Foraging theory, patch use, and the structure of a Negev Desert granivore community.
The ways competing species respond to opportunities for food and safety influence and reveal competitive interactions that determine species coexistence (Werner 1992). For this reason, foraging theory (e.g., Stephens and Krebs 1986) can be applied to study species coexistence and community organization (Emlen 1966, MacArthur and Pianka 1966). As an approach to community ecology, foraging theory provides a basis for deducing and testing mechanisms of coexistence (Tilman 1982, Price 1986, Kotler and Brown 1988). A mechanism of coexistence involves an axis of environmental heterogeneity along which species might exhibit niche partitioning and trade-offs among the species that give each species a competitive advantage along some portion of the axis. The joint consideration of an axis of environmental heterogeneity and a salient trade-off among the species defines a mechanism of coexistence.
Desert rodent communities have provided a model system for inspiring (e.g., Brown 1989b) and testing mechanisms of coexistence (for reviews see Brown 1975, Rosenzweig 1977, Price and Brown 1983, Kotler and Brown 1988). Studies on desert rodents have documented niche separation along axes of space, time, and diet (Rosenzweig and Winakur 1969, Brown and Lieberman 1973, Rosenzweig 1973, M'Closkey 1978, 1981, Price 1978, Wondolleck 1978). The application of foraging theory to assess community processes in desert rodents has emerged from research on habitat selection (Kotler 1984a, Price and Waser 1985, Rosenzweig and Abramsky 1986, Brown 1989a) and patch use (Reichman and Oberstein 1977, Hutto 1978, Price 1978, Bowers 1982, Behrends 1986, Brown 1988, Price and Longland 1989).
In a previous study (Brown 1989a), we measured rodents' giving-up densities (sensu Brown 1988) in depletable food patches to test four mechanisms of coexistence based on habitat selection in time and space. In a Sonoran Desert community of four granivorous species (Arizona pocket mouse, Perognathus amplus; Merriam's kangaroo rat, Dipodomys merriami; round-tailed ground squirrel, Spermophilus tereticaudus; and Harris's antelope squirrel, Ammospermophilus harrisii), we concluded that two mechanisms were jointly responsible for community organization. Under the first, the pocket mouse, kangaroo rat, and round-tailed ground squirrel exhibited habitat selection in time, with each possessing a time of year when it was the most efficient forager. Under the second, the kangaroo rat and the antelope squirrel exhibited habitat selection in space. The kangaroo rat appeared to have the competitive advantage in patches with low seed abundances, while the antelope squirrel had the advantage in rich patches.
Here, we investigated five potential mechanisms of coexistence in a community with three common species of rodents in the Negev Desert of Israel. These rodents (Allenby's gerbil, Gerbillus allenbyi; greater Egyptian sand gerbil, G. pyramidum; common jerboa, Jaculus jaculus) inhabit sandy habitats (Abramsky et al. 1985), share similar habits and seed diets (Bar et al. 1984), and have been shown to compete (e.g., density compensation and interspecific density-dependent habitat shifts; Rosenzweig and Abramsky 1985, 1986, Abramsky and Pinshow 1989, Abramsky et al. 1990, 1991, Mitchell et al. 1990). In the field, we used live-trapping to estimate population sizes, sand tracking to record how animals allocate activity time among different habitats and microhabitats, and giving-up densities of resources (GUDs) by rodents in manipulated resource patches (seed trays) to measure foraging efficiencies in the different habitats and microhabitats. These methods allow us to: (1) use patch use theory to consider five mechanisms of coexistence that should have broad applicability to other systems and taxa; (2) compare the organization of two communities on different continents using the same experimental techniques. Quadrupedal pocket mice (Perognathus and Chaetodipus) and bipedal kangaroo rats (Dipodomys) are morphologically similar to quadrupedal gerbils (Gerbillus) and bipedal jerboas (Jaculus), respectively. Does the morphological convergence of the communities indicate convergence in community organization as well? (3) Compare our results with those of several other studies on desert rodent communities in the Negev. Abramsky et al. (1985, 1990, 1991) suggest that habitat selection in space based upon degree of sand stabilization promotes the coexistence of these two gerbil species. By providing only limited support for the five mechanisms of coexistence, our results bolster those of two short-term studies (5-10 d) that implicate temporal partitioning within nights as a mechanism promoting coexistence between G. allenbyi and G. pyramidum (Kotler et al. 1993, Ziv et al. 1993).
GIVING-UP DENSITIES AND SEED TRAYS
We employed manipulated resource patches (seed trays) to quantify the foraging behavior of desert rodents. Because some of the predictions from the mechanisms of coexistence will be couched in terms of data from these patches, we first discuss the rationale behind the experimental patches.
Resource patches can be used to measure quitting harvest rates of desert rodents (Brown 1988). The resource patches were created by mixing a measured amount of seeds into a large volume of sand placed in a tray. Desert rodents foraging in these resource patches experience diminishing returns to harvest effort, that is, the longer a forager spends in the resource patch, the lower the rate it finds new seeds (Kotler and Brown 1990). As a rodent continues to exploit a tray, its harvest rate eventually declines to a point where it is no longer worthwhile to continue foraging. At this point, the rodent should quit the patch and seek another or pursue an alternative activity (Charnov 1976). Because the rodent's harvest rate in the tray is a function of the seeds remaining in the tray, the giving-up density (GUD) of seeds left behind in a tray after a forager has quit exploiting the patch provides a surrogate for its quitting harvest rate.
The quitting harvest rate estimated by GUDs provides an estimate of foraging costs and benefits (see Stephens and Krebs 1986). In Brown (1988, 1992), we showed that an optimal forager should leave a patch (or seed tray) when the value of its harvest rate, H, equals the sum of its metabolic, C, predation, P, and missed opportunity costs of foraging, MOC, i.e., H = C + P + MOC. By considering the marginal rates of substitution of different inputs into fitness, all of the terms, H, C, P, and MOC, can be expressed as a common currency (Caraco 1979, Brown 1988, 1992).
GUDs can be used to measure foraging efficiency. Efficiencies are ratios of output to input. In regard to foraging, the output is the gain to the forager while exploiting a resource patch, and the input is the summation of the foraging costs. Therefore, foraging efficiency can be defined as the ratio between the benefits and costs of foraging, i.e., H/(C + P + MOC). Since a forager should exploit a patch until H = C + P + MOC, its foraging efficiency will equal 1 at its GUD. A species with a higher harvest rate or lower foraging costs will be more efficient and will reach the point where H = C + P + MOC at a lower resource density. Thus, the species with the highest foraging efficiency has the lowest GUD (Tilman 1980, Vance 1985).
MECHANISMS OF COEXISTENCE
The five mechanisms of coexistence that we examined have been suggested or supported by previous studies of desert rodents, and they may have broad applicability to many other ecological communities. These mechanisms are not mutually exclusive, and TABULAR DATA OMITTED several may operate together in diverse communities. Here, we discuss each mechanism in terms of the axis of environmental heterogeneity and the trade-off among the species that allows coexistence. We also discuss the predictions of each mechanism in regard to changes in population density, patterns of activity, and GUDs of resources.
