* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download Experimental Manipulation of a Desert Rodent Community: Food
Survey
Document related concepts
Biodiversity action plan wikipedia , lookup
Unified neutral theory of biodiversity wikipedia , lookup
Ecology of Banksia wikipedia , lookup
Introduced species wikipedia , lookup
Occupancy–abundance relationship wikipedia , lookup
Latitudinal gradients in species diversity wikipedia , lookup
Ecological fitting wikipedia , lookup
Reconciliation ecology wikipedia , lookup
Storage effect wikipedia , lookup
Molecular ecology wikipedia , lookup
Transcript
Experimental Manipulation of a Desert Rodent Community: Food Addition and Species Removal Author(s): James H. Brown and James C. Munger Source: Ecology, Vol. 66, No. 5 (Oct., 1985), pp. 1545-1563 Published by: Ecological Society of America Stable URL: http://www.jstor.org/stable/1938017 . Accessed: 22/06/2011 13:03 Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at . http://www.jstor.org/action/showPublisher?publisherCode=esa. . Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]. Ecological Society of America is collaborating with JSTOR to digitize, preserve and extend access to Ecology. http://www.jstor.org Ecology, 66(5), 1985, pp. 1545-1563 © 1985 by the Ecological Society of America EXPERIMENTAL MANIPULATION OF A DESERT RODENT COMMUNITY: FOOD ADDITION AND SPECIES REMOVAL1 JAMES H. BROWN AND JAMES C. MUNGER2 Department of Ecology and Evolutionary Biology, University of Arizona, Tucson, Arizona 85721 USA Abstract. Since 1977 we have been conductingexperimentsin which we add supplementalseeds or remove certaincombinationsof species of seed-eatingrodents and ants from 0.25-ha plots in the ChihuahuanDesert of southeasternArizona. These experimentsevaluate the extent to which food availability and interspecificcompetition influence rodent populations. Monitoringwith live traps revealedthat:(1) the addition of seed at the rate of 96 kg plot-' yr-' resultedin an increaseddensity of the largestgranivorousrodentspecies (Dipodomysspectabilis),decreasesin the densities of the two next-to-largestspecies (D. merriamiand D. ordii),and no detectablechangesin the densities of other rodents;(2) the removal of D. spectabilis,as well as other experimentallyinduced changes in the abundanceof this species, resultedin reciprocalshifts in the densities of the two congenericspecies, D. merriamiand D. ordii,and no significantchangesin densitiesof otherrodents;and (3) the removal of all three Dipodomysspeciesresultedin largeincreasesin densityof fourof the five speciesof smaller seed-eatingrodents,but had no effect on two species of insectivorousrodents.Taken together,these resultsindicatethat limited food resourcesand interspecificcompetitionplay majorroles in regulating the density of rodent populations and determiningthe organizationof desert rodent communities. However,the responsesof the rodent populationsto our manipulationswere unexpectedlycomplex; they includedlong time lags, asymmetricalinteractions,and little compensationin energyconsumption. This indicateshow much remainsto be learnedabout the processesthat determinethe structure and function of even this relativelysimple and well-studiedcommunity. Key words: community; competition; desert; Dipodomys; experiment; food limitation; granivory; population; rodent; seeds. INTRODUCTION A fundamental question of population and community ecology is what determines the absolute and relative abundances of species within local regions. Local population density is a consequence of the survival, reproduction, and dispersal of individuals. These processes are affected by the capacity to tolerate the physical environment, the availability of essential resources, and interactions with other individuals of the same and different species. One problem with assessing the roles of these environmental factors in limiting population density is that often they are not independent of each other. Thus, for example, both fluctuations in climate and the presence of competing species can affect the availability of resources, and both climate and parasite infections can change requirements for resources. In foraging to meet these varying resource requirements, individuals may differ in their risk of predation. The independent and interacting effects of such limiting factors can be studied by controlled field experiments in which some variables are held constant while others are manipulated systematically. We used such experiments to analyze the effects of two factors, food availability and the presence of other rodent species, on the population densities and community organization of desert rodents, and we discuss the possible I Manuscriptreceived 20 October 1983; revised 20 June 1984; accepted31 August 1984. 2 Present address:Departmentof Zoology, University of Wisconsin, Madison, Wisconsin 53706 USA. relationships between these factors and climatic variation and predation. Communities of North American desert rodents may be comprised of as many as 16 species, of which about half are usually primarily granivorous (Table 1). Most ecological studies of desert rodents have focused on these seed-eaters, because they provide an excellent system for assessing the effects of food limitation and interspecific competition on a guild of ecologically similar species (e.g., see Brown et al. 1979b and references therein, Munger and Brown 1981, Thompson 1982a, b, Munger et al. 1983, Price and Brown 1983, Kotler 1984, Brown 1984). All of these species feed primarily on the seeds of annual plants, a particulate resource that is produced in pulses following infrequent and unpredictable precipitation. Although seeds appear to be harvested (or at least to disappear from the soil surface) rapidly following flushes of seed production, they can persist for many months or years in the soil and in the stored caches of rodents. Potential competition for these food resources occurs not only among rodent species, some of which differ substantially in morphology, physiology, and behavior, but also between rodents and other kinds of desert granivores such as birds and insects (Brown and Davidson 1977, Brown et al. 1979b). One advantage of the desert rodent system is that it lends itself to experimental manipulations of such potentially important factors as availability of food (Abramsky 1978, Hay and Fuller 1981), predation risk (Thompson 1982b, Kotler 1984), habitat structure JAMESH. BROWN AND JAMESC. MUNGER 1546 TABLE 1. Ecology, Vol. 66, No. 5 Some characteristicsof the 16 species of rodentscapturedon our study site. Species Family Dipodomysspectabilis Dipodomysordii Dipodomysmerriami Perognathuspenicillatus Perognathusflavus Reithrodontomysmegalotis Reithrodontomys fulvescens Peromyscusmaniculatus Heteromyidae Heteromyidae Heteromyidae Heteromyidae Heteromyidae Cricetidae Cricetidae Cricetidae Peromyscuseremicus Cricetidae Onychomysleucogaster Onychomystorridus Neotoma albigula Sigmodonhispidus Spermophilusspilosoma Ammospermophilusharrisi Thomomysbottae Cricetidae Cricetidae Cricetidae Cricetidae Sciuridae Sciuridae Geomyidae Body mass Diet Seasonalactivity (g) active all year 125.0 granivorous 47.0 active all year granivorous active all year 40.8 granivorous hibernates 15.4 granivorous 7.2 hibernates granivorous active all year 10.4 granivorous active all year 13.2 granivorous 21.7 granivorousor active all year omnivorous 21.6 granivorousor aestivates? omnivorous insectivorous active all year 32.6 insectivorous active all year 24.9 folivorous active all year 173.5 folivorous active all year 109.5 folivorous hibernates 92.7 omnivorous active all year 107.6 herbivorous active all year 118.2 Total * Based on captureson the control plots with the gates closed. (Rosenzweig 1973, Price 1978, Thompson 1982b), and interspecific competition (Schroder and Rosenzweig 1975, Brown and Davidson 1977, Price 1978, Munger and Brown 1981, Abramsky and Sellah 1982). We tested the hypothesis that food is a limiting resource by adding seeds to experimental plots. We also tested the hypothesis that these species compete by experimentally removing certain combinations of species from other plots. We assessed the effects of these manipulations by comparing the population densities, energy consumption, and life histories of rodents on the experimental and control plots. METHODS Study site The study was conducted on the Cave Creek Bajada (a deposit of alluvial soil) 6.5 km east and 2 km north of Portal, Cochise County, Arizona, at an elevation of 1330 m. The soil at this site is a fairly homogeneous mixture of alluvial boulders mixed with and overlaid by finer particles. The terrain is relatively flat except where it is dissected by several temporary watercourses. The vegetation is primarily upper elevation Chihuahuan desert scrub, but the habitat structure varies from open grassy areas to stands of widely spaced shrubs (primarily Gutierrezia, Ephedra, Flourensia) to dense stands of arborescent Acacia and Prosopis along the usually dry watercourses. The entire study site, encompassing - 20 ha, has been enclosed since July 1977 with a barbed-wire fence to exclude domestic livestock. Experimental plots Experiments were conducted on 24 plots, each 0.25 ha in area (50 m on a side). These plots were placed Total Population number density* Biomass of cap(individtures uals/ha) (g/ha) 7.92 990.0 1207 0.64 30.1 231 19.84 809.5 1612 1.52 23.4 93 0.64 4.6 266 0.60 6.2 199 1 rare insignificant rare 57 insignificant 0.28 6.1 164 1.80 2.24 2.76 58.7 55.8 231 279 259 4 27 7 0 478.9 rare insignificant rare insignificant rare insignificant not unknown measured >38.24 >2463.3 4637 to include the flattest areas of most homogeneous vegetation but to leave at least 25 m between adjacent plots (Fig. 1). Twenty-three plots were established in July 1977. The final plot was added in July 1979. All plots were fenced with 6-mm wire mesh. Fencing 90 cm wide was buried 20 cm in the ground and bent outward 