Advances in spectroscopic and machine learning approaches for estimating soil organic carbon and nitrogen: a comprehensive review

Soil organic carbon (SOC) and nitrogen (N) are fundamental indicators of soil fertility, productivity, and environmental sustainability. Traditional wet chemistry techniques, although precise in their determination, are resource-demanding, time-consuming, and, in most cases, not economical for large-scale monitoring. The growing need to adopt fast, economical, and dependable methods has seen soil spectroscopy come into the limelight as an alternative revolution. The paper is a review of up-to-date developments in soil spectroscopy to estimate SOC and N, including visible-near infrared (VNIR) and mid-infrared (MIR) spectroscopy. The use of unmanned aerial vehicle (UAV) multispectral imaging, satellite remote sensing, and the combination with machine learning model is critically discussed. It has been reported that MIR spectroscopy offers more accurate and superior spectral characteristics of SOC and N than VNIR and that several studies have found higher prediction accuracies (R2 often in the range of 0.94–0.98 with MIR compared to 0.49–0.83 with VNIR in relation to SOC, with slightly lower but comparable trends for nitrogen) and substantially lower RMSE values because of an enhanced fundamental molecular absorption features. When used in conjunction with chemometric and machine learning models, remote sensing platforms can be used to improve the scalability of SOC and N mapping. Nevertheless, there are some issues of transferability of calibration, soil heterogeneity, and barriers to adoption in developing nations like India, where the spectrum libraries of soils and the infrastructure are not well established. The review highlights the importance of preprocessing techniques, calibration policies, and advanced algorithms, including random forests, support machines, and deep learning, to enhance the accuracy of prediction. Future studies should focus on creating massive spectral libraries, incorporating multimodal streams of data, and deploying accessible digital platforms to facilitate precision agriculture. On the whole, soil spectroscopy can transform the concept of nutrient monitoring by transforming scientific innovation into field-level implementation.