Digital mapping of soil health card parameters and nutrient management zones in the Thar Desert regions of India using quantile regression forest techniques

A soil health card (SHC) aims to enhance productivity sustainable through wise use of fertilisers. The current work focused on delineating management zones (MZs) using SHC data by digital soil mapping and fuzzy clustering data mining techniques, as well as developing a simplified approach for evaluating zone maps in Thar Desert. The soil maps were predicted using a Quantile regression forest model which included 19 environmental covariates. The pH, EC, and OC varied from 6.20 to 10.6, 0.01 to 50 dS m−1 and 0.01 to 0.98%, respectively, in the study area. The available P and K varied from 2.23 to 87.0 kg ha−1 and 30 to 669 kg ha−1. The mean values of Zn, Cu, Fe, and Mn were 0.60, 0.65, 3.41, and 6.26 mg kg−1, respectively. While predicting SHC parameters, the model was captured 28 to 65% of variability. Concordance correlation was ranged from 0.48 to 0.75, showing that the predicted and observed values were in good agreement. The prediction interval coverage probability values were 89.7–92.1%. Furthermore, MZs were delineated using principal component analysis and fuzzy k-means clustering approach based on optimum clusters determined using the fuzzy performance index and normalised classification entropy. The area was divided into two zones, each of which could be managed differently. The MZ map, which meets the criteria of management zones to be simple, easy to understand that might be the guide to farmers adopting better nutrient management.