Consequences of ignoring variable and spatially autocorrelated detection probability in spatial capture-recapture

Abstract Context Spatial capture-recapture (SCR) models are increasingly popular for analyzing wildlife monitoring data. SCR can account spatial heterogeneity in detection that arises from individual space use (detection kernel), variation the sampling process, and distribution of individuals (density). However, unexplained unmodeled detectability may remain due to cryptic factors, both intrinsic extrinsic study system. This is case, example, when covariates coding variable effort probability general incomplete or entirely lacking. Objectives We identify how magnitude configuration unmodeled, spatially influence parameter estimates. Methods simulated data with autocorrelated probability. then fitted an model ignoring this assessed impact misspecification on inferences. Results Highly-autocorrelated (Moran’s I = 0.85–0.96), modulated by heterogeneity, lead pronounced negative bias (up 65%, about 44-fold decrease compared reference scenario), reduction precision (249% 2.5-fold) coverage 95% credible intervals associated abundance estimates 0. Conversely, at low levels autocorrelation (median Moran’s 0), even severe did not only caused slight reductions Conclusions Unknown liable be norm, rather than exception, studies. encourage practitioners consider has their inferences urge development methods take structured, unknown partially variability into account.