Improved neural network scatterometer forward models

Current methods for retrieving near-surface winds from scatterometer observ ations over the ocean surface require a forward sensor model which maps the wind vector to the measured backscatter. This paper develops a hybrid neur al network forward model, which retains the physical understanding embodied in CMOD4, but incorporates greater flexibility, allowing a better fit to t he observations. By introducing a separate model for the midbeam and using a common model for the fore and aft beams, we show a significant improvemen t in local wind vector retrieval. The hybrid model also fits the scatterome ter observations more closely. The model is trained in a Bayesian framework , accounting for the noise on the wind vector inputs. We show that adding m ore high wind speed observations in the training set improves wind vector r etrieval at high wind speeds without compromising performance at medium or low wind speeds.