Forman, B. A., and R. H. Reichle:
"Estimating Passive Microwave Brightness Temperature over Snow-covered Land in North America Using a Land Surface Model and a Support Vector Machine"
Presentation at the AGU Fall Meeting, San Francisco, CA, USA, 2013.

Abstract:
Recent research by Forman et al. [2013, IEEE] demonstrated the ability of an artificial neural network (ANN) to predict passive microwave (PMW) brightness temperatures (Tb) over snow-covered land as measured by the Advanced Microwave Sounding Radiometer (AMSR-E). The eventual goal is to use the ANN as an observation operator within an ensemble-based data assimilation framework where model estimates are merged with AMSR-E Tb measurements in order to improve snow water equivalent (SWE) estimates at regional and continental scales. The results from this current study suggest an alternative form of machine learning the support vector machine (SVM) outperforms the ANN for all frequencies and polarizations evaluated. During SVM development, the NASA Catchment Land Surface Model is used to define snowpack properties. The SVM is then trained on a split-sample of PMW Tb measurements from AMSR-E. Improvements in the SVM predictions (relative to the ANN) were found in both forested and non-forested regions as well as in regions where the snowpack is relatively thin and ephemeral. Additionally, the SVM predictions (relative to the ANN) captured much more of the daily temporal variability present in the AMSR-E Tb observations. These finding suggest a SVM could serve as an effective observation operator within a land data assimilation framework.


Home

NASA-GSFC / GMAO / Rolf Reichle