Distinguishing Ice from Snow for Melt Modeling Using Daily Observations from MODIS

Karl Rittger1,2, Mary J. Brodzik3,4, Adina Racoviteanu5, Andrew P Barrett2,6, Siri Jodha S Khalsa2,5, Thomas H Painter7, Richard L Armstrong2,8 and Annie Bryant Burgess9, (1)University of Colorado Boulder, Institute of Arctic and Alpine Research, Boulder, United States, (2)Univ Colorado, Boulder, CO, United States, (3)University of Colorado Boulder, CIRES, National Snow and Ice Data Center, Boulder, United States, (4)University of Colorado at Boulder, CIRES, National Snow and Ice Data Center, Boulder, CO, United States, (5)National Snow and Ice Data Center, Boulder, United States, (6)National Snow and Ice Data Center (NSIDC), CIRES, University of Colorado Boulder, Boulder, United States, (7)NASA Jet Propulsion Laboratory, Pasadena, CA, United States, (8)National Snow and Ice Data Center, Boulder, CO, United States, (9)University of Southern California, Computer Science, Los Angeles, CA, United States
Abstract:
In Earth’s mountainous regions, melt from both seasonal snow and glacier ice contributes to streamflow. Few in-situ observations exist that can help distinguish between the two components of melt, particularly across large mountain ranges. In this study, we analyze daily time series of MODIS data products to distinguish ice from snow as the seasonal snowpack recedes revealing firn and glacier ice surfaces. We run a temperature index melt model for the Hunza, a sub-basin of the Upper Indus basin using the MODIS data to discriminate between glacier ice and snow and partition the corresponding streamflow.

 During the ablation period, this high elevation mid-latitude snowpack receives intense incoming solar radiation resulting in snow grain growth and surface albedo decreases. To explore snow grain growth, we use estimates of grain size from both the MODIS Snow Covered Area and Grain Size Model (MODSCAG) and MODIS Dust Radiative Forcing in Snow (MODDRFS). To explore albedo reduction we use 2 standard albedo products from MODIS, the Terra Daily Snow Cover algorithm (MOD10A1) and Surface Reflectance BRDF/Albedo (MOD43). We use a threshold on the grain size and albedo products to discriminate ice from snow.

We test the ability of the 4 MODIS products to discriminate snow from glacier ice using higher resolution data from the Landsat 8 sensor from July 5th and July 21st, 2013 for a subset of the study area in the Karakoram region of the Himalaya that includes the Yazghil and Hopper Glaciers that drain north and northeast in the Shimshall Valley, part of the Hunza River basin. Snow and glacier ice are mapped using band ratio techniques, and are then separated on the basis of broadband albedo values calculated from Landsat bands for comparison with MODIS-derived snow and glacier ice pixels.

We run a temperature index melt model that uses gap filled snow covered area from MODSCAG and interpolated station temperature data for the Hunza River basin. The model outputs daily melt volume from three surface classifications: area of snow over land, snow over glacier ice, and exposed glacier ice. The partitioning is done using a combination of MODICE and 1 of 4 MODIS products. Melt from glacier ice accounts for 25% of the combined melt on average across the 4 MODIS products for the Hunza basin, with fluctuations based on the seasonal variability of snowfall.

Back to: Remote Sensing of the Cryosphere III Posters