Spatial Capture-recapture with Memory: A New Hidden Markov Model Perspective

Spatial capture-recapture (SCR) surveys are widely used to answer questions in population ecology, for species ranging from slow worms to snow leopards. They are used to estimate the total size (or density) of the population within the given area, for example, or to estimate relationships between density and environmental drivers of density, or to estimate spatial and temporal trends in abundance. However, existing statistical methods for analysing SCR data ignore the fact that detections of animals that are made close together in time, are also likely to be close together in space. Existing SCR methods neglect this spatio-temporal correlation, and this can result in substantial negative bias in estimates of population abundance, sometimes by as much as 50%.

We will develop and test a new spatial capture-recapture method that explicitly takes into account the spatio-temporal dependence in detections, by modelling the probability that an individual is observed in any trap as a function of the distance in both space and time from its last detection. This will provide more robust estimates than can be obtained using current SCR methods. The new method and associated software will be developed in collaboration with the Global Snow Leopard Ecosystem Protection Programme (GSLEP), which is coordinating the world's first range-wide survey of snow leopards. The new methods will be applied to camera trap surveys of snow leopards that form a key part of the GSLEP survey initiative. The new SCR models we develop will, however, have much wider utility than this survey alone. They are applicable across a vast range of species ranging from, for example, large mammals (e.g. tigers, bears) and primates (e.g. gibbons, chimpanzees) to small reptiles and amphibians (e.g. salamanders, frogs, slow worms) and birds (e.g. songbirds, grouse).