10 cm at the bottom to discourage rodents from digging under. The remaining 60 cm of wire was supported vertically by metal posts (13-mm steel reinforcing rod). A 15 cm wide vertical strip of aluminum flashing was riveted to the top of the fence to prevent rodents from climbing over. This fencing rendered all plots potentially rodent-proof, but 16 holes (gates) of varying sizes cut in the fences (4 gates equally spaced along each side of each plot) allowed access of selected rodent species to appropriate plots (see Experimental Manipulations, below). The fencing uniformly excluded jackrabbits (Lepus californicus) from all plots, but did not restrict access by cottontail rabbits (Sylvilagus auduboni), which regularly jumped over the fences. Experimental treatments were assigned to the plots at random, with the three exceptions mentioned below. There were two replicates of each of 12 treatments (Table 2): 4 kinds of seed addition, 3 kinds of rodent removal, 2 kinds of ant removal, 2 kinds of rodent and ant removal, and 1 control. A 3-mo pretreatment census period began in July 1977. During this time, large (3.7 x 5.7 cm) gates in the newly constructed fences allowed free movement of all rodents into and out of all plots. Experimental manipulations were initiated in September 1977. With the exception of the experimental manipulations, every effort was made to treat all plots identically, including subjecting them to the same intensity of rodent trapping and human traffic October 1985 MANIPULATIONOF A DESERTRODENT COMMUNITY 1547 (which was confined as much as possible to designated east-west pathways along the permanent grid stakes in each plot). There were three exceptions to this schedule and to random assignment of treatment to plots. One of the two plots (Plot 5) initially designated for removal of the ant species Pogonomyrmex rugosus was found to contain no colonies of this species, and in July 1979, Plot 12 (previously unassigned) was assigned to this treatment. On the same date, Plot 5 was reassigned for removal of the rodent species Dipodomys spectabilis (a new experimental manipulation) and another plot (Plot 24) was constructed to provide a second replicate of this treatment. Experimental manipulations Seed addition experiments. -Eight 0.25-ha plots were assigned to four seed addition treatments: (1) large size, constant rate; (2) small size, constant rate; (3) mixed sizes, constant rate; and (4) mixed sizes, pulsed. In all cases we added seed to each plot at a rate of 96 kg/yr. This should have approximately doubled the availability of seed biomass, because our estimates of native seed production are 400 kg ha-' *yr-'. From September 1977 to August 1980 the supplemental seed was milo (Sorghum vulgare Pers.), but when we found that granivores, especially ants, do not avidly take this species, we switched in September 1980 to millet (Panicum miliaceum L.). "Large" seeds were added whole (Xmass = 6 mg for millet), whereas "small" seeds were cracked (X mass _ 1 mg). A "mixed" addition was simply an equal mixture of whole and cracked seeds. In the "constant rate" additions, 8 kg of seed were supplied once per month, whereas in the "pulsed" treatment the entire yearly allotment of 96 kg was added in 2-6 equal installments over the 2-mo period (September-October) corresponding to peak seed production by the summer annuals. Seeds were scattered by hand over the entire area of each plot. Dipodomys removal experiment.--Four plots were assigned for removal of all Dipodomys species (D. spec- FIG. 1. Aerial photographof the study site, showing the natureof the vegetationand the locations of the 24 experimental plots, each 50 m on a side (0.25 ha in area). The experimentaltreatmentin each of the numberedplots can be determinedby referringto Table 2. tabilis, D. merriami, and D. ordii; see Table 2). In September 1977 the gates in the fences were reduced in size to prevent passage by Dipodomys, and all individuals of this genus were removed as they were captured in monthly censuses. The reduced gate size (1.9 x 1.9 cm) effectively prevented colonization by Dipodomys but allowed the free passage of five species of small granivorous rodents (Perognathus [hereafter abbreviated Pg.] penicillatus, Pg. flavus, Peromyscus eremicus, P. maniculatus, and Reithrodontomys meg- Outline of the 12 experimentaltreatments,including control. Note that some of the rodent and ant removal experimentshave a factorialdesign, and duplicatetreatmentsare listed underboth headings. TABLE 2. Control Treatment Plots Unmanipu- 11, 14 lated Seed addition Treatment Plots constant seeds, 6, 13 Large rate Small seeds, constant 2, 22 rate Rodent removal Treatment Plots 5, 24 Dipodomysspectabilis Mixed sizes, constant rate All Dipodomysspecies 3, 19 and Pogonomyrmex rugosus All seed-eating 7, 16 rodents All seed-eating 10, 23 rodentsand ants 9, 20 Mixed sizes, temporal 1, 18 pulse All Dipodomysspecies 15, 21 Ant removal Treatment Plots Pogonomyrmexrugosus 8, 12 Pogonomyrmexrugosus 3, 19 and all Dipodomys species All seed-eatingants 4, 17 All seed-eatingants and rodents 10, 23 1548 JAMESH. BROWN AND JAMESC. MUNGER alotis), as well as two insectivorous species (Onychomys leucogaster and 0. torridus). We placed wire-mesh baffles (15 x 15 cm) perpendicular to the fences on either side at the locations of the small gates to enable the rodents to find the gates easily and to move freely both into and out of the plots (well-worn runways attested to the fact that rodents had no difficulty in finding the much larger gates on nonremoval plots, and therefore baffles were not used on these plots). Two of the Dipodomys-removal plots, as well as two others (see Table 2), were assigned for the removal of the large harvester ant, Pogonomyrmex (hereafter abbreviated Po.) rugosus, as part of a 2 x 2 factorial experiment. Consequently, results on four experimental plots could be compared to those on four "control" plots to analyze for the effects of Dipodomys removal with the effect of ant removal held constant. D. spectabilis removal experiment. -Similar procedures were used to exclude only D. spectabilis from two plots. Gates and baffles were placed as described above, except that the gates were large enough (2.6 x 3.0 cm) to allow the free passage of D. merriami and D. ordii as well as the small rodent species. Removal of D. spectabilis began in July 1979 on Plot 5 and in February 1980 on Plot 24. Ant removal experiments. -Four plots were assigned for removal of all granivorous ant species and four were designated for selective removal of Po. rugosus. Removal and continued exclusion of ants was accomplished by the use of poisoned bait (Myrex through 1980 and AMDRO thereafter), which was applied beginning in September 1977. Studies of toxicity to mammals and our experimental results suggest that neither poison had any significant effect on the rodent populations. On plots from which all seed-eating ants were to be removed, bait was both broadcast and applied to individual colonies. On Po. rugosus removal plots, bait was applied only to colonies of this species. Colonies of Po. rugosus and Po. desertorum that persisted after poisoning were asphyxiated by first pouring 250 mL of gasoline down the entrance hole, then sealing the entrance. (As described above, there was a change in the assignment of one Po. rugosus removal plot.) On two of the ant removal plots, rodent removals were also carried out. Measurement of responses We used a monthly trapping regime to assess changes in the densities and total biomass of rodent species and in the life history attributes of individual rodents in response to our experimental manipulations. Trapping was conducted on all 24 experimental plots during a period of two or three successive nights at approximately monthly intervals, at a time corresponding as much as possible to the time of the new moon. During each monthly trapping period, each plot was trapped for a single night with 49 Sherman live traps (23 x 8 x 9 cm) baited with mixed birdseed or millet and Ecology, Vol. 66, No. 5 placed at permanent grid stakes spaced at 6.25-m intervals. The standard procedure was to close the gates through the fences on the night of trapping to insure that only individuals residing on the plots were captured. However, during three of the monthly trapping periods (March 1980, April 1981, and October 1982) the plots were trapped with the gates open in order to capture both residents and those individuals that regularly foraged on the plots but had their home burrows outside. All rodents captured were individually marked for subsequent identification. The three smallest species (Pg. penicillatus, Pg. flavus, and R. megalotis) were marked by toe clipping; all other species were marked with monel fingerling tags attached to the ears. Whenever an individual was captured it was weighed and checked to determine reproductive condition. The latter was scored by classifying males as either having fully scrotal testes, showing evidence of recently descended testes (distended scrotum but abdominal testes), or being reproductively inactive. Females were scored as being either in estrus (swollen vaginal orifice, with or without vaginal plug), pregnant (embryos palpable), lactating (nipples reddish and enlarged), or reproductively inactive. From these data on captured individuals, five life history attributes were assessed for each species on each plot: average adult body mass, percent of population exhibiting reproductive activity, and two measures of residence status. Average adult body mass of a species was calculated as the average mass of individuals that exceeded a threshold mass for their species and sex; the threshold used was the lowest mass at which the majority of individuals of a given species and sex showed reproductive activity. Rate of energy consumption was calculated using the number of individuals, the average individual body mass (M, in kilograms), and the allometric equation E = 