Estimating the spatial distribution of soil volumetric water content in an agricultural field employing remote sensing and other auxiliary data under different tillage management practices
David Kwesi Abebrese,James Kobina Mensah Biney,Kara Recep Serdar,Kamila Báťková,Jakub Houška,Svatopluk Matula,Nasem Badreldina,Lemma Adane Truneh,Tewodros Assefa Shawula
DOI: https://doi.org/10.1111/sum.12981
2023-11-01
Soil Use and Management
Abstract:Knowledge of soil volumetric water content (VWC) on agricultural soils as influenced by different soil management practices is important, but the measurement outputs of frequently used traditional sampling techniques are restricted to point‐based measurements with limited spatial coverage. Remote sensing (RS) techniques are therefore being explored due to their greater spatial and temporal availability as well as their ability to cover large‐scale areas. But a general limitation for RS data is the presence of vegetation cover, cloud cover, and the effect of topography. To minimise the effects of these factors during field sampling and the use of spectral indices, terrain attributes have proven successful in the estimation of several soil properties; however, the impact of these approaches for the estimation and mapping of soil VWC, an important factor in crop growth and development, especially where different soil management practices are employed, remains limited. To contribute to the knowledge base of RS under varying soil systems, this study explores the possibility of combining in‐situ measurements with remotely sensed (explanatory variables) data obtained under four different tillage practices to produce an estimated soil VWC that represents the entire study field. The tillage practices used include reduced till (RT), strategic till (ST), no till (NT), and conventional till (CT). From these tillage plots, three explanatory datasets, namely Sentinel‐2 (S2), spectral indices (SI), and terrain attributes (TA), were collected as predictors. In addition, each of the explanatory variables was structured into four groups, representing each of the four tillage methods. The eXtreme Gradient Boosting (XGBoost) model was used, and the best results were selected based on the root mean squared error (RMSE), the coefficient of determination (R2), and the mean absolute error (MAE). Prior to soil VWC prediction, the Pearson correlation matrix was used to determine the linear relationship between each of the selected explanatory variables (S2 bands, SI, and TA) and soil VWC. Furthermore, spatial distribution maps of soil VWC were constructed using the inverse distance weighting (IDW) interpolation technique. For soil VWC estimation, the TA outperformed the SI and S2 datasets (R2 = 0.84). Similarly, the spatial distribution maps obtained from the TA data show a study area with a large concentration of VWC compared to the other datasets. The study also identified CT as the tillage approach that most impacts soil VWC estimation because all predicted results were poor without the addition of data from the CT plot. According to our findings, using TA data collected from various tillage management systems to estimate soil VWC is very promising because they can be used as predictors to improve soil VWC estimation and mapping. This study demonstrates conclusively that remote sensing data collected from different tillage management systems can be used as predictors to enhance the estimation and mapping of soil VWC, complementing the bases for development of reliable and consistent precision irrigation management systems.
soil science