Soil Organic Carbon (SOC) plays a key role in soil health and climate regulation. Accurately measuring SOC is especially important in dry regions, where soil degradation is a concern. In this study, we used advanced lab-based hyperspectral imaging to predict SOC levels from soil samples. We improved our prediction models by applying fractional-order derivatives (a type of mathematical transformation), creating new 2D and 3D spectral indices, and including soil pH as an extra input. Our best model explained 87% of the variation in SOC, showing strong performance under controlled lab conditions.

To explore how well this method might work with satellite data, we simulated multispectral satellite bands from our lab data. These lower-resolution models still achieved solid results, explaining about 70% of SOC variation. This suggests that while lab methods are more accurate, simplified satellite-based approaches could still be useful for monitoring SOC over large areas. Our findings highlight the importance of using the right features and preprocessing steps to successfully scale up SOC estimation from the lab to real-world landscapes.