S., P. U. Alkon, U. N. Safriel and E. Randi, 1999 Genetic diversity across an ecological gradient: conservation priorities for the chukar partridge in Israel. Conservation Biology 13: 542-552.

Lesica, P., and F. W. Allendorf, 1995 When are peripheral populations valuable for conservation? Conservation Biology 9: 753-760.

Lewontin, R. C., 1974 The genetic basis of evolutionary change. Colombia University Press, NY.

Mayr, E., 1963 Animal Species and Evolution. Belknap Press/Harvard University Press, Cambridge MA.

Mayr, E., 1970 Populations, Species, and Evolution. Belknap Press, Cambridge MA.

Nevo, E., 1988 Genetic diversity in nature - patterns and theory, pp. 217-246 in Evolutionary Biology, edited by M. K. Hecht, and B. Wallace. Plenum Press, New York and London.

Parsons, P. A., 1989     Environmental stresses and conservation of natural populations. Annual Review of Ecology and Systematics 20: 29-49.

Parsons, P. A., 1990 The metabolic cost of multiple environmental stresses: implications for climatic change and conservation. Trends in Ecology and Evolution 5: 315-317.

Parsons, P. A., 1991 Evolutionary rates: stress and species boundaries. Annual Review of Ecology and Systematics 22: 1-18.

Safriel, U. N., S. Volis and S. Kark, 1994 Core and peripheral populations, and Global Climate Change. Israel Journal of Plant Sciences 42: 331-345.

Volis, S., S. Mendlinger, L. Olsvig-Whittaker, U. N. Safriel and N. Orlovsky, 1998 Phenotypic variation and stress resistance in core and peripheral populations of Hordeum spontaneum. Biodiversity and Conservation 7: 799-813.

Vrijenhoek, R. C., M. E. Douglas and G. K. Meffe, 1985  Conservation genetics of endangered fish populations in Arizona. Science 229: 400-402.

White, M. J. D., 1951 Structural heterozygosity in natural populations of the grasshopper Trimerotropis sparsa. Evolution 5: 376-394.

Yom-Tov, Y., and E. Tchernov, 1988 The zoogeography of Israel. Dr W. Junk Publ., Dordrecht.

Introduction

For the purposes of this conference, biodiversity is defined as "the full range of variety and variability within and among living organisms, their associations, and habitat-oriented ecological complexes. The term encompasses ecosystem, species, and landscape as well as intraspecific (genetic) levels of diversity (Fielder and Jain 1992)". Microorganisms are important to the three components of biodiversity included in this definition. They represent a huge and relatively unexplored source of organismal biodiversity, they play a central role in ecosystem-scale nutrient cycling functions, and they contribute to the structural and functional differences between patch types that create landscape scale biodiversity. In this chapter, we present an overview of the role that microbes play in deserts and discuss how diversity at organism, ecosystem and landscape scales influences this role.

What do microbes "do" in deserts?

Microbes carry out a group of functions common to all ecosystems (Groffman and Bohlen, 1999, Table 1). The most universal of these functions is the degradation of organic matter. In terrestrial ecosystems, including deserts, more than 50% of the organic material fixed by primary producers flows to the detrital, rather than trophic food chain (Odum 1983). As organic matter is processed by biota, nutrients are released to the environment and become available for recycling back to primary producers (Figure 1). Organic residues from decomposition become part of the stable pool of soil organic matter that plays an important role in moderating soil physical and chemical conditions (Paul and Clark 1996). Nutrient cycling is perhaps the most well studied microbial process and will be the focus of much of the discussion in this paper.

In addition to the degradation of organic matter and nutrient cycling, microorganisms carry out other functions that are important at micro, ecosystem, landscape, and global scales (Coleman et al. 1992). These functions include the production of a variety of "trace" gases (CO₂, N₂O, CH_4, CO, S gases) that influence the chemistry and physics of the atmosphere (Mooney et al. 1987), processing water, soil, and air pollutants, and maintenance of the physical structure of the soil.

Although the functions listed in Table 1 are carried out by microorganisms in all ecosystems, there is great variation in the ways that they are expressed and regulated in different ecosystems. There is also great variation in the importance of biodiversity as a regulator of these functions in different ecosystems. Deserts are ideal for exploring relationships between biodiversity and microbial function because they are easy to study and manipulate at organismal, ecosystem and landscape scales. Critical questions that are readily addressed in deserts, and that we will explore in this paper include:

0. Do extreme physical constraints (e.g., low rainfall) limit microbial biodiversity in ecosystems?
1. Does the presence of high complexity/biodiversity of nutrient cycles at ecosystem and landscape scales facilitate nutrient cycling in the presence of extreme physical constraints?
2. How does such complexity influence the nature and extent of nutrient loss at ecosystem and landscape scales?

Soil microbial diversity at the organismal scale in deserts

Interest in soil microbial biodiversity has increased in recent years due to recognition of the vast number of species in these habitats (Giller 1996, Kennedy and Gewin 1997, Groffman 1997). However, there has been little analysis of the factors that regulate soil biodiversity and of the role that diversity plays in soil functions (Schimel 1995, Beare et al. 1995, Freckman et al. 1997, Brussard et al. 1997, Groffman and Bohlen 1999).

A major question in soil ecology is if extreme physical constraints (e.g., low rainfall, extreme cold) limit biodiversity and soil ecosystem function. Our work in the Negev suggests that this is not the case. Much of the microbial activity that occurs in deserts is concentrated in brief periods of a high activity following wetting events (Parker et al. 1983, Schlesinger et al. 1987, Peterjohn 1991, Fliebach et al. 1994, Zaady et al. 1996a,b). During these periods, conditions (warm, wet) are nearly optimal for microbial activity. Our data show that the intensity of soil respiration during these pulses is high enough to deplete soil O₂ concentration, providing the anaerobic conditions required for denitrification. We hypothesize that anaerobic respiration processes like denitrification are actually more important in arid ecosystems than in more temperate areas (Peterjohn and Schlesinger 1990, Groffman et al. 1993, Frank and Groffman 1998).

These results are interesting because they suggest that low rainfall does not prevent the presence of an entire class of microorganisms (anaerobes) and that the presence of this group is important to the function of the soil ecosystem. It is important to note that we have not measured true organismal diversity in these studies. It would be interesting to evaluate the organismal scale diversity of desert soils relative to more temperate zone soils (e.g., is denitrifier diversity higher in temperate soils). It will also be interesting to explore the relationship between soil biodiversity and ecosystem function in deserts. The presence of anaerobic respirers like denitrifiers may be critical to the function of desert ecosystems, maximizing the extent of nutrient cycling activity that can occur during the brief periods of water availability following rain events. On the other hand, the presence of denitrifiers fosters gaseous N loss from these systems, which may be a critical constraint to primary production in these ecosystems.

In addition to questions about physical constraints, deserts are an excellent venue for asking questions about relationships between plant, ecosystem and microbial diversity. For example, in the Negev, plant diversity is extraordinarily high (Evenari 1985). High plant diversity likely creates a high diversity of organic matter quality that may lead to high microbial diversity. Negev desert ecosystems also have a large number of ecosystem components and complex nutrient cycling patterns (see discussion below). This ecosystem-scale diversity may also lead to high microbial diversity. Application of new tools to quantify microbial diversity (Jansson and Prosser 1997, Amman and Kuhl 1998) to Negev ecosystems may shed light on these fundamental questions.

Ecosystem diversity and microbial processes in deserts

In its simplest form, nutrient cyclinginvolves the movement of nutrients between plants and inorganic forms in the soil (Figure 1). However, as shown in Figure 1, ecosystem biodiversity (i.e. the diversity of habitat-oriented ecological complexes) can considerably complicate nutrient cycles. The presence of different plants, that produce different amounts and types of organic matter (Pastor et al. 1984, Scott and Binkley 1997, Finzi et al. 1998) and the importance of fauna (Beare et al. 1992) can vary greatly from place to place. The importance of ecosystem diversity to the maintenance of nutrient cycling activity under extreme conditions and to nutrient conservation are topics of extreme interest in ecology (Tilman 1998).

For example, snails have long been recognized as an important component of Negev desert ecosystems (Shachak et al. 1987, Jones and Shachak 1990, 1994). Due to their high density and high feeding rates they produce large amounts of feces, which greatly facilitate the cycling of nutrients from plants back to inorganic soil pools, creating a pool of nutrients that is readily mobilized when a wetting event occurs (Zaady et al. 1996a). However, this mobilizable pool of nutrients is subject to gaseous and leaching loss. Thus snails have both positive and negative effects on ecosystem nutrient cycling and retention. Other studies that we have carried out (Zaady et al. 1996b) show that plant litter acts as an important sink for nutrients released during wetting events, possibly counteracting some of the effects of snail feces, which often mixed in with litter.

The snail example suggests that arid conditions do not inherently inhibit the development of high ecosystem diversity and the emergence of complex nutrient cycling activity. Indeed, complex nutrient cycles may greatly facilitate nutrient availability and plant growth under arid conditions. However, these complex nutrient cycles may also stimulate nutrient loss. In all ecosystems, there is an inherent tension between processes that facilitate nutrient cycling and availability and those that act to prevent nutrient loss (Likens and Bormann 1979, Vitousek 1998). Determining if and how ecosystem-scale biodiversity affects this balance is not clear, and represents an important challenge for ecology over the next decade or so.

Landscape diversity and microbial processes in deserts

Many desert areas are characterized by high landscape diversity, i.e. marked patchiness in the distribution of vegetation (Barth and Klemmedson 1978). In the Negev Desert of Israel, there are two main types of patches (Figure 2, Shachak et al. 1993, Shachak and Boeken 1994); 1) macrophytic patches consisting of herbs and shrubs on slightly raised mounds (West 1989, Schlesinger et al. 1990, Allen 1991) and 2) microphytic patches consisting of algae, cyanobacteria, lichens and mosses that are characterized by the presence of a relatively impermeable soil crust (Friedmann and Galun 1974, Skujins 1984, West 1990, Zaady and Shachak 1994). Plant growth in the macrophytic patches is thought to be strongly limited by N (Evenari 1985), while N fixing cyanobacteria are thought to contribute significant amounts of N to the microphytic patches (West 1990, Evans and Ehleringer 1993). Runoff following rainfall events may transfer N from the crust-covered microphytic patches to the macrophytic patches (Zaady and Shachak 1994).

There is a clear sense that landscape diversity contributes strongly to overall productivity and nutrient retention in deserts. For example in the Negev, there is a strong assumption that both major patch types are important components of the landscape nutrient cycle. A major area of research is to determine the optimal juxtaposition and abundance of different patch types for specific ecosystem restoration objectives. However, some of our research raises questions about the importance of the soil crust areas to nutrient cycling. While there is a general assumption that the crust areas fix N at relatively high rates, and that this N contributes to productivity in the macrophytic mounds, we have found equal or higher rates of N fixation in the mounds than in the crusts (Zaady et al. 1998). Moreover, there is no direct evidence that N from crusts moves into the mounds. There is a clear need for more specific evaluation of the functional role of the different patch types to determine if landscape biodiversity really is important to water and nutrient cycling functions in deserts.

Bibliography

_______________________________________________________

---

 
 

Figure 1. Conceptual diagram of a nutrient cycle in a Negev desert ecosystem highlighting the importance of plant diversity, animals as regulators of the transformation of organic matter and the importance of microbes (MO) as producers of inorganic nutrients.
 
 
 
 

Figure 2. Conceptual diagram of the Negev desert landscape showing macrophytic patches dominated by shrubs and microphytic patches dominated by algae, cyanobacteria, lichens, mosses and soil crust. From Zaady et al. (1998).

Relating Biodiversity and Ecosystem Function in Dryland Systems

Theory as it pertains to the relationship between productivity and diversity is an excellent example of a body of knowledge rich in terms of the number of mechanistic or phenomenological hypotheses proposed to affect patterns (21 according to the assessment of Mittelbach et al., in litt.). It is equally rich with regard to the number of observational and manipulative experiments that have been conducted to confirm patterns or infer causal mechanisms. At the same time, results from simulation experiments are available to understand the way in which particular mechanisms should affect patterns in natural systems (see Rosenzweig 1995).

In our assessment, the theory relating diversity to productivity is currently at the consolidation stage of development (sensu Pickett et al. 1994). Most, or all, of the components of theory are in place, but their refinement and interconnections are not sufficient. This has resulted in a perplexing and contentious literature that does not instill confidence in the application of theory. Improvements in empirical studies relating productivity and diversity, as well as incisive testing of hypotheses, are the initial steps towards consolidating this body of theory.

Part of the confusion regarding the relationships between diversity and productivity may derive from (1) a general failure to recognize that patterns may change with scale, (2) theimprecise definition of scales of measurement relevant to the gradients under examination, (3) a lack of standardization in data collection, and (4) a mismatch between the scales at which patterns are observed and the scales at which purported causal mechanisms operate in nature (e.g., ecological vs. biogeographic vs. evolutionary explanations). When such imprecision and confusion permeate the literature and our thinking, substantive conceptual development of the discipline is stymied considerably (Pickett et al., 1994).

To address these issues, we organized a workshop at the National Center for Ecological Analysis and Synthesis in Santa Barbara, California USA. The purpose of the workshop was to assemble data that would allow us to identify problems inherent in previous approaches and to advance the development of theory. Because primary data spanning a broad range of productivity and diversity are available and accessible through the Long-Term Ecological Research (LTER) Network, many of the workshopparticipants represented LTER sites. The willingness of these sites to share data in an open forum contributed to our ability to develop comparative and quantitative analyses of the scale-dependent relationship between diversity and productivity.

Our approach to understanding the relationship between diversity and productivity was multifaceted. It included analyses of data from terrestrial and aquatic systems, plants and animals, and observational and manipulative experiments. Two important results from the NCEAS meeting will be presented at this workshop. First, we willemphasize the importance of scale to understanding patterns between productivity and species diversity. In their review of the literature, Mittelbach et al. (in litt.) documented that, in general, no particular pattern holds hegemony over the others. Positive linear, negative linear, unimodal, and non-significant relationships occurred in most categories of research based on geographic extent, ecological extent, or taxonomic hierarchy. Scheiner et al. (in litt.) refined the concept of scale, integrated it with species-area relations, and showed that patterns are clearly functions of scale. In addition, they provided methodological recommendations for combining data collected at different scales, so as to facilitate a broad and comprehensive understanding of relationships. Gross et al. (in litt.) acquire, standardize, and analyze data from a variety of biomes in North America to quantify the relationship between species density and annual above-ground net primary productivity (ANPP) in low stature vascular plants. They show that both focus and extent (sensu Scheiner et al. in litt.) potentially determine the pattern between diversity and productivity. Clearly, pattern is scale-dependent, and recognition of this phenomenon is a critical prerequisite to integration and synthesis.

