Fine-scale microclimate data improve species distribution models of forest plant species

In recent decades, species distribution models (SDMs) have become pivotal in forecasting how changing environmental conditions impact species distributions across space and time. Most SDMs rely on correlations, utilizing statistical or machine-learning techniques to infer links between species occurrences and their environment. Typically, these models rely on a traditional set of bioclimatic variables, often available at a coarse spatial resolution of 30 arc seconds or less. These macroclimatic data are derived through interpolating weather station data, essentially reflecting the free-air temperature conditions in open ecosystems. However, a significant portion of terrestrial life on Earth and many critical ecological processes respond to climate conditions at much finer scales beneath the canopies of trees. Neglecting this mismatch might lead to inaccurate predictions, misinterpretations, and potentially flawed conservation decisions. Hence, there's a pressing need to incorporate finer-scale microclimate data in ecological modelling to ensure more accurate assessments and informed conservation strategies.

By developing an innovative spatial machine learning model capable of quantifying the temperature buffering capacity of European forests at very fine resolutions, we crafted the ForestClim database, containing a novel set of bioclimatic variables. These variables unveil the intricate microclimatic temperature variations within forest ecosystems, marking a significant scientific breakthrough that shows promise to enhance ecological models and predictions. Furthermore, by freely sharing this data, we lower the threshold for fellow ecologists to incorporate pertinent microclimatic information into their research.

Leveraging the open-access ForestClim database, we further assessed how large-scale, gridded microclimate temperature data affect the accuracy of SDMs of European forest plant species and how their modelled environmental niches and projected geographic ranges differ from conventional SDMs. The study's findings demonstrate that SDMs based on microclimate significantly outperform their macroclimate-based counterparts. They also reveal the introduction of a systematic bias in thermal response curves when relying on macroclimate-based models, potentially leading to inaccuracies in forecasting range shifts. Furthermore, the inclusion of microclimate data in these models enables the identification of microrefugia within the landscape - areas where species can find a stable and suitable climate amid unfavourable, changing macroclimatic conditions. This newfound information holds particular significance in the realm of conservation science, as microclimate-based SDMs prove to be valuable tools for gaining insights into biodiversity conservation in the face of climate change. This is especially pertinent given the increasing policy and management emphasis on conserving refugia worldwide.