One page summary of the biology research

CHAPTER 1. COYOTE ENDOZOOCHORY OF PROSOPIS: CONSEQUENCES OF GUT PASSAGE, GERMINATION SUBSTRATE, AND SPATIAL PATTERNS OF DISPERSAL

ABSTRACT

Effective endozoochory requires that seeds maintain germinability after gut passage, that there is a suitable substrate for germination, and that animal dispersers deposit seeds in environments suitable for establishment. We sought to determine if coyotes (Canis latrans) are effective dispersers of western honey mesquite (Prosopis glandulosa var. torreyana) and screwbean mesquite (P. pubescens). Mesquites have increased their ranges over the past two centuries in southwestern North America and are among the dominant tree species at Ash Meadows National Wildlife Refuge, Amargosa Valley, NV, USA. We performed a laboratory experiment examining the effects of gut passage on seed germinability, conducted a field experiment comparing emergence from feces and soil, and developed spatial models of the distribution of coyote feces. Gut passage positively affected screwbean mesquite seeds through high rates of removal of seeds from their legumes and of germination, but had limited effect on honey mesquite seeds. However, only two screwbean mesquite seedlings and no honey mesquite seedlings emerged from 81 feces, while many seedlings emerged from soil. Coyotes deposited feces near water and human land use in riparian areas, savannas, and dunes, environments that are likely suitable for establishment. Although feces are an unsuitable substrate for germination, coyote primary dispersal may still contribute to mesquite establishment if seeds from feces are incorporated into the soil seedbank or undergo secondary dispersal. Further research is necessary to determine the potential of these

5

mechanisms. This study highlights the need to simultaneously consider the effects of gut passage, germination substrate, and spatial patterns of dispersal to determine the effectiveness of endozoochory.

INTRODUCTION

Endozoochory, or seed dispersal via ingestion by animals, is a widespread dispersal syndrome that can influence seed germination through gut passage, suitability of feces as a substrate, and the spatial pattern of dispersal. Seeds in large, edible fruits have adaptations for rapid passage through the gut and resistance to digestion (Traveset 1998). Gut passage can scarify seeds, which breaks their physical dormancy (Howe 1980; Gardener et al. 1993; Or and Ward 2003), and can separate seeds from their fruits, which can inhibit germination (Ortega Baes et al. 2001, 2002; Robertson et al. 2006). Furthermore, gut passage may increase germination rates for seedbanks infested with bruchid beetles due to higher digestibility of beetle-infested seeds and reduced spread of beetles after ingestion (Miller 1994; Or and Ward 2003).

Feces can provide a suitable substrate for germination through increased moisture and nutrient availability (Ocumpaugh et al. 1996) or can inhibit germination and water uptake by seeds (Brown and Archer 1989). Seeds that have not broken dormancy during gut passage may be incorporated into the seedbank or may be moved by secondary dispersers into suitable environments and germinate at a later time (Janzen 1981; Ortega Baes et al. 2001; Vander Wall and Longland 2004). Animals may provide directed dispersal by disproportionately depositing feces in habitat that is suitable for germination (Howe and Smallwood 1982; Chambers and MacMahon 1994) and removing seeds from

6

distance- or density-related mortality near the mother tree, including that caused by bruchid beetles (Janzen 1970; Howe and Smallwood 1982). Conversely, feces may be deposited in habitat that is unsuitable for seed germination or contain many seeds from the same mother plant, increasing intraspecific competition (Howe 1989; Hulme 2002).

Determining whether endozoochory is advantageous to plant establishment requires examining the cumulative effects of gut passage, deposition in feces, and dispersal patterns on seed germination. Seeds of many plants in the Mimosoideae (Fabaceae) subfamily (e.g., Prosopis, Acacia) have physical dormancy requiring scarification and are contained within indehiscent legumes, which require decomposition, ingestion, or other mechanisms to release seeds.

