Competition between birds and mammals: a comparison of giving-up densities between crested larks and gerbils

Competition between birds and mammals:

A comparison of giving-up densities between crestedlarks and gerbils

JOEL S. BROWN,* BURT P. KOTLER and WILLIAM A. MITCHELLz

Department of Biological Sciences, University of Illinois at Chicago, 845 W. Taylor St., Chicago, IL 60607, USA

and Mitrani Center for Desert Ecology, Blaustein Institute for Desert Research, Ben-Gurion University,

Sede Boqer Campus, 84 993 Israel

Summary

We combined the concept of mechanisms of co-existence with the approach of giving-up densities to study

inter-taxon competition between seed-eating birds and mammals. We measured feeding behaviour in foodpatches to de®ne and study the guild of seed-eating vertebrates occupying sandy habitats at Bir Asluj, NegevDesert, Israel. Despite a large number of putatively granivorous rodents and birds at the site, two gerbilspecies (Allenby's gerbil, Gerbillus allenbyi, and the greater Egyptian gerbil, G. pyramidum) dominated noc-

turnal foraging, and a single bird species (crested lark, Galerida cristata) contributed all of the daytimeforaging. We used giving-up densities to quantify foraging behaviour and foraging e�ciencies. A low giving-up density demonstrates the ability of a forager to pro®tably harvest food at low abundances and to pro®tably

utilize the foraging opportunities left behind by the less e�cient forager. Gerbils had lower giving-up densitiesin the bush than open microhabitat, and lower giving-up densities in the semi-stabilized than stabilized sandhabitats. Crested larks showed the opposite: lower giving-up densities in the open than bush, and on the

stabilized than semi-stabilized sand habitats. Despite these patterns, gerbils had substantially lower giving-updensities than crested larks in both microhabitats, all sand habitats, and during each month. Several mech-anisms may permit the crested lark to co-exist with the gerbils. Larks may be cream skimmers on the high

spatial and temporal variability in seed abundances. Larks may rely on insects, fruit or smaller seeds. Or, larksmay rely on adjacent rocky habitats.

Keywords: foraging theory; Galerida cristata; Gerbillus allenbyi; Gerbillus pyramidum; giving-up density;

granivory; habitat selection; inter-taxon competition; mechanism of co-existence; Negev Desert; patch use;predation risk; sand dunes

Introduction

MacArthur and Pianka (1966) envisioned foraging theory as a conceptual tool for understandingpopulation and community level phenomena. The expectation that organisms should feed sensiblygreatly restricts the set of observable behaviours (prey are unlikely to exhibit behaviours of nuz-zling up to their predators) and greatly increases the likelihood of successfully using feedingbehaviours to investigate population interactions (Rosenzweig, 1981), population dynamics(Fryxell and Lundberg, 1994) and communities (Kotler and Brown, 1988; Werner, 1992). Com-bining optimal patch use behaviour from foraging theory (MacArthur and Pianka, 1966; Charnov,1976) with mechanisms of co-existence from resource theory (Tilman, 1982) generates the giving-

* Address all correspondence to J.S. Brown, Department of Biological Sciences, University of Illinois at Chicago, 845 W.

Taylor St., Chicago, IL 60607, USA.zPresent address: Department of Zoology, University of Wisconsin, Madison, WI 53706, USA.

Evolutionary Ecology 1997, 11, 757±771

0269-7653 Ó 1997 Chapman & Hall

up density (GUD) approach to studying feeding behaviours, predation risk, habitat selection andspecies co-existence (Brown, 1988, 1992). Conceptually, a forager should abandon a depletablefood patch when the bene®t derived from its present harvest rate no longer exceeds the sum ofmetabolic, predation and missed opportunity costs of foraging. Empirically, measuring GUDs (theamount of food remaining in a patch when abandoned by a forager) from experimental foodpatches measures the forager's costs and bene®ts of feeding. At the community level, the GUD, asthe density of food to which a forager can pro®tably exploit the patch, provides one measure ofcompetitive ability (Tilman, 1982). When species co-existence involves the ability of feeding ani-mals to thoroughly exploit the resources of particular environments, the predictions of the asso-ciated mechanisms of co-existence can often be couched in terms of GUDs and other measures ofactivity and feeding behaviour.The giving-up density approach has been applied to examine the mechanisms of co-existence in

desert rodents (Brown, 1989; Kotler et al., 1993a,b; Brown et al., 1994a; Hughes et al., 1994;Bouskila, 1995), squirrels (Smith, 1991; Jedlicka, 1993) and bark-feeding insectivorous birds(Morgan, 1994). In these examples, as in most studies of species co-existence, members of a feedingguild (Root, 1967) are close taxonomic relatives. This eases the task of studying the mechanismsthat promote or inhibit co-existence. Close taxonomic relatives share most traits in common anddi�er in just a few salient features such as body size (in birds and mammals) or bill size (in birds).Close taxonomic relatives are generally amenable to the same sets of experimental procedures.And, close taxonomic relatives generally respond to experimentation on similar temporal andspatial scales.In reality, feeding guilds contain more distantly related taxa (e.g. Paine, 1966; Dayton, 1971;

