A two‐source approach for estimating turbulent fluxes using multiple angle thermal infrared observations

Abstract

A recently developed two‐source model for computing the surface energy balance with directional radiometric surface temperature observations can be simplified with the use of radiometric surface temperature observations at two substantially different view angles. Such observations have been available from the Along Track Scanning Radiometer (ATSR) aboard the First European Remote Sensing Satellite (ERS‐1) and therefore could be used operationally with this simplified model. Compared to the original model formulation, the new version, 2ANGLE, does not require an estimate of the fraction of “green” or active vegetation, fg, and use of the Priestley‐Taylor parameterization for partitioning of net radiation absorbed by the canopy into latent, LE, and sensible, H, heat in order to obtain a solution. However, if one uses the Priestley‐Taylor parameterization with two radiometric observations at different view angles, then the model, 2ANGLE_PT, does not require a measurement of air temperature. The model‐derived fluxes using multiple angle radiometric surface temperature observations collected during the First International Satellite Land Surface Climatology Project (ISLSCP) Field Experiment (FIFE), which was conducted in the tall grass prairie in eastern Kansas during the summers of 1987 and 1989, are compared to measured surface fluxes. Differences between predicted and measured H using the 2ANGLE and 2ANGLE_PT models were generally significant (i.e., >50%), while the original version, 1ANGLE_PT, yielded differences of 45–50%, on average. The 2ANGLE and 2ANGLE_PT model predictions of LE yielded differences of 25–40% with measured values, while the 1ANGLE_PT model produced differences of 15–20%, on average. Using a simple technique to estimate daytime total LE from the “instantaneous” values with the three versions of the model resulted in differences generally between 20 and 25%. A review of studies analyzing FIFE flux measurement techniques indicates that differences in half‐hourly to hourly observations of H and LE can reach 50% or greater, but typically are between 20 and 30%. Therefore the performance of the models in predicting H ranged from marginal to poor, but was generally satisfactory in estimating LE because in many cases LE was significantly larger than H. The performance of the 2ANGLE_PT model in predicting Hand LE was significantly better than the 2ANGLE model, but not as good as the original 1ANGLE_PT model. A sensitivity analysis with the three models indicated that not only were the 2ANGLE_PT model predictions unaffected by a 3 K uncertainty in the estimation of air temperature using satellite data, but also showed little sensitivity to a 1.5 K uncertainty associated with deriving an atmospherically corrected surface temperature from ATSR observations. Therefore the 2ANGLE_PT model has potential for computing regional‐scale LE operationally with a satellite‐based sensor such as the ATSR.