753M0-67, where E is the individual rate of energy consumption (in kilojoules per day) (King 1974). Percent reproductive was defined, for males, as the percentage of individuals above the adult-mass threshold with scrotal or recently descended testes, and, for females, the percentage of adult individuals either in estrus, pregnant, or lactating. The first measure of residence status was average residence time of individuals on a plot, which was defined as the average time between the first and last captures of each individual of a given species captured on that plot. This measure provides an estimate of minimum survival time of an individual on a permanent home range. The second measure, percent residents on a plot, was calculated as the number of individuals that were captured more than once on that plot divided by the total number of individuals of that species captured on the plot. The experimental manipulations and trapping program are ongoing. For purposes of the present paper, however, we shall in general limit our consideration to MANIPULATIONOF A DESERTRODENT COMMUNITY October 1985 1549 = Small Seedeaters = Dipodomys l. merriami + D. ordii etbi = D. spectabilis 0 'I, 0 0. CL V) 0 I0. V) L z C NON-SEEDEATERS NNSEE3.E0NNEEE z Onychomys spp. R= =Neotoma albigula LL 2u 2.0- S- YEAR TREATMENT C I§ 1 3 4 5 Control 1 2 41 Small Seed Constant 4 5 Large Seed Constant 1 l21 3 14 Mixed Seed Constant 5 Mixed Seed Pulse FIG.2. Effectsof the various seed addition treatmentson the mean population densities of seed-eatingand non-seedeating rodents. Data are shown by treatment(four seed addition manipulationsplus control), two plots per treatment,and by year (five years;year 1 = 1977-1978). Note that as a group, seed-eatingrodents did not change in density over the 5 yr in responseto supplementalseed, althoughD. spectabilisincreasedand D. merriamiand D. ordii decreasedon most experimental plots but not on the controls. data for the 5-yr period from June 1977 to June 1982. In analyzing the effects of trapping with the gates open, however, we also consider captures for the 3-mo period September-November 1982. Statistical analyses Despite the large spatial scale and long temporal duration of these experiments, replication of the particular treatments was limited. This created problems for statistical analysis. Nevertheless, we have tried to be statistically rigorous in our treatment of the data. Because the number of replicates for each test was small we have used nonparametric analysis (specifically, the Mann-Whitney Utest; Siegel 1956) wherever possible. In those cases where the number of replicates was so small as to preclude any possibility of significance with nonparametric tests, we have used parametric procedures (i.e., analysis of variance). Because a U test typ- ically has less statistical power than an analysis of variance, our reliance on the former test could have led to an increased frequency of Type II errors. This does not appear to have occurred, however. For each of the 56 U tests presented in the Appendix, we also tested the same data by an analysis of variance. For only two of the tests was there a significant difference; in both instances the U test gave the smaller probability value (see Appendix). To analyze for the effect of experimental treatment on the four life history parameters, we compared results for experimental plots (where variables had been manipulated) with those for control plots (where the same variables had been held constant). To determine the effect on densities and biomass, we used the magnitude of the change on each plot from the period before manipulation to the period after manipulation. This helped to control for any chance differences in Ecology, Vol. 66, No. 5 JAMES H. BROWN AND JAMES C. MUNGER 1550 600~7 500- 400- 300- "I ^1 11-1 "I^ 11- ." - ~ . 5~ 1-,. 7Sc=5-~ 200- O 100- I I YEAR "I >0 TREATMENT ci zz 0 NON-SEEDEATERS (_ = Onychomys LJ z LJ spp. = Neotoma albigula Fl TREATMENT Control Small Seed Constant uiL LI.LLihi. Large Seed Constantl Mixed Seed Constant Mixed Seed Pulse FIG. 3. Effects of the various seed addition treatments on the mean energy consumption of seed-eating and non-seedeating rodents. Data are plotted by treatment and year as in Fig. 2. Note that provision of supplemental seed resulted in little if any increase in total seed-eating rodent biomass, although D. spectabilis increased at the expense of D. merriami and D. ordii. relevant variables between plots; such differences can lead to misleading results when comparisons are made among a small number of randomly assigned and somewhat heterogeneous (in habitat and rodent population, in this case) plots. For these comparisons, the results included the entire 1st yr before-manipulation of data (three pretreatment sampling periods and eight posttreatment samples; July 1977 to June 1978) because there was no indication of any significant response to the manipulations during this period and because this larger data base permitted a more precise characterization of the initial state of each plot than if only the three pretreatment samples had been used. By attempting to be rigorous and conservative in our statistical analyses, we ran the risk of failure to detect significant results that would have been revealed if replication had been greater or if less conservative tests had been employed. For this reason we emphasize that the results we report in the next section may not represent the only responses of rodents to our experimental manipulations, but they are certainly the most dramatic responses and the only ones we were able to document with statistical confidence. RESULTS Over the entire period, July 1977 to June 1982, we recorded 4637 captures of 15 rodent species in a total of 68 159 trap-nights (Table 1). The low frequency of captures insured that sufficient empty traps were alshould not ways available so that trap competition have influenced the results. In performing the following analyses we frequently combined rodent species into functional groups whose member species were likely to respond similarly to our manipulations. Seed-eating rodents included: (1) the kangaroo rat, Dipodomys spectabilis, the largest granivorous species; (2) D. mer- 1551 MANIPULATIONOF A DESERTRODENT COMMUNITY October 1985 a) Seed addition b) Control, ant removal, D. spectabilis removal plots 6- plots KEY 1981-1982 plot no. 5C) 1977-1978 c 4- ' 3-5 2- -- 0 1c QL 0- I I I I I 0 1 2 3 4 I I I I Density of D. spectabilis (no./plot) I 0 1 2 3 Density of D. spectabil/s (no./plot) FIG.4. Changesin the mean densitiesof D. merriamiand D. ordiirelativeto changesin the mean densityof D. spectabilis in responseto various experimentalmanipulations.Each line representsthe changeson one plot between the 1st yr of the study(1977-1978) and the last 2 yr (1980-1982). Treatmentsand plot numbersareexplainedin Table 2. Note the pronounced tendency for reciprocaldensity shifts, with D. spectabilisincreasingat the expense of D. merriamiand D. ordii on seed addition plots and the reversetrend on other plots, includingcontrol and D. spectabilisremoval plots. riami and D. ordii, medium-sized kangaroo rats; and (3) five species of small granivores, Perognathus penicillatus, Pg. flavus, Peromyscus maniculatus, P. eremicus, and Reithrodontomys megalotis. Non-seed-eating rodents included: (1) Onychomys leucogaster and 0. torridus, small insectivorous forms; and (2) Neotoma albigula, a large folivore. Effects of seed addition The primary responses to the availability of supplemental seeds were: (1) a substantial increase in the density and energy consumption of the largest granivorous rodent species (D. spectabilis); (2) a significant decrease in the total energy consumption of all other seed-eating species; the latter effect can be attributed to (3) a large decrease in the combined density and energy consumption of the two medium-sized granivorous species (D. merriami and D. ordii) (Figs. 2 and 3, Appendix). These responses appeared to occur on all of the seed addition treatments. The size of seed particles provided and the temporal schedule of seed addition had little apparent effect on the rodents. The only exception was an apparently longer residence time and a higher percentage of adult individuals in reproductive condition for D. spectabilis on the plots where seeds were added in a seasonal pulse rather than evenly throughout the year. None of the smaller rodent species other than Dipodomys showed any significant response to the seed addition treatments (Figs. 2 and 3). On the eight plots to which seeds were added, D. spectabilis increased (between year 1 and years 2-5) in mean density from 0.97 to 1.61 individuals per plot per period (a 66% increase) and in energy consumption from 170 to 288 kJ plot- 'd-1 (69%) (Fig. 4). During this same period D. spectabilis on control plots decreased in density from 1.33 to 0.98 individuals per plot per period (26% decrease) and in energy consumption from 242 to 165 kJ plot- d- (32%). Effects of seed addition were significant for both measures (Mann-Whitney Utests: U = 1, P = .044 for both variables). The corresponding decreases in both density and energy consumption of D. merriami and D. ordii, in response to the addition of seeds, were even more significant (U = 0, P = .022 for both variables): density decreased from 2.35 to 1.37 individuals per plot per period (58%) and energy consumption decreased from 211 to 126 kJ plot- -d-l (60%) on plots to which seeds were added, while on control plots density increased from 2.54 to 2.65 individuals per plot per period (4%) and energy consumption increased from 228 to 244 kJ plot- d-l (7%). We analyzed attributes of individual kangaroo rats that may have caused or responded to these changes in population density with increased food availability. 