Snow Depth Estimation
My primary research project at The City College of New York was working with Professor Naresh Devineni (CCNY), Sean Helfrich (NOAA), and Dr. Cezar Kongoli (UMD/ESSIC) on generating a novel historical, spatially resolved map of snow depth. These maps were generated to include as climatological background for the Interactive Multisensor Snow and Ice Mapping System (IMS) Blended Snow Depth (BSD), which fuses in-situ and remote sensing observations of snow depth to forecast Northern Hemisphere snow depth twice-a-day. The BSD output is given to numerical weather forecasters for their land surface models, because SD is a significant factor in albedo, or surface reflectivity, and thus plays a significant role in the surface energy balance. Multiple techniques for generating the maps including a k-Nearest neighbors scheme and a LOESS for multiple predictors (location, elevation, air temperautre) were evaluated. Shown on the right are validation plots for each month over Eurasia for the kNN (left) and LOESS (right) approaches. Both techniques failed to capture the extremes in the snow depth estimates due to their inherently smooth nature.
Performance of snow depth estimation for Jan-April over the Eurasian region.
Related presentations and papers:
Lawrence M. Vulis, Naresh Devineni, Sean R. Helfrich, and Cezar Kongoli, Development of Gridded Snow Depth Climatology in Eurasia. AMS, January 22–26, 2017
Lawrence Vulis, Naresh Devineni, Sean Helfrich, and Cezar Kongoli, K-Nearest Neighbor Approach for Developing Gridded Snow Product in Russia. 8th Biennial NOAA EPP Education and Science Forum, August 28-30, 2016
Sean R. Helfrich, Cezar Kongoli, Lawrence Vulis, Milton Martinez, Christopher Grassotti, and Naresh Devineni, Evaluation of Algorithm Alternatives for Blended Snow Depth in the IMS. 73rd Eastern Snow Conference, June 14-16, 2016