The influence of localized habitat features on mongoose (Herpestes auropunctatus) trapping success in eastern Puerto Rico

Background/Question/Methods

Small mammals are considered one of the most detrimental biological invaders in island ecosystems. Conservation organizations and government agencies allocate a substantial amount of resources to manage these invasive species. However, lack of published information on the effectiveness of trapping and control techniques makes it difficult for conservation managers to devise population control strategies. I studied the introduced small Indian mongoose (Herpestes auropunctatus) population in eastern Puerto Rico to facilitate the implementation of trapping programs.  I quantified the influence of localized habitat features on individual trap success in 5 different forest types (4 in El Yunque National Forest and 1 in the Northeast Ecological Corridor). At each trap we collected vegetation information (e.g., overstory canopy, downed wood, understory cover) and calculated distances (m) to: coastal shoreline, trails, roads, rivers, and recreation areas.  I also included elevation (m) for each trap location. I developed a candidate model set (each model contained uncorrelated variables) and estimated the likelihood of capturing a mongoose at a trap location using logistic regression.  I included a random effect for each trapping grid to account for spatial autocorrelation among traps within the same grid.

Results/Conclusions

Cover estimates differed among locations (understory cover, F₄=8.4, P<0.001; overstory canopy cover, F₄=13.1, P<0.001; and woody debris, F₄=14.3, P<0.001) but that within a grid variability was low (SE<10%). On average, traps were located closer to roads, recreational areas, coastal shorelines, and trails when compared to the broader landscape, but farther from rivers however these measurements for any given covariate were highly variable.  I trapped 34 mongooses and recaptured 4 marked individuals. I found 4 competing models for describing the likelihood of capturing a mongoose at a trap location that included positive relationships to distances from rivers and recreational areas and canopy cover but negative relationships for distances to trails and coastal shoreline. The top-ranking model (27% AIC_wt) included proximity to rivers and this parameter was significant (P=0.003). Models revealed that vegetation features in the vicinity of traps had no influence on the likelihood of catching a mongoose. Rather, I found significant support that distance metrics were the best predictors of mongoose capture probability within a trap grid. The ability to predict where to place traps and monitoring trapping outcomes are important for reducing efforts and costs and measuring progress towards the management goal.