The primary objective of project RC-201205 was to demonstrate how noninvasive genetic sampling (NGS) could be combined with capture-recapture modeling (NGS-CR) to evaluate the status of species of conservation concern. A secondary objective was to demonstrate the combination of NGS with occupancy modeling (NGS-OM) to estimate the proportion of area occupied (i.e., occupancy) and patterns of local extinction and colonization. The researchers evaluated the efficacy of NGS as a viable, long-term monitoring approach for two species on the Department of Defense (DoD) installations: the kit fox (Vulpes macrotis) (Dugway Proving Ground [DPG]), a species of concern for western installations, and Sonoran pronghorn (Antilocapra americana sonoriensis) (Barry M. Goldwater Range [BMGR]), an endangered subspecies of North American pronghorn that occurs in southern Arizona. For both species, researchers developed a spatio-temporal sampling design for acquiring noninvasive genetic data (via fecal scats), genotyped samples for individual ID, analyzed genotypes with capture-recapture methods to obtain estimates of population parameters, and developed a protocol for long-term monitoring in the future. Researchers also quantified expenditures to examine cost efficiency of the approach. Additionally, researchers evaluated NGS-OM only for kit foxes, and its sympatric intraguild predator, the coyote (Canis latrans). Researchers monitored kit foxes and coyotes simultaneously, used genetic analyses (via scats) to confirm species, and employed dynamic occupancy modeling to obtain estimates of detection, proportion of area occupied, and local colonization and extinction for each species, and to evaluate the influence of coyotes and landscape features on kit fox space-use. 

The performance objectives for this project were to (1) improve monitoring protocols for kit foxes and Sonoran pronghorn based on NGS-CR, (2) obtain reliable estimates of demographic parameters from NGS-CR for each species, (3) improve efficiency of current monitoring programs, (4) evaluate ease of use, (5) obtain estimates of occupancy and dynamic parameters (i.e., local colonization and extinction) from NGS-OM for kit foxes, and (6) facilitate transference of monitoring programs for kit foxes and Sonoran pronghorn based on NGS-CR. 

Capture-recapture modeling has been commonly used for estimating wildlife population parameters. The theory is based on modeling capture and recapture probabilities of populations or individuals as a function of population size, survival, reproduction, and movements among populations. The process involves capturing individuals and marking them, such that on subsequent capture occasions, marked individuals can be identified. Using the observed capture histories, demographic parameters, such as abundance, survival, reproduction, immigration, or emigration, can be estimated. Researchers employed Pollock’s robust design (Pollock et al. 1990, Kendall et al. 1997) capture-recapture models. Additionally, they employed single session ‘capture with replacement’ (CAPWIRE) models. CAPWIRE exploits repeat detections of individuals within a single sampling occasion to generate abundance estimates (Miller et al. 2005). For kit fox, they also employed spatially explicit capture-recapture (SECR) models.

Occupancy modeling utilizes information from repeat surveys to account for imperfect detection and produces unbiased estimates of occupancy (MacKenzie et al. 2002, 2003, 2006). Unlike NGS-CR, the unit of analysis in occupancy studies is the survey site (or patch) not the individual. Consequently, patterns of occurrence can be modeled as a function of patch characteristics, such as habitat or landscape features. Replication is required to estimate probability of detection for occupancy models and may be accomplished through temporal or spatial replicates within a site.  One benefit of NGS-OM is that it requires only species ID of noninvasively collected genetic samples, and subsequently may offer a more affordable monitoring strategy if estimates of abundance and survival are not required. 

All performance objectives were met. Across sessions, 109 kit foxes were identified.  Researchers captured 36–50 kit foxes each session. They captured more males (60%) than females. Male kit fox survival (S_M) was slightly lower than female survival (S_F) across intervals and overall. Model-averaged kit fox survival was high in the period between winter 2013 and summer 2013 (S_M = 0.82, 95% CI = 0.26–0.98; S_F = 0.87, 95% CI = 0.28–0.99), high between summer 2013 and winter 2014 (S_M = 0.81, 95% CI = 0.19–0.98; S_F = 0.87, 95% CI = 0.24–0.99), and lower in the interval from winter 2014 to summer 2014 (S_M = 0.59, 95% CI = 0.11–0.94; S_F = 0.67, 95% CI = 0.16–0.96). Estimates of kit fox density from SECR models were similar across sessions (0.018–0.022 animals/km²); these estimates were among the lowest reported in the literature and at DPG. Derived estimates of kit fox abundance from SECR models were generally higher than those from robust design non-spatial models. The model-averaged abundance estimates from robust design non-spatial models indicated that there were 60.1–73.2 kit foxes in the study area and 95% confidence intervals suggested that population abundance was stable across sessions. Naïve estimates of coyote occupancy were >0.7 in all but the first session and probability of occurrence was not significantly different from 1. For kit foxes, naïve estimates of occupancy were ≤0.3, with the probability of occurrence estimated to be ≤0.5. Coyote occupancy was unrelated to water availability, but was positively related to the proportion of shrubland and woodland habitat. Kit fox occupancy displayed an inverse relationship, being negatively related to shrubland and woodland habitat. Kit fox probability of local extinction was positively related to site-level coyote activity, and within an occupied site, the probability of kit fox detection was positively related to transect-level coyote activity.

Researchers estimated abundance for Sonoran pronghorn in 2013 and 2014 and annual survival between 2013 and 2014. The population using developed water holes (drinkers) was 116 (95% CI: 102–131) and 121 (95% CI: 112–132) in 2013 and 2014. The combined population estimate for drinker and non-drinker locations was 144 (95% CI: 132–157). Adults had higher annual survival probabilities (0.83, 95% CI: 0.69–0.92) than fawns (0.41, 95% CI: 0.21–0.65). Simulations were used to evaluate empirical estimates and evaluate study design tradeoffs. The simulation results indicate the empirical estimates are reliable. Cost per individual monitored in 2014 was ~$184 USD for NGS-CR methods and $599 USD for aerial sight ability methods. However, the results indicate that at the current estimated abundance (~200), the same level of precision (aerial CV ~ 21%) can be obtained using NGS-CR methods for ~$5800, or an annual cost savings of over $4000 (Woodruff et al. In Review).

In transferring this technology to other installations, the researchers see the following challenges: 1) unpredictable weather and land access limitations can lead to insufficient sampling, 2) laboratories that can do these analyses need to be identified, and 3) experts will need to be identified to conduct quantitative analyses if the necessary expertise is not present within the DoD management team at the implementing installation.