Quality Management for Research Weather Data: USDA‐ARS, Bushland, TX

Abstract

Core Ideas Several redundant sensors for the same variable are required for quality control. Weighing lysimeters record rainfall more accurately than do tipping bucket gauges. Southern High Plains rainfall is variable, even within a 1‐km radius. Scatterplots and time‐series graphical and statistical analyses aid quality control. Weather mediates the complex interactions of soil, water, plants, atmosphere, and agricultural and environmental interventions. Accurate and representative weather data are thus important to agricultural and environmental research, but particularly to simulation model development and testing. USDA‐ARS research using large weighing lysimeters at Bushland, TX, has produced 30 yr of evapotranspiration and ancillary data related to different methods of irrigation application and management, dryland cropping, crop choice, and agronomic management and weather. To support crop simulation modeling, weather data were gathered on a 15‐min basis at a research weather station adjacent to the weighing lysimeters since the inception of the project in 1987, as well as at nearby stations. The goal of this article is to review and demonstrate Conservation and Production Research Laboratory (CPRL) weather data quality management procedures. Data quality management has involved both quality assurance planning and operations as well as quality control methods. Data analysis revealed the importance of having multiple sensors replicating the same weather measurement because sensor failure or degradation often created gaps in data that required another data source to fill, but also because replicate sensors aided in detecting sensor error. Data from multiple stations also revealed issues of datalogger time keeping and spatial variation of weather. And, comparisons of data on the same weather measurement from various sensors that used different operating principles revealed systematic errors in some sensors. Methods of quality assurance and control involved long‐term planning for sensor calibration, maintenance and replacement, and daily and longer‐term quality control efforts based on data stream analysis.