Application of statistical and machine learning approach for prediction of soil quality index formulated to evaluate trajectory of ecosystem recovery in coal mine degraded land

Analytical tools for evaluating reclamation success in terms of soil quality are the prime requisite in order to address sustainable mining issues. Identification of the influential parameters of reclamation success and prediction of soil health in terms of soil quality indexing and modelling remain obscure. Therefore, the present study aims to formulate reclaimed mine soil quality index (RMSQI) using the indicator selection method and scoring function analysis. Furthermore, for identification of the relationship between RMSQI and predictor variables, statistical (Multiple linear regression; MLR) and machine learning (Random Forest; RF) approaches were applied to determine the suitable soil quality assessment technique for chronosequence afforested post-mining site in India. The results indicated that the organic carbon, exchangeable potassium, cation exchange capacity, sand percentage, microbial biomass carbon, dehydrogenase activity and fluorescein diacetate acitivity were the most influential variables impacting soil quality. The linear scoring-minimum dataset (LSF-MDS) was considered as the most suitable approach to estimate RMSQI due to its highest correlation coefficients F, and coefficient of variance (CV) values. The lower error matrices and higher R2 for RF than MLR indicating the aptness of RF model in predicting soil quality in terms of evaluation of reclamation success. The present study concluded that the superiority of RF is associated with its ability to address hierarchical and non-linear relationship between RMSQI and its predictors.