Fig. 1: Conceptual definition of species shadow and expected distribution and an overview of our XAI workflow applying SHAP values to species distribution models.

The top panel (1) shows the spatial relationship of an ecological property (e.g., species occurrence, presence-absence, or abundance) across a geographic domain depending on various environmental variables (with environmental suitability indicated from red to green). Locations labelled a-d represent contrasting locations within a species’ geographic range, each with a different explanation for the local environmental suitability scores \({{{\rm{y}}}}_{i}\) defined by model function \(f\left({x}_{i}\right)\). The locations inside the expected distributions (\(E\)) and shadow distribution (\(S\)) are indicated by coloured borders. The middle panel (2) shows a theoretical biplot of species distributions depending on abiotic and anthropogenic contributions to environmental suitability. The binary expected distribution is defined where abiotic factors have positive SHAP values (grey dashed lines), while the observed distribution is where both abiotic and anthropogenic contributions are positive. This observed distribution is typically identified by SDMs that do not account for anthropogenic effects. The shadow distribution comprises areas where abiotic factors have positive SHAP values, but anthropogenic factors have negative SHAP values, as defined by Eqs. 1–4. Conservation strategies are suggested for each distribution. Quantitative definitions on the right of the panel (2) indicate that locations c and d fall within the observed distribution, but anthropogenic impacts reduce the expected distribution to the shadow distribution. Lower panel (3) shows how SHAP values categorise areas as inside or outside of shadow and expected distributions. a shows a site where negative SHAP values for natural abiotic factors reduced environmental suitability from the expected value (\(E\)), indicating it is outside of the species’ expected distribution. b shows a site with a positive SHAP value for natural abiotic factors, indicating the species is naturally present. c, d represent a site where the positive effects of natural abiotic factors are countered by the negative effects of anthropogenic factors placing the site in the ‘shadow distribution’. Notably, (a) and (b) fall outside of the shadow distribution because (a) is not inside the expected distribution and (b) does not have negative anthropogenic SHAP values.