Forest fragmentation alters winter microclimates and microrefugia in human‐modified landscapes

With over half of earth's terrestrial biota living beneath forest canopies, our ability to accurately capture organism–climate relationships in forested ecosystems is imperative for predicting species' vulnerability to future climate change. Assessing the vulnerability of forest dependent species, however, hinges on quantifying microclimates that exist below the forest canopy and might be influenced by varying levels of disturbance in human‐modified landscapes. The goal of our study was to examine the multi‐scaled predictors of subcanopy microclimate variability across a heterogeneous landscape in Midwestern USA during winter, and to further evaluate whether a widely available interpolated climate model accurately captures this variability. By deploying a network of temperature sensors along a fragmentation gradient, we found that forests in more fragmented landscapes with greater amounts of forest edge and increasing distances between forest patches, experienced colder minimum and average daily temperatures throughout the winter than forests in less fragmented landscapes. We found that greater tree densities and higher elevations led to warmer microclimates while increasing distances from urban centers led to colder microclimates. The negative effect of forest edge on minimum temperatures was lessened by the effect of increasing basal area, highlighting the importance of local‐ and landscape‐scale features on microclimate heterogeneity. Temperature discrepancies between subcanopy microclimates and climate interpolations were influenced by many of the same features, and could be of a similar magnitude as those predicted by future climate change scenarios. Using a biological threshold based on metabolic and demographic constraints for winter birds, we found that the variability in microclimates along our forest fragmentation gradient (50 km) was comparable to the magnitude captured by weather stations across a latitudinal gradient spanning more than 650 km. Our results suggest that biophysical properties of landscapes can alter spatial gradients of microclimates and should be considered when assessing species' vulnerabilities to future climate change.