Our second point involves the generality of the unimodal pattern between species richness and productivityoften cited in the literature. A comprehensive survey of the literature by Mittelbach et al. (in litt.) showed that unimodal patterns were more prevalent as the extent of analysis increased for plants, but not animals, and that negative relationships disappeared at the greatest ecological or geographic extents for both taxa. Gross et al. (in litt.) also detected a unimodal pattern for species densities examined at continental scale. Paralleling that result, Dodson et al. (submitted) show, for a variety of lacustrine biota, that unimodal patterns exist across broad extents of productivity when community type (i.e., lake) is the focus. Thus, unimodal relationships do emerge as dominant patterns, but only at some spatial scales.

A key strategy for improving our understanding of the interaction of biodiversity and productivity (or other ecosystem characteristics) involves the integration of two common experimental approaches: the manipulation of productivity and the alteration of the number of species or functional groups. A synthesis of ideas that have developed around these two approaches is a prerequisite for the advancement of a general theory that will direct the next generation of hypotheses and experiments. Conceptual models being developed by Grace (submitted) and M. Shackak (personal communication) foreshadow this synthesis. These emerging models incorporate disturbance, plant biomass (productivity), resource heterogeneity, colonization, and the available species pool as primary factors controlling species density. Consequently, they emphasize the importance of multivariate approaches to understanding patterns of species density. We propose a similar model in which the composition of a local assemblage is derived from a regional species pool via the action of five filters: available energy, nutrient availability, and the species already in the community, as well as structural characteristics of production and litter distribution. Similarly, the production of the ecosystem is affected by available energy as processed by species assemblages and constrained by nutrient availability. Because energy flow and species flow are affected by some of the same regulators, and in fact affect each other reciprocally, the relationship between them may be complex. We will present this model with illustrative examples from dryland systems.

Summary

There have been several attempts to generate and test hypotheses on the species diversity-ecosystem function relationship. A fundamental problem is that the hypotheses are not driven by a comprehensive theory of the relationship between species properties and ecosystem processes.

In the workshop, will present a conceptual model that links species properties and ecosystem processes in water-limited systems. The model is based on the differences among species in their use of water for biomass production and the non-uniformity of the soil moisture distribution.

The model identifies the species’ ecosystem related properties and the relationships among species in assemblages along a soil moisture gradient. This enables us to show that the; diversity-stability, rivet, redundant species and idiosyncratic response hypotheses are scale and perspective dependent.

Our model is an attempt to explicitly define the differences among species in terms that mechanistically link resource utilization by species with its ecosystem consequences. This introduces individual species properties that relate to ecosystem function, into the concept of species diversity. In the model, the concept of species diversity includes; the number of species, the differences among them in water utilization for biomass production, and their distribution and local abundance along a soil moisture gradient.

Introduction

The potential relationship between species diversity and ecosystem processes are described by five hypotheses:

1. The null hypothesis claims that there is no effect.of species diversity on ecosystem processes.

2. The diversity-stability hypothesis predicts that ecosystem productivity and recovery increases as the number of species increases.

3. The rivet hypothesis predicts a threshold in species richness, below which ecosystem function declines steadily and above which changes in species richness are not reflected in changes in ecosystem.

4. The redundant species hypothesis states that species loss has little effect on ecosystem processes if the losses are within the same functional group.

5. The idiosyncratic response hypothesis suggests that as diversity changes so do ecosystem processes without a general pattern.

There have been several attempts to test these hypotheses, in the field and the laboratory.

However, the interpretation and the generality of the results remains contentious. A fundamental problem is that the hypotheses are not driven by a more comprehensive theory of the relationship between species properties and ecosystem processes.

We propose that the foundations for the theory are in models of the distribution of resources and their utilization by organisms. This is because ecosystem processes, such as primary production, decomposition, mineralization, evapotranspiration etc., are dependent on the processing of resources by producers, consumers and decomposers

Within a community perspective, a theory of resource utilization is based on two assumptions:

1. The rates of ecosystem processes are determined by a few species that are most efficient in using and converting resources. For example: species that are proficient in using water for biomass production or in converting organic matter into inorganic materials.

2. The rates of ecosystem processes are determined by complementarity in resource use by different species. i.e. the rate of ecosystem processes of an assemblage of species should be higher than the rate determined by a single species.

Ecological studies of natural communities provide some empirical evidence that productivity is determined by few species. However, agricultural practice suggests that species complementarity can be of importance in determining ecosystem processes.

Here, we would like to present a conceptual model based on the differences among species in using water for biomass production and the nonuniformity of soil moisture distribution. The model demonstrates that the hypotheses on the relationship between species diversity and ecosystem function are not mutually exclusive.

The model, which integrates dryland hydrology with the idea of individual species responses along environmental gradients enables us to show that the diversity-stability, rivet, redundant species and idiosyncratic response hypotheses are scale and perspective dependent.

The Model

Water, species and productivity

We suggest that the relationship between species diversity and primary productivity in drylands is intimately related to water flow. We see this relation within a framework of a multiflow system which integrates species, water and energy flows. In our model we refer to water flow as processes that determine the spatial and temporal distribution of soil moisture.

Species flow refers to the colonization and extinction of species in a specific area. Colonization is possible due to the availability of a species pool larger than the number of species in the specific area. Energy flow refers to biomass production by all species that occupy the specific area. The interactions among soil moisture, species diversity and biomass is shown in Fig. 1.
 
 

Fig 1. The relationship among species diversity and ecosystem processes; water flow and primary production

We see the relationship among species diversity and ecosystem processes; water flow and primary production at two scales. The input scale, which consists of a species pool and rainfall. The species pool provides the species for the assemblage and rainfall provides the water for the soil moisture. The feedback scale consists of a web of interactions among soil moisture, the species assemblage and plant biomass.

Species assemblage

We see ecosystem processes as a product of the organization of species in a web of interactions. There are two steps in the organization of species in a web of interactions. The first step is an assemblage of species in an area due to colonization. This is a process of filtering a subset of species from the overall available species pool. This stage is dependent on propagule arrival from the species pool and on site environmental conditions for germination, establishment and growth. In water limited systems, the environmental conditions are usually related to soil moisture availability. The second step is assemblage organization by the formation of a web of interactions. Here, species form a web of interactions which will determine their ability survive by producing, using and sharing resources. The assemblage organization is the main controller over ecosystem processes (controllers 2b and 2c in the system Fig. 1)

Species diversity along a soil moisture gradient

We adopt the individualistic approach to describe the distribution and abundance of species along a soil moisture gradient. This approach enables us to describe species diversity of a species assemblage (numbers and difference in distribution and abundance) on various ranges of soil moisture. To evaluate the ecosystem function (primary production) of a given assemblage after organization, we combine the species distribution with a property that controls its ecosystem function. For a species in a water-limited system we assume that primary production will be higher in areas with species assemblages that are more efficient in using water for biomass production (controller 2b). We can draw a species diversity diagram along a soil moisture gradient (Fig. 2). In this diagram the differences among species are: their ability to use water for converting solar energy into biomass (biomass production per unit water under a given soil moisture content), their distribution range and abundance along the soil moisture gradient.
 

Fig 2. Mapping species diversity along a soil moisture gradient. The scheme shows three species that differ in their distribution, range, abundance and ecosystem related properties.

Formation of soil moisture gradient

As shown in Fig. 2 our concept of species diversity is defined in relation to soil moisture heterogeneity. Thus, factors controlling the spatial and temporal distribution of soil moisture are essential to our conceptual scheme on species-ecosystem relationships.

Hydrological studies show that the spatial and temporal distribution of soil moisture is controlled by several factors:

1. Spatiotemporal variability in rainfall on various scales (rainfall patchiness on a scale of kms. Rainfall patchiness on a slope. Rainfall distribution during a single rainfall event. Rainy season rainfall distribution and perennial distribution (related to the physics of the atmosphere).

2. Physical patchiness on various scales that control input and output of water into the soil (within watershed - rock to soil, slope direction, watershed patchiness - the spatial arrangment and size of watersheds). This controls the frequency and magnitude of evaporation and runoff as well as source-sink relationships at various spatial scales.

3. Biological patchiness at various scales and its effects on input and output of water into the soil(changes in vegetation cover from vegetation on a slope to vegetation on a high order watershed).

We can classify species assemblages along a soil moisture gradient by dividing the gradient into parts of equal range in soil moisture. We see many possible relationships, between soil moisture range along a gradient and the number of species that live in this range. For example, a large number of species can occupy a narrow range of soil moisture (Fig. 3). In this case the number of species per unit of soil moisture range is high. Another possibility is that along a wide range of soil moisture only a few species occupy each unit of soil moisture range (Fig. 3).
 
 

Fig 3. Arrangement of species assemblages (A to G) in two constrasting landscapes. a) A landscape with low variability in soil moisture and high number of species per assemblage. b). A landscape with high variability in soil moisture and a low number of species per assemblage.

In the first case we expect a low degree of specialization in relation to soil moisture utilization and high degree of overlap among species along the gradient. In the second we expect the opposite trend.
 
 

Species overlap and ecosystem processes

The implications of specialization and overlap in water utilization by species along a soil moisture gradient, in relation to the species diversity- ecosystem processes hypotheses, may be understood by analyzing a model of two species and their preformances (Fig. 4). When the two species of the assemblage specialize on two different ranges of the soil moisture gradient, the hypothesis that predicts that ecosystem productivity increases as the number of species increases is applicable. The situation is more complex when the two species overlap in part of their distribution range. Looking over all their distribution ranges we still see that ecosystem productivity increases as the number of species increases. However, if we look only at the soil moisture range of species overlap there are three ways to view the relationship between species diversity and ecosystem function:

1. Soil moisture range over which species a is better than species b in water utilization for biomass production (diagonal striping in Fig. 4). We suggest that in this range the loss of species has no effect or a positive effect on ecosystem productivity.

2. Soil moisture range over which species b is better than species a in water utilization for biomass production (horizontal striping in Fig. 4). We suggest that in this range the losses of species a have no effect or a positive effect on ecosystem productivity.

3. Soil moisture range over which the two species overlap (Diagonal and horizontal striped area). We suggest that in this range if either species a or b become extinct they are replaceable for each other.

The area along the soil moisture gradient over which the two species overlap may be viewed in two ways. In one view, the overlap, i.e. increase in species diversity, is insurance of a continuation of ecosystem function in case of species extinction. This is a future oriented view. At present there is a cost of species overlap. A species less efficient in water utilization is part of the assemblage and reduces the potential rate of primary production by the more proficient species. We conclude that there is a cost to ecosystem function associated with species diversity. This cost derives from the tendency of species to overlap in their distribution range.

Implications for research

The prevailing experimental approach for studying species-ecosystem processes is to manipulate the number of species as an independent variable and to measure ecosystem processes as a response variable (productivity, decomposition, mineralization etc). The results of this type of study show how species number controls the rate of ecosystem processes. The approach is an attempt to predict ecosystem responses to the number of entities i.e. species which are involved in the processes. However, when the properties of the species in relation to resource processing is of importance, then manipulation should take into account the differences among species in resource processing as well. A step toward an integration of the differences in resource processing among species into species- ecosystem studies are the experimental studies of the relationship between functional groups and ecosystem processes. In this approach we group species according to their properties that may control ecosystem processes.

Our model is an attempt to explicitly define the differences among species in terms that mechanistically links resource utilization by species with its ecosystem consequences. i.e. the relationship among water availability, water use and biomass production. This introduces an ecosystem property of the individual species into the concept of species diversity. Our concept of species diversity includes; the number of species, the differences among them in resource utilization, spatial distribution and local abundance along a soil moisture gradient. This approach requires an expansion of the experimental approach for studying species-ecosystem relationship in drylands.

The experimental approach for testing species richness-ecosystem process relationship is basically a comparative study of the differences in the rates of ecosystem processes among species assemblages over a uniform environment. Our model requires an extension of the experimental approach into two new directions. The first is the study of "individualistic" ecosystem effects. This implies experimentally studying water use and biomass production of some common species along a soil moisture gradient. The second extension requires testing simultaneously the effect of changes in the environment and species diversity on ecosystem outcomes.

Ecologists generally agree that species diversity is linked to landscape features. In the workshop we will present a basis for a theory connecting species and landscapes, by studying how changes in species assemblages and in landscape structure coincide. The observation that 1) individuals and populations of organisms are affected by environmental variation in the landscape, and 2) that communities consist of species populations (or parts of them) are intuitively connected. We propose a theory to explore this relationship quantitatively as well as qualitatively, by viewing species assemblages as collections of populations interacting with the heterogeneity of the landscape. We use the neutral term "assemblage" to preclude assumptions about interactions and proximity or encounters among the organisms. There are also practical reasons to link species assemblages and landscapes, as manipulating the former may be most efficiently done by modifying the latter, as has been demonstrated in semiarid shrubland and elsewhere. Such ecological management can be a powerful tool in conservation and restoration of biodiversity.

Species and landscape diversity

A species assemblage changes through time (Fig. 1A) by means of two separate processes, one internal and one external, affecting 1) the number of species, and 2) their properties (abundance, biomass, functional group assignment, etc.). A portion of the species in the assemblage at time t1 passes on to the assemblage at time t2, and new species may enter from the outside species pool. Following Gaston (1996 - Biodiversity: A biology of numbers and difference, Blackwell Science, Oxford), we express the diversity of the species assemblage as the number of species and the differences in their properties, which are defined from the population point of view. The processes of assemblage dynamics are controlled by external factors such as changes in the landscape, in addition to interactions between the species.

As another form of diversity, we quantify landscape diversity as numbers of landscape components and the difference in their properties. The components are patches, which have measurable properties (size, shape, etc.). Analogous to assemblage dynamics, landscapes change through time in frequency and properties of patches (Fig. 1B), but in contrast, newly created patches stem from existing patches. The external controls of landscape dynamics are resource inputs (control 5 in Fig. 1) and agents of disturbance (control 6). Both landscape and assemblage dynamics (Fig. 1 A and B) are connected by a number of control processes (Fig. 1): 1) Patch properties and configuration affect species composition by filtering colonization as incoming propagules, and 2) local conditions and resources determine species-specific recruitment and survivorship. In addition, the assemblage itself also controls recruitment through biological interactions (Fig. 1, control 3). The assemblage affects the landscape as species combinations differentially alter the properties of patches, (Fig. 1, control 4).

