Algorithms to process weighing lysimeter data

Authors: Colaizzi, Paul; Marek, Gary; Evett, Steven; Copeland, Karen; Ruthardt, Brice

Submitted to: Journal of the ASABE
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 7/21/2025
Publication Date: 8/1/2025
Citation: Colaizzi, P.D., Marek, G.W., Evett, S.R., Copeland, K.S., Ruthardt, B.B. 2025. Algorithms to process weighing lysimeter data. Journal of the ASABE. 68(5):729-752. https://doi.org/10.13031/ja.16404.
DOI: https://doi.org/10.13031/ja.16404

Interpretive Summary: Irrigation of crops is vital to maintain food and fiber production and to mitigate costs to consumers. Irrigation requires a large portion of available water resources, and careful management of irrigation is increasingly important to ensure water resources are not overexploited and to maintain rural economies. This is particularly important in the US Southern High Plains where irrigation is from the declining Ogallala Aquifer. Large weighing lysimeters offer the most accurate method to measure crop water use, which is the first step in effective irrigation management. Weighing lysimeters generate large volumes of data that require extensive processing, which can be very cumbersome and time consuming when done by spreadsheet. Scientists at USDA-ARS, Bushland, Texas, developed computer algorithms that greatly sped up the process, while increasing the accuracy of crop water use data. This result will reduce water and energy required to produce crops, increase crop yields, and increase farm profitability.

Technical Abstract: A set of algorithms written in MATLAB were developed and tested for weighing lysimeters to process 5-minute raw relative storage and event flag data to determine water balance components. Water balance components included non-rainfall water (NRW; i.e., dew or frost), precipitation (P), irrigation (I), drainage (D), and evapotranspiration (ET). The MATLAB program produced output files and graphs for quality assessment of ET results. The program reduced processing time required compared with a spreadsheet that had been used for the same purpose. The program was tested by comparing daily and cumulative NRW, P, I, and ET calculated using the algorithms to those calculated with the spreadsheet. Comparisons included three full calendar years and four lysimeters located at the USDA Agricultural Research Service in Bushland, Texas, USA. Two years (2016 and 2018) included maize seasons and one year (2019) included a soybean season. P and I amounts calculated from mass data were identical for the spreadsheet and program. The program reproduced the general trend of cumulative NRW, with most NRW occurring in the winter and early spring and very little during the summer at the study location for both MESA and SDI lysimeters. Annual NRW totals calculated using the spreadsheet and the algorithms were 42 to 78 mm. Spreadsheet vs. algorithm root mean squared error (RMSE) for daily NRW were 0.048 to 0.083 mm (~35 to 57% of spreadsheet mean NRW). The algorithms tended to overestimate NRW compared with spreadsheet calculations in 2016 (mean bias error, MBE were 0.033 to 0.050 mm) and underestimate in 2018 and 2019 (MBE were -0.015 to -0.030 mm). Annual ET totals calculated by the spreadsheet and the program were 952 to 1083 mm in 2016; 880 to 1016 mm in 2018, and 1058 to 1111 mm in 2019. Spreadsheet vs. program RMSE for daily ET were 0.11 to 0.25 mm (4.3 to 7.9% of spreadsheet mean ET) and MBE were -0.070 to 0.074 mm. The program was able to identify uncertain or spurious ET results; examples included mis-flagged events, prolonged precipitation over three days, and a leaking drainage line from a lysimeter.