This file was created by scanning the printed publication. Errors identified by the software have been corrected; however, some errors may remain. SEED USE BY DESERT GRANIVORES William S. Longland ABSTRACT and high-elevation deserts, because most bird species are resident only from spring through fall, and they eat insects during much of this time, while ants are relatively inactive above ground when temperatures are cold. By contrast, nonhibernating rodent species are active foragers for seeds all year. Furthermore, birds and ants glean seeds largely from the soil surface; rodents dig for buried seeds as well (Johnson and Jorgensen 1981). Thus, rodents are often the main consumers of desert plant seeds. In addition to consuming seeds, though, large numbers of seeds are cached by desert rodent species in three families (Heteromyidae-the kangaroo rats, kangaroo mice, and pocket mice; Muridae-the New World mice, voles, and woodrats; and Sciuridae-the ground squirrels and chipmunks). The heteromyids are a diverse group that is well represented in most desert localities. Heteromyids and other arid-land rodents can harvest substantial fractions of a given plant species' seed production. The degree of granivory probably depends both on seed density and on desirability of particular seed types. Thus, seeds of some plant species are greatly reduced by rodents, while others may be barely touched (table 1). From a plant's perspective, it is what the rodent does with the harvested seed that is important. After seeds are harvested, rodents may consume them or cache them in one of two ways (Price and Jenkins 1986). "Larderhoarding"-practiced by most desert rodentsrefers to placing large caches in a centrally located burrow or nest. "Scatterhoarding"-practiced mainly by heteromyids and to a lesser degree by sciurids-refers instead to placing numerous smaller caches in shallow depressions on the ground surface and covering them with soil. Three western Great Basin study areas that have undergone extensive disturbance were monitored for species composition ofgranivorous rodents. Disturbed habitats at two of these sites (Red Rock and Flanigan) recovered with native or introduced perennial plants and were dominated by heteromyid rodent species, which are important seeddispersal agents for many desert plants. At one of these sites, Indian ricegrass-a native perennial grass-rapidly dominated the disturbed area, perhaps because of strong interactions between this grass and local heteromyid rodents. The third site (Noble), which lacks heteromyids, has become infested with an introduced annual weed (medusahead). The seed dispersal activities of heteromyids may be important in obtaining desirable responses to disturbance on desert rangelands. INTRODUCTION Granivorous, or seed-eating, animals are the most abundant and diverse herbivores in North American deserts. This high abundance and diversity is probably due to the generally greater availability and suitability of seeds as food compared to other types of plant materials in deserts. Desert plants produce large seasonal flushes of seeds, which remain dormant and retain their nutritional quality for substantially longer periods than other aboveground plant parts (Janzen 1971). Even with the seasonal nature of desert seed production, seeds are available to desert granivores in the soil seed pool year-round; this is well illustrated by occasional spring flushes of annual plants from seeds lying dormant in the soil over one or more winters. These properties of desert plant seeds (seasonal production in massive quantities and nutritional retention over time) also make them an ideal food for storing for future use. As a consequence, granivorous diets have been adopted by various groups of desert animals, and desert plants, having coevolved with these granivores, often exhibit adaptations that either reduce levels of seed predation or allow them to capitalize on granivore activities for dispersing their seeds. Here, I concentrate on the latter coevolutionary relationship-seed dispersal. Various groups of rodents, birds, and seed-harvester ants comprise the granivore guild in North American deserts (Brown and others 1979). Birds and ants are to a large extent seasonal granivores, especially in northern Table 1-Previous studies documenting percentages of plant seed production harvested by rodents Plant type (site) Seeds harvested Source Percent Annual grasses (California Central Valley) Annual grasses (Southern California) Erodium cicutarium (Mojave Desert) Larrea tridentata (Chihuahuan Desert) Oryzopsis hymenoides (Great Basin Desert) Desert plants (Mojave Desert, NV) Forbs (Arizona grassland) Paper presented at the Symposium on Ecology, Management, and Restoration of Intermountain Annual Rangelands, Boise, ID, May 18-22, 1992. William S. Longland is Research Ecologist, U.S. Department of Agriculture, Agricultural Research Service, Conservation Biology of Rangelands, Reno, NV 89512. 