1552 JAMES H. BROWN AND JAMES C. MUNGER Dipodomys spectobilis a D. ordii merriami Dipodomys 1400 Ecology, Vol. 66, No. 5 1400*2 *13 1200 1200*13 *~ *2 1000- 1000 *22 04 014 5 A5 20 20@ I 800 800 012 A24 *1 *6 - *9 225 20 08 *6 *18 0 17 011 Oil 600 - 600 - 017 / 08 / / *18 012 / / / Q. (O 400- / 04 / / 400- / / Lu I- 014 / / Oil / / / / 200- 200- / 13 / / / z 0 / 0- 0 I 200 I 0 CL U) Z 0 400 / 0 600 Small Seedeaters I 400 200 I 600 Onychomys (-9 w zZ 250 - 250- LU *13 A24 014 200- 200- 011 / *6 / *9 / 0 19 / 150- 150- / 03 019 *18 *2 / / 0 15 / 100- 100- / A5 / / d21 6 5020 / 08 / A 24 017 o4 / / /0|7 * I e1 20. 50- 015 210 @22 o12 / / / 5 · 228 0 - / ·i 0 04 / *18 13 014 0- , , 50 / 03 , , 100 150 200 0 ENERGY CONSUMPTION (kJ/d)-- 50 100 GATES CLOSED 1I 1 150 200 October 1985 MANIPULATIONOF A DESERTRODENT COMMUNITY Although none of the species exhibited a detectable increase in reproductive success in response to additional seeds, mean body mass of D. spectabilis was greater on seed addition plots than on controls (U = 0, P = .022; Appendix), a result that could be attributed either to increased growth of D. spectabilis or to movement of large individuals onto enriched plots. Conversely, D. merriami had a significantly greater proportion of individuals captured only once (not recaptured) on the seed addition plots than on the controls (U = 0, P = .022). This apparent reduced ability of D. merriami to maintain permanent home ranges in the face of increased densities of D. spectabilis implies competition between these kangaroo rat species. Additional insights into this interaction are afforded by the results of trapping with the gates in the fences open. Because individuals with home burrows off the plots moved onto the plots to forage, we found that more individuals of all Dipodomys species were captured when the gates were left open than in the preceding and following trapping periods with the gates closed (Fig. 5). However, the increase was to 2.09 times as many D. spectabilis and 1.85 times as many D. merriami and D. ordii on the seed addition plots as on the controls (U = 0, P = .022 for both groups). The net result was that when the gates were open, 3.36 times as many individuals of D. spectabilis foraged on seed addition as on control plots (U = 0, P = .022), whereas D. merriami and D. ordii activity on the seed addition plots increased only to the approximate level on control plots (actually it was 1.21 times the control level, but U = 5, P = .267). Clearly individuals of all the kangaroo rat species were attracted onto the plots that received the supplemental seeds, but only D. spectabilis was able to increase the density of permanent residents; the increase in D. spectabilis apparently inhibited the establishment of its smaller congeners. Effects ofD. spectabilis When preliminary results of the seed addition experiments suggested that D. spectabilis increased at the expense of D. merriami and D. ordii, we initiated an experiment in which D. spectabilis was removed from two plots. Such a manipulation was expected to yield interesting results for two reasons. On the one hand, in terms of total food consumption D. spectabilis is one of the two most important rodent species in the community; it accounted for 53.0% of energy con- 1553 sumption by granivorous rodents on the control plots. Thus, removal of D. spectabilis should have made available large quantities of seeds for consumption by other seedeaters. We had already observed (Munger and Brown 1981) and now confirmed (see Effects of Dipodomys Species, below) that small granivorous rodents increased several-fold in density following the removal of all three Dipodomys species, and much of this response could potentially be attributed to removal of D. spectabilis, because of its large energy consumption. On the other hand, D. spectabilis is much more similar in morphology (including body size), physiology, and behavior to its congeners, D. merriami and D. ordii, than to any other seed-eating rodents on the study site. On this basis, D. spectabilis might be expected to have a larger competitive effect on the other Dipodomys species than on the granivorous rodents in other genera. The removal of D. spectabilis from two plots was begun z 2 yr after the other experiments had been initiated, and no effects can yet be demonstrated with statistical confidence because of the small number of replicates per treatment. Clearly there was no noticeable effect during the 1st yr (1979-1980). During the following 2 yr (1980-1982), combined average densities of D. merriami and D. ordii increased by a factor of 2.89 to 3.14 individuals per plot where D. spectabilis had been removed. However, they also increased by a factor of 1.62 to 2.66 individuals per plot on the control plots. Removal of D. spectabilis had no significant effect on any of the granivorous rodent species in other genera. Much more statistical power can be used to evaluate the effects of D. spectabilis on other rodents by considering the data for other plots. Densities of D. spectabilis changed substantially in response to other manipulations, and we examined the correlation between these changes and the response of the populations of other rodent species. Such an analysis provides strong evidence for a significant competitive interaction between D. spectabilis and the other two kangaroo rat species. Fig. 4 shows that in general D. spectabilis increased and D. merriami and D. ordii decreased when seeds were added, whereas the opposite trend occurred on the other plots where all three Dipodomys species were present and also on the two plots where D. spectabilis was removed. These overall reciprocal density shifts are highly significant (Fisher's exact test, P = FIG.5. Resultsof trappingin the indicatedplots with the gates in the fences open, comparedwith resultsof trappingwith the gates closed in the previous and followingtrappingperiods. O control and ant removal plots, 0 seed addition plots, A Dipodomysspectabilisremoval plots, and D Dipodomysspp. removal plots. Treatmentsand plot numbersare explained in Table 2. Note that all of these functionalgroupsexcept the small seedeatershad highercapturesin all cases with gates open than with gates closed (points above dashed line indicatingequal energyuse). Also, the seed-eatingDipodomysspecies, but not the insectivorousOnychomysspecies, showedgreaterincreaseson plots where seeds had been added than on controlsor where ants had been removed. JAMESH. BROWN AND JAMESC. MUNGER 1554 Dip o do mys 7- Ecology, Vol. 66, No. 5 spp. 6\ /k 4- 0 0 I \\ a 5! I I ^ I,e,,-e . I'" ^^^-^ \ .A *^/ 3- v / · - '-. ' f .'~ · , \ I4 I x 2- LLI a. _/ - < / z - - \ r ^^ ^? I- C9 < - ' "^ // ^ / \ - o - t* ^ 0- a. a. H Co 7- I- 6- Small Seedeaters Dipodomys 0--0 411 -- -- Control plots removal plots 5a. 4- ILii 3- Q. 2c, I- 0 z>Lii ULJ 0- z0 PI~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ I \ A, 3 -Y I I I II I 2 Onychomys 3- ~~~~Ii, V ' I I I I I -· 1 I I I I I or I I I m-A I 0~ J JA 1 I o- , -* y' I I I I I I -0 I I 'I I I I I I i/ I I I I I I I :\ Pt~~~~~~~~~~~~~~~~~~~~~~~· / \ i 1-1-1 SO'N'DIJ'F'M'AMJ 1977 I I I I I spp. 2- I- ·0·" I 1 -1 JA'SON 1978 DIJFMAMJ -1 -1 JASONDIJFMAMJ'JA'SONDIJ'F'M'A'M'J'J'A'S'O'N'DIJ'F'M'A'M'J 1979 1980 1981 1982 FIG.6. Changesover time in the densities of each of the three functionalgroupsof rodents on four experimentalplots from which all three Dipodomys species were removed beginning in October 1977, compared to the densities on four control plots. Each point represents the mean of four plots. .0055). Although interpretation of these results is complicated by the fact that other variables besides the density of D. spectabilis differed among plots, the results are consistent in suggesting that the various manipulations affected the densities of D. spectabilis, which in turn had direct negative effects on populations of D. merriami and D. ordii. Although our results do not show that D. spectabilis by itself has significant effects on the small granivorous rodents in other genera, all of the trends suggest that there may be some competitive interaction. Densities of small granivores increased by a factor of 4.97 to 0.83 individuals per plot on the two plots from which D. spectabilis had been removed, compared to an increase of 2.05 times to 0.57 individuals per plot on control plots with D. spectabilis present. An analysis that included these four plots and all other plots with D. spectabilis present revealed a weak trend toward reciprocal density shifts (Fisher's exact test, P = .16). Interestingly, both of these trends run counter to what would be predicted if D. spectabilis had an indirect positive effect (through its negative effect on the medium-sized kangaroo rat species) on the small granivores. Clearly, more data are required to determine whether D. spectabilis has significant interaction with the granivorous rodents other than its congeners. Effects of Dipodomys species We removed all three species of Dipodomys from four plots and monitored the populations of the small rodents on these plots and on four control plots to test for competition among rodent species. As expected under the hypothesis that competition for food is important in this community, small seed-eating rodents, but not small nongranivorous species, increased in density in response to this manipulation. This result was well established within 3 yr and was reported briefly elsewhere (Munger and Brown 1981). Here we include two additional years of data and present a more detailed analysis. The results continue to support the competition hypothesis. During the 3-mo pretreatment period and for an additional 8-9 mo after Dipodomys was removed, there were no significant differences in the densities of small rodents between the removal and control plots; in fact, densities of small seed-eating rodents averaged 2.17 times as high on the control plots (Fig. 6). Beginning in mid-1978, however, numbers of small granivorous rodents increased on Dipodomys removal plots, which from then on consistently maintained densities that averaged 2.8-4.5 times as great as those on control plots. During