Data quality control for stationary infrared thermometers viewing crops

Authors: Colaizzi, Paul; Oshaughnessy, Susan; Evett, Steven - Steve; Marek, Gary; Brauer, David; Copeland, Karen; Ruthardt, Brice

Submitted to: Applied Engineering in Agriculture
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 7/27/2023
Publication Date: 10/2/2023
Citation: Colaizzi, P.D., O'Shaughnessy, S.A., Evett, S.R., Marek, G.W., Brauer, D.K., Copeland, K.S., Ruthardt, B.B. 2023. Data quality control for stationary infrared thermometers viewing crops. Applied Engineering in Agriculture. 39(4):427-438. https://doi.org/10.13031/aea.15642.
DOI: https://doi.org/10.13031/aea.15642

Interpretive Summary: Infrared thermometers can measure the surface temperature of plant leaves. For irrigated crops, this is useful for scheduling irrigation timing and amounts. This is also useful for detecting crop diseases. These applications have been shown to conserve water. This is important because most agricultural regions have limited and declining water available for irrigation. The recent availability of wireless infrared thermometers has resulted in more widespread adoption by farmers. This has resulted in much greater volumes of plant temperature data. However, the large volumes of data required a new method to test for data quality. Scientists at USDA-ARS, Bushland, Texas, developed a computer model to rapidly test the quality of large volumes of plant temperature data. Ensuring data quality will improve the effectiveness of irrigation management. This can result in increased water conservation without compromising crop yield or farm income.

Technical Abstract: The increased adoption of infrared thermometers (IRTs) for irrigation management of crops has resulted in increasingly large surface temperature datasets, resulting in a need for data quality assurance and control (QA/QC) procedures similar to those developed for agricultural weather station data. A QC procedure was developed to test for seven common data conditions, including sensor not deployed, missing, too high, too low, upward spike, downward spike, or stuck. The conditions of “too high” or “too low” used a simple energy balance procedure similar to the crop water stress index, where the theoretical lower and upper temperature limits of a surface were calculated, accounting for the vegetation view factor appearing in the IRT field-of-view. After passing the seven tests, data were assigned as plausible, and further tested as confirmed or confirmed+. The confirmed test compared each IRT to the median of the other IRTs during 2 h before sunrise and applied a threshold of +/- 0.5 degrees C. The confirmed+ test compared each IRT to the median of the other IRTs during +/- 2 h of solar noon and applied a threshold of +/- 2.0 degrees C. The set of tests was applied to an IRT dataset that included six seasons of crops and fallow periods with 15-min time steps. Temperature differences greater than the thresholds (i.e., assigned plausible but not confirmed or confirmed+) could detect anomalies including ice, dirty or obscured lenses, or biased data that other tests did not detect. Of the ~1,500 days when 20 IRTs viewing a crop were deployed, 19% of the days resulted in plausible but not confirmed or confirmed+, 5% of the days resulted in confirmed but not confirmed+, and 56% of the days resulted in confirmed+. The procedure can be easily customized and can increase the value of IRT datasets used for irrigation management.