Bush/open microhabitat selection
In this mechanism, the axis of environmental heterogeneity is either vegetation structure or some factor associated with it (e.g., seed availability, predatory risk, substrate), and the trade-off is between the ability to forage efficiently in the bush microhabitat vs. the open TABULAR DATA OMITTED microhabitat. Coexistence is possible provided each species has a microhatbitat in which it is a more efficient forager than its competitors. For example, if seed distributions vary between the bush and open microhabitats, coexistence is possible if there is a trade-off in the ability to exploit the different distributions (Reichman and Oberstein 1977, Reichman 1981, 1984, Price and Reichman 1987). If predatory risk differs in intensity or in type (e.g., snakes vs. owls) between microhabitats, coexistence is possible if there is a trade-off between competitive ability and the ability to avoid predation or between the ability to avoid one type of threat or another (Rosenzweig 1973, Thompson 1982, Kotler 1984a, Kotler et al. 1988). Finally, if substrates differ between microhabitats, coexistence is possible if there is a trade-off in foraging efficiency on the different substrate types (Price and Waser 1985). Body size and/ or locomotory morphology may contribute to the critical trade-offs, with bipedal or larger sized individuals being better at avoiding owl predators in the open (Kotler et al. 1988, 1991, Longland and Price 1991), at interference (Ziv et al. 1993), and at exploiting rich seed patches (Kotler and Brown 1990). Smaller individuals may have lower energy foraging costs and higher values for H/C (Morgan and Price 1992).
In regard to rodents in the Negev Desert, bush/open microhabitat selection predicts that quadrupeds and smaller species, relative to bipeds and larger species, will have lower CUDs in the bush microhabitat and will be most active there, while the opposite will be true for bipeds and larger species (Table 2; Kotler 1984b).
Habitat selection in a mosaic
Habitat selection need not be restricted to the bush/ open dichotomy and can occur anywhere along a continuum of spatial scales from microhabitat to macrohabitat. As in bush/open microhabitat selection, coexistence along a habitat axis can occur if each species has a habitat in which it is a more efficient forager than its competitors. Habitat selection has been documented extensively for the psammophilous gerbils of the Negev Desert (Abramsky et al. 1985, 1991, Abramsky and Pinshow 1989). G. pyramidum predominates on patches of semistabilized dune, and G. allenbyi predominates on patches of dune stabilized by vegetation (Abramsky et al. 1985). If habitat selection provides the basis for coexistence, then G. allenbyi, relative to G. pyramidum, will bias its activity towards and have lowest GUDs in habitats with stabilized sand, and G. pyramidum will bias its activities towards and have lowest CUDs in habitats with semistabilized sand.
Spatial variation in resource abundance
This mechanism considers a patchily distributed resource. When the abundance of a resource varies spatially, at least two attributes of the forager contribute to its success: (1) travel cost or speed and (2) foraging efficiency. Travel cost or speed determines how inexpensively an individual can move from one place of high resource abundance to the next. Foraging efficiency determines whether an individual can profitably harvest resources from patches with low resource abundance. The trade-offs that can promote coexistence are either travel cost vs. patch foraging efficiency (Reichman 1981, Brown 1986) or foraging efficiencies at high vs. low food abundances (Stewart and Levin 1973, Abrams 1984). The species with low travel cost tends to exploit the rich end of the distribution of patch resource abundances; it requires richer patches because it cannot profitably harvest them down to low resource levels. The species with high foraging efficiency can exploit less rich patches and it can remain in patches longer as it can profitably harvest them down to lower levels. The clump size selection hypothesis for the coexistence of mobile vs. efficient species is an example of this mechanism (Reichman and Oberstein 1977, Reichman 1981).
This mechanism predicts that the smaller, more efficient foragers should visit fewer patches, but harvest them to lower CUDs.
Temporal variation in resource abundance
This mechanism considers the seasonal renewal and subsequent depletion of seed resources common in deserts. When the abundance of a resource varies temporally, at least two attributes of a forager will contribute to its success: (1) maintenance efficiency (the ratio of potential harvest rate to the cost of dormancy; Brown 1989b) and (2) foraging efficiency. Maintenance efficiency determines how inexpensively an individual can travel in time from one period of high resource abundance to the next; foraging efficiency determines to what extent an individual can profitably harvest resources from periods of low resource densities. Competitive coexistence is possible if there is a trade-off among species between maintenance efficiency and foraging efficiency (Brown 1989b). The species with high maintenance efficiency biases its activity towards periods of high resource abundance and avoids periods of lower abundance by inexpensively remaining dormant. The species with the higher foraging efficiency cannot remain dormant as inexpensively, but it can profitably harvest resources at abundances at which the other species is obliged to remain dormant.
This mechanism predicts that coexisting species have nested periods of dormancy. As resources become scarce, the least efficient forager with highest CUDs and lowest intensity of activity at tracking stations should be the first to become dormant; as seed resources renew, it should be the last to become active.
Temporal variation in foraging costs
Seasonal changes in climate or predatory risk may alter the costs of foraging. If changes in costs are not correlated among coexisting species, then each species may enjoy a period of the year when it is the most efficient forager. The trade-off is between the cost of foraging during different temporal periods. When seasonal rotation of foraging efficiencies is promoting coexistence, species densities will vary asynchronously due either to (1) the densities of the coexisting species changing out of phase or (2) overlapping, sequential patterns of dormancy (Brown 1986, 1989a). Asynchronous changes in species densities are important because changes in foraging costs only influence resource dynamics and competitive interactions via concomitant changes in forager densities (a population dynamic phenomenon) or activity (a behavioral phenomenon). Through dormancy or asynchronous changes in population size, each species exhibits greater activity during the period when it has the competitive advantage.
This mechanism predicts that: each species will possess a period of the year when it has the lowest GUD, the densities of the coexisting species will fluctuate asynchronously, and periods of dormancy (if present) should be distinct rather than nested.
We conducted experiments at Beer Asluj in the Holot Mashabim Nature Reserve, northwestern Negev Desert, Israel, The site contains mosaics of semistabilized and stabilized sand dunes. The dominant perennial plant species are Artemisia monosperma and Retama raetam (Abramsky et al. 1985). Rodent species on the sandy habitats include Gerbillus allenbyi (Allenby's gerbil, 25 g), G. pyramidum (greater Egyptian sand gerbil, 39 g), G. henleyi (pygmy gerbil, 11 g), Buxton's jird (Meriones sacramenti, 120 g), and common jerboa (Jaculus jaculus, 50-70 g). All of these species are nocturnal, burrow-dwelling, and either primarily or somewhat granivorous (Bar et al. 1984).
Within the sandy habitats, we established two 2.56-ha grids. Each contained 81 stations arranged 9 x 9 with 20 m spacing between stations. Both grids contained a mosaic of sandy habitats. We scored the habitat at each station from 1 to 4 based on the degree of sand stabilization within a 5 m radius of the station: 1 = sand completely stabilized by algal soil crust, 2 = sand partially stabilized by algal soil crust, 3 = mostly shifting sand with broken patches of soil crust, and 4 = shifting sand bare of any soil crust. Substrate habitat scores of 1 and 2 correspond to stabilized sand dune, and habitat scores of 3 and 4 correspond to semistabilized sand dune.
From 1986 to 1988, we completed nine rounds of our experimental protocol: November 1986, January, February, April, June, July, August, November 1987, January 1988. The 11-d protocol included live-trapping rodents on the first 2 d, quantifying spoor in tracking stations on days 3 and 11, and measuring GUDs in seed trays on days 4-10. Days of the protocol were consecutive, weather permitting.
Live-trapping: censusing species densities
We censused rodent populations by setting a Sherman live trap baited with millet seed at each station of the grids. Animals caught during the first night were marked on their stomachs with a black marking pen (this mark remains intact for at least several days). In this way, animals caught the second night could be identified as recaptures or new individuals. For all new individuals of a trapping period, we recorded trap station, mass, sex, and reproductive condition. Captured individuals of G. allenbyi and G. pyramidum were given species-specific toe clips by clipping the outer toe of the right or left hindfoot, respectively. This toe clip facilitated the identification of spoor in tracking stations and seed trays.
