Thermal sensitivity of western Canadian ecosystems: a remote-sensing based vulnerability analysis

We applied the approach of Ermida et al. (2020) to quantify land surface temperature (LST) using high-resolution satellite products (LANDSAT 5, 7, 8, 30-m resolution). We then calculated the historical LST trend during summer periods from 1985 to 2020 (beta coefficient associated with the year effect). To calculate the thermal sensitivity of a given location, we combined high-resolution LST with historical air temperature from ERA5 (Muñoz Sabater, 2019). Thermal sensitivity was calculated as the beta coefficient in a linear model relating daily ERA5 air temperature with LST over time.

We then fit a series of random forest models (Breiman, 2001) to identify the drivers of (1) LST trend and (2) thermal sensitivity using a collection of remotely sensed topographic, climatic, soil and land cover variables in Google Earth Engine (Gorelick et al. 2017). Model development was conducted using randomforest package (Breiman et al., 2018) in R version 4.4.1 (R Core Team 2024). Final models were used to create predictive maps of thermal refugia across western Canada