this period there was only one experimental plot in only one year (Plot 21 in 1980-1981) October 1985 MANIPULATIONOF A DESERTRODENT COMMUNITY that supported a lower density of small seed-eaters than any control plot. For all of the small granivorous species considered collectively, these results are highly significant (U = 0, P = .014). As shown in Table 3, none of the small rodents responded to the treatments during the 1st yr, but then four of the five small granivores (but neither of the nongranivores) showed significant increases in the absence of Dipodomys: Pg. flavus, R. megalotis, P. maniculatus, and P. eremicus. Only Pg. penicillatus showed no significant response, but this species occurred at low densities and exhibited a very patchy distribution among plots. We note parenthetically that P. eremicus, an excellent climber that sometimes scaled the "rodent-proof' fences, was also captured 2.7 times more frequently on the four plots from which all rodents had been removed than on the four control plots (U = 2.5, P = .08). The dramatic and consistent increase of small seed-eating rodents in the absence of Dipodomys supports the competition hypothesis. Additional evidence in support of this hypothesis and of the proposition that the competition is primarily for food came from monitoring densities of small nongranivorous rodents. Two species of primarily insectivorous rodents, Onychomys leucogaster and 0. torridus, passed freely through the small holes in the fences of the Dipodomys removal plots. Jointly the densities of these insectivorous species were not significantly different on experimental and control plots (densities were 26% greater on control plots, but U = 5, P = .43), and neither species showed any difference when considered individually (Table 3). Taken together, these data suggest that rodent species that eat significant numbers of seeds increase substantially in density when granivorous Dipodomys species are removed, but insectivorous rodents are unaffected. These results not only reinforce the conclusion that granivorous rodents compete for a limited food supply, they also help rule out alternative explanations, such as that the increase in small rodents was attributable to reduced losses to predators (e.g., snakes) that were differentially excluded from experimental and control plots by the differentsized holes in the fences. The data on individual life history attributes showed no significant differences between Dipodomys removal and control plots that might indicate the mechanism of the response of small seed-eating rodents to missing competitors. We emphasize, however, that the number of replicates was small and that there was considerable variation among experimental plots. We suspect that the increased density of small granivorous rodents on Dipodomys removal plots can be attributed largely to immigration. Effects of ants In contrast to an earlier study at another site where seed-eating rodents were shown to increase in density when granivorous ants were removed (Brown and Da- 1555 vidson 1977, Brown et al. 1979b), we noted virtually no significant response of rodents to removal of either Pogonomyrmex rugosus alone or all species of granivorous ants. The only exception was a significantly greater increase (F = 8.25, P < .05) in the combined densities of the small seed-eating rodent species on plots where both Dipodomys and Po. rugosus had been removed than on plots where only Dipodomys was absent; this result was obtained only for the 2nd yr of the study (1978-1979), which was also the 1st yr of the response of the small granivorous rodents (Fig. 6). Although we do not want to attach too much importance to this result, it is possible that the small granivorous rodents initially responded to a greater availability of seeds on plots where both Dipodomys and Po. rugosus had been eliminated, but, because the other ant species subsequently increased to compensate for the missing Po. rugosus, this response was ephemeral. Compensation in energy consumption Although experimental provision of additional seeds and removal of certain rodent species resulted in increased abundance of certain rodent species, it is of interest to ask to what extent these increases compensated for the changes in food resource availability. The degree of compensation was estimated from data on the usable energy content of supplemental seeds (11 637 kJ/kg for millet; from values in Altman and Ditman 1968) and from allometric equations relating the average daily energy requirements of a small mammal to its body mass (see Methods: Measurement of Responses). Results of these calculations are shown in Table 4. Compensation was surprisingly slight: virtually none for the added seed, only - 10% for removal of the three Dipodomys species, and perhaps as much as 34% for the removal of D. spectabilis. We emphasize, however, that statistical confidence cannot be placed in the last figure, because the response of the remaining rodents (primarily the increase of D. merriami and D. ordii) to removal of D. spectabilis was not statistically significant. Therefore we conclude that although adding seed or removing competing species made additional food resources available to the remaining rodent species, only a small fraction of these additional energy resources were actually converted into rodent biomass. Although certain species clearly profited from enhanced food availability and increased in density, these changes were inadequate to account for most of the millet seeds that we provided or the natural seeds that were made available by removing competing rodent species. DISCUSSION Our experiments provide three lines of evidence that interspecific interactions are important in determining desert rodent densities and community organization. 1) Small seed-eating rodents increased in density when three species of kangaroo rat were removed. JAMESH. BROWN AND JAMESC. MUNGER 1556 Ecology, Vol. 66, No. 5 3. Density response of small seed-eating and non-seed-eatingrodents in a 2 x 2 complete factorial experiment involving removal of the three large granivorousDipodomysspecies, as well as the large harvesterant Pogonomyrmex rugosusduring 1977-1982. TABLE Plots with Dipodomysremoved Po. rugosuspresent Plot 3 Year - 1 Po. rugosusabsent Plot 19 1 2-5 2-5 Plot 21 Plot 15 1 2-5 1 2-5 Rodent density (mean no. capturesper plot per census period) Small seedeaters Pg. flavus Pg. penicillatus P. maniculatus P. eremicus R. megalotis Total 0.25 0.08 0 0 0 0.33 0.78 0.04 0.02 0.13 0.57 1.54 0 0 0 0.08 0.08 0.16 0.27 0 0.40 0.32 0.67 1.66 0 0 0 0 0 0 0.79 0.37 0.19 0.33 0.20 1.88 0.17 0 0 0 0 0.17 0.80 0.07 0.07 0.04 0.29 1.27 0.50 0.08 0.58 0.24 0.20 0.44 0 0 0 0.11 0.02 0.13 0 0 0 0.42 0.09 0.51 0.92 0.33 1.25 0.32 0.11 0.43 Small non-seedeaters 0. torridus 0. leucogaster Total * Control. t Resultsof Mann-WhitneyU tests comparingthe fourDipodomys-removalplots and the fourplots with Dipodomyspresent, with respectto averagedensity for year 1 (July 1977 throughJune 1978), averagedensity for years 2-5 (July 1978 through June 1982), or the changein averagedensity betweenyear 1 and years 2-5. NS: not significant(P > .05). 2) The two medium-sized species of kangaroo rat showed reciprocal density shifts to changes (both induced and natural) in density of the larger kangaroo rat, D. spectabilis. 3) Use of plots by nonresidents of the two mediumsized species of kangaroo rat (as measured by increased trapping success when trapping with the gates open) was higher on plots where densities of D. spectabilis were high than on plots where densities were low, which indicates that although the smaller congeners could trespass and do some foraging on D. spectabilis home ranges, they were strongly inhibited from establishing residence there. In addition, our experiments, as well as other observations, indicate that these interactions are due, at least in part, to food limitation and competition for this scarce resource. This evidence includes: 1) densities of species of small rodents that are not highly granivorous did not respond to the removal of Dipodomys species; 2) the direction of the reciprocal density shifts between D. spectabilis and the two medium-sized Dipodomys species was determined primarily by seed availability: D. spectabilis increased on seed addition plots, D. merriami and D. ordii increased on all other plots; 3) use of plots by nonresidents was much higher on seed addition plots than on other plots; and 4) densities of large-seeded winter annual plants have increased as much as several thousand times in response to the removal of rodents, indicating that predation by these granivores substantially reduces the standing crop of their food resources (Brown et al., in press). We measured several life history parameters in an attempt to discern the demographic mechanism that was responsible for the density changes caused by our manipulations. D. merriami had a lower percentage of residents on seed addition plots, corresponding to its lower densities there. This may indicate that a reduction in successful immigration onto seed addition plots or an increase in emigration from these plots may have led to decreased densities there. D. spectabilis had a higher average adult body mass on seed addition plots; the demographic consequence of such a result is unknown. Our failure to find further differences in reproduction or residency that could account for the other changes in density (such as the increased populations of small seed-eaters in response to removal of Dipodomys) indicated that identifying the mechanisms that lead to a change in abundance is likely to be more difficult than simply documenting the shifts in density. This may be because very small adjustments in rates of reproduction, survival, immigration, and emigration can lead to rather large differences in standing densities. Although we have demonstrated food limitation and interspecific competition, we still know very little about how these processes operate to influence the abundance of species and the organization of the community. For example, we do not know to what extent species compete by interference and exploitation, exactly how resources are used differentially among species so that stable coexistence is maintained, or what roles antipredator strategies, life history traits, foraging behaviors, seed storage, and hibernation play in the interactions among species. It is important to work out these mechanisms, because in order to interpret our results MANIPULATIONOF A DESERTRODENT COMMUNITY October 1985 TABLE3. 