|Wood, E - PRINCETON UNIV|
|Pan, M - PRINCETON UNIV|
Submitted to: Journal of Geophysical Research Atmospheres
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: August 14, 2003
Publication Date: December 6, 2003
Repository URL: http://handle.nal.usda.gov/10113/59897
Citation: Crow, W.T., Wood, E.F., Pan, M. 2003. Multi-objective calibration of land surface model evapotranspiration predictions using streamflow observations and spaceborn surface radiometric temperature retrievals. Journal of Geophysical Research. 108(D23):3292-3330. Interpretive Summary: During the past twenty years, a large number of surface-vegetation-atmosphere transfer (SVAT) models have come into existence. The models are designed to predict fluxes of energy, momentum, and water between the surface of the earth and the lower part of the atmosphere. Some SVAT models are also designed to capture various hydrologic properties of the land surface such as runoff production and soil moisture storage. Consequently, output from SVAT models is of value for a wide range of agricultural applications - including drought monitoring, evaluating climate change scenarios, and improving precipitation forecasts obtained from a weather prediction model. However, a critical weakness of SVAT models is the large number of model parameters which must be pre-specified for a given landscape for the SVAT model to operate accurately. Finding appropriate values for many of these parameters is a difficult, and often highly ambiguous, activity. One strategy for improving SVAT model parameter selection is to calibrate model predictions against large-scale observations of the land surface. With the exception of streamflow observations, satellite retrievals of surface radiometric temperature offer perhaps the only output of SVAT models that can be accurately measured at large spatial scales. This paper explores the potential benefits of incorporating surface radiometric temperature and streamflow observations into a multi-objective calibration strategy. Comparisons are made to results obtained when taking a more traditional approach and calibrating against streamflow observations alone.
Technical Abstract: Physically-based models of surface water and energy balance processes typically require a large number of soil and vegetation parameters as inputs. Accurate specification of these parameters is often difficult without resorting to calibration of model predictions against independent observations. Along with streamflow observations from gauging stations, spaceborne surface radiometric temperature retrievals offer the only independent observation of land surface model output commonly available at regional spatial scales (i.e. > 502 km2). This analysis examines the potential benefits of incorporating spaceborne radiometric surface temperature retrievals and streamflow observations in a multi-objective calibration framework to accurately constrain regional-scale model evapotranspiration predictions. Results for the VIC (Variable Infiltration Capacity) model over the Southern Great Plains of the United States suggest that multi-objective calibration against radiometric skin temperatures and steamflow observations can reduce error in model monthly evapotranspiration predictions by up to 20% relative to single-objective calibration against streamflow alone.