233 93 Pearson 1964 30-65 Borchert and Jain 1978 95 Soholt 1973 87 Chew and Chew 1970 46 McAdoo and others 1983 30-80 Nelson and Chew 19n <1 Pulliam and Brand 1975 It is through scatterhoarding activities that desert rodents are most likely to have a positive effect on seedling recruitment; larderhoards are generally placed too deep underground for seedlings to successfully emerge if they germinate. Thus, the net effect of rodents on seedling recruitment of a particular plant species is largely determined by numbers of seeds that are removed from the germinable seed pool by consumption and larderhoarding versus numbers that persist in scatterhoards for later germination (Price and Jenkins 1986). Scatterhoards that are not recovered for future consumption may benefit seeds in three different ways: (1) buried seeds may have a higher probability of germination and establishment than unburied seeds (Vander Wall1990), (2) buried seeds are not vulnerable to consumption by nonscatterhoarding granivores (birds and ants) that harvest seeds only from the soil surface (Price and JenkinS 1986), and (3) seeds of certain desert plant species may have enhanced germinability when they have been handled by scatterhoarding rodents (La Tourette and others 1971; McAdoo and others 1983; Reynolds and Glendening 1949). In this regard, heteromyid rodents, being very common and avid scatterhoarders in North American deserts, are very important components of arid rangeland communities. Although heteromyids have been known to have important effects on range vegetation for some time (for example, Reynolds 1950), results oflong-term experiments in the Chihuahuan Desert have recently highlighted the ecological significance of these "keystone" granivores; rodent exclusion experiments have shown that heteromyids directly affect the species composition and physiognomy of the local plant community (Brown and Heske 1990). Table 2 lists a sample of plant species that have been found germinating from rodent scatterhoards; this list includes grasses, shrubs, and trees and both native and introduced species. The native species listed appear to be dependent on harvesting by animals for dispersal since their seeds lack external appendages that could facilitate dispersal by other means (for example, wind or adhesion to fur), while the introduced species (cheatgrass) has such appendages to facilitate dispersal. This is to be expected, since native plant species have coevolved with local granivores, while successful invasions of exotics are most likely with species that do not need to rely for dispersal on a granivore guild that is unfamiliar with them. In this paper, I present data on the species composition of granivorous rodent communities at three disturbed study sites in the western Great Basin Desert. I show that for two of these sites, where native plants have recovered well from disturbance, rodents include mainly scatterhoarding Table 2-Previous studies documenting germination of desert plant seeds from rodent scatterhoards Plant species Velvet mesquite Indian ricegrass Antelope bitterbrush Palo Verde Cheatgrass Location Source Arizona Nevada California Oregon Nevada Arizona Nevada Reynolds and Glendening 1949 McAdoo and others 1983 Hormay 1943 West 1968 Vander Wall1990 McAuliffe 1990 La Tourrette and others 1971 234 heteromyid species, while at the third site, which is dominated by an introduced annual weed species, heteromyids and other scatterhoarders are rare. I also discuss data from two experiments: (1) a field study illustrating strong interactions between heteromyids and a native animaldispersed grass species, and (2) laboratory seed preference tests with captive heteromyids showing that these rodents prefer seeds of this native grass to those of two introduced annual grasses. My aim is to illustrate the potential importance of scatterhoarding granivores for desert rangelands by showing examples of desirable successional responses to disturbance where these animals are common and undesirable responses where they are lacking. Certainly, numerous ecological and historical factors combine to determine such responses, but plant/animal interactions involving native granivores are one such factor that has received little attention. METHODS Field studies were conducted at two northwestern Nevada sites (Red Rock, Washoe Co., Reno NW Quad.: T21N.R18E.S14; and Flanigan, Washoe Co., Flanigan Quad. T27N.R18E.S2) and a northeastern California site (Noble, Lassen