1557 Continued. Plots with Dipodomyspresent Po. rugosus absent Po. rugosus present* Plot 11 1 Plot 14 2-5 1 Plot 12 1 2-5 Plot 8 1 2-5 Significancet 2-5 1 2-5 Change 0.014 Rodent density (mean no. capturesper plot per census period 0.08 0.13 0.08 0.06 0.08 0 0.42 0.26 NS 0.014 0.33 0.39 0 0.02 0.25 0.21 0 0.02 NS NS NS 0 0 0 0.41 0 0.02 0.14 0.68 0 0 0 0.08 0 0.06 0.06 0.20 0 0 0 0.33 0.02 0.08 0.24 0.55 0 0 0 0.42 0 0.02 0 0.30 NS NS NS NS 0.022 0.057 0.029 0.014 0.022 0.057 0.029 0.014 0.08 0 0.08 0.45 0.22 0.67 0.08 0.08 0.06 0.13 0.32 0.45 0 0 0 0.24 0.05 0.29 0.25 0 0.25 0.07 0.42 0.49 NS NS NS NS NS NS NS NS NS in a general conceptual framework it is necessary to know to what extent the observed processes of population regulation, interspecific interaction, and community organization depend on specific features of the biology of these particular rodent species, and to what extent these patterns and processes occur in other systems. Certainly interspecific competition appears to be an important phenomenon in other kinds of rodents (Grant 1969, 1971, 1972, Redfield et al. 1977) and other groups of terrestrial vertebrates (e.g., Hairston 1980, 1981, Pacala and Roughgarden 1982). The responses of rodents to our manipulations, additional responses of ants, plants, birds, and other organisms to these and similar experiments in desert ecosystems (e.g., Brown and Davidson 1977, Brown et al. 1979a, b, Inouye et al. 1980, Inouye 1981), and the results of other experimental manipulations of communities (e.g., in the intertidal zone: Connell 1961, Dayton 1971, Paine 1974, Lubchenco and Menge 1978, Sousa 1979, Menge and Lubchenco 1981) suggest that natural communities are organized by a complex web of interspecific interactions that cannot adequately be described or predicted by current ecological theory. The problems of interpreting the results of field experiments with respect to the specific predictions of mathematical models are beyond the scope of this paper, but see Schaffer (1981) and Bender et al. (1984). Here we simply note that although existing models of population regulation and interaction provide a basis for interpreting the simple qualitative trends that are caused by our manipulations, these models do not prepare us for the surprising complexity of several responses (including small degrees of compensation, asymmetries, and long time lags) revealed by the rodent populations. First, consider the extent of compensation. Although TABLE 4. Energeticcompensationby desert rodents to supplementalseeds and to removal of selected rodent species. (For details of calculation,see Methods:Measurementof Responses.)Note that rodent compensationfor the additional food energymade availablenever exceeded 34%. Experimentaltreatment Removal of all 3 Energymade available (kJ/d)* Energeticresponse (kJ/d)t Seed addition Removal of D. spectabilis Dipodomys species 3060 201.4 439.7 91 33.2 49.4 (by all 8 species of (by 5 species of small (by 7 species of smaller granivorousrodents) granivorousrodents) granivorousrodents) Percentcompensation: 33.8 9.5 2.9 * For seed additiontreatments,calculatedas the metabolizableenergyin the addedmillet. For othertreatments,determined by calculatinghow much the removed individualswould have consumed, based on their total biomass. t Determinedby calculatingthe consumptionof the indicatedgroups,based on their measuredincreasein total biomass. t For the rodent removal experiments,percentcompensationwas calculatedas: [(1977-1978 energy consumption minus 1978-1982 energyconsumptionfor the averageof the removal plots) minus (1977-1978 energyconsumption minus 19781982 energyconsumptionfor the averageof the control plots)] divided by (metabolizableenergy of added seeds or energy consumptionof the rodents removed). 1558 JAMESH. BROWN AND JAMESC. MUNGER manipulations that supplied additional food or altered the densities of particular rodent species resulted in substantial changes in densities of certain species, the magnitudes of the compensatory responses in terms of energy consumption were very small. This suggests that processes other than straightforward competition for food among rodent species must also play important roles in determining the structure and function of this community. Possibilities include: (a) intraspecific competition and aggression; (b) competition with other seedeating organisms, such as birds and ants, that might also compensate to unknown degrees for seeds made available by the manipulations; and (c) predation or foraging constraints that might make it unprofitable for certain rodent species to collect particular kinds of seeds or to forage in certain microhabitats. For example, the failure of small seedeaters to compensate for missing Dipodomys might be due in part to increased susceptibility of the small granivores to predation if they were to increase foraging in the open microhabitats favored by the missing kangaroo rats (see Rosenzweig 1973, Brown 1975, Price 1978, Hay and Fuller 1981, Thompson 1982a, Price and Brown 1983, Kotler 1984). Evidence for asymmetrical competition comes especially from the dependence of the outcome of interactions among kangaroo rat species on levels of food availability. Addition of seed did not result in an increase in the density of all granivorous rodent species, or even in an increase in some species and no significant change in others. Instead, regardless of the size of seed particles added and the temporal pattern of seed addition, the manipulation favored the largest granivorous rodent species, D. spectabilis, at the expense of its medium-sized congeners, D. merriami and D. ordii. This is consistent with reports that D. spectabilis is restricted to highly productive regions of deserts and arid grasslands, whereas D. merriami and D. ordii have wide distributions that include much less productive habitats (Rosenzweig and Winakur 1969, Brown 1975, Frye 1983). It is also consistent with evidence that D. spectabilis is behaviorally dominant and aggressively excludes small kangaroo rats (Frye 1983, M. A. Bowers et al., personal observation). Although our experiments provide no additional direct evidence on the mechanism, it seems reasonable that the asymmetry is due in large part to the size difference between the kangaroo rats. It is more likely that increased densities of D. spectabilis (in response to increased seed availability) caused the decrease in the densities of D. merriami and D. ordii, than that the converse is true. Unfortunately, because of the difficulty of designing fences that would allow free access of D. spectabilis while excluding the medium-sized kangaroo rats, it will be difficult to obtain direct experimental evidence to test this hypothesis. Similarly complex patterns are apparent in the response of the small seed-eating rodents to our manip- Ecology, Vol. 66, No. 5 ulations. These species showed no significant response to either seed addition or D. spectabilis removal, but they increased several-fold when all three Dipodomys species were removed. This suggests that the kangaroo rats, especially the medium-sized species D. merriami and D. ordii, have a strong competitive effect on the small granivorous rodents over a wide range of potential availability of seeds. On the other hand, had we been able to perform the reciprocal experiment and remove the small granivores, it is unlikely that we would have been able to detect any competitive effect on Dipodomys. Such noncomplementary results can be expected whenever coexisting species differ greatly in individual body mass (which can result in asymmetrical interactions based on aggression or differential foraging efficiency), population density, or both (as in the present study). Such differences can greatly complicate the interpretation of removal experiments to test for competition. For example, Abramsky and Sellah (1982) removed the smaller of two granivorous rodent species from a sand dune in Israel and found little evidence for competition, but it is questionable whether the expected response would have been of a detectable magnitude, especially given the lack of replication and short duration of their experiment. In all of our experiments we observed time lags of several months to a year before we noted any changes in the rodent populations, and several years were required to obtain statistically significant results (see also Hairston 1980). This observation also seems to have important implications for other field experiments in which supplemental food is provided or selected species are removed. Clearly the absence of an immediate response cannot be taken to indicate the lack of food limitation or competition. It is noteworthy, however, that other experimental studies of desert rodents that have failed to obtain any evidence of competition have been conducted for less than a year (e.g., Schroeder and Rosenzweig 1975, Abramsky and Sellah 1982). It remains unclear why we observed such long delayed responses to our manipulations. Production of desert seed crops and rodent reproduction both occur in temporal pulses associated with favorable conditions (adequate precipitation). Because at least some species of desert rodents (especially D. spectabilis) store seeds in caches and disperse primarily as juveniles, it is perhaps not surprising that availability of a new seed crop or of dispersing young rodents may be required