Colwell and Fuentes, 1975; Brown and Davidson, 1977; Lubchenco and Menges, 1978; Brownet al., 1979, 1981, 1986; Turkington and Harper, 1979; Inouye, 1981; Morse, 1981; Davidson et al.,1984; Thompson et al., 1991). For example, desert granivores (Brown et al., 1979, 1986) includemostly birds, rodents and ants. Any of these taxa can consume nearly all of the annual productivityof seeds (see Price and Jenkins, 1986, and references therein).In an e�ort to deduce mechanisms of species co-existence, we applied the giving-up density

approach to measure habitat selection and foraging e�ciencies of seed-eating birds and mammalsinhabiting sand dunes of the Negev Desert, Israel. We de®ned guild composition as those speciesfeeding from our food patches. Despite a high diversity of putatively granivorous birds in thevicinity (see Table 1), only crested larks (Galerida cristata) fed from our food patches. From among®ve rodent species, two gerbils, Allenby's gerbil (Gerbillus allenbyi) and the greater Egyptian gerbil(G. pyramidum), were responsible for most (> 95%) of the rodent GUDs in our patches (Brownet al., 1994a).Competition between the diurnal larks and the nocturnal gerbils occurs only through the joint

exploitation of seeds. Co-existence between gerbils and larks can occur if: (1) each species possessesa habitat in time or space in which it has the lowest GUD; (2) each species possesses a class offoods on which it has the lowest GUD; or (3) the species with the higher GUD has a compensatorycompetitive advantage in terms of priority at food patches (Kotler et al., 1993b), lower travel costs(Brown, 1986, 1989) or higher foraging e�ciency at high resource abundances (Stewart and Levin,1973). We tested from among the ®rst category of mechanisms of co-existence. Failure to supportany from this category will indicate a need for the second or third categories to understand co-existence of gerbils with larks.On the sand dunes at our study site, potentially relevant habitats in time and space include

season (habitat selection in time), the areas near and away from perennial shrub cover (bush andopen microhabitats), and the mosaic of habitats distinguished by di�erent degrees of sand stabi-lization due to algal soil crusts and vegetation. If larks and gerbils each possess seasons, micro-

758 Brown et al.

habitats or sand stabilization habitats within which they have the lowest GUDs, then the patch usebehaviour of the birds and rodents across the relevant habitats promotes co-existence. Here, wecompared the GUDs of gerbils and crested larks to examine how season, bush and open micro-habitats, and habitats varying in the degree of sand stabilization in¯uence GUDs.

Interpreting the giving-up density

Charnov's marginal value theorem can be generalized to show that a forager should leave adepletable food patch when (Brown, 1988):

H � C � P �MOC �1�where H, C, P and MOC are the quitting harvest rate, metabolic cost, predation cost and missedopportunity cost of foraging, respectively. In general, H and C can be thought of as having units ofenergy per unit time (e.g. J min)1). P can be converted into units of energy per unit time by

Table 1. The granivorous and partially granivorous birds and mammals in the habitats of Bir Asluj

Species Habitat

BirdsChukar (Alectoris chukar) rock and loessSand partridge (Ammoperdix heyi) rock

Quail (Coturnix coturnix) sand and loess b

Black-bellied sandgrouse (Pterocles orientalis) rock and loessRock dove (Columba livia) rock and loess

Turtle dove (Streptopelia turtur) rock and loess a

Palm dove (Streptopelia senegalensis) rock and loessDesert lark (Ammomanes deserti) rockShort-toed lark (Caladrella brachydactyla) loess a

Lesser short-toed lark (Calandrella rufescens) rock b

Crested lark (Galerida cristata) sand and rockSkylark (Alauda arvensis) loess b

Corn bunting (Emberiza calandra) loess b

Cretzschmar's bunting (Emberiza caesia) rock and loess b

Ortolan (Embiriza hortulana) sand, rock and loess b

Green®nch (Carduelis chloris) sand and loessTrumpeter ®nch (Rhodepechys githaginea) rock and loessHouse sparrow (Passer domesticus) rock and loess

RodentsAllenby's gerbil (Gerbillus allenbyi) stabilized sand

Greater Egyptian gerbil (Gerbillus pyramidum) semi-stabilized sandPygmy gerbil (Gerbillus henleyi) sand and loessWagner's gerbil (Gerbillus dasyurus) rock and loess

Common jerboa (Jaculus jaculus) sand and loessBuxton's jird (Meriones sacrementi) stabilized sandGentle jird (Meriones crasus) loess and rock

House mouse (Mus musculus) loess

a Present in spring, summer and autumn.b Present in winter and during spring and autumn migration.

Competition between birds and mammals 759

multiplying the predation risk (probability of death per unit time) by the marginal rate of sub-stitution of energy for safety. The marginal rate of substitution considers at what rate the animalcan sacri®ce energy for safety and still maintain the same level of ®tness (Caraco, 1979; Brown,1988). The cost of predation includes such terms as predation risk, marginal value of energy andthe animal's state (Brown, 1992).The missed opportunity cost gives the marginal value of time in units of energy per unit time. As

such, it measures the quality of the environment following exploitation by the feeding animals. Inan in®nitely repeating environment (one that does not deplete following some ®xed time period),MOC is positive, negative or zero depending upon whether the forager's expected per capitagrowth rate is positive, negative or zero, respectively. If the environment becomes depleted duringthe course of a day, night or other relevant time period, then the foragers should cease feeding andretire to a burrow, nest or den as a way of conserving energy and remaining safe. In this case,MOCbecomes the negative of the cost of resting in one's refuge (Brown, 1992; Brown et al., 1994b).A number of studies have indicated that the patch use behaviour of feeding animals conforms to

Equation (1). In appropriate experimental food patches, there is a monotonic relationship betweenthe forager's quitting harvest rate and its GUD (Kotler and Brown, 1990; Brown et al., 1994b;Smith, 1995). As predicted, GUDs increase with metabolic costs (Kotler et al., 1993a), predation(Brown et al., 1988; Kotler et al., 1988, 1991) and missed opportunity costs of foraging (Brownet al., 1992b). Giving-up densities in depletable food patches have been used to investigate animals'diets (Brown and Morgan, 1995), patch use strategies (Valone and Brown, 1989), predation risk(Kotler et al., 1991) and habitat preferences (Brown and Alkon, 1992; Bowers et al., 1993; Hughesand Ward, 1993; Hughes et al., 1995).As an approach to studying community level processes, GUDs measure foraging e�ciencies

(Tilman, 1982; Vance, 1985; Brown, 1988). The species with the lower GUD has an advantage overother species in that it depresses resource abundances below the other species' subsistence levels.The only exceptions occur when environments are unusually rich or poor. In a rich environment, ahigh GUD may manifest the forager's high state, or low marginal value of energy (both of whichraise the predation cost). Such an environment is not sustainable, in that either the forager'spopulation size will grow accordingly, or the environment's richness will deplete. Similarly, the lowGUD of animals in a poor environment may re¯ect the desperation of an animal's low state andhigh marginal value of energy (both lower the cost of predation). This `Stalingrad' e�ect (manifestin the behaviour of starving German soldiers in the winter of 1942 prior to their capitulation to theencircling Russians; Craig, 1982) is also not stable, in that the forager's population size will declineaccordingly, or the environment's quality will recover.In summary, for the purposes of this study:

· Giving-up densities in experimental food patches provide a surrogate for the forager's quittingharvest rate.