Quantifying diversity

The classical approach to species diversity provides a limited view, because of the tendency to either transform the two dimensions of diversity (numbers and differences) to a single index (Simpson’s or Shannon-Wiener’s diversity, for instance), or use only one dimension (richness and its derivatives). We propose to quantify species diversity as a phase plane of incidence and abundance of all species of the assemblage (Fig. 2A). This expresses diversity by showing 1) numbers of species as points in a single phase plane and 2) the differences between them in terms of their local abundance in patches and their landscape-wide incidence over all patches.

Landscape diversity can be defined similarly, since patches are landscape-level entities that can be assigned to a number of patch types with a frequency of occurrence in the landscape, with many differences in quality (properties and configurations) (Fig. 2B). Differences in quality may be the availability of a particular resource (water, nutrients, organic matter), composition of the substrate (soil types, etc), or any other relevant difference. Patch type differs from patch quality in a statistical sense, as quality varies less within patch types than among them, while patch types differ in some major way in overall structure. For instance, a forest stand differs from a treefall gap in the amount of light, and in other variables like soil moisture, soil organic matter content, etc., while stands or gaps may vary in light regimes or soil properties among themselves. In dryland systems, patch types can be recognized at two scales: 1) patches of "bare" or crusted soil, patches associated with large plants (shrubs, cacti, grass tussocks, etc.), litter-covered soil, diggings and mounds of animals or made by humans, or bare rock outcrops, and 2) watersheds and their components (water courses, slopes and hilltops and plains, in addition to 3) larger geographical patches with different properties, including rainfall and other gradients.

For quantification of diversity as defined by Gaston (1996) it is necessary and sufficient to measure numbers of entities (species or patches) and some inherent difference between them. The variable used should be relevant for the study, but there is no a priori reason to exclude any property on principle. However, in order to connect the species assemblage and the landscape together, a common criterion is required which relates both species and patches to the entire landscape. Frequency of occurrence of patches in the landscape and of species in the patches (incidence) provides such a common basis. Together, abundance and incidence reflect differences between species where they occur in the landscape, which is related to how often patches of different types occur in the landscape. Without incidence, abundance does not provide any spatial information on the species. This is why abundance, when used without incidence is usually calculated as the overall mean population density per patch over all patches, including unoccupied ones. In our case, null-populations are excluded from abundance, since incidence already expresses the spatial aspect.

Local and spatial processes

The positions of species in the incidence-abundance phase plane reflect their individualistic behavior in the landscape, which explicitly links population dynamics of all species to the patch dynamics of the landscape. Both incidence and abundance are net results of local and spatial population processes, but with entirely different emphasis. Abundance of a species (excluding unoccupied patches) focuses on local population growth and persistence, augmented by input from outside the patch. The latter include spatial processes in the form of immigration of individuals, but these are controlled by local properties affecting arrival. Incidence reflects colonization and extinction of the species in patches over the entire landscape, thus focussing more on spatial processes, with less emphasis on local processes. Movements of organisms among patches are no less important than local processes in shaping and maintaining local within-patch populations and community structure.

Abundance can be expressed by other quantities than density of individuals, such as cover or biomass, depending on the overall research question. Biomass is especially important for studies of productivity in grazed systems, for instance, or other land-use situations. This precludes a link with population dynamics, but establishes one with ecosystem processes.

The positions of patch types in the frequency-quality phase plane express landscape diversity by showing how many patch types there are, how often these occur in the landscape, and how different they are. It is often suggested that, because patch quality has a different meaning for all the species of an assemblage, it would be arbitrary to assign patches to particular types. However, if patches are defined as landscape components with their own dynamics, they can be assigned to types based on their measurable properties. This way landscape-level patch type descriptions are dictated by landscape dynamics, independently of the species’ perspectives. This is also necessary in order to use patch types as (statistically) independent variables in relation to species assemblage dynamics, and in order to compare all species on a common basis. (In fact, species perspectives can only be assessed against the landscape background, notwithstanding claims to the contrary.)

Patch type frequency is the net result of patch formation (controlled by biotic or abiotic disturbance or gradual long-term processes, Fig. 1B), while patch quality is the result of resource availabilities, conditions and biological interactions affecting the patches of different types. Because patches and species each have their own dynamics, the analogy between them should not be extended too far. In the first place, species abundances have fewer differentiating properties (density, biomass or cover, etc.) than patch types (size, shape, topographic structure, surface texture, soil structure, soil moisture, light regime, nutrient availabilities, microclimatic conditions, presence of herbivores and pathogens, proneness to disturbance, position within the landscape, distance from other patches, etc.). On the other hand, patch frequency is rigidly constrained (the sum of all patch types equals 1) while species’ incidences are not. In the relationship between landscape and assemblage dynamics, patch type quality controls both spatial and local population processes of the species. Frequency, in addition to controlling some of the same processes, assigns importance to the patch types, specifying their contributions to overall species assemblage dynamics.

Mapping trajectories of species and patches

As both patches and species change through time, both kinds of entities describe trajectories in their respective phase plane (Fig. 3). These reveal the local and spatial processes taking place in the species assemblage and the landscape mosaic from time t1 to time t2. Trajectories of patch types (patch type vectors) show how patches are formed and change in time, while species trajectories show how they change in their incidence and abundance as the landscape changes. Studying the relationships between species trajectories and patch type vectors can link assemblage and landscape dynamics.

As new patches are formed and disappear (changes in frequency), and as they change in quality (gradually and due to annual climatic variation), some species increase and others decrease in incidence and in abundance. Changes in incidence signify 1) new invasions into the assemblage, 2) colonization into more patches, 3) extinctions in some of the patches, or 4) landscape-wide extinction. Changes in abundance signify positive, negative or zero local population growth rates. Population growth can be density-independent (only limited by seed production and arrival) or density-dependent (resource or site limitation). Thus assemblage dynamics, reflects a spatial multi-species network of the local and spatial demographic behaviors of all species of the assemblage. These behaviors occur at the population level, and consist of the probabilities (transition rates) of individuals of the different species to pass from one state to another.

A change in the position of a patch type (Fig. 3) from time t1 to t2 (as in Fig. 1) and the resulting trajectory of a species can easily be viewed using incidence-frequency as a common axis. As patch type a changes in quality (for instance, soil moisture due to annual rainfall), species A only increases in abundance (from A(t1) to A(t2)) due to local increase in the rate of recruitment, or also in incidence (A’(t2)) as recruitment rate also increases in unoccupied patches of the same type from the few propagules that did arrive there. In another case, patch type b changes in frequency as more undisturbed patches are disturbed, for instance. As a result, we see that species B increases in incidence alone, apparently due to easy dispersal into the new patches.

In addition to species’ trajectories, the overall pattern of change of the assemblage as the landscape changes is informative as well. The exponential relationship usually found between incidence and abundance at all scales, may or may not change in slope, intercept or correlation, accompanied by smaller or larger species trajectories. It is still a point of controversy how to interpret the relationship. One of the questions a theory of assemblage dynamics will ultimately have to answer is whether the relationship is biologically meaningful, or merely useful as a null-model. For the time being we view any relationship as an emergent property, while acknowledging that species with high local abundance also have a larger chance of dispersing to unoccupied patches.

Applications of the theory

The proposed theory of species and landscape dynamics can be applied to a large number of ecological questions arising from changes in landscape structure. For instance, Boeken and Shachak (1998, Ecography 21: 63-73) applied it to colonization of annual plants in newly formed patches (soil disturbances). This showed that most species are more or less rare and sparse, and respond in a density-dependent manner to annual fluctuations in rainfall (suggesting site-limitation due to low soil moisture availability), while few common dense species had density-independent population growth after nearly instantaneous arrival. In another, unpublished study, addition of nutrients (changing patch quality) resulted in increased biomass production only of a few common (dominant) species.

Other theoretical questions that can be addressed with the theory, at least conceptually, are those related to species composition and species interactions (from competitive exclusions to facilitation). In this context, assemblage dynamics differ from more classical views of community dynamics, where the community is usually defined at a large scale with smaller-scale replicate samples. In our view a species assemblage is defined at the scale of the landscape with (replicate) samples of the constituent landscape patches. Because of the proximity and potential for interactions of the plants, samples come much closer to individual communities than the entire assemblage. What the proposed theory adds is that local species composition (within samples) is not only affected by local population dynamics depending on in situ conditions, resources and other organisms, but also by spatial population processes of colonization and extinction, determining and maintaining community structure at the landscape scale. Species composition, and scalars like species number, species richness (correcting number for density or area, as in rarefaction and Fisher’s a ), equitability and other diversity indices (e.g., Simpson’s and Shannon-Wiener’s) are then viewed as net results of these processes. Thus, species co-occurrences are strongly influenced by (co-)incidence, representing spatial stochasticity and causing non-equilibrium species combinations, and on the other hand by biological interactions among the species, representing equilibrium species associations.

The practical value of the theory is its application to questions related to human-induced changes in landscape structure at various scales. Desertification of semi-arid landscapes (which we define as a decline in productivity and/or diversity), or its counterpart, sustainable land-use, can be understood if we study the effects of grazing or of clear-cutting of woody plants on landscape patchiness and species incidence and abundance (density and biomass). In the workshop we will apply the theory to an empirical case study of annual plants in a shrubland landscape, where patches were manipulated to resemble grazing impacts of sheep (clipping and trampling). This highlights its use as a research tool for generating hypotheses and for designing and analyzing experiments. The theory is also applicable to larger-scale problems, such as habitat fragmentation, degradation and destruction, and for evaluating efforts of restoration and rehabilitation. In these cases GIS techniques are very useful in identifying and quantifying the landscape component.

Finally, the link between population processes and assemblage dynamics enables simulation modeling of the relationships between assemblage and landscape dynamics. In the workshop we will present a spatially explicit mechanistic model, in order to predict patterns of assemblage dynamics in changing landscapes from simple assumptions about species- and patch-type specific demographic transitions, species attributes, annual climate and species interactions.
 
 

Figures

Fig. 1. Flow chains of (A) species of an assemblage as they colonize landscape patches and grow in density (or dwindle and become extinct) and (B) patches in the landscape mosaic as they are formed and change in quality. The flow chains are connected by patch properties controlling colonization of species (1) and extinctions, recolonizations and population responses (2), and by species controlling formation, maintenance and modification of patches (4).

Fig. 2. A. Species in the incidence-abundance phase plane. B. Patch types a, b and c in the frequency-quality phase plane.

Fig. 3. Two species (A and B) and two patch types (a and b) changing from time t1 to t2 in their respective phase planes connected by their common spatial axis. See text for explanation.

What limits the productivity and biodiversity of deserts?

Deserts are environments in which water limits productivity. Photosynthesis, growth and reproduction of higher plant species in deserts are low and restricted to brief and often unpredictable periods when water is available. Rates of nutrient cycling are low because low plant production results in low soil labile organic carbon and nutrient accumulation, and because low soil moisture constrains soil microbial activity. Consequently nutrient availability can limit plant growth and reproduction during periods of high soil moisture. Similar arguments can be used to explain the lower biodiversity found in deserts compared to some less arid environments. Relatively few higher plant species have adapted to desert environments. The reduced diversity of plant resources that can be used by primary consumers and decomposers reduces the complexity and number of organisms in the food webs that are based upon these plants. For desert-adapted plant species, low productivity and rates of nutrient cycling place further energetic and nutrient transfer constraints on the length, hence diversity, of organisms in food chains.

Adding water to deserts increases primary production and rates of nutrient cycling, but what constrains the productivity and biodiversity of a desert for a given level of hydrologic input, or, in other words, what are the proximal constraints on productivity and biodiversity? During and after infrequent and sometimes intense rain events a substantial fraction of the hydrologic input is immediately lost via runoff into Wadis. This is particularly true for deserts with topographic relief and large amounts of exposed rock surface – such as the Negev Desert Highlands. Following rain and runoff events, high solar input leads to substantial further losses via evaporation, because of reduced higher plant cover and limited deep infiltration of moisture into soil. Production in the ecosystem is sustained primarily by stored water. It is not unreasonable to suppose that if all the water that fell in deserts was retained by the ecosystem that productivity and biodiversity would be substantially greater than is observed.

Hydrologic retention as a constraint on ecosystem productivity and biodiversity

In this sense, at a given and relatively low level of hydrologic input, desert ecosystems can be thought of as being limited by the number and appropriate distribution of internal hydrologic "sources" and "sinks". Hydrologic sources are locales within the desert that generate and redistribute runoff (as opposed to either uniformly distributed rain or runoff or no runoff at all). Hydrologic sinks are locales within the desert that trap runoff, allowing deep soil infiltration and storage (as opposed to no deep infiltration with high surface soil evaporative losses). An appropriate spatial distribution of sinks and sources is also critical. Sources must supply sinks, otherwise the concentrated runoff will leave the ecosystem via the Wadi or evaporation.

If deserts are hydrologically source- and sink-limited, then it logically follows that anything that increases the number of appropriately distributed internal sources and sinks in the desert will increase productivity and biodiversity. It also logically follows that there must be existing, internal sources and sinks in most deserts, otherwise productivity would be much lower, being limited to that sustained during the brief periods during and just after rainfall and runoff, or sustained via storage inside plants. So what determines the number and distribution of internal sources and sinks in deserts? Local geomorphologic forces certainly play a role. Crevices and cracks in rocks can channel runoff, retain soil and allow deep-water infiltration, and adjacent pits and mounds created by erosion can both generate and trap runoff. However, we do not think that geomorphological forces are the sole or even major factors determining the number and distribution of sources and sinks in deserts. Instead we postulate that physical ecosystem engineering by desert organisms – higher and lower plants and animals – is a primary determinant and major proximal control over desert productivity and biodiversity. What do we mean by "physical ecosystem engineering by desert organisms" and what evidence is there in support of such a hypothesis for the Negev?

Physical ecosystem engineering by desert organisms

The recognition that deserts are hydrologically source-sink limited is not new. Humans have been reducing this limitation by using runoff agriculture for over 4000 years in the Negev. Runoff agriculture is based on simple hydrologic engineering principles and involves the creation of an appropriate distribution of internal hydrologic sources and sinks by building physical structures in the landscape at different scales and juxtapositions. These physical structures all collect runoff from rain events over larger scales, redirect this runoff, and concentrate the water into smaller-scale collection areas where it can infiltrate deep into the soil and be stored. The stored water sustains higher levels of primary production over longer time periods than would be expected based solely on the amount of rainfall. It would be hubris to assume that Homo sapiens is the only species in the Negev to have taken advantage of hydrologic engineering principles. But before we examine these other engineers we need to define what we mean by hydrologic engineering.