for local populations to respond completely to altered conditions of food availability and density of competitors. This does not necessarily mean that periods of significant competition are infrequent, as Wiens (1977) has suggested. If this were the case, one would expect to see significant differences in the rodent populations among our experimental treatments only during such rare instances. But we observed that once the rodents had responded to the manipulations, the relative magnitudes of the differences in densities and biomass of October 1985 MANIPULATIONOF A DESERTRODENT COMMUNITY species between experimental and control plots remained surprisingly constant (e.g., Figs. 2 and 6). Both the low degree of compensation and the long time lags can probably be attributed at least in part to two complicating factors that are common to many communities: competitive interactions with distantly related organisms, and indirect effects of granivores mediated through their impact as predators on seedproducing plants. Preliminary results of ongoing experiments suggest that both are important in this system. Although much of the food made available by adding exogenous seeds or removing competing rodent species is not converted into rodent biomass, this seed does not persist for long in the environment. Preliminary data suggest that at this study site as elsewhere (e.g., Brown and Davidson 1977, Brown et al. 1979a, b) granivorous birds and ants consume many of the seeds that otherwise would be available to rodents. Also, rodents (and probably the other granivores as well) have significant long-term effects on the availability of different kinds of seeds because their selective predation on certain seed species importantly influences the composition of the plant community (e.g., Inouye et al. 1980, Inouye 1981). Another factor that may be extremely important in this community is the impact of rodent predators. We emphasize that although our experiments demonstrate that limited food resources and interspecific competition play major roles in the organization of the rodent community, this does not mean that predation is not also important. Indeed, our study site supports many predators (including coyotes, hawks, owls, and snakes), and we have little evidence to suggest that rodents die of starvation. We suspect that most individuals are killed by predators (and we have witnessed a few incidents). Probably food supply, interspecific competition, and predation interact to limit rodent populations. Competition among species decreases an already limited food supply and causes individual rodents to spend more time foraging and to forage in more exposed microhabitats, thereby increasing their susceptibility and losses to predators. Other processes, such as parasitism, may also be important, but these are even less well understood (Munger et al. 1983). Variation in the physical environment obviously also plays a major role in regulating the population size and determining the community structure of desert rodents. Although this has not been a focus of our experiments, the effects of climate are apparent in our data. The substantial increase in the average densities of both kangaroo rats and small granivorous rodents in the winter of 1981-1982 (Fig. 6) appears to have been a direct consequence of exceptionally heavy precipitation and high seed production in the preceding months. These temporal fluctuations are consistent with data showing correlations between precipitation, seed production, and granivorous desert rodent population density and species diversity over geographic gradients 1559 in southwestern North America (Brown 1975). This provides an example of how physical and biotic factors (climate and prey availability) can interact to limit population densities and affect community organization. Experimental manipulations are very much in vogue in contemporary ecology. Although we believe the present study provides another example of the valuable insights that can be obtained from replicated, controlled experiments conducted in the field, we caution against the uncritical acceptance of experimental results as an accurate representation of the patterns and processes that occur in unmanipulated systems. In this regard, we mention several limitations of our procedures that should be considered before extrapolating our results to make inferences about natural communities. 1) The spatial and temporal scales of our manipulations were arbitrarily limited. Our experimental plots were tiny islands in a vast sea of unmanipulated desert. Would we have observed the same kinds or magnitudes of density changes if we had been able to treat much larger areas, or do the responses we recorded simply reflect the choice of small patches of favorable habitat by individuals whose populations are regulated largely by other processes operating on a much larger spatial scale? 2) Although we tried to minimize or control for artifacts associated with the use of fences to limit immigration onto the plots, several possible effects remained. Perhaps the most serious was the differential exclusion of rodent predators, especially large snakes, which could pass through the large gates but not the small ones. To some extent, monitoring populations of non-seed-eating rodents, which presumably have the same predators, is a control for this. This control would be invalidated, however, if the non-seedeaters are not susceptible to predators (as is perhaps suggested by the clumsy locomotion and strong, unpleasant odor of the two Onychomys species). Another problem is that the rodents might ultimately be indirectly affected as a consequence of the differential exclusion of other kinds of organisms, such as insectivorous lizards. Also, to the extent that aggressive interactions among rodents are important, the fact that individuals must enter and leave the plots through the gates may facilitate the defense of space by dominant individuals and species. 3) The kind of seeds added and their method of distribution may have influenced the response to these manipulations. Most natural seeds are smaller and different in details of their packaging and chemical composition than the millet particles we added. In addition, native seeds fall to the ground beneath the parent plant and are disseminated gradually over a period of weeks, and many become buried in the soil where they may remain for many years (Tevis 1958). Although we cannot evaluate the effects of all of these factors, it is likely that we are supplying exceptionally attractive and available food particles. It seems likely that species JAMESH. BROWN AND JAMESC. MUNGER 1560 (such as D. spectabilis) that can efficiently harvest seeds and store them in large caches would benefit most from our seed additions, but perhaps not so much from more natural increases in native seed production. We do not believe that these potential problems detract unduly from the importance of our experimental results. Indeed, the creation of possible artifacts is the price of human intervention in any natural system; some form of this problem is common to all experiments. In conclusion, we wish to emphasize not only how much we have learned about the ecology of desert rodents from the present and previous studies, but also how much still remains to be done. Desert rodents probably have been studied more intensively by modern population and community ecologists than any other group of terrestrial organisms. These studies have been conducted by many independent scientists, who have used sophisticated observational, experimental, and statistical techniques to test current theories and investigate general questions. Although these studies generally agree in suggesting that food supply, interspecific competition, predation, habitat structure, and climate interact to regulate the abundance of particular species and to influence the composition of rodent communities, they have produced few other satisfying generalizations. Instead, they have revealed tremendous variation between sites as well as a wealth of complex interactions within local communities that serve only to underscore what challenging problems remain. ACKNOWLEDGMENTS This researchis part of a collaborative study supervised jointly by D. W. Davidson and J. H. Brown and sponsored by the National Science Foundation, most recently by Grant BSR-8021535. So many people have contributed to the project that we hesitate to list them all for fear of forgetting someone. We hereby express our gratitude to all of these good people, and we humbly acknowledge that if it were not for the help of many of them, this research could never have been done. LITERATURE CITED Abramsky, Z. 1978. Small mammal community ecology. Changes in species diversity in response to manipulated productivity. Oecologia (Berlin) 34:113-123. Abramsky, Z., and C. Sellah. 1982. Competition and the role of habitat selection in Gerbillus allenbyi and Meriones tristami: a removal experiment. Ecology 63:1242-1247. Altman, P. L., and D. S. Ditman. 1968. Metabolism. Biological handbook. Federation of American Societies for Experimental Biology, Bethesda, Maryland, USA. Bender, E. A., T. J. Case, and M. E. Gilpin. 1984. Perturbation experiments in community ecology: theory and practice. Ecology 65:1-13. Brown, J. H. 1975. Geographical ecology of desert rodents. Pages 314-341 in M. L. Cody and J. M. Diamond, editors. Ecology and evolution of communities. Harvard University Press, Cambridge, Massachusetts, USA. 1984. Desert rodents: a model system. Acta Zoo- logica Fennica 172:45-49. Brown, J. H., and D. W. Davidson. 1977. Competition between seed-eating rodents and ants in desert ecosystems. Science 196:880-882. Ecology,Vol. 66, No. 5 Brown,J. H., D. W. Davidson,J. C. Munger,and R. S. Inouye. In press. Experimental community ecology: the desert granivoresystem. In J. M. Diamondand T. J. Case,editors. Community ecology. Harper and Row, New York, New York, USA. Brown, J. H., D. W. Davidson, and 0. J. Reichman. 