· Giving-up densities provide an estimate of foraging costs.· The species with the lower GUD is the more e�cient forager.

Methods

Study site

We established two grids of 16 stations each on sand dunes at Bir Asluj, Holot Mashabim NatureReserve, northwestern Negev Desert, Israel. The grids were 4 ´ 4 arrays with 40 m intervalsbetween stations. The site o�ers a hierarchy of habitats. At the largest scale, distinct boundariesseparate extensive and continuous tracts of sandy and rocky habitats. Each of our grids lay entirely

760 Brown et al.

within the sandy habitat, with distances of greater than 100 m between grid margins and the closestrocky habitat.The sandy habitat o�ers a mosaic of sand stabilization types based on soil crusts and organic

matter. Zvika Abramsky, as an expert and independent observer, scored the sand habitat at eachstation on a scale of 1±4 based on the degree of sand stabilization within a 5 m radius of the station,where 1 � sand completely stabilized by algal soil crust (7 stations), 2 � sand partially stabilized byalgal soil crust (10 stations), 3 � mostly shifting sand with broken patches of soil crust (7 stations),4 � shifting sand bare of any soil crust (8 stations). Substrate habitat scores of 1 and 2 correspondto stabilized sand dune, and habitat scores of 3 and 4 correspond to semi-stabilized sand dune. Atthe smallest scale, each station o�ered bush and open microhabitats: bush � next to the canopy ofa perennial shrub, open � about 2 m from the nearest shrub. The dominant perennial plants areArtemisia monosperma and Retama raetam (Abramsky et al., 1985).

Seed-eating rodents and birds

The ecologies of Allenby's gerbil and the greater Egyptian gerbil at Bir Asluj are well known(Abramsky et al., 1985, 1990, 1991; Abramsky and Pinshow, 1989; Mitchell et al., 1990; Kotleret al., 1993a,b; Ziv et al., 1993). Allenby's gerbil (25 g) occupies all sand habitats found at the site,whereas the greater Egyptian gerbil (39 g) occupies primarily the semi-stabilized sand (sand hab-itats 3 and 4). Analysis of the rodent data from this project (Brown et al., 1994a) revealed that bothgerbils had lower GUDs on the semi-stabilized than on the stabilized sand habitats, and in the bushthan in open microhabitat. In all sand habitats and in both microhabitats, Allenby's gerbil had alower GUD than the greater Egyptian gerbil. Other rodents found on the sand dunes were in-consequential either because of their low abundances (pygmy gerbil, G. henleyi, and Buxton's jird,Meriones sacramenti), or because of low foraging e�ciencies that usually precluded them as the®nal forager in the food patches (common jerboa, Jaculus jaculus).Allenby's gerbil and the greater Egyptian gerbil co-exist via daily renewal of resource patches

and a trade-o� between foraging e�ciency at high versus low resource abundance (Kotler et al.,1993b). The greater Egyptian gerbil arrives at resource patches earlier than the smaller All-enby's gerbil, leaves earlier and, on average, forages on richer patches. Allenby's gerbil forageslater in the evening, but depletes resource patches to a lower GUD than the greater Egyptiangerbil.Bir Asluj o�ers a diverse community of birds as potential foragers in our food patches. All of the

species listed in Table 1 were seen on, over or within 500 m of our grids. Throughout the 15 monthsof the study, however, crested larks were the only birds known to forage from our food patches,based on visual spot checks of birds in the food patches and from footprints in the patches' sand.Of the birds at Bir Asluj, crested larks are distinct in having a long hindtoe, followed by a similarlylong claw. Crested larks range over most of Europe, North and East Africa, the Middle East andthe steppes of Central Asia. Their diet includes insects and fruits, but mostly seeds. In comparisonto other granivorous birds of the Negev, the crested lark (39 g) uses habitats opportunistically, digsvigorously with its beak to break soil crusts, and biases its diet towards larger seeds (Shkedy, 1990,1992; Shkedy and Safriel, 1991).At Bir Asluj, the crested larks moved in pairs or singly during the breeding season (February±

March). For the rest of the year, crested larks could still be found singly or in pairs, but morefrequently they associated loosely in ¯ocks numbering 4±10 individuals. On a grid, one or two ofthese ¯ocks would move from station to station when exploiting the food patches. Frequently, the¯ock would spread out so that individuals simultaneously exploited food patches at two or threestations.


Experimental protocol

From November 1986 until January 1988, we completed nine rounds of data collection (see Brownet al., 1994a). A round consisted of live-trapping small mammals, measuring rodent activity intracking stations, and measuring the GUDs of rodents and crested larks from food patches. Foodpatches consisted of aluminium seed trays (45 ´ 60 ´ 2.5 cm deep) ®lled with 3 g of millet seedmixed into 5 litres of sifted sand. Each station of each grid received a pair of food patches. One trayof 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, andrecharging the tray's sand with 3 g of millet. The GUD of the tray was credited to the one orseveral species whose tracks were visible in the tray. If a species' tracks were found adjacent to, butnot in, a foraged tray, then we assumed that it foraged in the tray, but that individuals of otherspecies came later to forage from the tray and covered the sand with their own tracks. Seedsrecovered from trays were cleaned of debris and weighed to measure the forager's GUD (Brown,1988). Data were collected at dawn and dusk. Seed trays set up at dawn and collected at duskmeasured the foraging behaviour of crested larks, and those set up at dusk and collected at dawnmeasured the foraging of the two gerbil species. For each round, we collected GUD data for 6±7mornings and afternoons. Days of data collection were consecutive, weather and logistics per-mitting.