Co., Shaffer Mtn. Quad. T30N.R15E.S27). The Red Rock and Flanigan sites were burned in 1985 and 1988, respectively, while the Noble site was disturbed by extensive grazing of sheep over several decades preceding this study. Prefire vegetation at Red Rock consisted of big sagebrush (Artemisia tridentata), Mormon tea (Ephedra viridis ), desert peach (Prunus andersonii), and various herbaceous species seeded after a previous fire, especially crested wheatgrass (Agropyron desertorum), which dominates the postfire vegetation. Prefire vegetation at Flanigan consisted largely of big sagebrush, scattered shrubs of other native species, and infrequent bunches of Indian ricegrass (Oryzopsis hymenoides); Indian ricegrass dominates the postfire community at Flanigan. The Noble site was dominated by low sagebrush <Artemisia arbuscula) before disturbance and remains so in adjacent undisturbed areas; the disturbed site has been dominated by an invasion of medusahead wildrye (Taeniatherum caput-medusae) but also has squirreltail (Sitanion hystrix) and rare Indian ricegrass bunches. Medusahead and cheatgrass (Bromus tectorum) have invaded undisturbed areas at Noble as well, but to a lesser extent. Soils are sand at Flanigan, heavy loam at Red Rock, and clay at Noble. I censused rodent populations by livetrapping twice monthly at Red Rock since the 1988 fire, and at irregular intervals at Flanigan and Noble since 1988 and 1991, respectively. I trapped in the disturbed areas and in adjacent undisturbed (or less disturbed) areas at all three sites. At Flanigan, I conducted experiments from 1989 to 1991 to quantify numbers of Indian ricegrass seeds harvested and cached by heteromyids. Seed harvest was quantified by planting single seeds and groups of 2, 10, and 100 seeds at various depths (0, 1, 2, 4, or 6 em) in randomly determined locations, which were checked after 7 days to determine if they had been removed by rodents. Seed caching was quantified by placing petri dishes with 40 g of Indian ricegrass seeds labeled with fluorescent dye on the ground and locating dye spots left at scatterhoards under UV light the following night. I tested for rodent preferences for Indian ricegrass, cheatgrass, and medusahead by offering food-deprived captive rodents from Noble 1.0 g of each seed type in separate petri dishes. Trials were videotaped in the laboratory. The seed types chosen first and eaten in the largest quantities were determined by replaying videos and used as criteria to assess potential preferences. kangaroo rat-D. deserti-and little pocket mouse-P. longimembris) were present as well (table 3). Heteromyids were equally or more abundant in disturbed habitats during early stages of plant succession as in undisturbed habitats at Red Rock and Flanigan, and at the latter site there was actually a pronounced increase in heteromyid species diversity during this recovery. By contrast, the only heteromyid species occurring at Noble was less common in disturbed than in undisturbed habitat. Although a cause-and-effect relationship cannot be inferred from these data, it is interesting to note that the disturbed habitat at Noble, where native perennial plants have been mostly replaced by a medusahead monoculture, is also depauperate in the scatterhoarding rodents that disperse the seeds of many native plant species. Heteromyids are generally able colonizers of disturbed areas, as indicated by the Flanigan and Red Rock data (table 3) and by an ongoing field experiment at Red Rock (USDA-ARS, unpublished data). In this experiment, I first demonstrated that three plots in the burned habitat had similar densities of three rodent species (Great Basin pocket mouse, Panamint kangaroo rat, and deer mouse), after which I removed vegetation on two of these plots by mowing. Heteromyids remained on the control and mowed plots in similar densities, but on the latter plots the nonheteromyid species (deer mouse) disappeared after vegetation removal. This suggests that the lack ofheteromyids at Noble is not simply attributable to past disturbance. The impressive abundance and diversity ofheteromyid rodents in the burned habitat at Flanigan is probably due to the dominance of Indian ricegrass in this area. This perennial grass species appears to have evolved a very close interaction with heteromyids. Its seed is highly preferred by these rodents; seeds of other plant species have been RESULTS AND DISCUSSION Proportions of various rodent species captured at the three study sites are shown in table 3. Four rodent species were captured at both the Red Rock and Noble study sites. Only one of the species occurring at Noble is a heteromyid (Great Basin pocket mouse--Perognathus parous), and this species was rare in the