Increased Bias in Evapotranspiration Modeling Due to Weather and Vegetation Indices Data Sources. (1st May 2019)

• Record Type: Journal Article
• Authors: Dhungel, Ramesh; Aiken, Robert; Colaizzi, Paul D.; Lin, Xiaomao; Baumhardt, R. Louis; Evett, Steven R.; Brauer, David K.; Marek, Gary W.; O'Brien, Dan
• Is Part Of: Agronomy Journal. Volume 111:Number 3(2019)
• Page Start: 1407
• Page End: 1424
• Publication Date: 2019-05-01
• Subjects: Agronomy -- Periodicals; 630
• DOI: 10.2134/agronj2018.10.0636
• Languages: English
• ISSNs: 0002-1962

Abstract

Abstract : Core Ideas: Increased bias in ET due to various data sources where generally utilized standard weather data poses similar biases. Increased understanding of the complex relationship among the various weather input and vegetation indices. Results showed local weather station data closely represented site condition indicating the difficulties and challenges in ET modeling using other commonly available and widely adopted data sources. ABSTRACT: Complex interactions among meteorological data and vegetation indices are incompletely understood in relation to evapotranspiration ( ET ) calculations for larger spatial domains with higher spatial and temporal resolution. Objectives of this study were to evaluate contributions of inputs to uncertainty in ET calculations and to enhance understanding of interactions among weather data, vegetative indices, and resistances utilized in biophysical ET model. We evaluated individual and combined effects of weather variables and vegetation indices using BAITSSS (Backward‐Averaged Iterative Two‐Source Surface temperature and energy balance Solution). Local weather station (LWS) data at a lysimeter site were obtained for irrigated corn (Zea mays L.) during the growing season (May to September, 2016) at Bushland, Texas. Gridded meteorological data were obtained from North American Land Data Assimilation System (NLDAS) (∼ 12.5 km) and remotely‐sensed vegetation indices (Landsat 30 m). Standard weather station (SWS) data were obtained from a grass reference near lysimeter site. The r 2 and RMSE of ET simulated using LWS data and measured vegetation indices were 0.90 and 0.85 mm for daily ET, and 0.90 and 0.10 mm, for hourly ET, compared to lysimeter ET (less than 4% cumulative error). However, r 2 and RMSE were 0.74 and 1.64 mm for daily ET, and 0.81 and 0.14 mm for hourly ET using gridded data, with positive bias (∼ 25% from NLDAS data). Simulated ET from SWS data exhibited similar behavior to gridded data with increased ET up to 21%. Results quantify difficulties in ET modeling using commonly available and widely adopted data sources.