The live-trapping data were used to estimate population abundances and detect seasonal trends in activity and population sizes.
Sand-tracking: measuring activity and habitat use
The sand-tracking procedure was modified from Kotler (1985) and follows Mitchell et al. (1990). At each station of both grids, we smoothed two plots of sand (45 x 45 cm), one at the margin of a perennial shrub (bush microhabitat) and one 2-3 m from the shrub's edge (open microhabitat). Following a night of activity, we examined sand-tracking plots and scored them for presence or absence of footprints and for activity level based on the fraction of the plot covered with tracks. The activity score of each species within a plot could range from 0 (0% coverage) to 4 (100% coverage). We identified rodent species based on the spoor's size, shape, and toe-clip.
We used sand-tracking data to measure per capita rates of activity, habitat selection, and intensity of patch use by gerbil species according to season, sand habitat, and bush/open microhabitat.
Seed trays: measuring relative foraging efficiencies
We used aluminum seed trays (45 x 60 x 2.5 cm deep) filled with 3 g of millet seed mixed into 5 L of sifted sand. Pairs of seed trays were placed at grid stations located at the intersection of evenly numbered rows and columns. This placement of trays required 16 pairs of trays per grid with 40-m spacing between stations with trays. One tray of a pair was placed in the open and the other in the bush microhabitat. Data collection consisted of identifying the forager species based on footprints in the tray's sand and in the surrounding sand, sifting the sand from the tray to recover the remaining seeds, and recharging the tray's sand with 3 g of millet. The GUD of the tray was credited to the one or several species whose tracks were visible in the tray. If a species' tracks were found adjacent to, but not in the tray, then we assumed that it foraged in the tray, but was not credited with the GUD. Seeds recovered from trays were cleaned of debris and weighed to measure the forager's GUD (see Brown 1988).
On both grids, G. allenbyi and G. pyramidum were abundant and readily trapped (9.3 and 4.3% trap success, respectively), G. henleyi and M. sacramenti were rare and trapped occasionally (0.4 and 0.3% trap success, respectively), and J. jaculus was never trapped inspite of frequent spoor in tracking plots and around seed trays. Using the incidence of recaptures on the second night of trapping to assess each species' trappability, we detected no significant differences in the trappability of G. allenbyi and G. pyramidum (G = 0.32, df = 1, P [is greater than] 0.5; summing over the different protocol rounds, we had 321 1st-d captures and 201 2nd-d recaptures of G. allenbyi, and 77 1st-d captures and 54 2nd-d recaptures for G. pyramidum). Because the gerbil species were equally trappable, we used the total number of different individuals captured during a protocol period as the relative estimate of a gerbil's population size.
Both mechanisms of coexistence based on habitat selection in time make predictions regarding patterns of dormancy and seasonal trends in population size. With respect to dormancy, there was no indication that G. allenbyi, G. pyramidum, or J. jaculus (based on trapping and/or sand tracks) used seasonal dormancy (our results are in accord with what is known of the biology of the two gerbil species, Abramsky et al. 1985). To measure the temporal covariation in population sizes between gerbil species apparent in Fig. 1, we used a Spearman rank correlation test. While the test statistic is not significant (Spearman rank correlation of 0.73, df = 8, P [is less than] 0.1), the trend suggests that population sizes of G. allenbyi and G. pyramidum fluctuated synchronously.
Sand-tracking: habitat selection
The spoor of G. allenbyi, G. pyramidum, and J. jaculus were present in 49.4, 15.8, and 5.8% of the tracking plots, respectively (combining over all protocol rounds). We used a four-way ANOVA without replication to determine activity by species, month, sand habitat, and bush/open microhabitat. Proportion of tracking stations visited (arcsine transformed) during sampling each month provided the dependent variable.
TABLE 3. ANOVA of the proportion of tracking stations with spoor for the factors species (G. allenbyi, G. pyramidum, and J. jaculus), sand habitat (1, 2, 3, and 4), microhabitat (bush vs. open), and month (all nine protocol periods). Variable df MS F Month 8 0.13 14.81(***) Species 2 6.24 712.82(***) Habitat 3 0.31 34.77(***) Microhabitat 1 0.03 3.35 Species x Month 16 0.09 10.01(***) Species x Habitat 6 0.20 22.92(***) G.a. 3 0.21 23.44(***) G.p. 3 1.79 203.74(***) J.j. 3 0.13 14.95(***) Species x Microhabitat 2 0.24 27.50(***) G.a. 1 0.17 18.90(***) G.p. 1 0.02 2.00 J.j. 1 0.33 37.30(**) Error 177 0.01 ... ** P [is less than] 0.01, *** P [is less than] 0.001.
The species differed in their utilization patterns of sand habitats, bush/open microhabitats, and months. G. allenbyi biased its activity towards intermediate habitats (2 and 3) and away from the extreme sand habitats (1 and 4) ([F.sub.3,69] = 23.44, P [is less than] 0.001). It also biased its activity towards the bush microhabitat ([F.sub.1,71] = 18.9, P [is less than] 0.001). In contrast, G. pyramidum and J. jaculus biased activity towards semistabilized sand habitats ([F.sub.3,69] = 203.74, P [is less than] 0.001, and [F.sub.3,69] = 14.95, P [is less than] 0.001, respectively), J. jaculus biased its activity towards the open microhabitat ([F.sub.1,71] = 37.3, P [is less than] 0.001), and G. pyramidum showed a weak, nonsignificant bias towards the open microhabitat ([F.sub.1,71] = 2.00, P [is less than] 0.4). Of the three species, J. jaculus showed the most extreme selectivity for the open microhabitat, and G. pyramidum showed the most extreme selectivity for the semistabilized sand habitat.
The ANOVA's main effects indicate several patterns. First, species differed in the number of tracking stations with spoor (significant species effect, Table 3). Not surprisingly, G. allenbyi, the most abundant species, occurred in the most tracking stations. Next, activity aggregated over species did not differ between the bush and the open microhabitat. Combined activity also differed by month and by sand habitat (significant month and habitat effects). Finally, more activity occurred in months with high gerbil population sizes, and more activity occurred in the semistabilized sand habitats than the stabilized sand habitats.
We looked for covariation in activity of G. allenbyi and G. pyramidum by month using a Spearman rank correlation. There was a nonsignificant positive covariation in activity of the two species across months (Spearman rank correlation of 0.62, df = 8, P = 0.11), suggesting that the gerbil species do not have separate times of the year when they are more active than their competitor. This result is consistent with and restates the positive covariance in population sizes born out by the trapping results.
Sand-tracking: per capita activity
We used the number of tracking plots per individual as a measure of per capita activity. For each combination of month and species (G. allenbyi and G. pyramidum only), we calculated per capita activity as the number of tracking plots with spoor divided by the number of individuals captured during the month's census period. To determine differences between G. allenbyi and G. pyramidum in per capita activity, we used a three-way log-linear model. We used species and month as two of the factors. The third factor consisted of two categories: number of individuals and number of tracking stations with spoor present (track/capture). This procedure yielded a 2 x 8 x 2 contingency table of species x month x track/capture. Summing over months, this table contained 462 and 114 captured individuals of G. allenbyi and G. pyramidum, respectively, and 1550 and 392 tracking plots with spoor of G. allenbyi and G. pyramidum, respectively.
