Bayesian species distribution models integrate presence‐only and presence–absence data to predict deer distribution and relative abundance

Abstract

Using geospatial data of wildlife presence to predict a species distribution across a geographic area is among the most common tools in management and conservation. The collection of high‐quality presence–absence (PA) data through structured surveys is, however, expensive, and managers usually have access to larger amounts of low‐quality presence‐only (PO) data collected by citizen scientists, opportunistic observations and culling returns for game species. Integrated species distribution models (ISDMs) have been developed to make the most of the data available by combining the higher‐quality, but usually scarcer and more spatially restricted, PA data with the lower‐quality, unstructured, but usually more extensive PO datasets. Joint‐likelihood ISDMs can be run in a Bayesian context using integrated nested laplace approximation methods that allow the addition of a spatially structured random effect to account for data spatial autocorrelation. Here, we apply this innovative approach to fit ISDMs to empirical data, using PA and PO data for the three prevalent deer species in Ireland: red, fallow and sika deer. We collated all deer data available for the past 15 years and fitted models predicting distribution and relative abundance at a 25 km2 resolution across the island. Model predictions were associated to spatial estimate of uncertainty, allowing us to assess the quality of the model and the effect that data scarcity has on the certainty of predictions. Furthermore, we checked the performance of the three species‐specific models using two datasets, independent deer hunting returns and deer densities based on faecal pellet counts. Our work clearly demonstrates the applicability of spatially explicit ISDMs to empirical data in a Bayesian context, providing a blueprint for managers to exploit unexplored and seemingly unusable data that can, when modelled with the proper tools, serve to inform management and conservation policies.

Bayesian areal disaggregation regression to predict wildlife distribution and relative density with low-resolution data

“…We found that our model predictions were highly correlated with the estimates provided by Morera-Pujol et al (2022) and with field survey density estimates from 61 sites across Ireland. We compared Morera-Pujol et al ( 2022) estimates of deer abundance at the pixel level with our model predictions for relative density and found correlation coefficients of 0.76 for red deer, 0.60 for sika deer and 0.79 for fallow deer.…”

“…Secondly, we obtained faecal pellet group count density estimates from 2019 by an expert surveyor, full details of data collection are detailed in Morera-Pujol et al (2022). All three species were target species for survey within 61 forest properties across Ireland.…”

“…Conflict species such as deer that have successfully navigated human modification and expanded out of their historic range, must be studied to understand and prepare for future land-use change scenarios e.g., increased afforestation, farm abandonment, rewilding and continued encroachment of human settlements that have been observed worldwide (Hoogendorm et al 2019;Castillo et al 2021). Information on relative population size and distribution is essential to the conservation and management of wildlife and can form the foundation of higher-resolution research programmes to further refine management objectives and decision-making protocols (Gillson et al 2019;Morera-Pujol et al 2022). Where high-resolution data are not available to analyse, it may be necessary to use lower-resolution data to obtain baseline information on species distribution and relative population density.…”

“…Since models were fitted in a Bayesian context through the INLA package, the prediction obtained at each location is not a point value but a distribution, from which we can produce the mean and corresponding 95% confidence intervals, thus obtaining a spatial estimate of the uncertainty of the prediction. Finally, to evaluate the performance of the model, we analysed the correlation (see below for full details) between the spatial predictions of our Bayesian areal disaggregation regression model and those from a concurrent analysis of deer distribution and relative density in Ireland, ran by Morera-Pujol et al (2022); as a second and independent evaluation method, we analysed the correlation between the spatial predictions from our year 2018 models against independent field survey data capturing relative deer density collected in forest properties across Ireland in 2019. The methodologies used to obtain the data we challenged our predictions against are described below.…”

“…Using habitat classification data from publicly available remote sensing data, we aimed to map the changing distribution and relative abundance of deer in Ireland from 2000-2018. To evaluate the robustness of the predictions and demonstrate the reliability of Bayesian areal disaggregation regression to analyse regional count data in ecology, we challenged our model predictions by assessing their correlation with two independent datasets of deer distribution and relative abundance in Ireland using (a) predictions of deer distribution and relative abundance modelled by Morera-Pujol et al (2022) using integrated species distribution models (therefore combining both presence-only and presence-absence data), and (b) faecal pellet group count data collected in field sites across Ireland.…”