Results

To test for the e�ects of season, microhabitat and sand habitat on the GUDs of larks and gerbils,we used a partially hierarchical ANOVA (Brownlee, 1965), with taxon (gerbil or lark), month (thenine rounds of data collection), microhabitat (bush and open), sand habitat (1±4; stabilized tosemi-stabilized sand dunes, respectively) and station (32 stations) as independent variables. Taxon,month and microhabitat were fully-crossed factors, and station was a factor nested within sandhabitat. As the dependent variable, we used the mean GUD at a station across the 6±8 days of around. We calculated separate means for the gerbils (morning data collection) and the larks(afternoon data collection). This analysis gave us 1152 data points (2 taxa ´ 9 rounds ´ 2 mi-crohabitats ´ 32 stations).By using the mean GUDs across days, we simultaneously lose information, streamline the

statistical analysis, and avoid possible non-independences associated with repeated measures of thesame tray. We lose two types of information. The ®rst is the e�ect of day within a round of datacollection; day is a variable nested within round. Including day as a variable when analysing datafor gerbils or larks separately may be useful. However, while it is possible to pair lark and gerbildata by station, it is arbitrary to do so for day because the taxa are temporally separated. Does onepair the previous night with the subsequent day, or vice-versa? The second loss of informationoccurs because the mean GUDs do not discriminate between the two gerbil species. We did notdiscriminate because we were interested in comparing the mammalian granivores with the aviangranivores in their e�ciencies at depressing seed abundances. Elsewhere, we report on the resultsfrom comparing the foraging behaviour and GUDs of Allenby's gerbil, the greater Egyptian gerbiland the common jerboa (Brown et al., 1994a).The largest e�ect in the ANOVA (Table 2) showed that crested larks had a signi®cantly higher

GUD than the gerbils (2.149 vs 0.593 g). The higher GUD of larks occurred across all months,microhabitats and sand habitats, despite the signi®cant interaction e�ects of taxon with month,microhabitat and sand habitat. The three-way interactions were not signi®cant. The main e�ects of

762 Brown et al.

month and microhabitat were signi®cant, but are not particularly interesting in light of the stronginteraction e�ects with taxon.Because of the strong interaction e�ect of taxon with month, we performed an array of post-hoc

tests (Bonferroni adjustments to experimentwise error rate) (Fig. 1). These tests compared theGUDs of the gerbils with those of the larks on a month by month basis (nine comparisons).Among gerbils and larks, respectively, we made all of the pairwise comparisons of months (anadditional 72 comparisons). During each month, gerbils had a signi®cantly lower GUD thancrested larks (P < 0:001 for all comparisons). For gerbils, months fell into two categories. Feb-ruary and April had signi®cantly higher GUDs than the other 7 months (P < 0:001 for the 14comparisons, and P > 0:5 for the comparison of February with April). For crested larks, monthsfell into four categories. February, April and June had similar GUDs (P > 0:4 for the threecomparisons) that were signi®cantly higher than the other months (P < 0:001 for the 18 com-parisons). November 1987 had the lowest GUD followed by January 1987 (November 1987 <January 1987, P < 0:001). Both of these months had signi®cantly lower GUDs than other months.A Spearman rank correlation revealed a positive, but non-signi®cant, relationship between theGUDs of gerbils and larks �R � 0:55; P > 0:1�. Most striking were the high GUDs of both gerbilsand larks during February and April. June was the most divergent month because larks had arelatively high GUD and gerbils had a relatively low GUD (Fig. 1). But, monthly variation withineach taxon paled in comparison to di�erences between taxa.We analysed the interaction e�ect of sand habitat with taxon in a similar fashion to that of

month (Fig. 2). We compared the GUDs of gerbils and larks for each of the sand habitats (fourcomparisons) and then compared among sand habitats separately for gerbils and crested larks (anadditional 12 comparisons). In all sand habitats, gerbils had signi®cantly lower GUDs than crestedlarks (P < 0:001, Bonferroni-adjusted post-hoc test). For gerbils, the habitats fell into two cate-gories. The least stabilized sand habitat (#4) had a signi®cantly lower GUD than the other threehabitats (4 and 1, P < 0:001; 4 and 2, P < 0:05; 4 and 3, P < 0:01), which did not di�er signi®cantlyfrom each other. For crested larks, habitats also fell into two categories. In contrast to the gerbils,larks had a signi®cantly lower GUD in sand habitat (#1) than in the other three (1 and 2,

Table 2. The ANOVA showing the e�ects of species (lark vs gerbil), microhabitat (bush vs open), sand habitat(four grades of stabilization from stabilized to semi-stabilized) and month on giving-up densities. Station is avariable nested within sand habitat. The model yields r2 � 0:89

Factor d.f. MS F

Species 1 683.11 7101.90 ***

Microhabitat 1 1.318 13.71 ***Sand habitat 3 0.375 1.87Month 8 13.45 139.80 ***

Species ´ microhabitat 1 6.91 71.80 ***Species ´ sand habitat 3 2.133 22.20 ***Species ´ month 8 2.757 28.70 ***Species ´ microhabitat ´ sand habitat 3 0.040 0.41

Species ´ microhabitat ´ month 8 0.120 1.25Species ´ sand habitat ´ month 24 0.071 0.74Station (nested within sand habitat) 28 0.201 2.14 *

Error 1063 0.094

* P < 0:05; *** P < 0:001.


