Assessing the elevational relationship with soil physico-chemical properties under traditional large cardamom agroforestry system in the Indian Eastern Himalayas: a case study using machine learning algorithms

Traditional large cardamom agroforestry system in the Indian Eastern Himalayas is known for their important role in soil conservation and fertility management. The present study evaluated the impact of altitude and soil depth on soil organic carbon (SOC) storage and soil nutrient status under this system. Standard analytical methods were adopted for quantifying the soil physico-chemical parameters. We used machine learning (ML) techniques and interpreted the result with a test of correlation followed by a test of statistical significance to evaluate the correlation employing XGBoost, SVR and RFR models. The study reveals altitudinal location significantly influences soil nutrient storage and fertility indices. Among the fertility indices, soil fertility index exhibited inconsistent trend with increasing altitude class while, soil evaluation factor consistently decreased with increasing altitude class. With an overall score of R-square = 0.95 and RMSE = 1.19, the XGBoost model outperformed all others in predicting SOC content. Present study suggests altitudinal gradient plays an important role in soil nutrient management. ML algorithms are crucial in quantifying estimating soil carbon stocks and predicting the carbon sequestering potential of diverse land uses and managements. However, there is need for integrating data from satellite imagery, ground-based sensors, and drone technology with adequate technical expertise for processing and analyzing these multi-modal datasets. Such insights will empower farmers to optimize fertilizer utilization, enhance crop productivity, and implement sustainable soil management practices.