Effectiveness of temporal matching in ecological niche models: Insights for a low-dispersing species

Abstract

Ecological niche models, crucial for estimating the potential distribution of species under global change, can face reduced accuracy when the timing of occurrence data does not align with the environmental data. One solution is to ensure a close temporal match between the environment and the observation date. While this approach is typically recommended for highly mobile species, a few findings support its use for species with limited mobility, whose distributions may be responding to climate change via local population changes. Additionally, it remains unclear what specific temporal resolution could improve model performance. This study assesses the effectiveness of temporal matching for a species with low mobility, the Mexican small-eared shrew (Cryptotis mexicanus), by evaluating different temporal resolutions (one-, five-, and ten-year averaged environmental data) against the standard method (30-year). Occurrences between 1981 and 2010 were used for model training and cross-validation, while those outside this range were used for independent evaluation. To address the temporal bias in occurrence data, dates were assigned to all background points through geographic interpolation of observation dates of species that can be captured similarly to the shrew. Based on the omission rate of the independent evaluation occurrences, the approaches that matched environmental data performed better than the standard 30-year average approach, while the rest of validation metrics (for any temporal resolution) were not different. Visual inspection indicated that the geographic predictions resulting from time-matched approaches were as realistic as the one from the standard 30-year approach. The improved prediction of temporally independent occurrence data (not used in model training) with time-matched approaches underscores the practical advantage of this methodology for low-mobility species, enhancing model performance and geographic predictions, which may also improve forecasts for future environmental conditions. Additionally, this approach identifies a potential time lag between climatic changes and population responses in this species. Studies can select the optimal temporal resolution by exploring several or using available information about population responses to climate change.