Remote sensing-based assessment of crop yield and water dynamics in the southern Great Plains (USA)

Abstract

Remote sensing has emerged as a crucial tool for estimating and understanding yield variability and water management in diverse cropping systems. This is particularly important in water-limited areas, such as the southern Great Plains, where water is the key to sustaining agricultural production. Therefore, the present research integrated remote sensing with weather variables and machine learning (ML) to predict crop yields, identify key variables affecting crop yields, and enhance overall water use efficiency.

The study began by highlighting the significance of satellite remote sensing in predicting soybean yield, utilizing various inputs that can impact soybean yield metrics. We evaluated different ML models to predict soybean yield and identify key variables, utilizing agronomic management factors such as planting dates, seeding rates, and maturity groups, with weather variables and remote sensing vegetation indices. Our findings showed that variability is accurately captured by the Random Forest and Adaptive Boosting models (R²: 0.77-0.79). The accuracy was higher during the late vegetative and late reproductive stages when crops require more water. Furthermore, it was found that crop evapotranspiration (ETc) was the most significant feature affecting the performance of the ML model among the various meteorological variables. The importance score was 0.73. These findings demonstrate that high-resolution satellite data can support early-season yield prediction and identify the key meteorological variable (ETc) under adaptive management practices.

Based on the knowledge that ETc was a key factor in measuring crop yields, the second chapter of the research compared and evaluated different precision irrigation systems using ETc. The overarching goal was to identify field-scale crop water stress and irrigation scheduling. Therefore, we have utilized and quantified satellite-derived ETc to evaluate different irrigation systems that can save irrigation water. In this chapter, we have compared a precision irrigation method (AI-integrated Ground Penetrating Radar; AI-Radar) with the conventional subsurface drip irrigation (SDI) system using spatial water stress maps. High-resolution satellite imagery, combined with meteorological data, was utilized to generate spatial water stress maps (Water Deficit Index; WDI), where lower WDI values indicate less crop water stress. This index is a key indicator of crop water stress. Results showed that fields irrigated with AI-Radar maintained considerably lower WDI values (0.15-0.16) than SDI fields (0.20-0.24) during critical reproductive stages, with 31-34% less irrigation water from SDI. These findings demonstrate the potential of the AI-Radar irrigation system for effective irrigation management, which can help to conserve irrigation water while maintaining crop productivity.

In the third chapter, we have expanded this study by moving from spatial water maps to field-scale water use efficiency (WUEf), which accounts for actual evapotranspiration (ETa) as a water variable and crop yield. Here, we have moved from satellite-scale monitoring to finer, field-scale observation using UAV-based measurements. At this stage, high-resolution UAV imagery was combined with OpenET satellite estimates to capture more accurate spatial variation in ET and its relationship to yield. Based on our results, fields irrigated with the AI-Radar irrigation system showed higher WUEf (6.7-7.17 kg ha⁻¹ mm⁻¹) and a 30.8% increase in yield compared to SDI.

The three chapters of this study demonstrate a clear progression from identifying ETc as an important factor influencing crop yield variability to estimating ETc for irrigation assessment using satellite data, followed by a UAV-based field-scale evaluation. Using satellite and UAV remote sensing together with machine learning provides a scalable method for linking yield prediction and water use efficiency. This framework supports efficient irrigation planning and improved productivity with reduced irrigation use in the Southern Great Plains.