The log-linear analysis yielded a nonsignificant three-way interaction (G = 7.43, df = 7, P [is greater than] 0.1 for species x month x track/capture). Considering G. allenbyi and G. pyramidum together, per capita activity varied by month (G = 38.7, df = 7, P [is less than] 0.001 for month x track/capture), with a peak in April (8.96 tracking plots per individual) and a low in August (2.67 plots per individual). The distribution of captures and activity between G. allenbyi and G. pyramidum varied by month (G = 49.0, df = 7, P [is less than] 0.001 for species x month; Fig. 3). This reflects the large temporal changes in population size of G. allenbyi as compared to the temporally more constant population size of G. pyramidum. The two gerbil species did not differ in their per capita activity (G = 0.1, df = 1, P [is greater than] 0.5 for species x track/capture). On average, individuals of G. allenbyi and G. pyramidum visited 3.36 and 3.44 tracking plots per night, respectively.
Sand-tracking: intensity of plot use
We used scores in those tracking plots with a species present to analyze intensity of plot use. We used a four-way ANOVA to determine the effect of species (G. allenbyi, G. pyramidum, and J. jaculus), sand habitat (1-4), microhabitat (bush and open), and month on plot score. Because of their relevance to the mechanisms of coexistence, we also included the three two-way interactions resulting from species x month, species x sand habitat, and species x bush/open microhabitat. Because sample sizes emerge from the species' use of plots, the cell sample sizes vary considerably with many empty cells. To remedy the unbalanced design, we combined data by calculating the mean tracking score for each combination of species, month, sand habitat, and bush/open microhabitat. This lumping may create some problems for a parametric ANOVA because the sample sizes used to determine each mean vary from 10 to 80 for G. allenbyi, 0-35 for G. pyramidum, and 0-15 for J. jaculus (the matrix of means contains 216 cells of which 16 were empty). Hence, following Conover and Iman (1981), we performed the ANOVA analysis on the rank transformation of cell means. This provides a nonparametric test that utilizes standard ANOVA procedures. Performing the ANOVA on ranks (non-paremetric test) provided a more conservative test than the corresponding ANOVA on the means (paremetric test).
TABLE 4. The effect of species (G. allenbyi, G. pyramidum, and J. jaculus), sand habitat (all four), microhabitat (bush and open), and month (all nine rounds) on score within tracking stations (for this analysis, only scores greater than zero were considered). The results are from a four-way randomized block ANOVA in which the dependent variable was the mean tracking score for all combinations of species, sand habitat, microhabitat, and month ([r.sup.2] = 0.717). We tested for the four main effects and for the three two-way interactions involving species. Of the 216 cells, 16 cells were missing data. Within species, we included the three pairwise comparisons (these tests are shown as subsets under the variables of the ANOVA). Because these planned comparisons were not orthogonal, Dunn-Sidak adjustments of experimentwise error rates were made. Variable df MS F Species 2 9.104 87.72(***) G.a. vs. G.p. 1 1.286 12.37(*) G.a. vs. J.j. 1 21.013 202.05(***) G.p. vs. J.j. 1 11.903 114.45(***) Habitat 3 0.162 1.56 Microhabitat 1 1.568 15.11(***) Month 8 1.386 13.35(***) Species x Habitat 6 0.310 2.99(**) Species x Microhabitat 2 1.212 11.68(***) Species x Month 16 0.258 2.48(**) Error 161 0.104 ... * P [is less than] 0.05, ** P [is less than] 0.01, *** P [is less than] 0.001.
Overall, G. allenbyi exhibited the highest use and J. jaculus the lowest use of tracking plots. Scores varied considerably by month with few detectable patterns; April saw the highest scores, and February saw particularly low scores. Combining over species, there was no significant effect of sand habitat on tracking score. Overall, plots in the bush saw more activity than those in the open microhabitat.
All three two-way interaction terms with species were significant. Both gerbil species had higher scores in the bush than open microhabitat, while the jerboa had a slightly higher score in the open than bush microhabitat. The interaction of species and habitat emerges from an erratic response of jerboas to habitat. All three species show somewhat irregular and erratic responses to month. Interestingly, G. pyramidum actually uses more open plots, but uses bush plots more intensively.
Seed trays: species differences in GUDs
We used GUDs in seed trays to measure relative foraging efficiencies and a four-way ANOVA based on rank transformed means (Conover and Iman 1981) to test for differences among species and effects of habitat, microhabitat, and month. As in the case of tracking data, sample sizes emerge from the species' use of seed trays, and cell sample sizes vary considerably, with many empty cells. To remedy the unbalanced design, we again used a nonparametric ANOVA based on ranks of the mean of GUDs for each combination of species, month, sand habitat, and bush/open microhabitat (Conover and Iman 1981). When compared to the corresponding parametric analysis on the means, the results from the nonparametric analysis were more conservative in every case.
TABLE 5. The effect of species (G. allenbyi and G. pyramidum only), sand habitat (Habitats 3 and 4 only), microhabitat (bush and open), and month (all months except January 1988) on giving-up densities (GUDs). The results are from a four-way randomized block ANOVA (nonparametric) in which the dependent variable was the rank of the mean of the logarithmic transformation of GUDs for all combinations of species, sand habitat, microhabitat, and month ([r.sup.2] = 0.943). We tested for the four main effects and for the three two-way interactions involving species. Of the 64 cells, one had missing data. Within the species x habitat interaction and the species x month interaction, we include planned comparisons of species by habitat and month, respectively (these tests are shown as subsets under the variables of the ANOVA). Because these planned comparisons are not orthogonal, Dunn-Sidak adjustments of experimentwise error rates were made. Variable df MS F Species 1 145.10 3.41 Habitat 1 2259.21 53.08(***) Microhabitat 1 293.26 6.89(*) Month 7 2229.57 52.38(***) Species x Month 7 44.78 1.05 Species x Microhabitat 1 17.10 0.40 Species x Habitat 1 393.78 9.25(**) G.a., Hab. 3 vs. 4 1 519.70 12.20(**) G.p., Hab. 3 vs. 4 1 30.04 0.71 Error 43 42.56 ... * P [is less than] 0.05, ** P [is less than] 0.01, *** P [is less than] 0.001.
Data for G. allenbyi were always ample (2733 GUDs). Data for G. pyramidum were ample for habitats 3 and 4 and all months except January 1988 (510 GUDs). GUD data for J. jaculus were sparse throughout the study except January 1988 (33 GUDs during that month). For this reason we selected only those means that contained at least five samples (species G. allenbyi and G. pyramidum: all months except January 1988, sand habitats 3 and 4, and bush and open microhabitats; we analyzed data from January 1988, including data from J. jaculus, separately). The analysis resulted in 64 cells with one missing value. Because of their relevance to the mechanisms of coexistence, we also included the three two-way interactions resulting from species x month, species x sand habitat, and species x microhabitat.
The main effects of sand habitat, bush/open microhabitat, and month were significant. GUDs fluctuated significantly throughout the year, with high GUDs in February and April and low GUDs during July and August. GUDs were lower in the habitat with the least stabilized sand (habitat 4). GUDs were lower in the bush than open microhabitat. There was also a significant two-way interaction of species and sand habitat. A comparison of species by habitat revealed that G. allenbyi had a significantly lower GUD than G. pyramidum in habitat 3 (0.463 vs. 0.613 g of millet and mean ranks of 33.75 vs. 41.81; F = 12.2, df = 1, 31, MS = 519.7, P [is less than] 0.01), and that GUDs of the two species did not differ significantly in habitat 4 (0.397 vs. 0.372 g of millet and ranks of 26.75 vs. 24.78; F = 0.71, df = 1, 31, MS = 30.04; Fig. 5). Both G. allenbyi and G. pyramidum had lower GUDs in the bush microhabitat; the two-way interaction of species and microhabitat was not significant.