Physical ecosystem engineers are organisms that directly or indirectly control the availability of resources to other organisms by causing physical state changes in biotic or abiotic materials. Physical ecosystem engineering by organisms is the physical modification, maintenance or creation of habitats. Ecological effects of engineers on many other species occur in virtually all ecosystems because the physical state changes directly create non-food resources such as living space, directly control abiotic resources, and indirectly modulate abiotic forces that, in turn, affect resource use by other organisms. Such a general definition encompasses the activities of humans in the Negev. Here, the two physical states are soil with low water storage capacity being converted to soil with high water storage capacity. The resource that is used by other organisms is water. The other organisms are the annual plants that comprise most of the higher plant species richness and productivity in the Negev. These plants are then consumed by humans and livestock. The physical structures built by humans from rocks, compacted soil and disturbed soil depressions directly control the flow of the resource by collecting, redistributing, concentrating and storing water.

Given this definition, we can recognize many other species in the Negev that build physical structures that could function as hydrological sources and sinks. We focus on structures built by cryptogamic communities, shrubs, and desert porcupines (Hystrix indica), as well as tree pits built by humans. Based on previous studies and our new studies we can summarize the likely hydrologic sink and source functions of these structures as follows. Cryptogamic communities produce a mucopolysaccharide secretion that binds soil particles into a crust that reduces water infiltration (i.e., has no sink function) but has a major source function of redistributing runoff. Much of the soil in the Negev is crusted. Perennial shrubs build soil mounds over 200-year periods via the local accretion of organic matter, snail feces and aeolian dust interception and deposition. Shrub soil mounds most likely function primarily as sinks because the loose, uncrusted soil of the mound allows infiltration and storage of water. Porcupines dig for the roots of geophytes creating soil pits and mounds. The pits and mounds most likely function as sinks that trap runoff via the effects of relief (pits) and loose, uncrusted soil that allows infiltration and storage of water (pits and mounds). Human-made tree pits serve as sinks.

Approach to quantifying hydrological physical ecosystem engineering in the Negev

But do these structures actually function as hydrologic sources and sinks? Do patches of the desert containing appropriate distributions of sink and source structures store more water than patches that do not contain these structures or do not contain both sources and sinks? If so, then does the increase in water retention result in increased productivity and diversity of annual plants? If so, how big are the effects? How much of the productivity and diversity of the Negev could be due to physical ecosystem engineering by these species? How do the effects of non-human engineering compare to human engineering?

Given that hydrological engineering depends upon the formation of physical structures, we took an "approach of the artifact" focusing on the physical constructs and their effects, rather than studying the engineering organisms themselves. This approach allowed us to clearly separate out the effects of the physical engineering per se from all the effects of the species in the ecosystem (e.g., engineering and trophic effects combined). We used a combination of both experimental and comparative studies. Data came from new experiments, reanalysis of our own prior experiments and reanalysis of previously published studies by others on the structures listed above, carried out over many years. Experimental studies compared water retention (soil moisture) or loss (runoff), annual plant productivity (number of annual plant individuals) and biodiversity (number of annual plant species) in studies where physical structures were removed or added to replicated patches of the desert, comparing the results with replicated control patches where the structure or lack of structure was not manipulated. Comparative studies took the same approach but compared replicated unmanipulated patches with contrasting degrees or types of structure. Our data analysis used traditional ANOVA, followed by a novel graphical representation. We plotted the amount of water retained, productivity and biodiversity in desert patches with both sink and source engineered structures (y axis) against the amount of water retained, productivity and biodiversity in desert patches with only the source engineered structures (x axis; i.e., primarily crust only). A one-to-one line represented no effect of engineering. Deviations from the line resulted in slope and/or intercept values that directly quantified the effects of the engineering on these 3 different response variables.

Summary of findings

Compared to patches of crusted, undisturbed soil, porcupine pits adjacent to crusted soil had 32% greater water retention (as soil moisture content) following runoff events. When annual plants were removed from pits and crusted soil, natural recolonization led to 4.1 times more annual plant individuals and 2.2 times higher annual plant species richness in porcupine pits compared to the same surface area of crusted soil. Runoff from crusted soil patches (0.5 m²) containing shrub mounds (75% area crusted; 25% area shrub mounds) was 4.1 times less that of crusted soil patches of the same size without shrub mounds. Shrub and mound removal from patches (90 m²) led to a 53% increase in runoff compared to undisturbed patches with shrubs and crusted soil. Shrub mound patches contained 4.5 times more annual plant individuals and had 3.0 times higher annual plant species richness. Human-made soil pits (0.3 m², 15cm deep) used for planting trees contained 2.9 times more annual plant individuals and 35% higher annual plant species richness compared to adjacent crusted soil of the same surface area. All effects were highly statistically significant (p<0.001).

Conclusions and implications

Patches of desert containing porcupine pits, shrub mounds or human-made tree pits proximate to crusted soil, resulted in substantial increases in water storage, annual plant production and annual plant species richness. Human-made structures of comparable size to structures made by other species in the Negev had similar effects. These data provide strong support for the postulate that physical ecosystem engineering by species in the Negev is a primary proximal control on water storage, productivity and diversity. Structures built by organisms result in a patch structure consisting of hydrologic sinks distributed next to hydrologic sources. Collectively, these structures result in large increases in water storage, productivity and diversity.

It is clear that engineered patches have very large local effects that certainly result in increased water storage, productivity and diversity at larger watershed and landscape scales.

However, the degree to which the patches collectively capture the entire hydrologic input to a watershed will depend upon the number, type, configuration and interaction among engineered patch types. Consequently, quantitative estimates of the overall effects of hydrologic engineering in the Negev will require data on engineered patch relationships at these larger scales.

Many other deserts have crusted soil, pits and mounds created by a diversity of species of plants and animals. Based on simple hydrologic engineering principles, it is reasonable to postulate that physical ecosystem engineering by organisms in other deserts may be a significant proximal control over water storage, productivity and diversity. However, testing this postulate will require comparative studies in a variety of deserts.

Ecologists still search for common principles that predict well-known responses of biological diversity to different factors. These include the physical spatial complexity of the environment, productivity, area, species’ body size, and habitat fragmentation. Here, we show that all these patterns can arise from the influence of landscape pattern on how organisms acquire resources in space. We use spatial scaling laws to describe complex landscapes where habitat, food, and food resource concentration vary, and explore how species of different size find food within this landscape. We then derive a mathematical rule for the minimum similarity in size of species that share these resources. This "packing" rule yields a theory of species diversity that generates left-skewed, rather than log-normal, distributions of diversity vs. size. It also predicts relationships of diversity vs. productivity better than most previous explanations. Size and diversity patterns for several locally coexisting dryland communities, including grazing mammals, insects, granivores and plants strongly support these predictions. The theory also explicitly predicts relationships of diversity vs. area and vs. habitat fragmentation. Thus, spatial scaling laws provide potentially unifying principles that may explain many important patterns of species diversity. These predictions have strong potential application in the conservation of diversity in arid environments.

The search for a "unified" theory of diversity has focused on the premise that more species can exist within a habitat whenever they can more finely divide up space and different-sized resource "packages". This premise assumes that landscapes in which species live are heterogeneous. Species are hypothesized to partition such landscapes according to body size, but the mechanisms by which organism size, resource availability and spatial structure of habitats control species diversity remain unclear. Here, we employ spatial scaling laws to describe landscape pattern and how species with different body sizes find resources within such landscapes. These scaling laws predict a limiting similarity in body size between any two species and therefore the potential number of species in a community. This yields an explicit analytical theory of how species diversity of some important taxa changes with the landscape extent (scale of observation) and pattern.

We assume that individual organisms must search within a space of suitable physical/chemical conditions (habitat) to find resources, which are often only available inside other material (food). Therefore, resources available to organisms are nested within food, and available food is nested within habitat. For example, insect herbivores move through suitable microclimates on terrestrial plants (habitat) to eat plant tissue (food), which contains digestible carbohydrates (resources). Predatory fish search macrophyte-free areas of lakes (habitat) to eat invertebrates or smaller fish (food) that contain protein (resources). More imaginatively, terrestrial plants extend roots into rock-free soil (habitat) to take up soil solution (food) that contains nutrients (resources). Within a habitat, different species of similar trophic position may harvest different sizes or types of food to obtain the same resources.

Spatial patterns of habitat, food, and resources often appear similar when viewed at different spatial scales across ecologically relevant ranges of scales (3-4 orders of magnitude). If so, their volume or area and spatial distribution can be described with fractal geometry, i.e., simple scaling laws. For example, the distribution of habitat on a landscape of extent x is hx^D, where D is the fractal dimension of the habitat and h is a prefactor that reflects the local density and contagion of habitat. Similar scaling laws apply to the distribution of food and resources in the landscape. In such an environment, body size critically determines the abundance of food and resources a species perceives. Individuals sample a volume of space at a particular scale of resolution, i.e., the length, w, of the "ruler" with which they perceive or sample the environment. This scale of resolution is presumably proportional to body size. If so, a species will subdivide its habitat into subvolumes of a particular size. The total amount of food available to the species is therefore the food contained in all subvolumes it perceives as completely filled with food, some of which are aggregated as larger food patches.

The model predicts that larger species will detect less total volume of food (only the larger patches) but can tolerate lower resource concentrations within their food, while smaller species detect more food (many small food patches), but require higher resource concentrations within it. These foraging rules can be incorporated into a simple model of population persistence to find thresholds of the smallest food patch size (P*) and resource concentration (R*) for species of different size. The model predicts that P* scales positively with size, while R* scales negatively, such that the two thresholds trade-off. Thus, species of different size each have exclusive niche of food patches of a particular size and resource concentration.

Applying these scaling laws to a group of species using similar resources, we find a "packing" rule for how close in size species can be, i.e., the size ratio g between species of adjacent size. The body size ratio g should is predicted to decline with increasing organism size, rather than be constant. Because the distributions of food and resources are fractal, the small resource-rich patches needed by smaller species occupy proportionately less total volume than larger, resource-poor patches used by larger species.

We tested this size ratio prediction with body size patterns of species guilds that use similar resources. These include co-occurring, East African grazing mammals that all eat primarily herbaceous plants, grasshoppers from a Minnesota prairie, desert granivores (ants, birds, and mammals) from southeastern Arizona, and vascular plants from a Great basin desert and Minnesota oak savanna. In every case, size ratios in these very different assemblages declined significantly with increasing size, and the relationships are consistent with the shape predicted by the model.

This "packing" rule dictates the number of species ranging in size from w_min to w_max that can coexist in a landscape of a certain size and pattern. The maximum size, wmax, is determined by whether there is at least one suitable patch of size P* and resource concentration R* in a finite space of extent x. A minimum resolution for a species within its environment, wmin may ultimately be set by physical constraints to a particular body plan or prey size (e.g. vertebrates, plankton, etc.). Species richness (S) is then the number of exclusive niches allowed between w_min and wmax, approximately

where w_avgis the mean body size in the guild. The functions implicit in w_max (Eq. 3) and g(w_avg) mean that the model also incorporates the effects on species richness of landscape extent and the quality and fragmentation of habitat.

This model yields two surprising predictions. First, it predicts a left-skewed, unimodal distribution of species richness vs. organism size. This distribution reflects the larger size ratios and thus looser species packing required for smaller species and the limitation of the largest species by the maximum habitat patch size in the environment. The species richness-size distributions of our herbivore, granivore, and plant communities are all significantly left-skewed and fit the pattern predicted by the model. These distributions also differ from the log-normal or right-skewed distributions most commonly reported for species grouped by taxa or biogeographic region. Our model may not apply to communities that include species that use different resources or different habitats. Virtually all observed log-normal distributions combine diversity-size distributions of separate guilds (e.g., nectarivores, granivores, herbivores, carnivores) or species from different habitats.

The model also predicts the most commonly observed unimodal pattern of species richness vs. productivity, namely that species richness should increase rapidly and then decline gradually in response to increased productivity. As resources become more abundant, maximum patch size rapidly increases to allow larger species to exist. However, further increases in resource abundance cause food patches to coalesce, eliminating small, resource rich patches and requiring greater size separation among smaller species. Once again, the model’s predictions are supported for the herbivore, granivore, and plant communities we examined. Although many models have predicted some sort of unimodal response of diversity to productivity, most do not predict this particular shape of curve. Interestingly, however, simulations of previous models that consider resource concentrations to be heterogeneous in space do predict these shapes. Our model and others suggest that the changes in spatial distribution of resources that may be a key component underlying diversity-productivity relationships.

A final model prediction is that species diversity should increase with the structural complexity of habitat, as judged by its fractal dimension. We tested this prediction with biological survey data for plants, dragonflies, butterflies, and birds in Dutch heathlands. Diversity of birds and butterflies strongly supported these predictions, but response of the diversity of dragonflies and plants to habitat fractal dimension was much weaker. Assuming that habitat complexity can be quantified unambiguously with fractal geometry, our model suggests an explicit hypothetical mechanism for well-known patterns of species diversity vs. measures of structural complexity, e.g., foliage height diversity. However, the dispersal capabilities of different species may influence the strength of its effect.

Our application of spatial scaling laws to predict species diversity suggests that many of the mechanisms controlling biodiversity may emerge from simple that organisms may use to find resources in space. The analysis formalizes earlier ideas that diversity depends on the number of spatial niches, and suggests that coexisting species cannot infinitely partition space. In addition, the model synthesizes recent ideas about how resource acquisition and landscape pattern influence diversity. Clearly, other factors, including diversity of resource types, disturbance, and biogeographical history are also important. Colonization limitation undoubtedly plays an important role for more sedentary species such as plants, and allows for coexistence of "ecological equivalents," or species of the same body size. Because habitats become increasingly differentiated and exclusively used food patches become increasingly isolated as landscape extent increases, colonization limitation may become an increasingly dominant factor in controlling diversity at larger landscape scales. Nevertheless, the spatial scaling of resource use by species of different body size may explain many species diversity patterns across a range of spatial scales and taxa. The implied mechanisms and patterns can be readily applied to dryland environments, where the distribution of water and primary productivity should be highly correlated with species diversity.

The International Arid Lands Consortium (IALC) was established in 1990 to promote research, education, and training for the development, management, and restoration of arid and semi-arid lands throughout the world. One activity of the IALC members and their cooperators is supporting research, development, and demonstration projects that leads to a better understanding of dryland biodiversity, and management of fragile dryland ecosystems for sustainable human use while sustaining or enhancing biodiversity properties. Topics of interest include those relating to nutrient cycling, habitat fragmentation, rehabilitation of degraded dryland environments, ecosystem restoration, and management.

