Retrieving Snowpack Properties From Land Surface Microwave Emissivities Based on Artificial Neural Network Techniques

The objective of this research is to better isolate the snow signature in microwave signals and to use this signature to retrieve the snowpack properties. The retrieval of snowpack properties is performed by inverting microwave emissivities using artificial neural network ANN-based technique. The microwave emissivities used in this study are derived from SSM/I passive microwave observations by removing the contributions of the cloud and atmosphere and then separating out the surface temperature variations using ancillary atmospheric, cloud and surface data. A time-anomaly of differences between effective emissivity at 19V and 85V enabled the constant effects of land surface vegetation properties to be removed to isolate the snow signature. An emission model (Microwave Emission Model of Layerd Snowpack, MEMLS) was used to produce the train dataset to the neural network. To retrieve the neural network the 7 channels of microwave emissivities (The pixels that were identified as snow), the IR skin temperature, and the ground emissivity (emissivity of snow-free of each pixel) were used as the inputs and the neural network retrieves snow depth, snow density, snow grain size, and volumetric water percentage in the snow. The resulting depth and SWE (depth * density) were compared with CMC snow depth and Chang algorithm. It was observed that the results obtained through ANN-based technique are lower by 8% than those obtained through other approaches. These comparisons have been more explored with different vegetation type, topography, and snow percentage coverage of each pixel.