Unlike the other two species, G. allenbyi had numerous GUDs in all four sand habitats. G. allenbyi had a lower GUD in the bush than open microhabitat for all four sand habitats. However, trends across sand habitats differed between microhabitats. GUDs in the open microhabitat declined steadily from the most stabilized sand (habitat 1) to the least stabilized sand (habitat 4). In contrast, GUDs in the bush microhabitat were lowest in the extreme sand habitats (1 and 4) and highest in the intermediate habitats (2 and 3).
In a second analysis, we compared the GUDs of G. allenbyi and J. jaculus for January 1988. The distribution of the 249 and 33 GUD measurements for G. allenbyi and J. jaculus, respectively, were fairly balanced across habitats (but not for microhabitats). To analyze these data we used a two-way ANOVA with species (G. allenbyi and J. jaculus) and sand habitat (including all four sand habitats) as the group variables and the logarithmic transformation of GUDs as the dependent variable. The ANOVA model provided a poor fit to the data ([r.sup.2] = 0.077), but the effect of species was significant ([F.sub.1, 274] = 12.09, P [is less than] 0.001), and the effect of habitat ([F.sub.3, 274] = 0.62, P [is greater than] 0.5) and the two-way interaction of species and habitat ([F.sub.3, 274] = 1.10, P [is greater than] 0.3) were not significant. G. allenbyi had a lower GUD than J. jaculus: 0.38 and 0.73 g of millet, respectively (back-transformed means). A separate one-way ANOVA found that J. jaculus had a lower GUD in the open (0.64 g) than bush microhabitat (0.95 g), although this difference was not significant ([F.sub.1,31] = 2.74, P = 0.11). Combining over all months, J. jaculus had GUDs of 0.92 and 1.36 g in the open and bush microhabitats, respectively.
Seed trays vs. tracking plots
Discrepancies between the relative frequency of a species' tracks within tracking plots and a species' relative frequency of GUDs in seed trays provided evidence for a species' willingness to be the last forager in a seed tray. We used a three-way log-linear model to analyze the frequency data in Fig. 7; the three levels of the analysis were species (G. allenbyi and G. pyramidum), habitat (including all four sand habitats), and GUDs vs. tracks (GUD/track). This last category distinguishes between frequency of footprints in tracking plots and frequency of GUDs in seed trays.
Relative to presence in tracking stations, G. allenbyi is over-represented in the seed trays (species x track/GUD interaction: G = 43.09, df = 1, P [is less than] 0.001; Fig. 7). This tendency for G. allenbyi to be over-represented in seed trays is more pronounced in the stabilized sand habitats, 1 and 2, and less pronounced in the less stabilized sand habitats, 3 and 4 (habitat x species x track/GUD interaction: G = 47.94, df = 3, P [is less than] 0.001). These results provide additional evidence that G. allenbyi is a more efficient forager than G. pyramidum and that this advantage increases with sand stabilization.
Seed trays: species' ability to out-forage other species
We were able to record spoor in and around seed trays. Spoor outside of a seed tray indicate which species visited the tray during the night, while spoor in the tray indicate the species responsible for the GUD. Because of the larger area and less intense foraging outside of seed trays, spoor outside of trays do not become obliterated by subsequent visitors to the tray, while spoor in trays do become obliterated by the activity of subsequent visitors. Cases where a species' spoor were present at a station but not present within the trays of a station revealed when another species had the lower GUD.
The paucity of J. jaculus GUDs appeared to result from their high GUDs rather than from their absence from seed trays; individuals of J. jaculus foraged from trays at a station without being wholly or partially credited with the GUD 279 times. G. pyramidum was present at seed trays but beaten out by G. allenbyi for the GUD 346 times. Rarely was G. allenbyi at a station and not at least partially credited with the GUD; only 38 times did G. pyramidum beat out G. allenbyi for the GUD.
Summary of results
We collected data from census trapping, sand-tracking plots, and seed trays for a community of seed-eating rodents in sand dune habitats of the Negev Desert, Israel. Census data revealed that Gerbillus allenbyi was [approximately equal to]4 times more abundant than G. pyramidum. Other species were either much rarer or, as in the case of J. jaculus, their population densities could not be estimated. The two gerbils' population densities changed in synchrony, with peaks in July and August and troughs in April.
Sand-tracking data provided three types of information: (1) habitat-specific levels of activity, (2) intensity of plot use, and (3) per capita activity levels. With respect to activity levels between the bush and open microhabitats: G. allenbyi: bush [is greater than] open, Result 1a; G. pyramidum: open [is greater than or equal to] bush, Result 1b; J. jaculus: open [is much greater than] bush, Result 1c. With respect to activity levels among the four levels of sand habitats (1 = most stabilized, 4 = least stabilized): G. allenbyi: habitats 2 and 3 [is greater than] habitats 1 and 4, Result 2a; G. pyramidum: habitat 4 [is greater than] 3 [is greater than] 2 [is greater than] 1, Result 2b; J. jaculus: habitat 4 [is greater than] 3 [is greater than] 2 [is greater than] 1, Result 2c. With respect to intensity of plot use between the bush and open microhabitat (species did not vary with respect to sand habitat): G. allenbyi: bush [is greater than] open, Result 3a; G. pyramidum: bush [is greater than] open, Result 3b; J. jaculus: open [is greater than] bush, Result 3c. When comparing the three species with respect to intensity of plot use (independent of microhabitat, sand habitat, or season): G. allenbyi [is greater than] G. pyramidum [is greater than] J. jaculus, Result 4. When comparing the two gerbil species with respect to per capita activity levels: G. allenbyi = G. pyramidum, Result 5.
The GUDs from seed trays measured relative foraging efficiencies among habitats (spatial and temporal) and among species. With respect to GUDs between the bush and open microhabitats: G. allenbyi: bush [is less than] open, Result 6a; G. pyramidum: bush [is less than] open, Result 6b; J. jaculus: open [is less than] bush, Result 6c. With respect to GUDs between the four sand habitats: G. allenbyi: habitat 4 [is less than] 3 [is less than] 2 [is less than] 1, Result 7a; G. pyramidum: habitat 4 [is less than] 3, Result 7b. In comparing GUDs among species, the data from Figs. 5 to 8 support the following conclusion. Independent of microhabitat, sand habitat, and season: G. allenbyi [is less than] G. pyramidum [is less than] J. jaculus, Result 8.
These data provide complementary information for testing the five mechanisms of species coexistence involving habitat selection in time and space. These mechanisms assume that the rodents behave optimally (exploit resource patches until the marginal value of patch exploitation equals the marginal cost), and each considers an axis of environmental heterogeneity and a corresponding trade-off among the species partitioning the axis. We consider, in turn, each mechanism and the evidence for and against it.
Bush/open microhabitat selection
Bush and open microhabitats strongly influenced the foraging behavior of the three rodent species (Results 1, 3, and 6). The lower foraging costs (as indicated by lower GUDs) of G. allenbyi and G. pyramidum in the bush microhabitat can be understood in terms of protection from owl predators (Kotler et al. 1991) and a more favorable microclimate (Goodfriend et al. 1991). In contrast to the gerbils, J. jaculus appeared to have lower foraging costs in the open than bush microhabitat.