Species of the Mimosoideae are widespread in arid ecosystems and provide crucial ecosystem services, including nitrogen fixation, high quality forage for livestock and wildlife, and wood products and food for humans (Felker 1981; Livingston and Nials 1990). Several of these plants have also increased their ranges and become invasive over the past two centuries, which may be linked to animal dispersal (Brown and Archer 1989; Cox et al. 1993).

Endozoochory is one possible mechanism by which these seeds become germinable and are dispersed across the landscape. Gut passage can have positive, negative, or no effect on seed germination, depending on the Mimosoideae species and the animal disperser (Cox et al. 1993; Campos and Ojeda 1997; Ortega Baes et al. 2002; Kneuper et al. 2003; Silverstein 2005). The suitability of feces as a substrate also differs between plant species and animal dispersers (Kramp et al. 1998; Sánchez de la Vega and Godínez-Alvarez 2010). The ability of animals to disperse Prosopis seeds to locations

7

suitable for germination and establishment after incorporation into the soil seedbank is relatively unknown.

The purpose of this research was to examine the effects of primary dispersal by coyotes (Canis latrans) on germination of western honey mesquite (Prosopis glandulosa var. torreyana [L. Benson] M. Johnson) and screwbean mesquite (P. pubescens Benth.). Honey mesquite seeds can seasonally constitute a substantial portion of coyote feces in the southwestern United States (Meinzer et al. 1975; Ortega 1987; Kramp et al. 1998; Silverstein 2005), suggesting that coyotes may play a significant role as dispersers as do several other animal species (Kneuper et al. 2003). Animal dispersal and germination ecology of screwbean mesquite seeds have not been examined to our knowledge. Coyotes preferentially use habitats based on vegetation structure, water availability, and other landscape factors (Kays et al. 2008; Kozlowski et al. 2008; Atwood et al. 2011). Consequently, deposition of feces is likely non-random as well. Seeds of both mesquite species exhibit physical dormancy, have hard seed coats that require scarification, and are contained within indehiscent legumes. Honey mesquite seeds are also contained within a woody endocarp, while screwbean mesquite seeds are not.

Our question is whether primary dispersal of mesquite by coyotes effectively promotes establishment of mesquite seedlings. Effective dispersal is dependent on germinability after gut passage and deposition of seeds on a substrate and in an environment suitable for establishment; failure of either of these processes indicates that coyotes are not likely to be effective dispersers. We performed a laboratory experiment to compare the effects of gut passage through coyotes on germination to those of mechanical removal from the legume and mechanical/chemical scarification. We

8

conducted a field experiment to examine seedling emergence from coyote feces and soil substrates in different microhabitats. Finally, we mapped the distribution of coyote feces at Ash Meadows National Wildlife Refuge, Amargosa Valley, NV, to develop spatial models of mesquite seed dispersal by coyotes. Together, these analyses provide a framework to describe the efficacy of coyote endozoochory for mesquite dispersal.

METHODS

We collected coyote feces and mesquite legumes at Ash Meadows National Wildlife Refuge, a 9310-ha wetland complex in the Mojave Desert in southern Nevada, USA (36°25’12’’N, 116°19’48’’W; ca. 670 m a.s.l.). Dominant vegetation varies with water availability and includes: emergent wetlands (Typha spp., Scirpus spp., and Eleocharis spp.) where surface water is present, mesquite bosques at intermediate water availabilities, and desert scrub on dry, upland sites (Atriplex spp., Lycium spp., and Larrea tridentata [DC.] Cov.). Honey mesquite generally occurs on drier sites than screwbean mesquite.