Figure 1. The monthly giving-up densities of the gerbil species and the crested lark. The species' bars have

been superimposed, not stacked. Hence, the top of each species' bar indicates its GUD. During each period,the crested lark had a signi®cantly higher giving-up density than the gerbils. The error bars indicate onestandard error around the mean.

Figure 2. The e�ect of sand habitat on the giving-up densities of the gerbil species and the crested lark. The

habitat scores of 1±4 represent a continuum from stabilized to semi-stabilized sand habitats. The crested larkhad a signi®cantly lower giving-up density in the most stabilized sand habitat. The gerbils showed theopposite, and had a signi®cantly lower giving-up density in the least stabilized sand habitat.

P < 0:001; 1 and 3, P < 0:01; 1 and 4, P < 0:001). A perfect negative rank-correlation existedbetween the GUDs of gerbils and larks across sand habitats.For the interaction e�ect between taxon and microhabitat, we used a Bonferroni-adjusted post-

hoc test to compare gerbils and larks in each microhabitat (two comparisons) and to compare bushand open microhabitats separately for each taxon (two additional comparisons) (Fig. 3). Gerbilshad a signi®cantly lower GUD in the bush than in the open microhabitat �P < 0:01�, while larkshad a signi®cantly lower GUD in the open than in the bush microhabitat �P < 0:001�. In bothmicrohabitats, gerbils had signi®cantly lower GUDs than larks (P < 0:001 for both comparisons).

Discussion

On the sand dunes at Bir Asluj, the two gerbil species (Allenby's gerbil and the greater Egyptiangerbil) were more e�cient granivores (as measured by giving-up densities) than the crested larkduring every month, on every sand habitat, and in both the bush and open microhabitat. Severalrelated issues merit discussion:

1. Is the lower e�ciency of larks more apparent than real?2. What mechanisms of co-existence might promote the presence of larks in the community?3. What is the role of predation risk in shaping the foraging behaviour of larks and gerbils?4. How does the foraging behaviour of gerbils and larks ®t into the larger desert landscape of

sandy, rocky and loessal habitats?5. Do birds and mammals restrict each other's distribution and diversity?

Figure 3. The e�ect of the bush and open microhabitat on the giving-up densities of the gerbil species and thecrested lark. The gerbils had a signi®cantly lower giving-up density in the bush than open microhabitat, and

vice versa for the crested lark. The error bars indicate one standard error around the mean.


Foraging e�ciencies of gerbils versus larks

Within our food patches, gerbils harvested millet seeds to a much lower level than larks. We feelthat this supports the conclusion that gerbils also deplete native seedbanks to a lower density thanlarks. Two concerns must be addressed. First, the experimental food patches might give an ad-vantage to gerbils that is not relevant under natural conditions. If larks cannot exploit the fulldepth of 2.5 cm of sand, then maybe larks are more e�cient foragers on surface seeds while gerbilscan access deeper recesses of the patch. This is unlikely. Larks swung their beaks from side to sideto dig thoroughly through the sand and, based on their lowest GUDs, larks are capable ofdepleting the seeds to less than one-third of their average GUDs. The high GUDs of the larks inour trays stem from an unwillingness to forage longer, rather than the inability to access all of theseeds in the patch.Secondly, larks may overtake gerbils in foraging e�ciency as seed size declines. Millet is a

relatively large seed (c. 6 mg). However, it has recently been demonstrated that the foraginge�ciency of the crested lark changes little with seed size, although that of the gerbil species dropssharply on smaller seeds (c. 1.5 mg; J. Garb, B.P. Kotler and J.S. Brown, unpublished results).Regardless, the gerbils had lower GUDs than larks on all seed sizes examined (1.5±6 mg).

The mechanism of co-existence between larks and gerbils

When a lark leaves a food patch, the patch remains a valuable feeding opportunity for a gerbil. Incontrast, when a gerbil leaves a food patch, the patch must experience considerable resourcerenewal before providing a pro®table foraging opportunity for a crested lark. The conditions forgerbils to persist in the presence of crested larks are obviously met, but less clear is how crestedlarks co-exist with gerbils. Larks forage during the hours of resource renewal by wind action(mornings and afternoons; Kotler et al., 1993b). In this way, a mechanism of co-existence wouldinvolve temporal and spatial variability in seed abundances, and a trade-o� among the taxa ine�ciency (higher in gerbils) and speed/mobility (greater in larks) (Brown, 1986, 1989). Larks andgerbils would function as `cream skimmers' and `crumb pickers', respectively. Alternatively,mechanisms such as diet selection may explain co-existence. Gerbils may be more e�cient foragerson seeds, whereas crested larks may be more e�cient foragers on insects (J.S. Brown and B.P.Kotler, unpublished results). Habitat selection at a larger scale may provide yet another mecha-nism of co-existence. While individual gerbils are restricted to the sand dunes, the mobility of larksand the landscape of Bir Asluj permits individual larks to exploit rocky, loessal and sandy habitats.Larks may require the other habitats to prevent competitive exclusion.

The role of predation risk for crested larks and gerbils

Crested larks had lower GUDs in the open than bush microhabitat, and lower GUDs on the moststabilized sand habitat. Because the same ¯ocks of larks had access to food patches in bothmicrohabitats and all sand habitats, the di�erences in GUDs should re¯ect predation costs ratherthan metabolic or missed opportunity costs of foraging. At Bir Asluj, predators of crested larks onthe sand dunes include the great gray shrike (Lanius excubitor), hen harrier (Circus cyaneus), pallidharrier (C. macrourus) and monitor lizard (Varanus griseus). Lima (1992, 1993), in his review ofescape tactics of birds, found that several granivorous sparrows and larks of arid and semi-aridhabitats should favour the open rather than the bush microhabitat. Similarly, for the crested lark,the bush microhabitat may o�er higher predation risk by constraining escape ¯ights and facili-tating ambushes by monitor lizards and perching shrikes.

