Integrating Soil Spectral Library and PRISMA Data to Estimate Soil Organic Carbon in Crop Lands

• Published in: IEEE geoscience and remote sensing letters Vol. 21; pp. 1 - 5
• Main Authors: Reddy, Bonthu Sandeep, Shwetha, H. R.
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 2024; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Agricultural land; Agricultural practices; Algorithms; Alluvial soils; Analytical methods; Calibration; Crop rotation; Crops; Data recording; Developing countries; Food security; Hyperspectral imaging; Hyperspectral remote sensing; india; LDCs; Libraries; Machine learning; machine learning (ML); Mapping; Organic carbon; Organic soils; Performance prediction; PRISMA; Reflectivity; Soil; soil organic carbon (SOC); soil spectral library (SSL); spectroradiometer; Spectroradiometers; Spectroscopy; Sustainable Development Goals; Synthesis; India
• ISSN: 1545-598X; 1558-0571
• DOI: 10.1109/LGRS.2024.3374824

The increasing demand for precise soil organic carbon (SOC) monitoring in croplands is crucial for food security (SDG 2), and has led to the exploration of fusing soil spectral libraries (SSLs) with hyperspectral sensing data for SOC estimation. However, the widespread adoption of SSL for SOC estimation faces challenges, particularly in developing nations, due to inconsistent calibration libraries and reliable estimation models. Furthermore, SSL rely on regular soil sample collection and spectral data recording using spectroradiometers, which is impractical in agricultural-predominant countries, such as India, with limited time for sample collection between crop rotations. To address this challenge, we developed synthesized SSL in laboratory conditions and integrated it with hyperspectral data using machine learning (ML) algorithms to bridge the gap between synthesized SSL and hyperspectral data for local-scale SOC mapping. This approach was tested by mapping SOC in Mysore, India, using spectroradiometer hyperspectral measurements and PRISMA sensor data. The proposed approach and synthesized SSL exhibited better performance prediction accuracies, <inline-formula> <tex-math notation="LaTeX">{R}^{{2}} </tex-math></inline-formula> of 0.92 and 0.79, and the RMSE values of 2.31 and 9.91 g/kg, respectively, for PRISMA and laboratory spectroscopy data. These results highlight the potential of synthesized SSL for SOC prediction in alluvial soils, leveraging local datasets, and hyperspectral data. Our future work will expand the synthesis approach to other study areas, particularly those with alluvial soil origins, further enhancing the applicability of this methodology for SOC estimation and aiding food security efforts.