Introduction

The IALC is a partnership of organizations dedicated to research, education, and training relative to development, management, restoration, and reclamation of arid and semi-arid lands throughout the world. Member institutions are the University of Arizona, New Mexico State University, the Jewish National Fund, South Dakota State University, the University of Illinois, Texas A&M University-Kingsville, and the Desert Research Institute, UCCSN, Nevada. Affiliate members are Egypt's Ministry of Agriculture and Land Reclamation, Undersecretariat for Afforestation, and Jordan's Higher Council for Science and Technology. Collaboration with cooperators from other institutions, including the USDA Forest Service and Cooperative State Research, Education, and Extension Service, and those from other countries that strengthens the mission of the Consortium, is also desirable. Resources and expertise brought together by the IALC are generally greater than those offered by a single institution, serving to facilitate efficient exchange of research technologies. The IALC brings together people and programs in the important fields of land reclamation, land use, water resources conservation, water quality, inventory technology, ecosystem processes supporting sustainable management, and enhancements of sustainable management of ecosystems. Funding research, development, and demonstration projects gives assurance that experts are researching and disseminating their results in the most efficient and effective manner possible.

Our perception of biodiversity

Biodiversity has many meanings. For this chapter biodiversity will be defined as the variety and abundance of life forms, processes, functions, and structures of plants, animals, and other living organisms, including the relative complexity of species, communities, gene pools, and ecosystems at spatial scales that range from local through regional to global. Related to biodiversity are questions as to what is the role of biodiversity in the function of drylands. Also, what are the critical components of biodiversity necessary for long term functioning of the system.

Research and development projects

Research and development projects lead to new knowledge and technologies for the management of sustainable forest, shrubland, grassland, and agroforestry in arid and semiarid ecological systems. Ecological systems include humans both as part of bio-physical components and socially. The IALC supports research and development projects in the following areas and their socioeconomic dimensions: land reclamation, land use, soil and water resources development, soil and water resources conservation, water quality, inventory technology, ecosystem processes supporting management for ecological systems, and enhancement of management of ecological systems.

Since 1993, the IALC has funded 44 research and development projects, of which 20 relate to biodiversity concerns. Nine examples of IALC-sponsored research and development projects provide a basis for increasing our understanding of dryland biodiversity are reviewed (Hegwood 1998).

1) The relationships among desert patchiness, resources, production and diversity: implications for rehabilitation of desertified environments

2) Patchiness and nutrient cycling in arid lands of the USA and Israel

3) Southwestern borderlands grassland ecosystem-restoration

4) Redistribution of resources in drylands of the USA and Israel

5) Role of seedbanks in the management of semiarid rangelands under grazing

6) Microbiotic crusts: their nature and establishment in Israel and New Mexico

7) Disturbance, seedbanks and vegetation dynamics of desert annuals

8) Effects of habitat fragmentation and patch alteration on desert lizards

9) Banner-tailed kangaroo rats and shrub invasion in desert grasslands.

Results have revealed the danger in underestimating the importance of any organism in the structure and function of dryland ecosystems. Discoveries concerning the roles and complex interactions of organisms in dryland ecosystems continue at a rate suggesting that our understanding of these systems is far from complete.

IALC-supported demonstration projects represent practical applications of available knowledge and technologies derived from research and development efforts for the management of sustainable ecological systems, clearly identifying the benefits derived and clientele served. Demonstration topics eligible for support are similar to those specified for research and development projects. Other conditions that demonstration projects must meet to be supported by the IALC have been outlined elsewhere by Ffolliott et al. (1998). A total of 16 demonstration projects have been funded by the IALC since 1993, with 4 oriented towards demonstrating knowledge and technology derived in part from research on biodiversity. Three examples of IALC-sponsored demonstration projects contributing to dryland management that enhances biodiversity are reviewed (Hegwood 1998).

1) Information visualization for the management of arid lands vegetation.

2) Savannization: linking ecological understanding and applications.

3) Savannization demonstration project in Chile: management implications.

A particular focus of the IALC has been woody plants in drylands. In order to properly manage drylands for wildlife and grazing, both the competitive and beneficial interactions among dryland trees and other components of their ecosystems must be understood. Overgrazing, unsustainable harvesting of trees for fuel and fodder, and range "improvement" via woody plant removal have altered the balance, structure and composition of dryland savannas. Research and demonstration projects are elucidating the role of dryland woody plants as nurse species for other plants and as critical habitat for a range of dryland animals.

Summary

The IALC works to achieve research and development, and demonstration projects, educational and training initiatives, workshops, and other technology-transfer activities applied to the development, management, restoration, and reclamation of arid and semi-arid lands in the world. All activities are supported by the IALC's member institutions through efforts aimed at sustaining arid and semi-arid land ecological systems and human populations inhabiting these systems. Included among these activities are efforts relating to dryland biodiversity.

Bibliography

Biodiversity is one of the principal pillars of natural ecosystems. It is a manifestation of the various biotic and abiotic components of the ecological systems and their mutual interactions. Biodiversity also expresses the impact of disturbances. In this paper we would like to present a third dimension that affects biodiversity: the human intervention through management.

Biodiversity can be treated at different ecological levels: genetic, organismal (species) and habitat. Using 5 case studies from the Middle East (Israel and Sinai) we attempt to demonstrate how water management and livestock grazing play major roles in creating patchy, productive and relatively stable environments. The perspective of biodiversity, in this paper, is related to the landscape structure or, more specifically, to the network of natural and human made habitats, that support the fauna and flora characterizing each location.

Three of the case studies: Mt. Sinai (SIN), Sede Boqer (BOQ) and Lehavim (LEH) deal with traditional subsistence, where contemporary activities carry on an ancient practice. The fourth and fifth case studies, Ramat Hanadiv Park (RHN) and Sayeret Shaked Park (SHQ), deal with modern society and one of its new needs: natural recreation sites. In all 5 locations the natural landscape is composed of a mosaic of habitats. However, human activities have promoted and accentuated this spatial pattern into a more complex and diverse environment, hence affecting biodiversity at all levels.

The Mt. Sinai area is located at the center of a mountainous region in the southern part of the Sinai peninsular. This is a high-altitude ecological island (few peaks are more than 2000 m above sea level). This island receives higher amounts of rainfall than the surrounding area, on average 100 mm per year compared with only 10 mm in the surrounding area. It is one of many peaks rising 2000 m or more above sea level. The area is mainly of 2 types of rocks - 'black' (volcanic and granite) and 'red' (mostly red granite). The cracked black rock is characterized by large, soil-covered slopes and creates a relatively dry environment. The adjacent red granite produces huge exposed rocky slopes that promote considerable amounts of runoff after a rainfall. Consequently, soils of the internal valleys and dry river beds enjoy an improved water regime. The higher soil moisture supports a unique flora with many species of northern climate origin (Irano- Turanian, Mediterranean and Alpine). The ample water table enables the local Bedouin to practice orchard agriculture, mostly of Rosaceae fruit trees. Apparently, the agricultural potential of the red granite outcrops facilitated the settlement of many monks in the Mt. Sinai area during the Byzantine period (400-600 A.D.). These settlements (chapels, farms and monasteries) are almost exclusively located on the red granite, as are the orchards. The orchards provide a favorable environment for non- desert fauna and serve as a stop-over site for migrating birds.

Goat herding, the second source of traditional subsistence, takes advantage of the heterogeneous environment by using the elevation-geological spatial changes in an adaptive temporal sequence that provides forage resources almost all year long. However, these spatial movements are strongly correlated with the random distribution of rainfall cells in the desert that determine the location of abundant pasture in each season. The long history of grazing has, most certainly, altered the natural patches by selecting for unpalatable perennial species (mostly shrubs) and for annuals with a short life cycle and adaptive mechanisms to cope with high grazing pressure.

The Mt. Sinai case study demonstrates how a heterogeneous environment is manipulated by humans, using limited technology, to support a productive agro-pastoral system. Patches of high and unique biodiversity in the Sinai area are maintained by the special water regime on the red granite landscape. Superimposed on this spatial structure are the rich, human-made patches (orchards) that enhanced the regional biodiversity. Skillful exploitation helps local people to maximally benefit from the goods provided by the natural ecosystem.

The Sede Boqer site is located in the central Negev Highlands where average annual rainfall is about 100 mm. A main factor controlling biological activity is the redistribution of rainfall water by runoff flow. Where runoff water infiltrates into the soil there is greater species diversity and productivity than in locations not receiving the additional water. An important factor in controlling the distribution of runoff water is the physical patchiness of the area. Patchiness is determined by the ratio of stoney outcrops to exposed soil. In general, the stoney outcrop patches supply runoff while the soil patches act as sinks. Studies in the area show that analysis of the landscape as a mosaic of soil and stoney outcrop patches may be an indicator of the species diversity and productivity of the area.

In this area there is evidence of extensive human activity for the past thousands of years. This activity is in the form of grazing by herds owned by nomadic tribes and by terraced runoff agriculture. We do not know enough about ancient grazing in these areas, but undoubtedly the herders coordinated the movement of their herds spatially and temporally in accordance with the spatial and temporal distribution of productivity in the area. Ancient runoff agriculture has been more intensively studied because of the archaeological evidence for runoff farming. All of the central Negev landscape is covered with the remains of ancient channels and dams for diverting runoff water to the agricultural fields. When studying the complex network of channels and dams it is clear that the ancient farmers understood the relationship among landscape mosaic, runoff generation and soil moisture. The remains of ancient agricultural systems are not a functional, but they still have an effect on natural biodiversity.

The Lehavim site is located in the northern Negev in a semi-arid area (~300 mm). This site represents the vast dry rangelands that have been used by the Bedouin pastoralists over, at least, the past 4 millennia. As a matter of fact, this traditional subsistence system is also an agro-pastoral one. However, agriculture such as growing barley is rainfall-dependent. The long history of grazing made the livestock impact (vegetation consumption and trampling) an integral part of the ecosystem rather than a source of disturbance. The patchy nature of the local environment, partially created by the grazing activity itself, helps to cope with and mitigate the grazing impact. The Lehavim landscape is composed of 3 types of patches: exposed rocks, shrubs and crusty soil. The rock and crusty soil function as a source for runoff while the shrubs' understorey act as a sink for most of the water. Seeds also flow from the herbaceous vegetation growing in the soil patches, into the shrubs or the dense cover of prostrate plants. Shrubs also act as a refuge for herbaceous plants, thus helping them escape grazing before seed set.

In Lehavim, we see again a spatially heterogeneous structure of the natural environment that has been manipulated and extended, in terms of complexity by the pastoral activity.

Sayeret Shaked Park is located in the northern Negev and is under an average precipitation regime of 200 mm/annum. In this area, as in most of the Negev, most of the vegetation production and diversity is determined by annual herbaceous plants. The landscape mosaic is characterized by a matrix of soil crust, covered with a microphytic community, and small patches of dwarf bushes. The microphytic matrix is composed of cyanobacteria, bacteria, algae, bryophytes and lichens. The crust patches are sources for soil materials and runoff water. Shrub patches are sinks and the principal loci of productivity and diversity. This is mainly because of the accumulation of soil, water and nutrients, which promote the growth of rich herbaceous vegetation under the shrubs.

This area has been utilized by humans for thousands of years. Humans have been using the ecosystem production mainly for livestock grazing, firewood and construction. Uncontrolled use of the ecosystem has caused reduction of shrub cover and soil erosion. This prolonged process of exploitation has resulted in increased crust cover and decreased shrub cover.

The Forestry Department of the Jewish National Fund (JNF) decided to carry out management activities in the desertified areas of the northern Negev, in an attempt to convert them into recreational parks. The ecosystem management objectives of the JNF are to develop management practices to reduce the leakage of resources, in order to increase productivity and diversity in desertified areas. The main technique is to build contour-line catchments which collect water from 10 – 15 m upslope of the catchment.

A management solution for decreasing leakage of resources is to create sink patches. Sink patches are characterized by two features. The first is a structure which prevents the flow of runoff water, soil, organic matter and nutrients from a landscape unit. The second is a storage unit, which maintains and absorbs water, organic matter and nutrients.

These management activities increased the value of the land for the local populations. These modified landscape systems, located near urban settlements, serve as green belts. Scenic roads, walking trails and observation points fulfil the needs of the nearby urban population for open landscapes. They also serve as rangeland for shepherds.

The Ramat Hanadiv Park is unique in this discussion since it is located outside of the desert. This park extends over the southern tip of Mt. Carmel in central Israel, a mediterranean climate (600 mm/year) region. Consequently, the characteristic vegetation is different than that of the other locations discussed. Ramat Hanadiv Park is covered, almost completely, by woody vegetation, some of low stature and some of moderate-sized shrubs that create dense stands.

The natural physical environment in the park is not homogeneous and the different rock-soil patches produce varying conditions, principally of soil moisture, over short distances. However, long-term human activity, fire, wood-cutting and land clearing for cultivation, enhanced the patchiness of the environment and in some cases determined the patchy nature (e.g., afforestation). The final outcome is a complexity of patches with differing vegetation formation.

The local fauna and flora respond to the network of patches making the spatial structure a principal management issue. Dense vegetation creates a fire hazard that may result in burning of the whole park (as occurred in 1980). It is also undesirable, where flowers, especially the attractive geophytes, are concerned. However, when examining the habitat preferred by the noctornal fauna -wild boar, small carnivores and the newly-introduced roe deer. it turns out the the patches of dense Mediterranean scrubland is of outmost importance. Therefore, park management should consider the maintenance of a patchy environment as its principal goal. In other words, biodiversity is maximized through the optimization of habitat complexity.

A comparative analysis of the 5 case studies presented reveals that they all are heterogeneous. The patchiness of these environments determines the nature of flows in the ecosystem, and most significantly in this dry region, the flow of water. Human activity identifies the natural hetergeneity and manipulates it for a better use. In principle, there is no difference whether such use is for subsistence: traditional agro-pastoralism or modern recreation and conservation (table 1).

Table 1: Comparison of the five research sites in relation to biodiversity and management.
 
 

	SIN	BOQ	LEH	SHQ	RHN
Patchiness	physical	Physical	physical & biological	Biological	physical & biological
Landscape function	source-sink relationships between slope and valley	source-sink relationships between upper and lower slope	source-sink relationships on small scale	source-sink relationships on small scale	source-sink relationship within the geological structure
Landscape modification by humans	Creation of source and sink patches for water	creation of source and sink patches for water	shrub clearing	creation of sink patches for water	shrub clearing and reintroduction of livestock grazing
Processes controlling landscape diversity	geological and geo- morphological	geological and geo- morphological	geological and geo-morphological. shrub – crust relationships	geological and geo-morphological. shrub – crust relationships
Processes controlling species diversity	water redistribution imported resources (orchards)	water redistribution	water redistribution, seed redistribution	water redistribution, seed redistribution	water redistribution, seed redistribution

Bibliography

Why do we need focal species?