766 Brown et al.

The GUDs of the lark suggest that predation risk increases as the degree of sand stabilizationdeclines. The soil crust surrounding the experimental food patches (the patches themselves wereidentical) may be important to crested larks walking on the ground. Soil crusts may provide bettertraction and support for initiating and directing escape ¯ights or for running across the surface.

The role of other habitats in the interaction between gerbils and larks

The primary interaction and co-existence of seed-eating birds and mammals may actually occur atthe larger scale of rocky and sandy habitats. Rodents may predominate on the sand by virtue ofhigh foraging e�ciencies and may be less abundant on rocky habitats by virtue of high predationcosts (e.g. escape substrate; Brown et al., 1992) and competition from birds. Allenby's gerbils hadmuch higher GUDs on sandy habitats than on rocky or loessal habitats. And, their GUDs in-creased steadily the farther into the rocky habitat the gerbils ventured from the sand (Brown et al.,1992). Interestingly, the GUD of crested larks on the sand (about 2 g) was similar to that of gerbilsin the same seed trays placed in the rocky habitat.The diverse community of granivorous birds (Table 1) may rely on and occupy the rocky and

loessal habitats. In this way, each major habitat (sand and rock) may possess a taxon-speci®c guildof vertebrate granivores. But, the ghost of competition past that drives species to occupy di�erenthabitats (Rosenzweig, 1981; Brown and Rosenzweig, 1986) may not be complete. Crested larks jointhe gerbil species on the sand; and some rodents (Wagner's gerbil, G. dasyurus, and common spinymice, Acomys caharhinus) join diverse species of birds on the various grades of rocky habitats.

Do birds and mammals restrict each other's distribution and diversity?

With a few exceptions (e.g. Thompson et al., 1991), studies of community organization in mam-mals or birds have not considered the in¯uence or role of the other taxon. Examples of potentialresource competition between birds and mammals include fox squirrels and blue jays consumingacorns, nutcrackers and red squirrels consuming pine nuts, frugivorous birds and primates intropical forests (Howe, 1980, 1990), sugar gliders and honeyeaters or lorikeets consuming nectar,shrews and thrushes consuming invertebrates among leaf litter, and mustelids and owls consumingvoles. In these examples, the similarities in diets between the mammal and bird species oftencontrast with di�erences in foraging tactic and foraging scale. Super®cially, mammals often appearto forage more intensively while birds forage more extensively.These features of tactic and scale that promote the co-existence of birds and mammals may

partially explain the absence of joint community studies of birds and mammals; the scale and toolsfor studying one taxon often fail to yield meaningful data for another taxon. In addition, ifmammals and birds do competitively exclude each other, then feeding guilds with high diversitiesof mammals may tend to be depauperate in birds and vice versa. On the other hand, there may besituations where birds and rodents, through their foraging behaviour, create opportunities for eachother. Thompson et al. (1991) suggest that granivorous birds and rodents may facilitate each otherthrough an indirect mutualism. By preferentially consuming large seeds, rodents promote plantsproducing the small seeds favoured by some birds.In general, the mechanisms of co-existence (diet choice, habitat selection, variance partitioning)

reported for mammal communities (e.g. Rosenzweig, 1966; Terborgh, 1983; Kreuger, 1986; Kotlerand Brown, 1988; Morris et al., 1989; McNaughton, 1993) are the same as those reported for birdcommunities (e.g. MacArthur, 1958; Cody, 1974, 1985; Dunning, 1986; Terborgh et al., 1990;Benkman, 1991; Sohonen et al., 1993). It is likely that these mechanisms also explain co-existence,and limits to co-existence, between birds and mammals.


Acknowledgements

We thank Charles Eesley, M.F. Eyphat, Matthew Goldowitz, Oren Hasson, Reuven Yosef, BerryPinshow, Jean Powlesland, Aziz Subach and Laurie Zaarur for assistance with ®eldwork and seedsorting. We thank Zvika Abramsky, John Fryxell, Douglas Morris, Michael Rosenzweig andJames Thorson for stimulating and valuable comments. B.P.K is a Bat-Sheva de RothschildFellow. This work was supported by United States-Israel Binational Science Foundation GrantNo. 86-00087 (to B.P.K., Z. Abramsky and M. Rosenzweig) and Grant No. 93-00236 (to B.P.K.and J.S.B.). The Jacob Blaustein International Center for Desert Studies provided ®nancial as-sistance for J.S.B. and W.A.M. This is publication #225 of the Mitrani Center for Desert Ecology.

References

Abramsky, Z. and Pinshow, B. (1989) Changes in foraging e�ort in two gerbil species with habitat type andintra- and interspeci®c activity. Oikos 56, 43±53.

Abramsky, Z., Brand, S. and Rosenzweig, M.L. (1985) Geographical ecology of gerbilline rodents in the sand

dune habitats of Israel. J. Biogeogr. 12, 363±372.Abramsky, Z., Rosenzweig, M.L., Pinshow, B., Brown, J.S., Kotler, B.P. and Mitchell, W.A. (1990) Habitat

selection: An experimental ®eld test with two gerbil species. Ecology 71, 2358±2369.

Abramsky, Z., Rosenzweig, M.L. and Pinshow, B. (1991) The shape of a gerbil isocline measured usingprinciples of optimal habitat selection. Ecology 72, 329±340.

Benkman, C.W. (1991) Predation, seed size partitioning, and the evolution of body size in seed-eating ®nches.

Evol. Ecol. 5, 118±127.Bouskila, A. (1995) Interaction between predation risk and competition: A ®eld study of kangaroo rats and

snakes. Ecology 76, 165±178.Bowers, M.A., Je�erson, J.L. and Kuebler, M.G. (1993) Variation in giving-up densities of foraging chip-

munks (Tamias striatus) and squirrels (Sciurus carolinensis). Oikos 66, 229±236.Brown, J.H. and Davidson, D.W. (1977) Competition between seed-eating rodents and ants in desert eco-

systems. Science 196, 880±882.