For bush/open microhabitat selection to provide a basis for coexistence, each species must have a microhabitat in which it has a lower GUD than its competitor. We found no evidence for this (Result 8). The evidence best supports the conclusion that G. allenbyi was the most efficient forager and J. jaculus was the least efficient forager in both microhabitats. Only one line of evidence supported this mechanism: G. allenbyi biased its activity towards the bush microhabitat while G. pyramidum and J. jaculus biased their activity towards the open (Result 1). While statistically significant, the microhabitat biases of the two gerbil species may be too small to promote coexistence.
In contrast, much evidence exists for this mechanism among the desert rodents of North America (Kotler and Brown 1988). In general, kangaroo rats (Dipodomys) and kangaroo mice (Microdipodops) exploit the open microhabitat, and pocket mice (Chaetodipus and Perognathus) and deermice (Peromyscus) exploit the bush (e.g., Rosenzweig and Winakur 1969, Brown and Lieberman 1973, Rosenzweig 1973, 1977, Hutto 1978, M'Closkey 1978, Price 1978, Wondolleck 1978, Bowers 1982, Kotler 1984a, b, but see Thompson 1982 and Brown 1989a). Responses of individuals and populations to manipulations of habitat structure, illumination, resource availability, and forage substrate suggest that these differences in foraging behavior help promote species coexistence (Rosenzweig 1973, Kotler 1984a, b, Price and Waser 1985, Brown et al. 1988, Kotler et al. 1988, Price and Longland 1989). In terms of microhabitat preferences and utilization, the two gerbil species behaved like a Namib Desert gerbil (Gerbillurus tytonis, Hughes and Ward 1993) and like pocket mice in North American deserts (Rosenzweig 1973, Price 1978, Kotler 1984a; for comparisons with GUDs see Brown et al. 1988, Brown 1989a). The jerboa, however, did not closely parallel the behavior of kangaroo rats (B. P. Kotler et al., unpublished manuscript). At a Sonoran desert site, kangaroo rats accounted for [is greater than] 50% of GUDs in seed trays, and their average monthly GUDs ranged from 0.318 and 0.348 g to 0.952 and 1.414 g in the bush and open microhabitats, respectively (Brown 1989a). In this study, jerboas accounted for [is approximately equal to]1% of the GUDs, and except for January 1988, their average monthly GUDs were never [less than] 1 g.
Habitat selection in a mosaic
Sand habitats strongly influenced the foraging behavior of all three rodent species (Results 2 and 7). Based on GUDs, all three species appeared to prefer the less stabilized sand dune habitats. Based on activity, G. pyramidum and J. jaculus predominated in the less stabilized sand habitats, and G. allenbyi predominated in the more stabilized sand habitats. Despite these divergent patterns of sand habitat use, we conclude that most of the data refute a mechanism of coexistence based on habitat selection in a mosaic. Independent of sand habitat, G. allenbyi was the most efficient and thorough forager (Results 4 and 8), and J. jaculus was the least efficient and thorough.
Our results accord with those of others for the Negev Desert and the Beer Asluj site (Abramsky et al. 1985, 1990, Rosenzweig and Abramsky 1985, 1986, Abramsky and Pinshow 1989). In particular, G. allenbyi's heavy use of the stabilized sand habitat (as evidenced by tracking plots) contrasted with its preference for less stabilized sand habitats (as evidenced by GUDs). This preference has been corroborated in removal experiments within 1-ha enclosures (Abramsky et al. 1990). In the presence of G. pyramidum, G. allenbyi used predominately the more stabilized sand habitat. In the absence of G. pyramidum, G. allenbyi at low density shifted all of its activity to the less stabilized and at high densities used both habitats equally.
Spatial variation in resource abundance
This mechanism predicts that the less efficient foragers will visit many more resource patches. Based on GUDs, G. allenbyi was the most efficient forager, and J. jaculus was the least (Result 8). As predicted by this mechanism, the species with the highest (J. jaculus) and lowest (G. allenbyi) GUDs also had the lowest and highest scores in tracking plots, respectively (Result 4). However, there was no difference between the average number of tracking plots visited by an individual of G. allenbyi and an individual of G. pyramidum (Result 5). This makes it unlikely that G. pyramidum coexists with G. allenbyi by finding rich patches, skimming the cream, and then moving on to search for other rich patches.
Unfortunately, we cannot evaluate the mechanism for J. jaculus because we lack census data for estimating the number of tracking plots visited per capita. Spatial variation in resource abundance remains a possible mechanism for promoting the coexistence of J. jaculus with the two gerbil species, particularly if there were relatively few individuals of J. jaculus on our study site (i.e., at least an order of magnitude less than G. allenbyi).
Some evidence exists for this mechanism in North America. At our Sonoran Desert site, this prediction was supported by Merriam's kangaroo rat and Harris's antelope squirrel (Brown 1989a). The kangaroo rat foraged patches to a lower resource density while the ground squirrel, on a per capita basis, traveled much farther and visited more experimental patches. Also, the clump size selection hypothesis for the coexistence of mobile vs. efficient species is an example of this mechanism (Reichman and Oberstein 1977, Reichman and Roberts 1994); movement data of pocket mice and kangaroo rats suggest that kangaroo rats may visit many more patches, but forage them less thoroughly than do pocket mice (Bowers 1982, Thompson 1982).
Temporal variation in resource abundance
Both G. pyramidum and J. jaculus could compensate for their lower foraging efficiencies if they used dormancy or torpor to travel inexpensively from one season of rich resources to the next. As predicted by this mechanism, the abundances of the two gerbil species tended to covary positively, suggesting that each does best during the same seasons. However, there was no evidence for torpor, at least among the gerbils in the community. While our data do not support this mechanism of coexistence on an annual basis, there is strong evidence to suggest that this mechanism is operating on a daily basis. G. pyramidum appears to visit food patches first, with G. allenbyi foraging later and beating out G. pyramidum for the GUD (Result 8).
Inspired by the results of this paper, two subsequent studies at Beer Asluj have found that G. pyramidum forages relatively earlier in the night than G. allenbyi (Kotler et al. 1993). Interference competition may give G. pyramidum a priority effect in food patches before the patches have become depleted. The higher efficiency of G. allenbyi may allow it to continue exploiting seeds from patches after they have been depleted by G. pyramidum. Removal of G. pyramidum results in G. allenbyi shifting its foraging activity to earlier in the night and shifting its activity towards the less stabilized sand (Ziv et al. 1993).
Temporal variation in foraging costs
None of our data support this mechanism of coexistence. Although GUDs changed across months, species did not differ in average GUDs either overall or from month to month. Species population sizes did not vary asynchronously, and the species rankings of per capita activity and scores in tracking plots did not rotate seasonally.
These results from the Negev Desert community contrast sharply with those obtained at a Sonoran Desert site using similar techniques of seed trays and census trapping (Brown 1989a). In that community, Merriam's kangaroo rat was the most efficient forager species in both the bush and open microhabitats from November until April, the round-tailed ground squirrel was most efficient in both microhabitats during June and July, and the Arizona pocket mouse was most efficient in both microhabitats from August until October. This seasonal rotation in foraging efficiencies was associated with a negative covariance in species population densities. From patterns of seasonal activity and dormancy, several authors have also suggested that coexisting North American desert rodent species temporally partition the year (MacMillen 1964, Reichman and Van de Graaff 1973, Petryszyn 1982, Brown and Zeng 1989).
Coexistence among gerbils and jerboas
The presence of G. allenbyi within the rodent community at Beer Asluj is easily understood. Within all microhabitats (bush and open) and within all sand habitats, it tended to be the most efficient forager. Therefore, the mechanism(s) permitting the presence of G. pyramidum and J. jaculus require explanation.
