Evaluating soil moisture content under maize coverage using UAV multimodal data by machine learning algorithms

Authors: ZHANG, YU - Northwest A&f University; HAN, WENTING - Northwest A&f University; Zhang, Huihui; NIU, XIAOTAO - Northwest A&f University; SHAO, GUOMIN - Northwest A&f University

Submitted to: Journal of Hydrology
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 1/5/2023
Publication Date: 1/9/2023
Citation: Zhang, Y., Han, W., Zhang, H., Niu, X., Shao, G. 2023. Evaluating soil moisture content under maize coverage using UAV multimodal data by machine learning algorithms. Journal of Hydrology. 617. Article e129086. https://doi.org/10.1016/j.jhydrol.2023.129086.
DOI: https://doi.org/10.1016/j.jhydrol.2023.129086

Interpretive Summary: To cope with scarce water supplies for agriculture, deficit irrigation is an important tool to achieve the goal of reducing agricultural water use. However, proper deficit irrigation management is facilitated by precise information about soil moisture content (SMC). In this study, we used UAV-based multimodal data (RGB, multispectral and thermal) to quantify SMC in a maize field under different irrigation treatments in 2018 and 2019 by using three machine learning algorithms. The results showed that the random forest regression (RFR) based model using multimodal data fusion at the vegetative stage generated the most accurate and robust SMC estimations for 10- and 20-cm soil depths. The RFR-based model produced the best estimation accuracy under mild to modest deficit irrigation treatments. The developed approach using UAV remote sensing produces surface soil moisture maps with high spatial-temporal resolution. An accurate estimate of SWC will benefit all irrigation scheduling tools that use soil moisture to determine irrigation timing and amount, such as the soil water balance method. It also allows precision irrigation application based on soil water status and eliminates the waste of water like run-off/deep percolation.

Technical Abstract: Timely and accurate estimation of soil moisture content (SMC) is essential for precision irrigation management at a farm scale. The advanced unmanned aerial vehicle (UAV) remote sensing technology with high spatiotemporal resolution has become a promising method for SMC monitoring. Many existing SMC models have only been tested at a specific crop growth stage using a single type of sensor and the effects of growth stages and irrigation variation on the SMC estimation accuracy remain unclear. To address these limitations, this study used UAV-based multimodal data to quantify SMC in a maize field under various levels of irrigation over two years by using three machine learning algorithms (MLA): partial least squares regression (PLSR), K nearest neighbor (KNN), and random forest regression (RFR). The results demonstrated that multimodal data fusion improves the SMC estimation accuracy regardless of MLA, especially the joint use of thermal data and multispectral data. Among three SMC regression models, the RFR model produced the most accurate SMC estimates in two maize growing seasons. The RFR model using RGB, multispectral, and thermal data generated the most accurate and robust SMC estimations at the vegetative stage with R2 of 0.71 and 0.78, and rRMSE of 19.94% and 19.98% for 10- and 20-cm soil depths, respectively. The RFR model using all three types of data produced the best accuracy of SWC estimation under mild to modest deficit irrigation treatments for both soil depths. Thus, the high spatial-temporal maps of SMC using UAV-based multimodal data show promising potential in supporting decision-making in irrigation scheduling at the farmland scale.