Between September and November 2009, during the season of maximum mesquite seed dispersal, we collected samples of 30-50 legumes per tree from the ground surface beneath 10 to 17 trees from 7 sites for honey mesquite and 9 sites for screwbean mesquite. We walked approximately 80 km along a random path across the refuge, encountering 147 coyote feces and collecting 79.6% (n = 117) of those encountered (Fig. 1). We did not collect feces that were deposited in previous years as evidenced by color or dissolution or that did not have mesquite seeds visible on the exterior. Coyote feces are conspicuous and easily detected in all vegetation types. For each feces, we recorded the

9

location, substrate type, canopy cover, and distances to water, mature mesquite trees, and roads.

For 30.8% of the feces (n = 36), we classified mesquite seeds by the presence or absence of the endocarp for honey mesquite and exocarp for screwbean mesquite, presence of bruchid beetles (Algarobius prosopis) or their exit holes, and damage due to ingestion. We compared the proportion of seeds in each category between species using two sample z-tests. We weighed the proportion of material in each category to estimate the distribution and amount of seed in feces across the landscape. For the remaining 69.2% of the feces (n = 81) used in the subsequent field experiment, we recorded the total weight and visually identified the species of mesquite seeds visible on the surface of the feces.

Laboratory experiment

The laboratory experiment was designed to parse the effects of removal from the legume, scarification, and beetle infestation on germination of honey and screwbean mesquite. Treatment factors included: collection substrate (feces or trees and ground), bruchid beetle infestation (present or absent), seed pericarp status (removed or intact), and seed scarification status (scarified or unscarified). The treatments were inherently nested (e.g., seeds cannot be scarified unless removed from their pericarps), resulting in a total of 10 treatment combinations. There were four replicates of 25 seeds for each treatment combination for each species. For seeds collected from trees and the ground, we scarified seeds by mechanically nicking honey mesquite (Vilela and Ravetta 2001) and by a 30-minute 98% H2SO4 bath for screwbean mesquite (Jackson et al. 1990).

10

We tested for seed viability using four replicates of 25 seeds of each species using tetrazolium staining (Moore 1973). Seeds were removed from their pericarps, scarified, soaked in deionized water for 24 hours, and placed in Petri dishes on filter paper soaked with a 1% solution of 2,3,5-triphenyl tetrazolium chloride. Dishes were wrapped in foil and incubated at 30°C for three hours, at which point staining was evident. Embryos stained red or pink were considered viable. Mean (±SD) viability proportions were 0.99±0.02 for honey mesquite and 0.97±0.02 for screwbean mesquite.

Seeds were incubated in a dark growth chamber in 100 x 100-mm square plastic Petri dishes containing 60 g of sand saturated with deionized water at 30°C for 16 hours and 12°C for eight hours. These correspond to the mean daily maximum and minimum temperatures in April and May, the beginning of the growing season. We recorded germination weekly for a minimum of 30 days or until no additional seeds germinated for seven days. Seeds were considered germinated when the radicle emerged at least 2 mm from the seed. We analyzed the effects of pre-germination treatment on germination proportion for each species using one-way ANOVA. Significant differences were evaluated using Tukey’s test.

Field experiment

We placed the 81 feces that were not dissected back into the field in January 2010 on two transects along water availability gradients defined by changing dominant vegetation. Along these same two transects and at the same locations, we planted 10 mechanically scarified seeds of each mesquite species 0.5-cm deep. Feces and seed placement was stratified by canopy cover (under tree canopy or in interspace) and

11

substrate (bare soil where vegetation was removed, grass, or litter). We expected these factors to affect feces moisture, and thus seed imbibition. We monitored feces and planted seeds for seedling emergence and survival every 1-2 months during the 2010 growing season (March to November). During this monitoring, we collected a surface soil sample to 10-cm depth from within 0.5 m of the seeds for measurement of gravimetric soil water content (GSWC) and electrical conductivity (EC), an indicator of salinity. Together, these provide a measure of water availability. Soil samples were weighed, dried at 105°C for 72 hours, and reweighed to calculate GSWC. Soils were wetted to saturation for measurement of EC.