Brown, J.H., Reichman, O.J. and Davidson, D.W. (1979) Granivory in desert ecosystems. Ann. Rev. Ecol.Syst. 10, 201±277.

Brown, J.H., Kodric-Brown, A., Witham, T.G. and Bond, H.W. (1981) Competition between hummingbirds

and insects for the nectar of two species of shrubs. Southwest. Nat. 26, 133±145.Brown, J.H., Davidson, D.W., Munger, J.C. and Inouye, R.C. (1986) Experimental community ecology: The

desert granivore system. In Community Ecology (J.L. Diamond and T.J. Case, eds), pp. 41±61. Harperand Row, New York.

Brown, J.S. (1986) Coexistence on a resource whose abundance varies: A test with desert rodents. PhD thesis,University of Arizona, Tucson, AZ.

Brown, J.S. (1988) Patch use as an indicator of habitat preference, predation risk, and competition. Behav.

Ecol. Sociobiol. 22, 37±47.Brown, J.S. (1989) Desert rodent community structure: A test of four mechanisms of coexistence. Ecol.

Monogr. 20, 1±20.

Brown, J.S. (1992) Patch use under predation risk: I. Models and predictions. Ann. Zool. Fenn. 29, 301±309.Brown, J.S. and Alkon, P.U. (1990) Testing values of crested porcupine habitats by experimental food

patches. Oecologia (Berl.) 83, 512±518.

Brown, J.S. and Morgan, R.A. (1995) E�ects of foraging behavior and spatial scale on diet selectivity: A testwith fox squirrels. Oikos 74, 122±136.

Brown, J.S. and Rosenzweig, M.L. (1986) Habitat selection in slowly regenerating environments. J. Theor.Biol. 123, 151±171.

Brown, J.S., Kotler, B.P., Smith, R.J. and Wirtz, W.O., II (1988) The e�ects of owl predation on the foragingbehavior of heteromyid rodents. Oecologia (Berl.) 76, 408±415.

768 Brown et al.

Brown, J.S., Arel, Y., Abramsky, Z. and Kotler, B.P. (1992a) Patch use by gerbils (Gerbillus allenbyi) in sandyand rocky habitats. J. Mammal. 73, 821±829.

Brown, J.S., Morgan, R.A. and Dow, B.D. (1992b) Patch use under predation risk: II. A test with fox

squirrels, Sciurus niger. Ann. Zool. Fenn. 29, 311±318.Brown, J.S., Kotler, B.P. and Mitchell, W.A. (1994a) Forging theory, patch use, and the structure of a Negev

Desert granivore community. Ecology 75, 2286±2300.

Brown, J.S., Kotler, B.P. and Valone, T.J. (1994b) Foraging under predation: A comparison of energetic andpredation costs in a Negev and Sonoran Desert rodent community. Austr. J. Zool. 42, 435±448.

Brownlee, K.A. (1965) Statistical Theory and Methodology in Science and Engineering, 2nd edn. John Wiley,New York.

Caraco, T. (1979) Time budgeting and group size: A theory. Ecology 60, 611±617.Charnov, E.L. (1976) Optimal foraging and the marginal value theorem. Theor. Pop. Biol. 9, 129±136.Cody, M.L. (1974) Competition and the Structure of Bird Communities. Princeton University Press, Princeton,

NJ.Cody, M.L. (ed.) (1985) Habitat Selection in Birds. Academic Press, New York.Colwell, R.K. and Fuentes, E.R. (1975) Experimental studies of the niche. Ann. Rev. Ecol. Syst. 6, 281±310.

Craig, W. (1982) Enemy at the Gates: The Battle for Stalingrad. Bantam Books, New York.Davidson, D.W., Inouye, R.S. and Brown, J.H. (1984) Granivory in a desert ecosystem: Evidence for indirect

facilitation of ants by rodents. Ecology 65, 1780±1786.

Dayton, P.K. (1971) Competition, disturbance, and community organization: The provision and subsequentutilization of space in a rocky intertidal community. Ecol. Monogr. 41, 351±389.

Dunning, J.B. (1986) Shrub-steppe bird assemblages revisited: Implications for community theory. Am. Nat.128, 82±98.

Fryxell, J.M. and Lundberg, P. (1994) Diet choice and predator±prey dynamics. Evol. Ecol. 8, 407±421.Howe, H.F. (1980) Monkey dispersal and waste of a neotropical fruit. Ecology 61, 944±959.Howe, H.F. (1990) Seed dispersal by birds and mammals: Implications for seedling demography. In Repro-

ductive Ecology of Tropical Plants (K.S. Bawa and M. Hadley, eds), pp. 191±218. Man and BiosphereSeries Vol. 7. UNESCO, Paris and Parthenon Publishing, Carnforth.

Hughes, J.J. and Ward, D. (1993) Predation risk and distance to cover a�ect foraging behaviour in Namib

Desert gerbils. Anim. Behav. 46, 1243±1245.Hughes, J.J., Ward, D. and Perrin, M.R. (1994) Predation risk and competition a�ect habitat selection and

activity of Namib Desert gerbils. Ecology 75, 1397±1405.Hughes, J.J., Ward, D. and Perrin, M.R. (1995) E�ects of substrate on foraging decisions by a Namib Desert

gerbil. J. Mammal. 76, 638±645.Inouye, R.S. (1981) Interactions among unrelated species: Granivorous rodents, a parasitic fungus, and a

shared prey species. Oecologia (Berl.) 49, 425±427.

Jedlicka, D. (1993) A foraging approach to habitat selection in time and space in three sciurid rodents. PhDdissertation, University of Illinois, Chicago, IL.

Kotler, B.P. and Brown, J.S. (1988) Environmental heterogeneity and the coexistence of desert rodents. Ann.