Ehrlich and Roughgarden (1987) defined ecosystem as "all the organisms in an area with the physical environment in which they interact". One objective of conservation is to conserve all organisms and processes within a system all at one time (Simberloff 1998). A key component for meeting this objective is maintenance of biodiversity. This may be achieved by monitoring and managing each species separately, or by using a holistic approach in which the entire system and the processes within it are evaluated and managed. Both approaches have severe shortcomings that make their feasibility doubtful. The first approach requires a prohibitive amount of effort in keeping track of all organisms within the system, while the second requires an unrealistic in-depth understanding of all processes and interactions in the system.

It is clear that we cannot (yet?) identify all the components of an ecosystem (biotic and abiotic), and quantify the relationships between them. At most, we can identify most of the multi-cellular species, and detect the major ecosystem processes. Therefore, we are searching for shortcuts, so that by monitoring and managing selected parts of the ecosystem, we will conserve the entire ecosystem. If the part(s) of the ecosystem we are managing (the focal parts) are central, we may, theoretically, assume the rest of the ecosystem is being conserved and, therefore, ignore it. While this approach is very appealing because it saves time and money, most critiques argue that a single focal species approach is not comprehensive enough. As a result, the status of many important components of the system is not reflected by the status of the focal species (Lambeck 1997, Niemi et al. 1997), and problems in the system may go unnoticed.

We may, however, choose between conserving a focal species, that will protect the entire ecosystem, or conserving an ecosystem process, that will do just same. The issue of which of these methods is preferable is yet unresolved. However, it is most likely that the answer to this varies depending on the system and how well it is known. In our discussion we focus on focal species only.

Definitions of focal species and management implications?

Focal species can belong to one of the following categories:

Indicator species are species that testify to the well being (health) of the entire ecosystem. These species are the first to respond to deterioration of the ecosystem, since they are more sensitive. By monitoring only the indicator species population, we can estimate the welfare of the entire ecosystem. We assume that as long as the status of the indicator species is satisfactory, the entire system is operating adequately. If changes are observed in the population of the indicator species (population size and distributions, age distribution, etc.) we can implement various management protocols directed at the species that will counter these changes and these actions will improve the status of the entire system.

It is not clear, though, what a healthy ecosystem is (Simberloff 1998), and what are the changes that the indicator species should be sensitive to. Also, to be sure that no part of the ecosystem is damaged before effects are exhibited by the indicator species, it is necessary to compare the sensitivity of all the species in the ecosystem to a set of environmental changes. In any case, management in such situations is reactive rather than proactive.

Keystone species are species that play an important role in the ecosystem by influencing many of the other species and/or processes of the ecosystem. If the population of a keystone species is removed from the ecosystem, the ecosystem changes dramatically. Therefore, by managing only the population of the keystone species, we are managing most of the ecosystem. Species that hold most of the biomass in the ecosystem may also be considered keystone. Although it is often argued that unless their biomass is associated with a considerable impact on the system, such species can not be considered keystone, we will consider such species in our discussion.

In many ecosystems, we still do not know what the keystone species are. Furthermore, it is not clear whether such species exist in all ecosystems. Many species may be keystones for only parts of the ecosystem. For example, a tree that its fruits are consumed by all frugivorous birds and does not have a major influence on other parts of the ecosystem (the rest of the avifauna). It can be argued that in similar cases, by definition, this species should not be treated as a keystone species.

Umbrella species are species that require a large area containing many types of habitats to sustain a viable population. Often these are large-bodied homeotherms with large home ranges. By securing a large enough tract of land to sustain a viable population of these species, many others will come under the same protection (UNEP 1995). The focus here is on enough area for the umbrella species and other members of the ecosystem, but there are no direct management implications for this approach.

Flagship Species are species that attract public attention by being charismatic, colorful, or behaviorally intriguing. These species are then used to gain public support to protect their environment and in doing so achieving the protection of the ecosystem they inhabit. However, if the flagship species cannot serve as an umbrella species as well, we might neglect many species and functions of the ecosystem, and eventually harm the flagship species itself. Furthermore, occasionally management of flagship species may come at the expense of other organisms with which they share resources.

How to select good focal species?

Lambeck (1997) suggested to select from the entire species pool in an ecosystem, those species that are not protected and secured, and to rank them according to the threats the ecosystem is facing. The best candidates for focal species are those that are most likely to suffer from these threats. This approach suffers from two major flaws. One, we usually do not have enough information on all species in the ecosystem and it is presumptuous that we can obtain it. Two, Lambeck considered only species that are not secured in the system, in contrast to the concept of using focal species for conservation. For example, the conservation of a keystone species is a proactive measure for protecting an ecosystem even if the species is currently secured.

Depending on the type of focal species, different sets of considerations should be applied. These considerations are not necessarily mutually exclusive. Unlike other types of focal species, the considerations for choosing a good flagship species are not scientific, but concern economic and public opinion issues.

The first step in selecting a good indicator species is defining a healthy system. Then, we should define the major threat the ecosystem is facing, and find the species that is most likely to be the first to respond to this threat. Another condition is that we can easily detect the response of the indicator species to changes in the environment.

We can identifya good keystone species usually after the ecosystem was manipulated, either experimentally, or as a consequence of the way we managed the ecosystem. There are no criteria that define how much of the ecosystem has to be changed when manipulating a population so that this species can be considered a keystone species, and it is out of the scope of this paper. Therefore, the definition of a keystone species remains the decision of the land manager. Once a keystone species is identified, its population must be monitored, and when necessary it must also be managed.

By definition, umbrella species should be used only when the major threat to the environment is habitat destruction and fragmentation. Species that requires wide and heterogeneous habitat are most likely to respond first to further fragmentation of the habitat, and in that respect umbrella species are a special case of indicator species. Usually, large bodied homeotherms will fit these criteria.

Focal species and drylands

The major question of whether conservation efforts directed at a single 'focal' species can protect and conserve entire communities or ecosystems, is yet unresolved. Clearly, the simplicity, in terms of evaluations and management protocols, of using focal species makes this approach attractive. However, as previously mentioned, the reliance on focal species is subject to criticism directed at the non-comprehensiveness of the methodology. The unique characteristics of dryland ecosystems makes them and populations inhabiting them (Saltz et al. 1999), in certain aspects, more amenable (relative to other ecosystems) to the focal species approach and less amenable in others. Here we discuss various characteristics of drylands and how they might affect the various focal species approaches.

Productivity - Drylands are typified by low productivity, making such systems 'simple' and 'understandable' (Kotler and Brown 1988). The relative simplicity of drylands may facilitate the identification of keystone species. Low productivity also dictates that larger animals either require large tracks of lands, or food generalists (and use various types of physical conditions), or both. As a consequence, umbrella species are easily identified and could be used effectively for conservation purposes.

Variability over time and space - Due to rainfall and runoff patterns, desert are highly variable over time and space. The high variance makes it difficult to discern trends from stochastic effects (Saltz et al. 1999) limiting the applicability of indicator or keystone species. Some of the variability and patchiness in drylands is due the creation to micro-sites formed by species that improve their micro-environment. Because of low rainfall and productivity typical of drylands, any modification induced by organisms (either for improving their own micro-environment or while foraging), remains intact over long periods of time. For example, porcupines (Hystrix indica) dig up to 2-3 holes m^-2, that are noticed in the ground for long periods, and create favorable sites for seed germination (Yair & Rutin 1981; Gutterman 1982). This type of ecosystem engineering by organisms (Jones et al. 1994) impact other species in the ecosystem. As a consequence, the impacts of engineer-species are expected to be longer-lasting in drylands than in wetter environments and, therefore, functionally more important. Thus, engineer-species in drylands are expected to be good keystone species.

Differentiation between ecosystems - The borders between different types of habitats in arid and hyper-arid regions are often sharp and clear. For instance, the green veins running along dry riverbed are easily distinguishable from their surroundings. In other environments it is often difficult to delineate the border between various plant communities. Thus, in desert ecosystems, at least along wadi-beds, it is relatively easy to define the range over which a specific species plays an important role.

Rarity - Because of the low productivity and patchiness of deserts, many desert organisms fall under at least one of the three types of rarity (limited range, low densities, and habitat specificity - Rabinowitz et al. 1986), and have evolved unique adaptations for surviving in deserts. For example, large vertebrates are found in densities that are much lower than in other regions, while most small vertebrates are restricted to the narrow green veins of the dry riverbeds. These traits make many desert organisms good candidates for indicator species.

Acknowledgments: We thank Dan Simberloff and Walt Whitford for their comments on earlier drafts of this abstract.

Bibliography

People who care about nature have divided it into two parts: Nature, and the rest of the sorry lot. But that is a serious mistake. We can -- and we must -- turn the habitats of the world into a continuum. At one end will be wilderness, at the other, the ugliest works of Homo sapiens. Then we can push habitats along that continuum in the direction of the beauty and sustaining power of nature herself.

Until we do, diversity will be doomed to suffer, because, for all practical purposes, we have actually shrunk the Earth. Wherever we have planted our crops, farmed our timber, settled our cities, we have created emptiness. Most native species cannot live with us.

An ominous example comes from the rainforests of southeastern Brazil, a place with the world's largest tree diversity (450 spp/ha). We have taken more than 95% of this area. Such loss typifies the world's ecosystems.

In the USA, some habitats are gone almost entirely. For instance, less than 0.1% of its prairie remains. The rest have become farms. Many USA forest types have suffered too, especially the longleaf pine ecosystem. It once covered some 36 million ha. Today only 2000 ha remain.

Some habitat types, such as desert, have not done so badly. The average amount of US forest remaining in natural condition is actually 4.6%. The average percentage of world-wide natural habitat that remains is about 5%. Thus, I will use species-area analysis to calculate the diversity loss caused by loss of 95% of the land. This analysis depends on three sorts of species-area curves.

Species-area curves within provinces. Larger areas within a biogeographical province contain more habitats. Because many species are restricted by habitat, the result is a species-area curve (slopes -- or z-values -- between 0.15 and 0.2).

Island species-area curves. Larger islands have more species than small ones, because smaller islands have fewer habitats. But all islands of area A have fewer species than a similar mainland patch of area A. The island has fewer because islands, by definition, are too isolated for species to rescue their sink populations with regular immigration. Islands thus lack sink species.

Interprovincial species-area curves. Larger biogeographical provinces have more species than small ones (slopes between 0.6 and 1.5). The interprovincial pattern depends on the dynamics of speciation and global extinction. At any diversity, S, a larger province will have a higher rate of speciation and a lower rate of extinction compared with a smaller province. At the provincial scale, habitat diversity does not limit species diversity. Instead, habitat diversity evolves in proportion to the steady-state value of species diversity.

So, the three patterns of SPAR (species-area curve) -- within a province, among islands, and among provinces -- exist at three different time scales. The pattern among continents depends on the slow pace of speciation itself. The pattern among islands depends on the immigration rates of species that occasionally go extinct, but ordinarily can reproduce successfully on an island. The pattern within provinces depends on the rapid and regular rescue of sink populations. That is why SPARs within provinces have the shallowest slopes, whereas SPARs among provinces have the steepest slopes. We can tell SPARs apart by their z-values.

The science of species-area relationships predicts a massive, three-phase extinction because we shrank the natural world.

First we lose the endemics. Endemics are the species whose entire territory vanishes under asphalt and farmfield. If we expropriate 95% of a biogeographical province, we expect to exterminate outright about 25% to 40% of its species. But fragmentation (!) and careful siting of nature reserves have reduced this loss considerably.

Second, we lose the sink species. These are the species whose few populations all live in habitat that is less than adequate to sustain generation-to-generation replacement. Sink species have diseased demographics. They are extinctions-in-waiting. We can closely estimate the fraction of sink species by using island SPARs. Assuming a loss of 95% of natural area, some 50% to 60% of species will be sink species waiting to disappear. We buy their extinctions on the installment plan. We shall pay out our bill in tens to hundreds of years.

Third, inevitable accidents will claim species that ordinarily have healthy demographics. No world of the past has escaped such accidents, and neither will any world of the future. Because mini-provinces have depressed speciation rates, most of these accidental losses will not be replaced. The third phase will end when the rate of accidental extinctions has diminished to match the very small rate of speciation that our remnant earth can support as a mini-province. Our world will support about the same proportion of species as she has saved land area. If we save 5% of the land, we will have 5% of its diversity.

The third phase of the mass extinction is the slowest phase, perhaps taking on the order of 100,000 years. But global warming may greatly speed up the effective accident rate. Global warming pulls the climatic rug out from under our reserves. Species will lose the habitats they need to maintain source populations. The decline to a depauperate steady state of diversity may not take very long at all.

Today, conservation biology battles to save species by using two dominant tactics: Reservation ecology and Restoration ecology. Given the power of area however, these two cannot do much by themselves. No conservationist seriously believes that we can reserve much more than the 5% or so that now remains. And many will admit that the human population is likely to continue expanding. So, conservation biology must learn to use the land we occupy. The species-area relationship teaches that the rest -- no matter how magnificent -- amounts only to crumbs, the leavings of a rich and technologically powerful society.

We must learn how to reconcile our uses of the land with its use by native species. Reconciliation ecology is the science of inventing, establishing and maintaining new habitats to conserve species diversity in places where people live, work or play. Reconciliation ecology is the third ‘R’ of conservation biology. It means meeting nature halfway. And there are already many examples of successful reconciliation ecology around the world.

Longleaf pine forests once covered about 36 million ha of the US coastal plain from Texas to Virginia. Longleaf pine defines a whole ecosystem supporting dozens of other species. Many of these can survive nowhere else.

Most of areas of longleaf pine have been timbered and changed to forest plantations. In 1992, as little as 1 ha in every 18000 ha of old-growth longleaf pine remained. Due to lack of fire, longleaf pine is disappearing even from its reserves where its competitors, especially hardwoods, are taking over.

Eglin Air Force Base, Florida, contains one of the few remaining stands of longleaf pine. Eglin encompasses 188,000 ha. Most of it used to be pineland.

