Answers are grounded in this paper and cite their sources.
Soil fertility plays a vital role in sustaining agricultural productivity, and accurately assessing soil fertility indices is essential for informed and effective land management. This study aimed to predict the Soil Fertility Index (SFI) by integrating soil physical and chemical properties with visible–near infrared (Vis–NIR) spectral data and environmental variables. Spectral preprocessing techniques including first and second derivatives, detrending, moving window average, standard normal variate (SNV), multiplicative scatter correction (MSC), and Savitzky–Golay filtering, were applied to enhance spectral features and reduce noise. Five machine learning algorithms (Partial Least Squares Regression (PLSR), Principal Component Regression (PCR), Bagging-PLSR, Support Vector Machine (SVM), and Random Forest (RF)) were employed to develop 35 hybrid models combining preprocessing methods and algorithms for SFI prediction. Environmental variables derived from a digital elevation model (DEM) and Sentinel-2 satellite imagery were selected using principal component analysis (PCA) and integrated with spectral data to investigate their impact on prediction accuracy. The results demonstrated that the first and second derivative preprocessing techniques significantly improved model performance by reducing baseline effects and emphasizing chemical and structural features of the soil, with the second derivative providing particularly precise detection of subtle spectral variations. Among the tested models, SVM outperformed, achieving the highest prediction accuracy. Integration of spectral and environmental data further enhanced model performance, yielding an R2 of 0.82, RMSE of 8.53, and full coverage within the 95% prediction interval, indicating robust and reliable predictions with reduced uncertainty. These findings highlight the complementary role of environmental covariates in capturing soil heterogeneity and demonstrate that combining optimized spectral preprocessing, advanced machine learning algorithms, and auxiliary environmental data offers a rapid, accurate, and cost-effective approach for soil fertility assessment and precision soil management. This approach provides valuable insights for sustainable agricultural planning and resource-efficient fertilization strategies.
No takes yet. Share an insight, caveat, or question.
Nazari et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: