The Elephant in the Room: Spatial Heterogeneity and the Uncertainty of Measurements and Models

Monday, 15 December 2014: 4:15 PM
Joseph G Alfieri1, William P Kustas1, John H Prueger2, Nurit Agam3, Christopher M U Neale4 and Steven R Evett5, (1)USDA ARS HRSL, Beltsvillle, MD, United States, (2)USDA ARS NLAE, Ames, IA, United States, (3)Jacob Blaustein Institutes for Desert Research, Sed-Boker, Israel, (4)University of Nebraska, Robert B. Daugherty Water for Food Institute, Lincoln, NE, United States, (5)USDA ARS, Bushland, TX, United States
Variations in surface conditions can significantly influence the exchange of heat and moisture between the land and atmosphere. As a result, measurements of surface fluxes using disparate methods not only may differ, they may fail to represent the surrounding landscape due to localized differences in surface conditions. To illustrate this, data collected over adjacent cotton fields during the Bushland Evapotranspiration and Agricultural Remote Sensing Experiment (BEAREX08) will be used. The evapotranspiration (ET) within each field was determined via lysimetry (LY), mass balance using neutron probe (NP) data, and a pair of eddy covariance (EC) systems. A comparison of the cumulative ET from each field showed that ET from LY was 20% to 25% greater than that derived from NP and 10% to 15% greater than those from EC. Additionally, the cumulative flux for the two fields collected using the same approach differed by 5% to 10%. These discrepancies can be explained, in large part, by the variations in vegetation density within the two fields. Not only were there substantial variations in the leaf area index (LAI) within the source areas of the different measurement systems – for example, the LAI within LY was, on average, 0.4 m2 m-2 greater than the LAI within the source area of NP – there were also significant differences in the LAI between the fields as a whole. The cumulative ET output by the remote sensing-based Two-Source Energy Balance (TSEB) model was also compared to the cumulative ET from each of the three measurement approaches. Depending on which measurement technique is used, the model either underestimated the moisture flux by approximately 5%, in the case of LY, or overestimated the flux by nearly 20%, in the case of NP. Comparison of the model output with EC data also indicated that the model overestimated ET, in this case, by approximately 10%. Clearly, the choice of which dataset is used to validate the model significantly effects the conclusions drawn regarding the model’s accuracy and utility in estimating ET. The results of this study also underscores the limitations of each of these measurement techniques and the need to understand those limitations when using observational datasets to make general conclusions about field scale ET and validating model output.