High-severity wildfire in arid regions has caused ecological state change, transforming previously forested areas into shrublands. This dramatically alters the climatic envelope for tree seedlings, rendering the likelihood of returning post-wildlife landscapes to their previous state relatively low. We used a combination of sUAS imagery, satellite data and in-situ microclimate data recordings, together with a machine learning approach, to model monthly near-ground minimum, mean and max temperature as well as relative humidity and vapor pressure deficit in a previously forested area, which is now dominated by shrubs species. Spatially explicit models predicted recorded microclimate well (r = 0.73 to 0.97), and projections of models highlighted the solar buffering capacity of existing vegetation to alter the maximum temperature in the hottest month by ~12^oC, increase relative humidity by ~20% and reduced vapor pressure deficit by 0.3mbar in locations 2m apart. By harnessing these microclimate refugia, the success rate of reforestation efforts in post-wildfire landscapes could be substantially increased and mitigate seedlings from climate warming at local scales.

These data were collected using iButton temperature and relative humidity data loggers, Watchdog meteorological stations, and small unmanned aircraft systems.  The code to process the data are included as part of this dataset.