Modelling microclimatic variability in Andean forests of northern Patagonia.

Information on microclimatic conditions beneath canopies is key to understanding small-scale ecological processes, especially concerning the response of biodiversity to climate change. In north-western Patagonia, where data on climate-driven species distribution are scarce, our study provides valuable insights by providing microclimatic models covering spatiotemporal dynamics at 30 × 30 m resolution. Applying in-situ data from 2022 to 2024, we employed a random forest-based regression to assess the impact of several biophysical predictor variables describing terrain and vegetation properties on four microclimatic response variables at three vertical levels within forests. We also interpolated this data spatiotemporally, using statistical downscaling of ERA5 data. Our analysis reveals that the influence of the predictor variables varies strongly by month and response variable. Moreover, significant variability was observed between the models and months regarding their explanatory power and error range. For instance, the model predicting maximum air temperature at a 2 m height achieved an R² of 0.88 and an RMSE of 1.5 °C, while the model for minimum air temperature resulted in an R² of 0.73 and an RMSE of 1.8 °C. Our model approach provides a benchmark for spatiotemporal projections in this data-scarce region, aligned with the climate normal from 1981 to 2010. Future refinement could benefit from data on snow cover, land use and land cover, soil, as well as structural information on vegetation over an extended period, to enhance the dynamical aspects of microclimatic modelling. We are confident that our present model will substantially enhance possibilities to analyse species distribution across vegetation types and terrain-related features within the area.