Temperature and Heat Stress in Forests

Temperature shifts as a result of climate change and deforestation are set to increase the risk of heat stress in many tropical forests[1].

Particularly in a context of deforestation where forest is less well connected, understanding the effect of increased temperature at a high resolution is key to predicting the level of heat stress such trees experience.

To this end the project will explore upscaling of remotely sensed temperature data to achieve better coverage in both spatial and temporal scale.

This will build on recent work by some of the supervision team on upscaling MODIS Aqua temperature data (1 km resolution, ~1:30pm pass when forests experience maximum stress) using a range of data sources at higher resolution but different pass time, e.g. Landsat at 30m and ~10am pass, using a machine learning model.

The project would add to this by extending the temporal and spatial extent studied using a similar upscaling and exploring other possible models for upscaling temperature (for example spatial models, GANs, or CNNs), and the potential for inpainting in scenes containing cloud cover[2].

Critically the project will then also relate this to field and UAV data to ground truth the maps, and match pixels to tree-level stress noting the challenges of doing this[3]