Introduction
Ecological niche models (ENMs) are commonly used to estimate the potential range of species based on the association between occurrence sites and their environmental conditions (Peterson 2001). Under the traditional ENM framework, occurrences are associated with a single long-term environmental average without considering the temporal match of the observation date and the period of the environmental baseline (Ingenloff and Peterson 2021, Stewart et al. 2021). This temporal occurrence-environment mismatch may result in less accurate distribution models when occurrences do not align with the temporal range of environmental data, thereby impacting estimations of how biodiversity will respond to environmental change (Roubicek et al. 2010, Nadeau et al. 2017). While constraining the temporal range of the occurrence data may help address this issue (Fig. 1a), practical applications are challenging due to several factors. Examples of these factors include the usually limited availability of occurrence data within long-term temporal ranges (Roubicek et al. 2010), the need for methodologies that incorporate temporal variability in model building (Milanesi et al. 2020, Ingenloff and Peterson 2021, Pang et al. 2022, Dobson et al. 2023), and the lack of guidelines to obtain environmental variables reflecting non-traditional temporal ranges.
ENMs commonly rely on climatic variables, such as temperature and precipitation, usually derived from the interpolation of monthly data obtained by weather stations (e.g., Climatic Research Unit dataset; Harris et al. 2020). Although these data are available for more than the past 50 years, climatic variables are typically summarized as a single period that corresponds to a climatological standard normal, a 30-year average in which the last year of the period ends with a zero (e.g., 1971-2000). Bioclimatic variables in global climatic datasets are created using these long-term periods as reference (Karger et al. 2017, Fick and Hijmans 2017, Abatzoglou et al. 2018). Unfortunately, using a single average can hide trends and variability in climate (Zimmermann et al. 2009, Ingenloff and Peterson 2021, Perez-Navarro et al. 2022, Pinilla-Buitrago 2023) or could fail to represent the current climate due to recent changes in temperature and precipitation (Livezey et al. 2007, Arguez and Vose 2011, Wilks and Livezey 2013). One approach to address these limitations is to incorporate explicit temporal variables, such as extremes during the reference period (Zimmermann et al. 2009, Moran-Ordonez et al. 2018, Stewart et al. 2021) and inter-annual variability variables (Zimmermann et al. 2009, Brodie et al. 2021, Gardner et al. 2021). However, this may not completely solve the issue of occurrence-environment mismatch since occurrences are associated with the climatic variability or extreme values that occurred after their observation date, especially for records from the beginning of the reference period (Fig. 1a).
ENMs that account for temporal mismatches have improved model performance by aligning the weather or climatic variability with the precise location and time of the occurrence data. These frameworks, known as dynamic ENMs (Milanesi et al. 2020, Dobson et al. 2023) or time-matched ENMs (Peterson et al. 2005, Ingenloff and Peterson 2021), establish correlations between occurrences and atmospheric conditions over temporal resolutions ranging from days (Abrahms et al. 2019), months (Reside et al. 2010, Welch et al. 2018, Andrew and Fox 2020, Ingenloff and Peterson 2021), seasons (Williams et al. 2017), years (Bateman et al. 2012, VanDerWal et al. 2013), or even a decade (Smith et al. 2019), and are designed to match the species’ response time to the changing variable (referred to in this study as a time-matched approach; Fig. 1b). For example, Abrahms et al. (2019) used tracked satellite data to create daily suitability maps of blue whales. In another study, monthly to yearly variability in temperature and precipitation improved predictions of geographic ranges in highly mobile birds (Reside et al. 2010, VanDerWal et al. 2013, Williams et al. 2017, Andrew and Fox 2020, Ingenloff and Peterson 2021).
While most studies using time-matched ENMs focus on species with high mobility (the ability of individuals to move or spread in the environment), only a few studies have addressed its utility for models of species ranges with limited dispersal capacity. This is particularly relevant for non-volant small mammals, which may exhibit greater sensitivity to environmental variation due to their slower ability to track suitable conditions (Schloss et al. 2012). For instance, Smith et al. (2019) used a time-matched approach of ten years to generate models of American Pika distribution. Similarly, Bateman et al. (2012) used a five-year matching to uncover competitive interactions between two bettong species, small marsupials in Australia, in which the time-matched approach was able to define better range edges. These studies are some of the few examples in which a temporal resolution longer than 12 months is used. Of these, Bateman et al. (2012) is the only study that compares against a long-term average. Despite the potential advances in time-matched ENMs, it remains unclear whether and in which kinds of cases different temporal resolutions of the environment (ranging from one year to less than 30 years) could perform better or as accurately as traditional ENMs that use long-term averages.
This study aims to gain insights into the effectiveness of matching the timing of occurrences and environmental data in the modeling framework in a species with low mobility. It evaluates the performance and ecological plausibility of different temporal resolutions of occurrence-environment matching to predict the potential distribution of a non-volant small mammal, the Mexican small-eared shrew Cryptotis mexicanus , in the current time. Additionally, the results of this time-matching method were compared with a standard 30-year average model and differences in the prediction of the current distribution ofC. mexicanus are explored.