disturbed (medusahead) habitat. By contrast, a nonheteromyid species (deer mouse--Peromyscus maniculatus) was caught significantly (P < 0.05) more frequently in the medusahead than in the sagebrush habitat at Noble. Heteromyids occurred in similar proportions in disturbed and undisturbed habitats at Red Rock, and were represented by two species (Great Basin pocket mouse and Panamint kangaroo ratDipoclomys panamintinus) in both habitats. These two species were the earliest colonizers of the burned area after the 1988 fire at Red Rock. Six rodent species occurred at Flanigan, and all but one of them (white-tailed antelope ground squirrel-Ammospermophilus leucurus) were heteromyids. While only one of the heteromyid species (Merriam's kangaroo rat-D. merriami) was common in unburned sagebrush habitat at Flanigan, three were quite common in the burned (Indian ricegrass) habitat (Merriam's kangaroo rat, Panamint kangaroo rat, and Ord's kangaroo rat-D. ordii) and two others (desert Table 3-Relative abundances (percent) of various rodent species occurring in disturbed and undisturbed habitats at three western Great Basin study sites. Fire caused the disturbance at the Red Rock and Ranigan sites; extensive use by domestic sheep disturbed the Noble site. Heteromyid rodents, the primary scatterhoarding species at these sites, include species in the genera Dipodomys and Perognathus Percentage of captures Undisturbed Study sHe Rodent species Flanigan: Dipodomys merriami Dipodomys ordil Dipodomys panamintinus Dipodomys desert/ Perognathus longimembris Ammospermophilus leucurus Red Rock: Dlpodomys panamintinus Perognathus parvus Peromyscus maniculatus Reithrodontomys mega/otis habitat Sagebrush Indian rlcegrass 54 27 12 3 3 89 2 1 0 1 7 Agropyron/shrub 2 55 14 29 Sagebrush Noble: Perognathus parvus Peromyscus maniculatus Reithrodontomys mega/otis Spemophilus latera/Is 235 Disturbed habitat 23 50 23 4 1 Agropyron 16 67 17 0 Medusahead 2 95 2 0 found to drop out of heteromyid diets in nature when Indian ricegrass produces seeds (McAdoo and others 1983). Indian ricegrass has a classical animal-dispersed seed morphology, and seedlings appear to come mainly from heteromyid scatterhoards, which generally persist longer after germination than single seedlings (personal observation). Seeds in these scatterhoards often have 100 percent germination, partly because heteromyids discriminate against empty and nonviable seeds when harvesting Indian ricegrass (McAdoo and others 1983). Finally, the germinability of Indian ricegrass is enhanced when the seeds have been handled by heteromyids; this may be due in part to breaking mechanical dormancy of the seed by removal of its coat by rodents, but even unshelled seeds that have been handled by heteromyids have improved germinability (McAdoo and others 1983). Results of field experiments at Flanigan indicated that up to 50 percent of Indian ricegrass seeds were harvested by heteromyids over a 7-day period. Depth of seeds in the sand had a significant effect on harvest rate (P < 0.05); seeds on or near the surface were harvested in the greatest quantities. Seed density, however, had no effect on harvest rate. Single seeds or pairs were harvested at rates similar to groups of 10 or 100, illustrating the efficiency of heteromyids at locating even low-density soil seed reserves. By locating caches containing seeds labeled with fluorescent dyes, I found that up to 25-35 percent of Indian ricegrass seeds harvested by heteromyids were initially cached in scatterhoards, and that scatterhoards contained about 250 seeds on average. The largest scatterhoard found had 1,427 seeds. Although the scatterhoarding rates I have found at Flanigan are fairly high, it was also apparent from dye traces around rodent burrows and from large amounts of dyed seeds that I could not locate that many seeds were being larderhoarded or consumed. While it is not yet possible to determine net effects of heteromyids at this site on the Indian ricegrass population, it is clear from the rapid domination of Indian ricegrass in response to removal of shrubs by fire that a large soil seed reserve was in place even though this grass was uncommon in the prefire plant community. Because most of the Indian ricegrass population at Flanigan consists of clumps of two or more sameaged individuals, it is likely that this rapid postfire response was due to rodent scatterhoards providing this seed reserve. In laboratory seed preference experiments, the only rodent species that was common in medusahead habitat at Noble (deer mouse) exhibited a significant (P < 0.05) preference for medusahead and Indian ricegrass over cheatgrass. Great Basin pocket mice, the only heteromyid species at Noble, also