Abstract

This research was intended to examine the use of radar data obtained from Sentinel-1 imagery with different combinations of topographic attributes and multispectral data to predict soil texture fractions (STFs) in some agricultural soils located in the southern part of Kurdistan province, in western Iran. For this purpose, by applying the stratified random sampling technique, 216 soil samples were collected from the upper soil layer (0–30 cm) of the studied area. The particle size fractions were measured using the hydrometer method. Then, a comprehensive set of environmental variables including Sentinel-1, Sentinel-2 and digital elevation model (DEM) were derived. These environmental variables were included in four datasets; these included: dataset A (Sentinel-2 and DEM), dataset B (Sentinel-1 and DEM), dataset C (Sentinel-1 and Sentinel-2), and data set D (Sentinel-1, Sentinel-2 and DEM). After selecting the parsimonious variables in each dataset based on Pearson correlation coefficients and variance inflation factor (VIF) analyses, three machine learning models including random forest (RF), support vector machine (SVM), Cubist, and the ensemble of individual models were used to create the statistical link between environmental variables and STFs. The results showed that the ensemble modeling method with the dataset D had the highest accuracy in comparison to individual models in the prediction of clay (RMSE = 4.07%, MAE= 3.22%, ME= −0.08, CCC= 0.53), silt (RMSE = 5.96%, MAE=4.63%, ME=−0.52, CCC=0.49), and sand fraction (RMSE = 8.32%, MAE=6.46%, ME= 0.52, CCC=0.55). In addition, the results of variable importance revealed that all three types of environmental variables contributed to the prediction of the spatial distribution of STFs. These results also showed that elevation, Band 11 and VH_1 (which is the one provided by the radar data obtained from Sentinel-1), respectively, were the most important variables in the prediction of clay and sand fractions. In addition, Band 11, elevation, and multi-resolution valley bottom flatness index (MrVBF) were identified as the most important variables contributing to the prediction of the silt fraction. This study also demonstrated the effectiveness of backscatter coefficients obtained from Sentinel-1 in digital soil mapping (DSM) in the prediction of soil properties. Finally, the map of each fraction with its uncertainty and also, the soil texture map were generated from the ensemble model and dataset D. These maps could be used as the decision tools for agricultural management.

Full Text
Paper version not known

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call

Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.