In 1992, Eglin's managers invited The Nature Conservancy to survey Eglin's rare plants and animals. More than 90 species live on the base. A number of these are of international concern, such as red-cockaded woodpeckers and Bachman's warblers. Eglin's four small areas of old-growth longleaf pine totalled 700 ha acres. They were heavily infested with various species of oak trees in the understory. The pines were doomed.

The Air Force began removing large numbers of other pine species and planting more than 3 million longleaf seedlings. And they began burning the understory of the forest. In 1996 alone, they burned more than 10,000 ha. Longleaf pine now dominates 81,000 ha of Eglin Air Force Base.

Meanwhile ecosystem managers have also been able to help the red-cockaded woodpecker. This woodpecker was declared rare and endangered in 1973, and its population keeps dropping. During the 1980s, it lost 23% of its few remaining individuals.

The red-cockaded woodpecker depends on mature, live, longleaf pine. Even in the presence of such pines, the woodpecker will actually abandon a forest that has too much oak growing under its old pines. It uses the pines for nest holes, often choosing old longleaf pine trees partly rotted by a fungus. Of all woodpeckers in North America, only the red-cockaded excavates its nest holes in live trees. It will not use a dead trunk or an artificial nest box.

But the woodpecker population could not wait until the new longleaf pine trees were old enough to get fungal rot. So Eglin's wildlife managers learned how to drill artificial nest holes in the pine trunks. The red-cockaded woodpecker now nests in 30% of the artificial holes. Its population has begun to grow. From 1995-1997 it increased 6% at Eglin Air Force Base where approximately 523 red-cockaded woodpeckers now live.

The Air Force has not made its property a wildland. Eglin is still an Air Force Base. It flies its planes, tests its munitions, and performs all the other functions for which it was set aside. Logging continues and recreation continues. But its human uses no longer ignore the wild species that can and do live inside its fences.

Moreover, the Air Force has to intervene to maintain the habitats. Fences alone will not work. People must start the fires that allow the pinelands to continue. If the managers relax their grip, all the old problems will return. The management activities at Eglin Air Force Base are reconciliation ecology writ large.

Sometimes reconciliation ecology is a happy accident. For instance, the healthiest population of American crocodiles in the USA lives in the cooling canals of an electricity generating station south of Miami, Florida. The power company has now hired biologists to protect the diversity of its property.

More often, reconciliation comes from good natural history, effort and a clever plan. One of these is Reuven Yosef's psuedo-salt marsh in Eilat, Israel. Eilat sits along a heavily used bird migration route. Without its salt marsh and its buzzing breakfasts, a crucial link in this route would be broken.

But Eilat is also a great resort city, a boom town with massive, luxurious hotels, shops and restaurants. Build, build. Birds versus builders. Today the natural salt marsh is nearly gone. But reconciliation ecology averted the evil decree.

Yosef took a tract of wasteland near the few remaining hectares of salt marsh. He designed a new environment with dikes and a channel to admit seawater. Under his supervision, volunteers planted various species of shrubs. As the construction shovels excavated the hotel sites, they moved the dirt into position to form the dikes. Soon there was an artificial salt marsh complete with open water areas, a rich oasis for migrant birds of many species.

Yosef's salt marsh teaches a crucial lesson about reconciliation ecology. Instead of costing money, it may make money. Eilat required builders to move their dirt far away into the desert. Expensive. But the city permitted the builders to dump the dirt locally if they would follow Yosef's plan. The builders saved. The birds were saved.

Reconciliation cannot work unless vast areas of land are reconciled. Thus, money is a fundamental consideration. It must not cost too much to initiate or maintain a reconciled hectare of forest, pasture or farmland. So much the better if reconciliation makes a profit. A number of examples suggest that it will. Shade coffee plantations support many more species than more 'modern' ones. Yet these plantations offer considerable profit now, and may be the key to sustainable profit in future. Extensive, profitable, old fish ponds in the Czech Republic create a large wetland where none existed before. They provide the last redoubt of otter, natterjack toad and many imperiled plant species.

The Earth is not really smaller than it was five centuries ago. We have just transformed it so much that most species have not yet evolved a use for it. So, we still have time to reverse our course and save a very large proportion of the world's species. To do so, we shall have to pay close attention to anthropogenic environments. Most cannot yet be called habitats. To change that, we must study the natural habitats of a wide variety of charismatic species and find out what allows these places to fulfill the needs of the species they harbor. We must then design a large variety of new habitats to support as many species as possible. We must even study the desires of humans -- more area will be reconciled if people actually like the new habitats that we design.

Introduction

As biodiversity, ecosystem function, and ecosystem services become more and more closely linked with human well-being at all scales, the study of ecology takes on increasing social, economic and political importance. However, when compared with other disciplines long linked with human well-being, such as medicine, physics and chemistry, the technical tools and instruments of the ecologist have tended to lag behind those of the others. This disparity is beginning to be overcome, especially with the increasing use of biotelemetric techniques, satellite and airborne imagery, geographic information systems (GIS) and both regional and global data networks. However, we believe that the value and efficiency of ecosystem studies can benefit significantly by more wide-spread use of existing technologies, or with the adaptation of technologies currently used in other disciplines to ecosystem studies.

Arid land grazing systems play critical roles in supporting human social systems and economies in regions all over the world, and can be expected to play increasingly critical roles as human populations increase. The economic and political implications of grazing in arid lands, the controversy over its impacts, and the pressing need for resolving the controversies and for creating sustainable grazing systems make arid land grazing systems a good example to illustrate potential applications of new technologies. We frame the discussion with the following questions:

• What kinds of data about arid land grazing systems, biodiversity and sustainability are currently needed by researchers?
• How are these data needs met currently?
• What are the gaps between data needed and data currently generated, and how could those gaps be filled by technological applications and innovations?

We will consider four categories of data needs for arid ecosystems: grazing animals and wildlife, vegetation, soils and climate. The ideas expressed in this chapter about the application of technologies to arid land grazing systems are intended to transfer easily to other ecosystems and to other research and management issues.

Data on grazing animals and wildlife

Critical data needed for maintaining biodiversity, ecosystem function and sustainable grazing systems in arid as well as mesic environments include data on the following questions: What routes and patterns do grazing animals follow as they move across a landscape, and how do they affect vegetation, soils and water? How do other species modify their movements and behavior in interaction with grazing animals? How are insect populations and behavior dynamics associated with grazing systems? How do grazing systems affect wildlife populations in neighboring regions?

The current standard for addressing these questions is in field observations. Though field studies provide the foundation for ecosystem studies, it is always constrained by various limitations on personnel, time and the spatial scale of study.

The rapidly advancing technologies of biotelemetry address these limitations by offering researchers the opportunity to collect huge amounts of data over great spatial scales on a surprisingly wide range of animal taxa. These technologies have been used on taxa from bumblebees and salamanders, to passerines and rodents, to snow leopards and humpback whales. Attachment of data storage or transmitting devices is made either surgically, or with collars and harnesses.

Data can be stored "on board", transmitted by radio frequencies to handheld or airborne receivers, or transmitted to the Argos satellite system and then back to researchers around the world. Miniature GPS units transmit precise locational data and allow tracking of movements. Harmonic radar records time-series locational data for insects mounted with <1mg metallic tags. Mini-video cameras mounted on large vertebrates record movements, habitats and behavior, and bioaccoustics uses microphone arrays to identify and track birds, elephants and whales.

The field of biotelemetry faces a number of challenges. Human commercial use of radio frequencies is resulting in a shortage of radio frequencies available for wildlife studies, and urban commercial use sometimes interferes with wildlife studies, especially near large cities. The improvement of attachment systems and the further miniaturization of microprocessors, transmitters, data loggers and cameras are all constant technological challenges. The effects of monitoring equipment on the behavior of the animals being monitored must be more well understood. Also, there is a need for the standardization of anesthetic and surgical attachment techniques, placement of monitors collecting physiological data, and animal welfare guidelines.

Data on Vegetation

Critical data on vegetation needed for maintaining sustainable arid land grazing systems includes data on the following: What are the rates of increase or decrease in vegetation biomass and/or net primary productivity associated with grazing? What are the changes in species composition that take place as a result of grazing? How are grazers affected by changes in species composition?

The standard methods for obtaining such data is still through field observations, however optical and multi-spectral remote sensing is being used increasingly to obtain spatially large-scale data on biomass, vegetation patchiness, changes in species composition and net primary production. The use of remote sensing for mapping distributions of microbiotic crusts is also being developed. Additionally, water quality and availability is often inferred from satellite data by vegetation distribution.

Common satellite imagery spatial resolution currently ranges from 20 -30 meters (SPOT, Landsat) and within a few years will be available from 1 - 4 m (Ikonos, Quikview, Orbview). Common satellite imagery spectral resolution currently covers from 1 band (B&W) up to 8 bands, including near IR and thermal IR. Prices per image range from approximately $500 to $5000, for images approximately 15,000 km². Airborne photography can provide sub-centimeter resolution (but costs are "per mission"), and hyperspectral sensing that covers hundreds of bands. Synthetic aperture radar (SAR) is unaffected by cloud cover and precipitation, and can collect data on vegetation and topography as well as soil surfaces and subsurfaces. Lastly, geographic information systems (GIS) allow construction of multi-layered images including data on vegetation, soils, geology, topography, precipitation, land uses, political boundaries, and more.

As the spatial and spectral resolutions of commercial systems improve, developing signatures for specific vegetation types will be a critical advance for ecosystem studies. More efficient, automated, and accurate in-situ techniques for field workers to identify signatures will become important tools. These tools could include optical and spectral analysis methodologies for near ground-level images to characterize vegetation type, tools for accurate assessment of vegetation volume, and nondestructive methods to accurately correlate biomass with vegetation type and volume.

Data on Soils

Critical data on soils needed for maintaining sustainable grazing systems includes data on the following: What are the soil structures, textures and chemistries, including salinity, pH, levels of organic matter, and nutrient levels? How are soil temperatures and moisture levels related to evapotranspiration? How are soil patches distributed across a landscape? How does grazing intensity affect soil characteristics, including microbial communities? How does grazing influence erosion rates?

Characterization of soil organic matter, soil texture, structure and chemistry (especially salinity and nutrient availability) currently involves laboratory analysis or laborious (and sometimes inaccurate) field techniques. In addition to being time-consuming and laborious, laboratory analysis entails sample collection, storage and transportation, which can introduce errors and remove site-specific contextual information from the analysis process.

The development of analysis techniques for near ground-level or aerial photography could be very useful in accurately characterizing soil rockiness and soil particle size. In addition, new micro-capabilities for chemical detection and analysis ("lab on a chip" technology) developed for applications such as in-situ chemical and biological weapons detection, could be adapted for assessing nutrient availability, soil organic matter, soil chemistry, and other ecological data needs. Such in-situ soil characterization could also be an important first step in developing signatures for remote sensing. These data – from existing meteorological stations and other in-situ technologies -- can be transmitted to local or satellite receivers and downloaded around the world. Lastly, ground penetrating radar and SAR can be used to map surface and near-surface soil structure and moisture content.

Data on Climate

Critical data on climate needed for maintaining sustainable grazing systems includes data on the following: How much precipitation occurs, how often, and over what spatial scales? How is net primary production correlated to precipitation in local landscapes? How are atmospheric data like barometric pressure, humidity, wind speed, wind direction, solar radiation and temperature correlated with precipitation patterns and evapotranspiration?

Current methods for collecting data on the questions above are already predominantly automated. Automated meteorological stations – including NEXRAD sites -- upload data to satellite systems such as the USGS Geostationary Observational Environmental Satellite (GOES). NEXRAD data in conjunction with lightning locator systems can be used to estimate precipitation amounts in the regions around the lightning strikes. Other satellite networks, mentioned below, also provide links to meteorological data. Integration of these data into user-friendly interfaces will make them more accessible for research and management of all kinds.

Data integration and analysis

Even with the widespread use of technology, better data integration and analysis will be critical for gaining better understanding at an ecosystems level. Current technologies might detect soil erosion and species composition changes, but early warning indicators for each are essential. New software applications will allow better data analysis and could help design studies that mesh with new data collection and processing technologies. Better integration and analysis of data can provide predictive power that can be used by land managers, ranchers, pastoralists and farmers for adjusting grazing load, rotating land uses, adding nutrients, building erosion barriers, fencing wetlands, and so on.
 
 

• long-term sustainability of land resource conservation;
• concomitant land use by protected or managed wildlife species;
• preservation of local floral and faunal biodiversity;
• other human concerns (residential, recreational, public health).

Baseline data needs for management of this hypothetical grazing land will require map coverages, using satellite imaging and GIS, for topography, soils, vegetation, political divisions, property boundaries, dwellings, transportation routes and agricultural fields. Baseline data will also include climatic records, including precipitation, temperature, humidity, solar radiation, wind speed and direction, and soil temperature and moisture at various depths.

At the Sevilleta, these data are collected at eight automated meteorological stations transmitting weather data every eight hours from several locations representing the major ecosystems and elevations. From these data, initial estimates of monthly potential evapotranspiration (PET) can be calculated and used in management models.

Estimates of NPP can be made from aerial or satellite images using plant cover or density estimates, or using normalized difference vegetation indices (NDVI). Some level of ground-truth initial sampling must be undertaken to develop these correlations for the sites, but once validated, the remotely sensed estimates could be used almost exclusively.

NPP is linked to precipitation; highly-localized summer thunderstorms can be tracked using lightning locator systems and NEXRAD radar data. Correlations between number and location of lightning strikes and total precipitation amounts can be established. (At the Sevilleta, 36,190 m³ of water falls within a 3 km radius of every cloud-to-ground lightning strike.)

Temporal and spatial changes in NPP can be monitored with frequent satellite images and correlated to climate with meteorological updates. Monthly satellite vegetation images can verify or correct model predictions, and remote weather stations can automatically download data every few hours. NPP calculations can be used to determine optimal livestock densities, and then stocking rates and pasturage can be adjusted frequently to optimize productivity and prevent overgrazing. Wildlife telemetry technologies can track livestock. Transmitters on a few dominant animals will track movements of an entire herd.

ADAR images from fixed-wing aircraft, satellite images and images from ground penetrating radar and SAR can monitor changes in mineral compositions and moisture levels in soils, identify surface erosion patterns, stream channel changes and arroyo formation, and provide early warning of erosion. Critical habitats (springs, marshes, ponds) can be monitored remotely for size, productivity, color, and if changes are noted ground teams can be dispatched or grazing regimes altered. Behavior and movements of wildlife and predators can be tracked via telemetry.

