Estimating and Projecting Environmental Indicators using Satellite Data and Machine Learning

Data collected by Earth observation satellites are important information sources about the environment and state of our planet. Deep learning techniques provide ways to use this data to model environmental indicators, thus are useful in systems that alert us to potential environmental disasters. This thesis advances deep learning techniques to monitor the moisture content of vegetation, which is important for wildfire risk monitoring, by designing models that make accurate, timely, and extensive predictions. These models can potentially assist wildfire management authorities to develop plans mitigating the effects of catastrophic wildfires, thus helping to protect our environment and save human lives.