preferred Indian ricegrass, but avoided medusahead seed. At Noble, established Indian ricegrass plants thrive quite well, but new recruits are seldom found. Although Indian ricegrass may show some promise for being established in areas vulnerable to medusahead infestation, the heteromyids which prefer Indian ricegrass seeds and act as its dispersal agents may be lacking, th~ limiting the persistence of an Indian ricegrass population. Other rodents, such as deer mice, may occur commonly in medusahead habitats, but their activities will not promote the establishment of native, animal-dispersed plant species. This may explain why squirreltail, which has seed appendages for dispersal by wind or adhesion to 236 fur, is more common in the medusahead habitat at Noble than other native plants dependent on animal dispersal vectors. In conclusion, I suggest that a stronger understanding of granivores' effects on native and introduced plant species may be an important element in understanding successional processes in desert rangeland plant communities. Such information may simply allow prediction of plant species responses following disturbance based on knowledge of granivore species composition in a given area. Eventually though, arid-land restoration efforts may include active management for granivores that play important roles in maintaining native plant communities. Although range restoration efforts involving active human intervention, such as artificial seeding, may produce more rapid results than "natural restoration" by native fauna, the latter carries the advantages of reduced financial expenditure, of not being logistically restricted by spatial scale, and of ultimately resulting in more "pristine" plant communities. ACKNOWLEDGMENTS I thank Steve Jenkins, Steve Vander Wall, and Jim Young for thoughtful comments that improved the manuscript. This study is a contribution of the USDAAgricultural Research Service, Conservation Biology of Rangelands Unit, Reno, NV. REFERENCES Borchert, M. I.; Jain, S. K. 1978. The effect of rodent seed predation on four species of California annual grasses. Oecologia. 33: 101-113. Brown, J. H.; Heske, E. J. 1990. Control of a desertgrassland transition by a keystone rodent guild. Science.250:1705-1707. Brown, J. H.; Reichman, 0. J.; Davidson, D. W. 1979. Granivory in desert ecosystems. Annual Review of Ecology and Systematics. 10: 210-227. Chew, R. M.; Chew, A E. 1970. Energy relationships of mammals of a desert shrub (Larrea tridentata) community. Ecological Monographs. 40: 1-21. Hormay, A L. 1943. Bitterbrush in California. For. Res. Note 34. Washington, DC: U.S. Department of Agriculture, Forest Service. 13 p. Janzen, D. H. 1971. Seed predation by animals. Annual Review of Ecology and Systematics. 2: 465-492. Johnson, T. K.; Jorgensen, C. D. 1981. Ability of desert rodents to find buried seeds. Journal of Range Management. 34: 312-314. La Tourrette, J. E.; Young, J. A; Evans, R. A 1971. Seed dispersal in relation to rodent activities in seral big sagebrush communities. Journal of Range Management. 24: 118-120. McAdoo, J. K.; Evans, C. C.; Roundy, B. A.; Young, J. A.; Evans, R. A 1983. Influence ofheteromyid rodents on Oryzopsis hymenoides germination. Journal of Range Management. 36:61-64. McAuliffe, J. R. 1990. Paloverdes, pocket mice, and bruchid beetles: interrelationships of seeds, dispersers, and seed predators. Southwestern Naturalist. 35: ·329-337. Nelson, J. F.; Chew, R. M. 1977. Factors affecting seed reserves in the soil of a Mojave Desert ecosystem, Rock Valley, Nye County, Nevada. American Midland Naturalist. 97: 300-320. Pearson, 0. P. 1964. Carnivore-mouse predation: an example of its intensity and bioenergetics. Journal of Mammalogy. 45:177-188. Price, M. V.; Jenkins, S. H. 1986. Rodents as seed consumers and dispersers. In: Murray, D. R., ed. Seed dispersal. Sydney, Australia: Academic Press: 191-235. Pulliam, H. R.; Brand, M. R. 1975. The production and utilization of seeds in plains grassland of southeastern Arizona. Ecology. 56: 1158-1166. Reynolds, H. G. 1950. Relation of Merriam kangaroo rats to range vegetation in southern Arizona. Ecology. 31: 456-463. Reynolds, H. G.; Glendening, G. E. 1949. Merriam kangaroo rat, a factor in mesquite propagation on southern Arizona rangelands. Journal of Range Management. 2:193-197. Soholt, L. F.1973. Consumption of primary production by a population of kangaroo rats Wipodomys merriami) in the Mojave Desert. Ecological Monographs. 43: 357-376. Vander Wall, S. B. 1990. Food hoarding in animals. Chicago: University of Chicago Press. 445 p. West, N. E. 1968. Rodent-influenced establishment of ponderosa pine and bitterbrush seedlings in central Oregon. Ecology. 49: 1009-1011. 237