Skip to main content
ARS Home » Northeast Area » Beltsville, Maryland (BARC) » Beltsville Agricultural Research Center » Hydrology and Remote Sensing Laboratory » Research » Publications at this Location » Publication #365323

Research Project: Integrating Remote Sensing, Measurements and Modeling for Multi-Scale Assessment of Water Availability, Use, and Quality in Agroecosystems

Location: Hydrology and Remote Sensing Laboratory

Title: Measurement uncertainty and model validation: An instructive case study of an irrigated cotton crop

item Alfieri, Joseph
item Kustas, William - Bill
item Anderson, Martha
item Prueger, John
item Evett, Steven - Steve
item NEALE, C. - University Of Nebraska

Submitted to: Meeting Abstract
Publication Type: Abstract Only
Publication Acceptance Date: 4/25/2019
Publication Date: N/A
Citation: N/A

Interpretive Summary:

Technical Abstract: Remote sensing-based models represent the only practical approach for monitoring evapotranspiration (ET) at the field and sub-field scales needed for range of agricultural applications including irrigation management and crop yield forecasts. However, the accuracy of these models relies on the quality of the in-situ measurements used for their development, calibration, and validation. Therefore, one key to properly assessing the utility of remote sensing-based models is understanding the limitations and uncertainty associated with the underlying observational data. To illustrate this, a case study comparing the output of the Two-Source Energy Balance (TSEB) model with a pair of independent datasets collected over irrigated cotton during the 2008 Bushland Evapotranspiration and Agricultural Remote Sensing Experiment (BEAREX08) was conducted. The ET measurements from the two datasets, which differed by as much as 4.5 mm d-1, were collected via eddy covariance (EC) and lysimetry (LY), respectively. When the uncertainty associated with the two observational datasets was neglected, the ET output from the TSEB model agreed with the EC data to within 2% while underestimating the LY data more than 15%. As a result, different conclusions could be drawn regarding the utility of the model depending on which observational dataset is used for the evaluation. However, when the sources of error (e.g. spatial variability within the measurement footprint) associated with the observational datasets were fully accounted for, the mean discrepancy between the two datasets decreased to 6% or 0.4 mm d-1. More importantly, the model output agreed with both in-situ datasets to within their measurement uncertainties, which was nominally 10%. Thus, the case study results demonstrate both the uncertainty inherent in field measurements and the need to carefully consider the complex factors affecting ET and its measurement when using observational data for model validation.