Sand habitat must play a role in explaining the presence of G. pyramidum, but in a less simple manner than postulated by the mechanism based on habitat selection in a mosaic. It appears that G. pyramidum persists because of the semistabilized sand habitats and in spite of the stabilized sand habitats dominated completely by G. allenbyi. In addition, the nightly foraging period of G. pyramidum is somewhat nested within that of G. allenbyi (Kotler et al. 1993, Ziv et al. 1993). It appears that in the semistabilized sand habitats, but not in the stabilized sand habitats, the mechanism of coexistence based on temporal variation in resource abundances on a daily basis may be operating.
Only circumstantial evidence exists for daily renewal and depletion of seed abundances. Seed densities in desert habitats vary greatly in time and space, as much as 70-fold over distances of a metre or two and from month to month (Reichman and Oberstein 1977, Reichman 1984, Price and Reichman 1987). But data are not available to evaluate fluctuations on a daily basis. Seeds become redistributed by wind and soil disturbances; blowing sand can cover, uncover, and aggregate seeds. At Beer Asluj, daily afternoon winds usually blow strongly enough to erase footprints and accrete or erode up to several centimetres of sand. The daily redistribution of sand together with the nightly emergence of rodents to forage indicate some sort of daily renewal process. If this process occurs, it should be more pronounced in semistabilized sand habitats than in stabilized sand habitats where soil crusts and vegetation inhibit wind erosion and accretion. Less stabilized sand habitats should promote higher variability in temporal seed abundances. This, in turn, favors the success of the cream-skimming species (Brown 1989b). By cream-skimmer, we mean a species that only harvests relatively rich patches due to a high GUD. Such a species usually compensates for its low foraging efficiency and high GUD by either traveling frequently to many patches (Reichman 1981, Brown 1986), by remaining dormant during periods of low resource abundance (Brown 1989b), or by gaining a priority effect on rich patches via interference competition.
J. jaculus's presence in the community may be promoted by two possible mechanisms. While we could not document per capita activity, jerboas can travel great distances during the night (Osborn and Hemly 1980; A. Subach, unpublished data). In this way, they may be "cream skimmers," taking advantage of spatially variable resources. Alternatively, jerboas may be far less granivorous and much more herbivorous (green vegetation and roots) than the two gerbil species (Osborn and Helmy 1980, Bar et al. 1984). During January 1988, jerboas foraged efficiently for millet in seed trays, but otherwise they showed relatively little interest or success in the seed trays. At best, J. jaculus at Beer Asluj is a highly mobile granivore with a very low foraging efficiency.
It seems appealing that convergence in community organization would follow from strong morphological convergencies of species within different communities. The independent evolution of desert rodents in Old and New World deserts and the striking similarity of gerbils to pocket mice and jerboas to kangaroo rats make the notion of community convergence even more attractive.
However, diet separation may be far more important for promoting the coexistence of J. jaculus with Gerbillus sp. than habitat selection in time or space. If so, then J. jaculus may not fill the same ecological roles as most kangaroo rats (also see Mares 1993). In this way, the rodents of Beer Asluj show little community convergence on those studied in North American deserts. Ecologically, J. jaculus may have more in common with the springhare (Pedetes pedetes) of Africa or woodrats (Neotoma sp.) in North American deserts.
If coexistence of the two gerbil species is based upon daily variation in seed abundances, then our community has even less in common with the community organizations that have been reported for rodents of North American deserts. Still, nightly temporal partitioning may be partly responsible for the coexistence of kangaroo rats and kangaroo mice on sand dunes in the Great Basin Desert of Nevada where kangaroo mice are energy efficient (Morgan and Price 1992), enjoy low rates of predation (Kotler 1985), but appear to suffer greatly from interference from larger rodents (Kotler 1989). Regardless, our community may be much more similar in organization to that of Schaffer et al. (1979) with three bee species and that of Brown et al. (1981) with hummingbirds and nectarivorous insects that appear to coexist through daily renewal and depletion of nectar resources. Not only may there be as much variability in the organization of communities within deserts as among deserts (see Kotler and Brown 1988), there may be as much variability within desert rodent communities as between desert rodents and the communities of other diverse taxa and biomes.
Does the lack of shared convergence of morphology and community organization bode poorly for understanding ecological communities? We think not. Our results simply suggest that the kinds of factors that shape the more conspicuous attributes of species may not be the same as those factors that promote or inhibit species coexistence. Because of this, mechanisms of coexistence should have broad generality and be independent of taxon or ecosystem. The diversity of organizations found among communities may be far less than would be suggested by the diversity of organisms' life histories and ecologies. Finally, if desert rodent communities exhibit many diverse mechanisms of coexistence, then they can indeed be a model system for understanding community organization in general.
We thank Charles Eesley, M. F. Eyphat, Matthew Goldowitz, Oren Hasson, Reuven Yosef, Berry Pinshow, Jean Powlesland, Aziz Subach, and Laurie Zaarur for assistance with field work and seed sorting. We thank Zvika Abramsky, B. Pinshow, Thomas Poulson, Michael Rosenzweig, and two anonymous reviewers for valuable comments and discussion. B. P. Kotler is a Bat-Sheva de Rothschild Fellow. This work was supported by the United States-Israel Binational Science Foundation grant Number 86-00087 (to B. P. Kotler, Z. Abramsky, and M. Rosenzweig), and grant Number 84-00065 (to Z. Abramsky, B. Pinshow, and Warren Porter). The Jacob Blaustein International Center for Desert Studies provided financial assistance for J. S. Brown and W. A. Mitchell. This is publication number 185 of the Mitrani Center for Desert Ecology.
Abrams, P. 1984. Variability in resource consumption rates and the coexistence of competing species. Theoretical Population Biology 23:106-124.
Abramsky, Z., S. Brand, and M. L. Rosenzweig. 1985. Geographical ecology of gerbilline rodents in the sand dune habitats of Israel. Journal of Biogeography 12:363-372.
Abramsky, Z., and B. Pinshow. 1989. Changes in foraging effort in two gerbil species with habitat type and intra- and interspecific activity. Oikos 56:43-53.
Abramsky, Z., M. L. Rosenzweig, and B. Pinshow. 1991. The shape of a gerbil isocline measured using principles of optimal habitat selection. Ecology 72:329-340.
Abramsky, Z., M. L. Rosenzweig, B. Pinshow, J. S. Brown, B. P. Kotler, and W. A. Mitchell. 1990. Habitat selection: an experimental field test with two gerbil species. Ecology 71:2358-2369.
Bar, Y., Z. Abramsky, and Y. Gutterman. 1984. Diet of gerbilline rodents of the Israeli desert. Journal of Arid Environments 7:371-376.
Behrends, P. 1986. Utilization of seed dispersions by two sympatric kangaroo rat species. Southwestern Naturalist 31:548-551.
Bowers, M. A. 1982. Foraging behavior in heteromyid rodents: field evidence for resource partitioning. Journal of Mammalogy 63:361-367.
Brown, J. H. 1975. Geographical ecology of desert rodents. Pages 315-341 in M. L. Cody and J. M. Diamond, editors. Ecology and evolution of communities. Belknap, Cambridge, Massachusetts, USA.
Brown, J. H., A. Kodric-Brown, T. G. Whitham, and H. W. Bond. 1981. Competition between hummingbirds and insects for the nectar of two species of shrubs. Southwestern Naturalist 26:133-145.
Brown, J. H., and G. Lieberman. 1973. Resource utilization and coexistence of seed-eating rodents in sand dune habitats. Ecology 54:788-797.
