MODIS Land Surface Temperature as an index of surface air temperature for operational snowpack estimation

Abstract

Regional operational modeling systems that support forecasters for the real-time warning of flash flood events often suffer from lack of adequate real-time surface air temperature data to force their accumulation and ablation snow model. The Land Surface Temperature (LST) product from MODIS, which provides four instantaneous readings per day, was tested for its feasibility to be used in real-time to derive spatially distributed surface air temperature (Ta) forcing for the operational snow model. The study was conducted in the Southeast region of Turkey using an atypically dense network of hourly Ta, daily snow depth, snow water equivalent (SWE), and rainfall datasets for the period: October 2002–September 2010. A comparison between the Ta and the corresponding LST grid-cell data indicated close associations that are different in nature for periods with and without snow on the ground. The LST-derived Ta was compared with that obtained from on-site gauge-based interpolation procedures and climatological time series. The LST-derived Ta was found inferior only to the Ta derived from the interpolation of the dense gauge network (31-gauges). Snow-pack simulations using estimated Ta time series were compared to simulations that were forced by the observed Ta at each site of 18 sites. The LST-derived Ta performed well in simulating snow mass and maximum SWE magnitude, while it did not represent well the timing of the annual peak of SWE and the duration of spring melt. Our study concluded that the MODIS/LST product can be a valuable additional source of real time forcing data for regional operational snow models, especially in remote mountainous areas with sparse telemetric data.