We compared seedling emergence from feces with emergence from seeds planted nearby in the soil to determine the suitability of feces as a substrate at each location. We then identified conditions necessary for seed germination of each species by comparing the distributions of environmental factors between locations where seeds in feces emerged, planted seeds emerged, and all locations using Pearson’s χ2 test for categorical factors and student t-tests for continuous water availability measures.

Spatial analysis of coyote seed dispersal

We analyzed the spatial pattern of coyote dispersal of mesquite seeds through deposition of feces at local and landscape scales. At local scales, we summarized the proportions of coyote feces deposited on different substrates (bare soil, grass, or litter), under tree canopies versus in interspaces, on or off roads, and near (<3 m) or far from water. We calculated the equivalent proportion of tree, road, and surface water cover within a 10-m belt surrounding the random walking paths using GIS data to determine

12

null conditions. Feces locations were compared to these null conditions using Pearson’s χ2 test. There was no equivalent test for substrate composition due to a lack of comparable GIS data.

At the landscape scale, we developed probability maps of coyote dispersal of honey and screwbean mesquite seeds using maximum entropy modeling (Maxent), a machine learning method for presence-only occurrence data (Phillips et al. 2006). Maxent estimates the probability distribution of a species or phenomenon (in this case, coyote feces) that has the largest spread (i.e., maximum entropy) subject to presence-only observations and in relation to a set of “feature” data that can include thresholds, gradients, or categorical environmental data (Phillips et al. 2006). Model gain, an indicator of the probability of occurrence, assesses how closely the model is concentrated around the presence observations in comparison to a uniform distribution. Maxent avoids making assumptions about what is unknown, including the meaning of observed absences, and weights each feature variable according to the complexity it adds to a model to prevent overfitting. Maxent generally performs well compared to other species distribution models (Phillips et al. 2006; Elith and Graham 2009).

We included eight predictor variables related to topography (elevation and slope), habitat (vegetation height, vegetation abundance, generalized vegetation types, and distance to surface water), and human land use (distance to roads and distance to buildings). Elevation and slope were derived from a digital elevation model constructed from 1-ft LiDAR data, and vegetation height was interpreted from the associated LiDAR canopy model (White Horse Associates 2008). Vegetation abundance was estimated by proxy of the soil adjusted vegetation index (SAVI) calculated from 30-m Landsat

13

Enhanced Thematic Mapper (ETM+) images from July 2009. Generalized vegetation types (emergent vegetation, riparian forest, mesquite bosque, grassland alkali scrub, dunes, and desert scrub), surface water, roads, and buildings were derived from photo- interpretation of true-color 1-m resolution digital orthographs (Sunderman and Weisberg 2012). Surface water and human land use variables were extrapolated across the landscape using Euclidean distance.

To build our dispersal suitability models (akin to habitat suitability models), we used Maxent 3.3.3 (Phillips et al. 2006; http://www.cs.princeton.edu/~schapire/maxent/) using the default settings (regularization multiplier = 1, maximum iterations = 500, convergence threshold = 10-5, maximum number of background points = 10,000). We ran ten replicates in which 80% of feces occurrences were randomly selected as training data, with the remaining 20% used for model testing. We used the average area under the curve (AUC) of receiver operating characteristic (ROC) analysis for the 10 replicates as a measure of predictive performance. We also used jackknife analyses of the model training gains to assess the importance of each predictor variable.

RESULTS

Honey mesquite seeds were present in 71.4% and screwbean mesquite seeds were present in 51.4% of dissected coyote feces (n = 36). Coyote feces contained large numbers of mesquite seeds (n = 2784 seeds for honey mesquite and n = 10341 seeds for screwbean mesquite). Screwbean mesquite seeds were more frequently freed from their pericarps (z = -32.19, P < 0.001) and were less frequently infested by bruchid beetles than honey mesquite seeds (z = 8.64, P < 0.001) (Table 1). No screwbean mesquite seeds

14

15 and 12% of honey mesquite seeds exhibited visible damage from coyote ingestion (z =

35.68, P < 0.001).

