Inference over radiative transfer models using variational and expectation maximization methods

Earth observation from satellites offers the possibility to monitor our planet with unprecedented accuracy. Radiative transfer models (RTMs) encode energy through atmosphere, and are used model understand system, as well estimate parameters that describe status of satellite observations by inverse modeling. However, performing inference over such simulators is a challenging problem. RTMs nonlinear, non-differentiable computationally costly codes, which adds high level difficulty in inference. In this paper, we introduce two computational techniques infer not only point estimates biophysical but also their joint distribution. One them based on variational autoencoder approach second one Monte Carlo Expectation Maximization (MCEM) scheme. We compare discuss benefits drawbacks each approach. provide numerical comparisons synthetic simulations real PROSAIL model, popular RTM combines land vegetation leaf canopy analyze performance approaches for modeling inferring distribution three key quantifying terrestrial biosphere.