• DRS-based digital soil mapping can enable soil testing in every smallholder farm. • Spiking improves DRS model accuracies both within and outside calibration zones. • Relative importance of covariates changes with measured and DRS-estimated DSM inputs. Diffuse reflectance spectroscopy (DRS) and digital soil mapping (DSM) offer opportunities to rapidly assess soil in large areas. Specifically, the combined DRS-DSM modelling pipeline may be used to create soil test recommendations for every smallholder farm in a given region although comprehensive testing of such a pipeline is rarely attempted. With multi-year and multi-site soil spectral data from the smallholder farms of the Bundelkhand region, we evaluated the DRS-DSM pipeline for estimating soil properties and making nutrient recommendation for every smallholder farm both within and outside the DRS calibration zones. Specifically, we compared both measured and DRS-estimated soil properties as inputs in DSM approaches using 1112, 607, and 407 soil samples collected during 2018 (T 2018 : calibration zone), 2021 (T 2021 : within the calibration zone), and 2022 (T 2022 : outside the calibration zone), respectively, for estimating 17 soil parameters and their soil test crop response (STCR) ratings. For T 2022 samples, DRS models calibrated within the calibration zone accurately predicted 7 out of 17 soil properties with Lin’s concordance correlation coefficients (LCCC) exceeding 0.6. Spiking these datasets with T 2022 data further improved predictions to 10 properties and reduced errors by 3–29%. In T 2021 dataset, both measured property- and DRS-based DSM approaches achieved comparable accuracy. Estimated STCR rating accuracies for the DRS-DSM pipeline exceeded 70% for 9 out of 13 properties suggesting that these two emerging technologies may be combined to make nutrient recommendations across smallholder farms within a given region.
Purushothaman et al. (Wed,) studied this question.