Microwave Remote Sensing for soil moisture estimation and vegetation characterization with Physics-Informed Machine Learning

Soil moisture and vegetation parameters are essential elements in understanding the land-atmosphere interactions. Traditional method such as in-situ measurements are confined in the spatial and temporal domains for monitoring purposes, and generally they are time- and labor-consuming. Synthetic-aperture radar (SAR) and microwave radiometers have been successfully applied in estimating soil moisture and vegetation properties, due to their sensitivities to soil moisture and plant growth and their working capabilities irrespective of cloud and day/night conditions.

The research aims of this study are:

• To develop an emulator and carry out backscattering coefficient at C-band and Brightness Temperature at L-band forward simulations using machine learning algorithm, evaluate the accuracy of the forward modeling via comparisons to ground-based and satellite-based observation,
• To retrieve the soil moisture and vegetation parameters including LAI, VWC, AGB and plant height using multi-target regression method and microwave signal simulation as input, evaluate the performance of the regression model against the in-situ measurements and inter-compare the soil moisture and vegetation parameters estimation with other global products.
• To evaluated the applicability and performance of these methods in some specific study areas.