Brown, J. H., and Z. Zeng. 1989. Comparative population ecology of eleven species of rodents in the Chihuahua Desert. Ecology 70:1507-1525.
Brown, J. S. 1986. Coexistence on a resource whose abundance varies: a test with desert rodents. Dissertation. University of Arizona, Tucson, Arizona, USA.
-----. 1988. Patch use as an indicator of habitat preference, predation risk, and competition. Behavioral Ecology and Sociobiology 22:37-47.
-----. 1989a. Desert rodent community structure: a test of four mechanisms of coexistence. Ecological Monographs 20:1-20.
-----. 1989b. Coexistence of a seasonal resource. American Naturalist 133:168-182.
-----. 1992. Patch use of predation risk. I. Models and predictions. Annales Zoologici Fennici 29:301-309.
Brown, J. S., B. P. Kotler, R. J. Smith, and W. O. Wirtz II. 1988. The effects of owl predation on the foraging behavior of heteromyid rodents. Oecologia (Berlin) 76:408-415.
Caraco, T. 1979. Time budgeting and group size: a theory. Ecology 60:611-617.
Charnov, E. L. 1976. Optimal foraging, the marginal value theorem. Theoretical Population Biology 9:129-136.
Conover, W. J., and R. L. Iman. 1981. Rank transformations as a bridge between parametric and nonparametric statistics. American Statistician 35:124-129.
Emlen, J. M. 1966. The role of time and energy in food preference. American Naturalist 100:611-617.
Goodfriend, W., D. Ward, and A. Subach. 1991. Standard operative temperatures of two desert rodents, Gerbillus allenbyi and G. pyramidum--the effects of morphology, microhabitat, and environmental factors. Journal of Thermal Biology 16:157-166.
Hughes, J. J., and D. Ward. 1993. Predation risk and distance to cover affect foraging behaviour in Namib Desert gerbils. Animal Behaviour 46:1243-1245.
Hutto, R. 1978. A mechanism for resource allocation among sympatric heteromyid rodent species. Oecologia (Berlin) 33:115-126.
Kotler, B. P. 1984a. Predation risk and the structure of desert rodent communities. Ecology 65:689-701.
-----. 1984b. Harvesting rates and predatory risk in desert rodents: a comparison of two communities on different continents. Journal of Mammalogy 65:91-96.
-----. 1985. Microhabitat utilization in desert rodents: a comparison of two methods of measurement. Journal of Mammalogy 66:374-378.
-----. 1989. Temporal variation in the structure of a desert rodent community. Pages 127-140 in D. W. Morris, Z. Abramsky, B. J. Fox, and M. R. Willis, editors. Patterns in the structure of mammalian communities. Texas Tech University Press, Lubbock, Texas, USA.
Kotler, B. P., and J. S. Brown. 1988. Environmental heterogeneity and the coexistence of desert rodents. Annual Review of Ecology and Systematics 19:281-307.
Kotler, B. P., and J. S. Brown. 1990. Harvest rates of two species of gerbilline rodents. Journal of Mammalogy 71:591-596.
Kotler, B. P., J. S. Brown, and O. Hasson. 1991. Owl predation on gerbils: the role of body size, illumination, and habitat structure on rates of predation, Ecology 72:2249-2260.
Kotler, B. P., J. S. Brown, R. J. Smith, and W. O. Wirtz, II. 1988. The effects of morphology and body size on rates of owl predation on desert rodents. Oikos 53:145-152.
Kotler, B. P., J. S. Brown, and A. Subach. 1993. Mechanisms of species coexistence of optimal foragers: temporal partitioning by two species of sand dune gerbils. Oikos 67:548-556.
Longland, W., and M. V. Price. 1991. Direct observations of owls and heteromyid rodents: can predation risk explain microhabitat use? Ecology 72:2261-2273.
MacArthur, R., and E. Pianka. 1966. On optimal use of a patchy environment. American Naturalist 100:603-609.
MacMillen, R. E. 1964. Population ecology, water relations, and social behavior of a southern California semidesert fauna. University of California Publication in Zoology 71:1-66.
Mares, M. A. 1993. Desert rodents, seed consumption, and convergence. BioScience 43:372-379.
M'Closkey, R. T. 1978. Niche separation and assembly in four species of Sonoran Desert rodents. American Naturalist 112:683-694.
-----. 1981. Microhabitat use in coexisting desert rodents--the role of population density, Oecologia (Berlin) 50:310-315.
Mitchell, W. A., Z. Abramsky, B. P. Kotler, B. P. Pinshow, and J. S. Brown. 1990. The effect of competition on foraging activity in desert rodents: theory and experiments. Ecology 71:844-854.
Morgan, K. R., and M. V. Price. 1992. Foraging in heteromyid rodents: the energy cost of scratch-digging. Ecology 73:2260-2272.
Osborn, D. J., and I. Helmy. 1980. The contemporary land mammals of Egypt (including Sinai). Fieldiana Zoologica 5.
Petryszyn, Y. 1982. Population dynamics of nocturnal desert rodents: a nine year study. Dissertation. University of Arizona, Tucson, Arizona, USA.
Price, M. V. 1978. The role of microhabitat specialization in structuring desert rodent communities. Ecology 58:1393-1399.
-----. 1986. Structure of desert rodent communities: a critical review of questions and approaches. American Zoologist 26:39-49.
Price, M. V., and J. H. Brown. 1983. Patterns of morphology and resource use in North American desert rodent communities. Great Basin Naturalist Memoirs 7:117-134.
Price, M. V., and W. S. Longland. 1989. Use of artificial seed patches by heteromyid rodents. Journal of Mammalogy 70:316-322.
Price, M. V., and O. J. Reichman. 1987. Spatial and temporal heterogeneity in Sonoran Desert soil seed pools, and implications for heteromyid rodent foraging. Ecology 68:1797-1811.
Price, M. V., and N. W. Waser. 1985. Microhabitat use by heteromyid rodents: effects of artificial seed patches. Ecology 66:211-219.
Reichman, O. J. 1981. Factors influencing foraging of desert rodents. Pages 196-213 in A. Kamil and T. Sargant, editors. Foraging behavior: ecological, ethological, and psychological approaches. STPM, New York, New York, USA.
-----. 1984. Spatial and temporal variation in seed distributions in desert soils. Journal of Biogeography 11:1-11.
Reichman, O. J., and D. Oberstein. 1977. Selection of seed distribution types by Dipodomys merriami and Perognathus amplus. Ecology 58:636-643.
Reichman, O. J., and E. Roberts. 1994. Computer simulation analysis of heteromyid rodent foraging in relation to seed distributions: implications for coexistence. Australian Journal of Zoology, in press.
Reichman, O. J., and K. Van de Graaff. 1973. Seasonal reproductive and activity patterns of five species of Sonoran Desert rodents. American Midland Naturalist 90:118-126.
Rosenzweig, M. L. 1973. Habitat selection experiments with a pair of coexisting heteromyid rodent species. Ecology 62:327-335.
-----. 1977. Coexistence and diversity in heteromyid rodents. Pages 84-99 in B. Stonehouse and C. Perrins, editors. Evolutionary ecology. MacMillan, London, England.
|Printer friendly Cite/link Email Feedback|
|Author:||Brown, Joel S.; Kotler, Burt P.; Mitchell, William A.|
|Date:||Dec 1, 1994|
|Previous Article:||Analysis of larval trematode communities.|
|Next Article:||Local adaptation in the armored scale insect Pseudaulacaspis pentagona (Homoptera: Diaspididae).|