Laboratory experiment

Pre-germination treatments had significant effects on germination of both honey (F9,31 = 78.27, P < 0.001) and screwbean mesquite (F9,31 = 19.11, P < 0.001). In all cases, beetle-infested seeds failed to germinate (Table 2), so the following results focus on seeds that were not infested by beetles. For honey mesquite, gut passage improved germination rates through both removal from the exocarp (even if the woody pericarp remained) and scarification relative to unscarified seeds within intact endocarps (Table 2; Fig. 2). However, gut passage resulted in significantly lower germination rates than seeds that were manually removed from their endocarps and manually scarified (Table 2; Fig. 2), suggesting loss of seed viability.

In contrast, gut passage appeared to improve germination rates of screwbean mesquite seeds primarily through removal from the pericarp. No seeds germinated from within intact pericarps, regardless of treatment agent (Table 2; Fig. 2). Manual scarification did not result in significantly higher germination rates than unscarified seeds, but did reduce variability (Table 2; Fig. 2). Screwbean mesquite seeds did not experience loss of viability after gut passage, with some replicates experiencing higher germination rates than manually treated seeds (Table 2; Fig. 2). Gut passage also resulted in less variable germination rates compared to unscarified seeds (Fig. 2), indicating a potential role of scarification.

Field experiment

The field experiment suggested that coyote feces provide a poor substrate for mesquite seedling emergence. We observed only two screwbean mesquite seedlings and no honey mesquite seedlings emerge from the 81 coyote feces. Both screwbean mesquite seedlings emerged from feces under the canopies of screwbean mesquite trees where the soil surface was visibly moist in January. The two seedlings emerged in May 2010 from feces within 25 m of each other and survived only through August 2010.

The range of field conditions under which mesquite seeds germinated in soil was substantially greater than the range in which they germinated in feces. Screwbean mesquite seedlings emerged from seeds planted in soil at 44.4% of locations, with no significant difference in emergence based on substrate or canopy cover (χ2 = 5.077, P = 0.079). Similarly, honey mesquite seedlings emerged from planted seeds at 33.3% of locations, also with no significant difference in emergence based on substrate or canopy cover (χ2 = 1.029, P = 0.598).

Soils near feces with emergence of screwbean mesquite seedlings had intermediate water content, with GSWC at the start of the growing season in March of ~0.35 g g-1 (Fig. 3). Soils where planted screwbean mesquite seeds emerged generally had GSWC between 0.12 and 0.67 g g-1, more than the minimum measurement of 0.03 g g-1 and less than the maximum measurement of 0.73 g g-1 (Fig. 3). Honey mesquite seedlings emerged where seeds were planted across the entire range of GSWC (Fig. 3). Neither screwbean mesquite nor honey mesquite seedlings emerged where EC exceeded 31 dS m-1 (Fig. 3).

16

Spatial analysis of coyote dispersal

At local scales, coyotes preferentially deposited feces on roads and near water (Table 3). Of those not located on roads, 12.5% were located on coyote trails. Feces were deposited proportionally under tree canopies and in interspaces (Table 3). The majority of feces were deposited on bare soil, not grass or litter (83.0%, 10.2%, and 6.8%, respectively).