Rev. Ecol. Syst. 19, 281±307.Kotler, B.P. and Brown, J.S. (1990) Harvest rates of two species of gerbilline rodents. J. Mammal. 71, 591±

596.

Kotler, B.P., Brown, J.S., Smith, R.J. and Wirtz, W.O., II (1988) The e�ects of morphology and body size onrates of owl predation on desert rodents. Oikos 53, 145±152.

Kotler, B.P., Brown, J.S. and Hasson, O. (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., Brown, J.S. and Mitchell, W.A. (1993a) Environmental factors a�ecting patch use in two speciesof gerbilline rodents. J. Mammal. 74, 614±620.

Kotler, B.P., Brown, J.S. and Subach, A. (1993b) Mechanisms of species coexistence of optimal foragers:

Temporal partitioning by two species of sand dune gerbils. Oikos 67, 548±556.Kreuger, K. (1986) Feeding relationships among bison, pronghorn, and prairie dogs: An experimental

analysis. Ecology 67, 760±770.


Lima, S.L. (1992) Strong preference for apparently dangerous habitats? A consequence of di�erential escapefrom predators. Oikos 64, 597±600.

Lima, S.L. (1993) Ecological and evolutionary perspectives on escape from predatory attack: A survey of

North American birds. Wil. Bull. 105, 1±47.Lubchenco, J. and Menge, B.A. (1978) Community development and persistence in a low rocky intertidal

zone. Ecol. Monogr. 59, 67±94.

MacArthur, R.H. (1958) Population ecology of some warblers of northeastern coniferous forests. Ecology 39,599±619.

MacArthur, R. and Pianka, E. (1966) On optimal use of a patchy environment. Am. Nat. 100, 603±609.McNaughton, S.J. (1993) Grasses and grazers, science and management. Ecol. Appl. 3, 17±20.

Mitchell, W.A., Abramsky, Z., Kotler, B.P., Pinshow, B.P. and Brown, J.S. (1990) The e�ect of inter- andintra-speci®c competition on foraging e�ort: Theoretical development and a test with granivorous desertrodents in Israel. Ecology 71, 844±854.

Morgan, R.A. (1994) Using giving-up densities to investigate search images, functional responses and habitatselection. PhD thesis, University of Illinois, Chicago, IL.

Morris, D.W., Abramsky, Z., Fox, B.J. and Willig, M.R. (eds) (1989) Patterns in the Structure of Mammalian

Communities. Special Publication No. 28 of the Museum, Texas Technical University, Lubbock, TX.Morse, D.H. (1981) Interactions among syrphid ¯ies and bumblebees on ¯owers. Ecology 62, 81±88.Paine, R.T. (1966) Food web complexity and species diversity. Am. Nat. 100, 65±75.

Price, M.V. and Jenkins, S.H. (1986) Rodents as seed consumers and dispersers. In Seed Dispersal (D.Murray, ed.), pp. 242±261. Academic Press Australia, North Ryde.

Root, R. (1967) The niche exploitation pattern of the blue±grey gnatcatcher. Ecol. Monogr. 37, 317±350.Rosenzweig, M.L. (1966) Community structure in sympatric carnivores. J. Mammal. 47, 602±620.

Rosenzweig, M.L. (1981) A theory of habitat selection. Ecology 62, 327±335.Shkedy, J. (1990) Niche breadth and geographical distribution of the Desert Lark (Ammomanes deserti) and

the Crested Lark (Galerida cristata). PhD dissertation, Hebrew University, Jerusalem.

Shkedy, J. (1992) Niche breadth of two lark species in the desert and the size of their geographical range. OrnisScand. 23, 89±95.

Shkedy, Y. and Safriel, U.N. (1991) Fat reserves of an opportunist and of a specialist species in the Negev

Desert. Auk 108, 556±561.Smith, R.J. (1991) Predation risk and the community organization of montane ground squirrels and a phy-

logenetic test of association between diurnal activity and gregarious behavior in mammals. PhD dis-sertation, University of Arizona, Tucson, AZ.

Smith, R.J. (1995) Harvest rates and escape speeds in two coexisting species of montane ground squirrels. J.Mammal. 76, 189±195.

Sohonen, J., Halonen, M. and Mappes, T. (1993) Predation risk and the organization of the Parus guild.

Oikos 66, 94±100.Stewart, F.M. and Levin, B.R. (1973) Partitioning of resources and the outcome of interspeci®c competition:

A model and some general considerations. Am. Nat. 107, 171±198.

Terborgh, J. (1983) Five New World Primates: A Study in Comparative Ecology. Princeton University Press,Princeton, NJ.

Terborgh, J., Robinson, S.K., Parker T.A., III, Munn, C.A. and Pierpont, N. (1990) Structure and organi-

zation of an Amazonian forest bird community. Ecol. Monogr. 60, 213±238.Thompson, D.B., Brown, J.H. and Spencer, W.D. (1991) Indirect facilitation of granivorous birds by desert

rodents: Experimental evidence from foraging patterns. Ecology 72, 852±863.Tilman, D. (1982) Resource Competition and Community Structure. Princeton University Press, Princeton, NJ.

Turkington, R. and Harper, J.L. (1979) The growth, distribution, and neighbor relationship of Trifoliumrepens in a permanent pasture: Fine-scale biotic di�erentiation. J. Ecol. 62, 245±254.

Valone, T.J. and Brown, J.S. (1989) Measuring patch assessment abilities of desert granivores. Ecology 70,

1800±1810.Vance, R.R. (1985) The stable coexistence of two competitors for one resource. Am. Nat. 126, 72±86.Werner, E.E. (1992) Individual behavior and higher-order species interactions. Am. Nat. 140, S5±S32.

770 Brown et al.

Ziv, Y., Abramsky, Z., Kotler, B.P. and Subach, A. (1993) Interference competition and temporal partitioningin two gerbil species. Oikos 66, 237±246.


Competition between birds and mammals: a comparison of giving-up densities between crested larks and gerbils

Documents