Location: Soil and Water Management ResearchTitle: Increased bias in evapotranspiration modeling due to weather and vegetation indices data sources Author
|Dhungel, Ramesh - Kansas State University|
|Aiken, Robert - Kansas State University|
|Lin, Xiaomao - Kansas State University|
|Baumhardt, Roland - Louis|
|Brauer, David - Dave|
|Evett, Steven - Steve|
|O'brien, Dan - Kansas State University|
Submitted to: Agronomy Journal
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 1/29/2019
Publication Date: N/A
Interpretive Summary: As irrigation water available from the Ogallala Aquifer decreases, farmers need to better match irrigation applications to water needs of the crop. Maps of crop water use across large agricultural regions are useful for drought detection and prediction, and management of water resources. Maps of crop water use can be produced by combining satellite images and weather data. The satellite images usually have large pixels, from 100 m to 1 km. The weather data are obtained continuously from weather stations which are located as discrete smalls. The spaces between the weather station points can be filled using models so that weather data can have the same locations as the satellite pixels to produce the crop water use maps. However, filling the spaces between the weather stations causes loss of accuracy. Therefore, scientists at USDA-ARS, Bushland, Texas and Kansas State University in the Ogallala Aquifer Program developed and tested a new method to improve the accuracy of crop water use maps. The improved accuracy will provide earlier detection of crop water stress, and will improve farm profitability.
Technical Abstract: Complex interactions among meteorological forcing including vegetation indices are still poorly understood while calculating evapotranspiration (ET) in larger spatial and temporal resolution in a time series. We evaluated North American Land Data Assimilation System (NLDAS) (12.5 km) meteorological forcing and Landsat (30 m) based vegetation indices to understand these interactions using energy balance model (BAITSSS). Influence of solar radiation, wind speed, air temperature, ambient vapor pressure, precipitation, leaf area index, and fraction of cover were investigated. The study was conducted during growing to maturity period (May to September 2016) in a highly advective environment in Bushland, Texas for drought-tolerant corn. The coefficient of determination and root mean square error using hourly measured weather data and vegetation indices were 0.91 and 0.85 mm for daily, respectively, and 0.90 and 0.10 mm for hourly, respectively, time steps compared to lysimeter ET (less than 3 percent cumulative ET error). However, utilizing gridded data decreased the coefficient of determination and increased root mean square error (0.76 and 1.56 mm for daily, respectively, and 0.81 and 0.14 mm for hourly, respectively). The combined effect of NLDAS meteorological forcing primarily induced positive bias (up to 27 percent) when compared to lysimeter cumulative ET. Calculated leaf area index had a negative bias, resulting in 14 percent positive bias in cumulative ET from entire gridded data.