Using Biophysical Models to Improve Survey Efficiency for Cryptic Ectotherms

Inefficiencies in monitoring programs waste resources. Ideally, we would predict when and where target species are most detectable and place our effort accordingly. Statistical models can generate predictor functions relating survey conditions to detectability but are phenomenological; they do not incorporate biological constraints and so using them to predict into unsampled time and space is risky. Biophysical ecology allows us to place constraints on detection by identifying abiotic conditions in which the target species cannot be present. We show how such constraint can be incorporated into standard detection models. We use the striped legless lizard (Delma impar), a threatened cryptic species, in southeastern Australia, as a case study, using fortnightly monitoring data collected between June 2016 to May 2017. These lizards are monitored by searching under tiles placed in arrays; we used a biophysical model to simulate the thermal microclimate under tiles and combined this with the striped legless lizard's thermal physiology to predict when tiles could be used. When compared against a large monitoring dataset, lizards were rarely observed at times the model predicted they should not be present, but the model also over-predicted presences. A statistical occupancy model showed that much of this over-prediction is explained by a generally low detection probability, and a seasonal trend in detection beyond that captured in the biophysical model. We then replaced the temperature covariates of detection in the statistical model with our biophysical prediction (a categorical variable). The resulting model explained almost the same amount of variance in detections as the original model but using a single variable that captured biological constraints. The biophysical model allowed us to place mechanistic constraints on detection, but it still needed to be informed by observed aspects of the species' behavior that were not influenced by temperature. Hybrid statistical-biophysical models such as ours offer a powerful tool for forecasting optimal survey conditions for a wide array of species. © 2020 The Wildlife Society.