Landscape changes due to grazing can affect population dynamics of various wild species. For instance, grasshoppers and locusts can increase markedly in grazed systems, presenting pest threats to nearby agricultural regions. Populations of small mammals (rodents and rabbits) which might carry diseases (plague, hantavirus, Tularemia, rabies) can increase in grazing systems as well. Satellite imagery can detect changing vegetation and moisture that are correlated with pest population changes and provide data for predictive models of insect or disease outbreaks. Harmonic radar can track swarms of locust, and Landsat TM data have been used to construct GIS models of patterns of hantavirus infection in humans, and grasshopper outbreaks in rangelands.

Existing Networks

Various regional and international data sharing networks like the ones described in the hypothetical management scenario are currently accessible on the World Wide Web. The US Geological Survey’s GOES provides national near-real-time hydrology and international cloud cover data. The UN Environmental Programme’s Global Environment Monitoring System (GEMS) assembles global water quality data. The Global Climate Observing System, sponsored by the World Meteorological Organization (WMO), the UN Education, Science and Cultural Organization (UNESCO), the UN Environmental Programme (UNEP) and the International Council for Science (ICSU) assembles data on global atmospheric, oceanic and terrestrial observations. NOAA’s Argos system transmits biotelemetry and environmental data, and the Species Analyst is a network that assembles data on species populations and migrations from museums around the world.

The Global Terrestrial Observing System (GTOS) is a network of networks intended to link ecological and environmental data bases from all over the world to provide policy makers, resource managers, and researchers with data to detect, quantify, locate, and understand changes in the capacity of the land to support sustainable development. GTOS is an international system led by various UN agencies and organizations of the world scientific community that links existing monitoring sites and networks with satellite systems. Eleven international scientific networks and four space agencies are collaborating to develop requirements and standards for acquiring and integrating data.

GTNet, a GTOS pilot project, is designed to assess global primary land productivity. Data and imagery from satellites will be provided free of charge to countries in exchange for ground-truthing observations. The information provided by the countries will be used to verify models of primary productivity metrics such as land cover, net primary productivity and leaf area indices (LAI), and shared on GTOS web sites.

Conclusions

The timely and efficient collection, organization, analysis, dissemination and application of landscape-scale data is critical for the maintenance of sustainable ecosystems and the human social, economic and political system that rely upon them.

New applications of the technological advances of the last half-century, used in conjunction with conventional techniques, can dramatically increase the quality and quantity of data that can be collected and analyzed in ecosystem studies. The integration of new technologies into ecosystem studies will require new hardware, new software, new study designs, new funding programs and new training programs at colleges and universities. Finally, innovations that can increase data sharing and reduce costs of data collection and analysis are critical.

Heavy grazing is widely cited as the major cause of desertification in semi-arid and arid areas of Africa, Asia and Australia (see e.g. Acocks 1953, Jarman and Bosch 1973, Sinclair and Fryxell 1985). Nowhere is this effect of heavy grazing more apparent than in the Sahel of Africa (Sinclair and Fryxell 1985). It is now generally agreed that a combination of the following parameters has been the primary cause of land degradation in the Sahel:

(1) settlement of pastoralists around wells,

(2) the restriction in grazing area due to the expansion of agriculture into pastoralists' grazing lands, and

(3) the provision of supplementary feed during drought periods, which resulted in increased stock survival and greater depletion of plant resources during the post-drought recovery phase (Mainguet 1991, Sinclair and Fryxell 1985). This land denudation has resulted in a negative feedback loop via decreased soil nutrient status and increased soil albedo (due to lower vegetation cover). This caused increased evaporation and decreased precipitation, which in turn reduces the stocking capacity of the land, further exacerbating the negative effects of grazing (Schlesinger et al. 1990). However, this example of the negative effects of grazing in the Sahel is not the norm for arid ecosystems (reviewed by Forse 1989, Behnke and Abel 1996). Indeed, it is generally agreed that where pastoralists are able to maintain their activities on a large spatial scale by migrating to areas where key high resources can be exploited, such negative situations do not develop because previously-used resources have sufficient time to recover (Sinclair and Fryxell 1985, Ellis and Swift 1988, Behnke and Abel 1996). Moreover, even where pastoralists are forced to settle in small areas, abiotic variables such as precipitation may be of such over-riding importance that these negative effects of grazing can not be detected (see e.g. Ward et al. 1998).

A consensus has developed in recent years that arid grazing ecosystems are non-equilibrial, event-driven systems (see e.g. O’Connor 1985, Venter et al. 1989, Milchunas et al. 1989, Parsons et al. 1997). Ellis and Swift (1988), Tapson (1993), Werner (1994), and Sullivan (1996) contend that rainfall in arid regions is the major driving variable and has the ability to "recharge" a system that suffers heavy grazing pressure. Mammalian herbivore numbers are kept low during bottlenecks caused by droughts. Hence, their numbers seldom reach levels at which they can limit vegetation biomass (Ellis and Swift 1988). Furthermore, the cost of regrowth subsequent to herbivory to plants in arid ecosystems is high due to low levels of precipitation and low soil nutrient availability (Coley et al. 1985). This has resulted in selection in plants for resistance (or tolerance) to herbivory, thereby minimizing the impacts of organ removal on fitness. Thus, because of low herbivore numbers and resistance in plants in arid ecosystems, the effects of grazing in these ecosystems is expected to be low. A global review of the effects of grazing on ecosystems bears this out: effects on species composition and plant biomass increases with increasing precipitation (Milchunas and Lauenroth 1993). However, there are many exceptions to this overall pattern and much that still requires explanation, as we outline below:

(1) Grazing usually has greater effects on ecosystems with a long evolutionary history of grazing. Milchunas et al. (1988) predicted that a long evolutionary history of grazing results in selection for regrowth following herbivory and for prostrate growth forms. In such communities, grazing causes rapid shifts between suites of species adapted to either grazing avoidance/tolerance or competition. In their global review, Milchunas and Lauenroth (1993) found that increasing evolutionary history of grazing produced increasing dissimilarity in species composition between grazed and ungrazed sites regardless of the level of precipitation. However, in the Middle East and North Africa, where heavy grazing has occurred for thousands of years, grazing has seldom been shown to affect species composition (Noy-Meir et al. 1989, Perevolotsky 1994, Ward et al. 1999). A possible reason for this lack of grazing impact is the Narcissus effect, viz. selection in the past has resulted in the extinction of all non-resistant/tolerant genotypes. Thus, all extant species are similarly resistant to herbivores, resulting in the absence of an effect of current herbivory on biodiversity (Ward and Olsvig-Whittaker 1993, Perevolotsky 1994). Presumably, in such ecosystems, conditions seldom favor growth-dominated genotypes so that only one (resistant/tolerant) genotype exists in these populations. What are the conditions under which selection will favor a shifting mosaic of species adapted to either grazing avoidance/tolerance or competition, or under which selection will favor a homogenous community of species that are all adapted to grazing avoidance/tolerance?

(2) The period of rest after drought and heavy grazing is more important for ecosystem persistence than mean stocking rate. It has widely been observed that plants require a period of rest after drought and heavy grazing. In ecosystems where plants are not allowed to recover, land degradation may occur as quickly as in ecosystems where stocking rates are 5-10 times as high but plants are allowed to recover. This may be one reason why comparisons of the effects of commercial and communal ranching seldom show large differences in impacts on vegetation in spite of far higher mean stocking rates on communal ranches (Archer et al. 1989, Scoones 1993, Tapson 1993, Ward et al. 1998). That is, commercial ranchers can afford to (and do) restock their lands soon after drought-induced mortality, thus not allowing the vegetation time to recover lost resources. Communal ranchers can seldom afford to restock and thus must allow their herds to enlarge through natural reproduction, much in the way that wild herbivore populations recover after drought-induced mortality. During herd regrowth, stocking rates are well below capacity, allowing the plants to recover lost resources. Controlled studies of the effects of rates of post-drought restocking and herd regrowth on plant biodiversity have not been conducted thus far. Traditional range-management classifications of species as increasers (which increase in cover and abundance after grazing or disturbance) or decreasers (which decrease in cover or abundance with grazing) are associated with abilities to withstand grazing, and were an important first step towards understanding and predicting the effects of grazing on biodiversity. However, increaser-decreaser species descriptions are rooted in Clementsian perceptions of ecological succession and have hampered understanding in this field, particularly because such classifications have been shown to depend on genotype-by-environment interactions and on the particular competitors in the environment (see Milchunas and Lauenroth 1993, among others). Eco-physiological studies of the effects of grazing on regrowth, competitive ability and fitness of key species under a variety of abiotic conditions and with different suites of competitors are needed. Because grazing may affect the plant quality and the induction of physical and chemical defenses in plants, and because that will affect the probability of subsequent grazing, more studies of the effects of grazing on palatability of plants under natural conditions are required. Mechanistic understanding of the responses of increaser and decreaser genotypes (not species) to grazing should prove particularly important for furthering our understanding and predicting the impacts of grazing.

(3) Heavy grazing may cause very slow degradation of ecosystems. Several studies have shown that degradation as a result of overgrazing may take 80-100 years to be manifested, particularly in shrub-dominated habitats (Dean and MacDonald 1994, Ward and Ngairorue 1999, Ward and Kapofi in prep.). Thus, some claims that grazing may have little effect on ecosystems may depend on the short time-scale of most ecological studies. Appropriate monitoring schemes, coupled with historical studies and spatially-explicit models, are necessary to determine whether degradation is occurring.

(4) The impact of heavy grazing on soil resources is poorly understood in arid ecosystems. The popular concept of arid ecosystems as fragile has frequently been exaggerated. Some arid ecosystems are capable of recovering after years of heavy grazing and apparently suffer no long-term effects (Venter et al. 1989, Ward et al. 1998). Thus, vegetation cover is a poor measure of ecosystem condition. A key factor in the recovery of ecosystems following heavy grazing may lie in the differential ability of certain soils to withstand vegetation removal without suffering undue losses of nutrients. Another important factor in ecosystem change following grazing is the spatial redistribution of soil nutrients (Schlesinger et al. 1996). More emphasis needs to be placed on the development of appropriate indicators of soil degradation and on the study of nutrient cycling under different levels of grazing, and the effects thereof on plant biodiversity.

(5) Heavy grazing causes denudation in some habitats and bush encroachment in others. In the latter case, this leads to the paradox of enrichment whereby a negative effect (removal of plants by grazing) causes an increase in overall plant biomass. Bush encroachment is ascribed to the outcome of competition between tree and grass layers. That is, grasses usually outcompete trees for water and soil nutrients, thereby limiting the opportunity for tree seedling germination and recruitment. Once heavy grazing of grasses occurs, more water and nutrients and space becomes available for the trees to germinate, which often occurs en masse, resulting in bush encroachment. The switch between denudation and bush encroachment usually occurs at the higher end of the rainfall gradient in semi-arid/arid ecosystems because trees can only germinate and survive at higher rainfall. However, examples exist where heavy grazing results in bush encroachment in one habitat and not in an adjacent one, in spite of similar rainfall (Ward et al. 1998). In these cases, it seems that soil type is important: in deep sandy soils, water percolates deeper into the ground following rainfall than on schists, where water is trapped close to the soil surface. Thus, less rainfall is required on the latter substrate to facilitate germination of tree seeds en masse following heavy grazing. In the Middle East and in arid parts of Asia, bush encroachment is relatively rare, in spite of appropriate rainfall and soil conditions. A possible reason for the absence of bush encroachment may be the species pool capable of encroaching.⁷ Greater understanding of the interactions between rainfall, soil nutrients and species pool effects is necessary to understand this phenomenon.

(6) Termites may be more important herbivores than mammals in arid ecosystems. Comparisons of areas grazed by mammalian herbivores with plots where these animals are excluded sometimes shows little or no difference in vegetation biomass or species composition. This has frequently been ascribed to the effects of termite removal of grasses and/or wind and other abiotic factors. Also, after the mammalian herbivores have been removed or have left an area, the vegetation remains in a short-grass statex.⁸ Can termites create a new stable equilibrium and maintain an arid ecosystem in an "overgrazed" state after mammalian herbivores have been removed? Is grazing by mammalian herbivores necessary to create the conditions (short grass) for termites to establish, or can abiotic factors such as droughts facilitate their establishment in arid ecosystems? Adequately controlled studies are lacking to demonstrate the relative importance of mammalian and insect herbivores on the vegetation and biodiversity of arid ecosystems.

(7) Grazing results in the replacement of perennials by annuals or invasive aliens in some ecosystems. In many arid ecosystems, annual species or invasive aliens replace native perennials following heavy grazing owing to their ability to quickly invade open spaces and utilize soil resources. However, in the Middle East and in southern Africa, the prevalence of annual species in the herb layer cannot be unequivocally ascribed to the effects of grazing. Furthermore, species that invade in some arid ecosystems (e.g. Prosopis glandulosa in southern Africa) do not so under virtually identical abiotic conditions in other ecosystems. There are many life history models that predict the conditions under which certain strategies are optimal. For example, several optimization models produced in the 1960's and 1970's indicated that an annual lifestyle may be an optimal one in such ecosystems, regardless of the level of grazing. Unfortunately, such models have been abandoned and experimental tests of their predictions all but ignored. Appropriate tests of these life history models may help to understand the probability of invasion by annuals and/or alien plants following grazing.⁹

(8) Nurse plants dominate in some arid grazing ecosystems. Many plants in arid ecosystems depend on the presence of nurse plants to protect them from grazing and/or for the improved soil conditions in their vicinity (e.g. Ward and Rohner 1997). Although frequently reported, little is known of the effects of these nurse plants on the maintenance of biodiversity in arid ecosystems, and whether grazing can have negative effects on ecosystem persistence in cases where nurse plants are grazed.

(9) Positive effects of herbivores on biodiversity may be common in some arid ecosystems. Herbivores can alter the competitive interactions among plant species, sometimes removing a small number of dominant species and thereby allowing a larger number of subordinate species to become more abundant. Another instance of the promotion of greater species diversity by herbivores is in areas where they frequently defecate, creating nutrient-rich patches for plant establishment, or where they disturb the ground by rolling and sandbathing, thereby facilitating the invasion of disturbance-tolerant species. Also, herbivores may increase the germination of many plant species with hard-seed coats such as Acacia species via passage through their digestive tracts and subsequent defecation (Rohner and Ward 1999). In ecosystems where mammalian herbivores have been removed or have declined, the absence of recruitment facilitation by herbivores may have serious impacts on the maintenance of biodiversity (Reid and Ellis 1995, Ward and Rohner 1997, Rohner and Ward 1999). Further examination of the relative importance of facilitation by herbivores on ecosystem biodiversity, as opposed to a local-scale description of the phenomena involved, is necessary.

Bibliography