Estimation of soil organic carbon in arid agricultural fields based on hyperspectral satellite images

This study introduces a remote sensing approach to estimate soil organic carbon in arid agricultural fields, emphasizing sustainable land management. The United Arab Emirates (UAE) serves as the case study, representing a region where soil organic carbon dynamics have not been previously assessed. A total of 186 topsoil samples were collected and analyzed for soil organic carbon. Spectral data from field measurements, the DLR Earth Sensing Imaging Spectrometer (DESIS), and Sentinel-2 were integrated, marking the first application of this combination for soil organic carbon prediction. To address the challenges of arid environments, the study introduced specialized preprocessing techniques, including a novel vegetation index (UAEVI) for masking vegetation, principal component analysis for filling missing attributes, area normalization, and Savitzky-Golay smoothing to reduce noise and enhance spectral data. Soil organic carbon exhibited significant spectral correlations, with negative relationships observed in the wavelength ranges 401–416, 670–698, and 926–957 nm, and strong positive relationships in the ranges 519–560, 744–785, 937, and 1610 nm. A ridge regression model was developed and validated, achieving an Coefficient of Determination (R²) of 0.671, Root Mean Squared Error (RMSE) of 0.120 %, and Ratio of Performance to InterQuartile distance (RPIQ) of 2.271. The model demonstrated reliable performance in mapping soil organic carbon, achieving results comparable to studies in non-arid climates. Seasonal analysis highlighted the influence of meteorological parameters on soil organic carbon trends, and the model was successfully applied to monitor temporal changes in soil organic carbon within a sub-region from June 2022 to December 2023, revealing a slight increase in soil organic carbon over this period. This research emphasizes the effectiveness of integrating hyperspectral (DESIS) and multispectral (Sentinel-2) data with advanced preprocessing techniques for soil organic carbon estimation in arid environments. This study offers a scalable framework for more accurate and timely soil assessments, promising significant improvements in the management of arid soil ecosystems.