1979a. An experimental study of competition between seed-eating desert rodents and ants. American Zoologist 19:1129-1143. Brown, J. H., O. J. Reichman, and D. W. Davidson. 1979b. Granivory in desert ecosystems. Annual Review of Ecology and Systematics 10:201-227. Connell, J. H. 1961. Effect of competition, predation by Thais lapillusand other factors on naturalpopulationsof the barnacle Balanus balanoides. Ecological Monographs 31:61-104. Dayton, P. K. 1971. Competition, disturbance and community organization: the provision and subsequent utilization of space in a rocky intertidal community. Ecological Monographs 41:351-389. Frye, R. J. 1983. Experimental field evidence of interspecific aggressionbetween two species of kangaroorat (Dipodomys). Oecologia(Berlin)59:74-78. Grant,P. R. 1969. Experimentalstudies of competitive interactionin a two-species system. I. Microtusand Cleth- rionomys species in enclosures. Canadian Journal of Zoology 47:1059-1082. 1971. Experimental studies of competitive interaction in a two-species system. III. Microtus and Peromyscus species in enclosures. Journal of Animal Ecology 40:323-350. 1972. Interspecific competition among rodents. Annual Review of Ecology and Systematics 3:79-106. Hairston, N. G. 1980. An experimental test of an analysis of field distributions: competition in terrestrial salamanders. Ecology 61:817-826. 1981. An experimental test of a guild: salamander competition. Ecology 62:65-72. Hay, M. E., and P. J. Fuller. 1981. Seed escape from heteromyid rodents: the importance of microhabitat and seed preference. Ecology 62:1395-1399. Inouye, R. S. 1981. Interactions among unrelated species: granivorous rodents, a parasitic fungus, and a shared prey species. Oecologia (Berlin) 49:425-427. Inouye, R. S., G. S. Byers, and J. H. Brown. 1980. Effects of predation and competition on survivorship, fecundity, and community structure of desert annuals. Ecology 61: 1344-1351. King, J. R. 1974. Seasonal allocation of time and energy resources in birds. Pages 4-85 in R. A. Paynter, Jr., editor. Avian energetics. Publication Number 15 of the Nuttall Ornithological Club, Cambridge, Massachusetts, USA. Kotler, B. P. 1984. Predation risk and the structure of desert rodent communities. Ecology 65:689-701. Lubchenco, J., and B. A. Menge. 1978. Community development and persistence in a low rocky intertidal zone. Ecological Monographs 48:67-94. Menge, B. A., and J. Lubchenco. 1981. Community orga- nizationin temperateand tropicalrockyintertidalhabitats: prey refuges in relation to consumer pressure gradients. Ecological Monographs 51:429-450. Munger, J. C., and J. H. Brown. 1981. Competition in desert rodents: an experiment with semipermeable exclosures. Science 211:510-512. Munger, J. C., M. A. Bowers, and W. T. Jones. 1983. Desert rodent populations: factors affecting abundance, distribution, and genetic structure. Great Basin Naturalist Memoirs 7:91-116. Pacala, S., and J. Roughgarden. 1982. Resource partitioning and competition in two-species insular Anolis lizard communities. Science 217:4/11 116. October 1985 MANIPULATION OF A DESERT RODENT COMMUNITY Paine, R. T. 1974. Intertidal community structure: experimental studies on the relationships between a dominant competitor and its principal predator. Oecologia (Berlin) 15:93-120. Price, M. V. 1978. The role of microhabitat in structuring desert rodent communities. Ecology 59:910-921. 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. Redfield, J. A., C. J. Krebs, and M. J. Tait. 1977. Competition between Peromyscus maniculatus and Microtus townsendii in grasslands of coastal British Columbia. Journal of Animal Ecology 46:601-616. Rosenzweig, M. L. 1973. Habitat selection experiments with a pair of coexisting heteromyid rodent species. Ecology 54: 111-117. Rosenzweig, M. L., and J. Winakur. 1969. Population ecology of desert rodent communities: habitats and environmental complexity. Ecology 50:558-572. Schaffer, W. M. 1981. Ecological abstraction: the consequences of reduced dimensionality in ecological models. Ecological Monographs 51:383-401. 1561 Schroder, G., and M. L. Rosenzweig. 1975. Perturbation analysis of competition and overlap in habitat utilization between Dipodomys ordii and Dipodomys merriami. Oecologia (Berlin) 19:9-28. Siegel, S. 1956. Nonparametric statistics for the behavioral sciences. McGraw-Hill, New York, New York, USA. Sousa, W. P. 1979. Experimental investigations of disturbance and ecological succession in a rocky intertidal algal community. Ecological Monographs 49:227-254. Tevis, L., Jr. 1958. Interrelations between the harvester ant Veromessor pergandei (Mayo) and some desert ephemerals. Ecology 39:695-704. Thompson, S. D. 1982a. Microhabitat utilization and foraging behavior of bipedal and quadripedal heteromyid rodents. Ecology 63:1303-1312. 1982b. Structure and species composition of desert heteromyid rodent species assemblages: effects of a simple habitat manipulation. Ecology 63:1313-1321. Wiens, J. A. 1977. On competition and variable environments. American Scientist 65:590-597. 1562 JAMES H. BROWN AND JAMES C. MUNGER Ecology, Vol. 66, No. 5 APPENDIX Summary of the most important results of seed addition and species removal experiments on desert rodents. For each entry, top line gives mean values for plots subjected to the same treatment, and bottom line gives value of F or MannWhitney U, from a comparison of treatment means, followed by the probability level (underlining indicates P < .05). Treatment: Comparison: Plots: Dipodomys spectabilis Change in density (mean no. captures per plot per period) Change in energy consumption (kJ plot-' d-1) Adult body mass (g) Percent reproductive Residence time (mo) Percent residents (% multiple captures) Dipodomys merriami and D. ordii Change in density (mean no. captures per plot per period) Change in energy consumption (kJ plot-. d-l) Adult body mass (D. merriami only; g) Percent reproductive (D. merriami only) Residence time (D. merriami only; mo) Percent residents (D. merriami only) Small seedeaters Change in density (mean no. captures per plot per period) Change in energy consumption (kJ plot-I d-1) Onychomys species Change in density (mean no. captures per plot per period) Change in energy consumption (kJ plotd-l) Adult body mass (Onychomys leucogaster only; g) Percent reproductive (O. leucogaster only) Residence time (mo) Percent residents Seed addition Seed size Control Added Small Mixed 11, 14 1, 2, 6, 9, 13, 18, 20, 22 2, 22 9, 20 -0.355 0.638 U= 1, P=.044 0.462 0.823 F= 0.344, P Seed schedule Large Constant Pulse 6, 13 2, 6, 9, 13, 20, 22 1, 18 0.871 .740 0.718 U=4, 0.396 P=.321 117.89 -76.83 U= 1, P=.044 142.63 81.91 F= 0.352, P 162.46 .729 129.00 U= 5, 84.54 P=.429 127.1 U=0, 43.8 U= 6, 5.67 U=4, 56.6 U= 7, 135.2 P=.022 44.8 P=.356 7.11 P=.200 57.1 P=.444 137.3 136.0 F = 0.692, P 38.6 32.0 P F= 1.57, 7.24 6.40 F= 0.602, P 58.3 50.0 F= 0.387, P 130.1 .566 44.1 .341 6.43 .603 56.6 .709 134.5 U= 3, 38.2 U=0, 6.69 U=0, 54.9 U= 3, 137.3 P= .214 64.5 P=.036 8.36 P=.036 63.6 P= .214 0.113 U= 0, -0.984 P=.022 -1.151 -0.854 F= 0.385, P -0.683 .710 -0.896 U= 4, -1.250 P= .321 15.75 U= 0, -85.57 P=.022 -55.43 .611 -78.46 U= 5, -106.89 P=.429 45.4 46.4 U= 3, P=.133 39.8 36.4 U= 6, P=.356 5.26 6.50 U= 4, P= .200 41.9 55.2 U= 0, P =.022 -104.39 -75.57 F= 0.584, P 44.5 46.5 F= 1.216, P=.410 40.0 35.1 39.0 F= 0.048, P=.953 6.47 5.73 4.21 P=.013 F=25.281, 41.7 44.1 37.3 F= 0.268, P=.782 45.4 U= 5, 38.0 U= 5, 5.47 U= 4, 41.2 U= 5, 45.5 P=.429 31.7 P=.429 4.63 P= .321 44.2 P=.429 0.022 U= 4, 0.231 P=.321 2.55 U= 4, 12.40 P=.321 45.2 0.193 U= 7, 0.074 P=.444 0.113 -0.218 0.170 F= 0.401, P =.701 8.39 U= 8, 5.01 P=.556 -7.91 0.446 U= 2, 0.123 P=.089 0.242 0.092 -0.186 F= 1.520, P=.350 31.25 U= 2, 9.53 P=.089 -9.18 17.65 6.58 F= 1.756, P= .313 23.07 5.02 U= 1, P=.071 36.2 U= 7, 36.6 P=.444 35.9 36.4 36.1 F= 0.173, P=.849 36.2 U= 4, 38.1 P= .321 40.5 U= 8, 4.80 U= 7, 29.2 U= 8, 37.3 P=.556 4.91 P=.556 29.8 P=.556 56.7 26.7 F= 0.710, P=.559 3.00 3.90 7.50 F= 2.021, P=.331 43.0 21.0 31.3 F= 0.219, P = .815 32.2 U= 4, 4.28 U= 2, 30.8 U= 5, 52.5 P=.321 6.50 P= .190 25.4 P=.429 13.2 7.47 8.10 F= 0.559, P=.622 0.345 0.049 U= 1, P=.071 APPENDIX. 1563 MANIPULATION OF A DESERT RODENT COMMUNITY October 1985 Continued. D. spectabilis removal Dipodomys spp. removal Po. rugosus removal Ant removal Control Removal Control Removal Control 11, 14 5,24 8, 11, 12, 14 15, 19, 21 11, 14 Po. rugosus removal 8, 12 Ant removal 4, 17 Control 11, 14, 15, 21 Po. rugosus removal 3,8, 12, 19 -0.233 -0.523 -0.355 F= 0.178, P = .845 -76.83 -37.09 -79.07 P=.845 F=0.172, 133.0 131.9 F= 2.389, P=.240 52.3 53.2 43.8 F= 0.151, P= .866 10.43 5.67 7.69 F= 1.966, P=.285 52.8 56.9 56.6 P=.936 F=0.067, 127.1 -0.278 1.300 0.113 P=.541 F=0.760, 1.174 F= 0.402, 1.806 P= .591 113.16 F= 0.224, 157.45 P= .683 -10.34 129.62 15.76 P=.530 F=0.790, 46.4 F= 0.014, 39.8 F= 0.393, 46.2 P= .916 44.1 P = .595 46.4 55.2 F=0.101, 52.1 P= .780 0.231 F = 0.984, 0.662 P = .426 0.198 U=0, 1.456 P=.014 -0.42 0.203 0.193 P=.682 F=0.435, 0.858 U=8, 0.796 P=.557 10.17 F= 0.769, 33.45 P= .473 8.71 U= 0, 54.21 P=.014 0.70 9.02 8.39 P=.667 F=0.446, 31.98 U= 8, 30.94 P=.557 0.077 F= 4.031, -0.483 P= .183 0.312 U=3, -0.143 P=.100 0.411 0.177 0.446 F= 2.846, P=.203 0.103 U= 7, 0.066 P= 0.443 5.89 F= 5.073, -30.42 P=.153 21.98 U=3, -8.50 P=.100 12.70 29.94 31.25 P=.149 F=3.834, 7.76 U=7, 5.71 P=.443 36.2 F= 12.066, 38.2 P=.074 37.0 U= 5, 33.9 P=.429 36.2 34.5 U= 6, 37.2 P=.571 40.5 F= 0.140, 4.80 F = 0.003, 29.2 F= 0.021, 30.0 P= .745 4.50 P = .959 32.1 P= .898 64.8 49.9 U= 4.5, P= .371 2.83 4.31 P=.500 U=4, 23.3 20.0 P=.443 U=7, 57.7 U=3, 3.40 U= 2, 28.0 U= 6, 54.4 P=.200 3.92 P=.267 15.4 P=.343 46.3 45.8 F= 0.062, P= .941 24.8 51.0 39.8 P=.0065 F=41.61, 6.39 6.97 6.50 F= 0.223, P=.812 66.7 56.9 55.2 P=.498 F=0.887, 35.8 37.8 P= .712 F=0.382, 32.1 59.4 P=.793 F=0.251, 3.33 2.62 4.80 P=.856 F=0.169, 10.7 56.4 29.2 F= 7.795, P=.065 40.5