Answers are grounded in this paper and cite their sources.
Demonstrates accurate soil quality prediction in temperate forests, indicating a shift towards cost-effective assessments.
Soil quality is a key indicator of ecosystem functioning, resilience, and sustainability in forest landscapes; however, its assessment often relies on labor-intensive field and laboratory measurements. To address this limitation, this study integrates a principal component analysis (PCA)–based soil quality index (SQI) with machine learning models to enable accurate and cost-effective soil quality prediction in the Hyrcanian forests of northern Iran. Using soil physical, chemical, and biological data from 40 sampling points, the performance of random forest (RF) and support vector machine (SVM) models was evaluated for predicting SQI derived from a minimum dataset. The RF model outperformed the SVM, achieving higher predictive accuracy (R2 = 0.93 for training and 0.77 for testing) and lower prediction errors (root mean square error = 0.03 to 0.05; mean absolute error = 0.02 to 0.04). Permutation-based variable importance analysis identified soil organic carbon (SOC) as the most influential predictor, followed by bulk density and clay content. Partial dependence analysis revealed nonlinear relationships, showing that moderate SOC and clay levels enhance soil quality, whereas excessive clay content, higher bulk density, elevated pH, and electrical conductivity reduce SQI. Overall, this study demonstrates the effectiveness of RF models for scalable and cost-efficient soil quality assessment, supporting sustainable forest management in temperate forest ecosystems.
No takes yet. Share an insight, caveat, or question.
Tafazoli et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: