BIMr2 : Inference of migration rates and their drivers from multilocus genotypes

Understanding how environmental factors drive recent migration is crucial for predicting species' responses to habitat change and for informing conservation strategies. There is one method available (BIMr) aimed at identifying the environmental factors associated with migration patterns using genetic data. While the Bayesian approach implemented in BIMr is theoretically robust, it can encounter difficulties estimating migration rates in datasets with many populations, hampering its application in many empirical scenarios. To address this limitation, we developed BIMr2, which implements a revised methodology while retaining the core advantages of the original program. Simulated datasets generated under the inference model and considering a range of scenarios support the robust performance of BIMr2 when the number of populations increases (from 5 to 20), given sufficient genetic differentiation between populations (pre‐migration ). The identification of environmental factors affecting migration and the accuracy of migration parameter estimates did in fact improve for larger simulated population numbers when holding the number of sampled individuals per population constant. Applying BIMr2 to a blue‐tailed damselfly dataset with 25 populations revealed that both geographic distance and negative tree‐cover gradients significantly reduced migration. Crucially, a positive distance tree cover interaction showed that the asymmetry in tree cover from source to recipient populations interacts with geographic separation to shape migration. Individuals tend to migrate less from high‐tree‐cover to low‐tree‐cover populations, but such effect is mitigated as distance grows, and conversely, the distance penalty is amplified when the source is less forested. BIMr2 resolves the convergence issues of its predecessor, enabling reliable inference of migration and its drivers even for tens of populations. In addition, by providing per–individual inbreeding estimates and better model fit metrics, the method offers a more insightful and robust statistical tool for landscape genetics, conservation planning, and studies of how rapid environmental change reshapes population structure and evolutionary dynamics.