Page Banner

United States Department of Agriculture

Agricultural Research Service

Research Project: MANAGEMENT TECHNOLOGIES FOR ARID RANGELANDS

Location: Range Management Research

Title: An analysis of MODIS fractional snow cover estimates for snowmelt runoff modeling

Authors
item Steele, Caiti -
item RANGO, ALBERT
item Bleiweiss, Max -

Submitted to: American Geophysical Union
Publication Type: Abstract Only
Publication Acceptance Date: November 1, 2009
Publication Date: December 14, 2009
Citation: Steele, C.M., Rango, A., Bleiweiss, M. 2009. An analysis of MODIS fractional snow cover estimates for snowmelt runoff modeling [abstract]. 2009 American Geophysical Union, December 14-18, 2009, San Francisco, California. C44A-04.

Technical Abstract: Rio Grande streamflow is fed primarily by water from the melting snowpack in mountainous basins in the Southern Rockies. The snowpack accumulates mass in winter months and melts continuously through Spring and Summer. It is estimated that snowpack contributes 50% of the annual water supply of Colorado and New Mexico. A warmer winter climate anomaly, earlier snowmelt and declining snowpack in mountainous basins has been observed since the 1950s throughout the Western United States. The effect of this warming climate pattern on snowpack mass, extent and duration has serious implications for water resources in the entire Rio Grande basin.<br /> <br /> We use the Snowmelt Runoff Model (SRM: Martinec et al., 2008) to provide seasonal estimates of runoff contributions from snow cover in the upper Rio Grande basin and to predict runoff under different future climate scenarios. Sub-basin snow cover at a range of dates through the melt season is a key input variable driving SRM. Remotely sensed estimates of fractional snow cover, such as the MODIS snow cover product (MOD10A1 and MYD10A1, are the most available sources of snow cover but the accuracy of these estimates is highly variable. The aim of our current research is to compare existing snow cover products against alternative methods for estimating sub-pixel snow cover. Because we intend to share our snow mapping method with local water resource managers, we also aim to identify the most accurate, least complex algorithm to use for estimating snow cover from remotely sensed data.<br /> <br /> In this paper we compare four methods for estimating fractional snow cover from 500 m MODIS data: (i) the empirical model of Kaufman et al., (2002) (ii) linear mixture modeling (Andersen, 1982), (iii) the Normalized Difference Snow Index (NDSI; Salomonson and Appel, 2002) calculated from MODIS daily reflectance products (MOD09GA and MYD09GA) and (iv) the MODIS daily fractional snow product (Salomonson and Appel, 2002) available from the National Snow and Ice Data Center. To evaluate accuracy of each method, all snow cover estimates from MODIS data were compared with snow cover estimates derived from 30 m Landsat TM data for a range of dates and snow conditions between 2001 and 2008. We also compare the results for both the Aqua and Terra sensors. Further, to ascertain the degree of error that could be introduced into snowmelt runoff predictions by each mapping method, we run SRM for the snowmelt season of 2009.

Last Modified: 9/10/2014
Footer Content Back to Top of Page