At the landscape scale, the most important predictor of coyote feces presence in the final Maxent model was distance to surface water (52% contribution), with elevation (16.4% contribution), generalized vegetation types (13.2% contribution), distance to buildings (10% contribution), and distance to roads (6.6% contribution) also contributing (Fig. 4). Vegetation abundance, vegetation height, and slope contributed minimally to the final model (Fig. 4). The probability of feces presence decreased dramatically with increasing distance to surface water, with the probability falling below 0.5 for areas over 500 m from surface water (Fig. 5). Feces were more likely to occur in the riparian, savanna, and dunes vegetation types (Fig. 5), all of which are located relatively close to surface water and include mesquite trees as co-dominants. The probability of feces presence was also higher within a narrow 2150 to 2330 m elevation band and closer to human land use (Fig. 5). The Maxent model AUC of 0.859±0.001 for the training data and 0.621±0.071 for the test data substantially exceeded the random prediction of 0.5, indicating good predictive accuracy. Model predictions identified several hotspots as highly suitable for coyote feces deposition (Fig. 6); these areas were located where tree cover is high, suggesting they are also suitable environments for establishment.

17

DISCUSSION

Our approach provides a framework towards a functional understanding of endozoochory that considers independently the effects of gut passage, germination substrate, and the spatial pattern of seed dispersal. We found positive or neutral effects of gut passage and dispersal of seeds into suitable environments, but that feces were a poor germination substrate. The positive effects of gut passage and the frequency of mesquite seeds in feces suggest coyotes may contribute to mesquite dispersal. Gut passage through coyotes enhanced germination for both honey and screwbean mesquite, primarily through the removal of seeds from indehiscent legumes (Ortega Baes et al. 2001, 2002; Villagra et al. 2002; Robertson et al. 2006). For mesquite and other species with hard-coated seeds, seed liberation appears to be the main benefit of ingestion by many animals, including goats (Ortega Baes et al. 2002; Kneuper et al. 2003), donkeys (Ortega Baes et al. 2002), horses (Janzen 1981; Campos and Ojeda 1997), cattle (Cox et al. 1993; Campos and Ojeda 1997; Kneuper et al. 2003), sheep (Cox et al. 1993; Kneuper et al. 2003), rodents (Campos and Ojeda 1997), foxes (Campos and Ojeda 1997), and bears (Auger et al. 2002).

We found coyote gut passage to be more effective for enhancing germination of screwbean mesquite than honey mesquite. Only 5.2% of honey mesquite seeds were freed from their woody endocarps by coyote ingestion, while 36.9% of screwbean mesquite seeds were freed from their pericarps. Similarly, germination rates after coyote gut passage were substantially higher for screwbean mesquite than honey mesquite (Table 2; Fig. 2). Other studies have found similarly low germination rates of honey mesquite seeds after coyote gut passage (Kramp et al. 1998; Silverstein 2005). Coyote gut passage

18

also decreased viability of honey mesquite seeds, but not screwbean mesquite seeds. All together, low rates of seed liberation and germination and decreasing seed viability after gut passage suggest that coyotes may primarily act as seed predators rather than effective dispersers of honey mesquite. Conversely, high seed liberation and germination rates after gut passage suggest that coyotes may be effective dispersers of screwbean mesquite if seeds are later buried.

In contrast to the positive effects of gut passage, our results suggest that coyote feces themselves are not suitable for mesquite germination, as also observed by Kramp et al. (1998). Under field conditions, only two screwbean mesquite seedlings and no honey mesquite seedlings emerged from feces. The screwbean mesquite seedlings emerging from feces then died before the end of the growing season. Feces contained abundant seeds (Table 1), with many seeds remaining visibly ungerminated on the surface at the end of the growing season (pers. obs.). In comparison, 53.7% of screwbean mesquite seedlings and 40.9% of honey mesquite seedlings that emerged from soil survived until the end of the growing season. Seedling emergence was high in 2010 (pers. obs.), likely due to precipitation that was 52% higher than normal (Amargosa Valley weather station, Western Regional Climate Center), indicating that successful emergence from feces is likely a very rare event.

