Search

Predictive species distribution models (SDMs) are becoming increasingly important in ecology, in the light of rapid environmental change. However, the predictions of most current SDMs are specific to the habitat composition of the environments in which they were fitted. This may limit SDM predictive power because species may respond differently to a given habitat depending on the availability of all habitats in their environment, a phenomenon known as a functional response in resource selection. The Generalised Functional Response (GFR) framework captures this dependence by formulating the SDM coefficients as functions of habitat availability. The original GFR implementation used global polynomial functions of habitat availability to describe the functional responses. In this study, we develop several refinements of this approach and compare their predictive performance using two simulated and two real datasets. We first use local radial basis functions (RBF), a more flexible approach than global polynomials, to represent the habitat selection coefficients, and balance bias with precision via regularization to prevent overfitting. Second, we use the RBF-GFR and GFR models in combination with the classification and regression tree CART, which has more flexibility and better predictive powers for non-linear modelling. As further extensions, we use random forests (RFs) and extreme gradient boosting (XGBoost), ensemble approaches that consistently lead to variance reduction in generalization error. We find that the different methods are ranked consistently across the datasets for out-of-data prediction. The traditional stationary approach to SDMs and the GFR model consistently perform at the bottom of the ranking (simple SDMs underfit, and polynomial GFRs overfit the data). The best methods in our list provide non-negligible improvements in predictive performance, in some cases taking the out-of-sample R2 from 0.3 up to 0.7 across datasets. At times of rapid environmental change and spatial non-stationarity ignoring the effects of functional responses on SDMs, results in two different types of prediction bias (under-prediction or mis-positioning of distribution hotspots). However, not all functional response models perform equally well. The more volatile polynomial GFR models can generate biases through over-prediction. Our results indicate that there are consistently robust GFR approaches that achieve impressive gains in transferability across very different datasets. [ABSTRACT FROM AUTHOR]

In ecology, the observation process (how we collect data) can be as complex as the biological process we are investigating. Failure to account for complex observation processes leads to uncertainty, biased inference and poor predictions, resulting in misleading research results. Often, field scientists are best placed to describe observation problems that occur but are excluded from discussions about how to tackle these problems statistically. Statisticians are often unaware of the nuances of observation processes leading to the problems being ignored, or tackled on a case-by-case basis. We propose a typology of observation problems and inferential solutions, hence facilitating the linkages between field protocols and statistical treatments. Advances in statistics mean that it is now possible to tackle increasingly sophisticated observation processes. The intricacies and ambitious scale of modern data collection techniques mean that this is now essential. Methodological research to make inference about the biological process while accounting for the observation process has expanded dramatically, but solutions are often presented in field-specific terms, limiting our ability to identify commonalities between methods. We suggest a typology of observation processes that could improve translation between fields and aid methodological synthesis. We propose the LIES framework (defining observation processes in terms of issues of Latency, Identifiability, Effort and Scale) and illustrate its use with both simple examples and more complex case studies. [ABSTRACT FROM AUTHOR]

High pathogenicity avian influenza virus (HPAIV) caused the worst seabird mass-mortalities in Europe across 2021–2022. The northern gannet (Morus bassanus) was one of the most affected species, with tens of thousands of casualties in the northeast Atlantic between April–September 2022. Disease outbreaks can modify the movement ecology of animals by diminishing spatial consistency, thereby increasing the potential for disease transmission. To detect potential changes in movement behaviour, we GPS-tracked breeding adults following the initial HPAIV outbreak, at three of the largest northern gannet breeding colonies where major mortality of adults and chicks occurred (Bass Rock, Scotland, UK; Grassholm, Wales, UK; Rouzic, Brittany, France). We also gathered background epidemiological information and northern gannet colony dynamics during the outbreak. Our data indicate that HPAIV killed at least 50 % of northern gannets, and suggest the presence of HPAIV H5N1 antibodies in juveniles. GPS-tracked adult northern gannets remained faithful to their breeding sites despite the HPAIV outbreak and did not prospect other breeding colonies. They performed regular foraging trips at sea, similar to their behaviour before the outbreak. Comparison with GPS-tracking data gathered in 2019, i.e. before the HPAIV outbreak, suggested lower foraging effort in birds which survived HPAIV in 2022, potentially as a consequence of reduced intra- and interspecific food competition. Breeding colony fidelity of surviving adult northern gannets following HPAIV mass-mortalities indicates limited capacity for viral spread during our study. This may contrast with the behaviour of adults during the initial disease outbreak, and with that of younger individuals. [ABSTRACT FROM AUTHOR]