Diffuse reflectance spectroscopy is gaining traction as a rapid, low-cost method for soil fertility testing. A new critical review in Discover Soil examines decades of research on the technique, which analyzes how soil reflects visible and infrared light to infer properties like organic carbon, clay content, and moisture. The review concludes that some properties are well predicted, while others like plant-available phosphorus and micronutrients still require traditional lab analysis. The paper stresses that spectral preprocessing and model choice matter as much as the instrument. Partial least squares regression remains a strong baseline, and deep learning offers gains mainly when trained on large datasets. Transfer learning from global libraries to local regions shows promise but needs independent validation. For soil carbon monitoring and precision agriculture, spectroscopy could enable landscape-scale assessments at a fraction of the cost. However, the review cautions against overreliance on global models and points to privacy issues as digital soil maps link fertility data to specific landholdings. The technology is ready for complementary use, not replacement.