Alternatively, seeds that remain viable and dormant after gut passage could be incorporated into the seedbank and germinate at a later date. Mesquite seeds have physical dormancy and can maintain some viability after multiple years in the soil seedbank, even under conditions suitable for germination (Tschirley and Martin 1960). Coyotes preferentially deposited feces in environments that could be suitable for tree

19

establishment if seeds came into contact with a suitable substrate and dormancy was broken. In particular, we found that coyotes disproportionately deposited feces near surface water (e.g., Kays et al. 2008; Atwood et al. 2011), where soils are more likely to be moist and thus promote germination. Coyotes also deposited feces in vegetation types that include mesquite trees, indicating that these environments have been suitable for establishment in the past. For this landscape scale primary dispersal to result in successful establishment, some form of secondary dispersal is likely necessary to incorporate seeds into soil.

Janzen (1981) and Kneuper et al. (2003) suggest that secondary abiotic dispersal may result from fluctuations in environmental conditions, like moisture or temperature, or environmental processes, like erosion or chemical etching, that can decompose woody tissues and move seeds along the soil surface and within the soil profile. However, the efficacy of these abiotic processes for secondary dispersal is unknown and likely to be low. Because seeds must be freed from their woody pericarps for abiotic dispersal to occur and microbial decomposition is unlikely to free seeds of these species (Ortega Baes et al. 2001), animal ingestion by coyotes or other species may be a necessary first step towards germination. Further research is necessary to determine the efficacy of this pathway to germination.

Secondary animal dispersal is the more likely pathway by which viable seeds freed from their legumes could be incorporated into the soil seedbank and deposited in suitable substrates (Villagra et al. 2002; Vander Wall and Longland 2004). In particular, ant or rodent dispersal of seeds from coyote feces or trees to more suitable microsites could lead to high rates of establishment (Vander Wall and Longland 2004; Beck and

20

Vander Wall 2011). Kangaroo rats are particularly effective dispersers of honey mesquite and the morphologically similar velvet mesquite (P. velutina) through their caching of seeds in suitable microsites (Reynolds and Glendening 1949; Reynolds 1954). To our knowledge, no studies have examined rodent dispersal of screwbean mesquite. Seeds dispersed by kangaroo rats are cached in clumps, resulting in emergence of multiple seedlings within a small area (Reynolds and Glendening 1949). However, we did not observe a clumpy distribution of naturally regenerating seedlings at Ash Meadows. If coyotes participate in diplochory with other animal dispersers, an additional benefit of coyote dispersal could be reduced predation from bruchid beetles (Howe and Smallwood 1982; Chambers and MacMahon 1994; Miller 1994; Or and Ward 2003; Rodríguez-Pérez et al. 2011). We observed high rates of beetle predation in coyote feces (Table 1) and found that seeds infested with bruchid beetles were not able to germinate, regardless of ingestion by coyotes. Because beetles are likely to eventually infest all seeds remaining in legumes under trees (Miller 1994; Ortega Baes et al. 2001; Or and Ward 2003), coyote ingestion may reduce seed loss due to beetles. More research is needed to determine if secondary dispersal by ants and rodents of seeds from coyote feces to suitable microsites increases the efficacy of coyote dispersal of mesquite (e.g., Cox et al. 1993; Villagra et al. 2002; Vander Wall and Longland 2004).

CONCLUSIONS

To determine the functional effectiveness of endozoochorous seed dispersal mechanisms, it is critically important to consider the effects of gut passage, germination substrate, and the spatial pattern of seed dispersal with respect to environmental

21

suitability for seedling emergence. Yet few studies of endozoochory have done so. Here, we present methods that link mechanistic experiments on mesquite germination to spatial models of coyote dispersal. As our results demonstrate, positive effects of gut passage on germination are not necessarily indicative of effective animal dispersal because seeds may be deposited in locations where establishment is unlikely. Although coyotes preferentially deposit feces in environments suitable for establishment, coyote feces are not a suitable substrate for germination. Collectively, our results suggest that coyotes may be effective long distance dispersers of seeds away from beetle predation near the mother plant and into suitable environments at the landscape scale. More research is needed to determine if secondary dispersal moves seeds into suitable substrates and microsites for germination.