Assessing and transferring soil health information in a hilly terrain

Spatial variability of soil health related variables in a hilly terrain may be high, and its characterization may require many samples. Our research compares deterministic and geostatistical interpolation methods in two hilly areas in India. The soil in the study area was acidic, without salts and with sufficient organic carbon content. Hence, three soil parameters – pH, Electrical Conductivity (EC) and Total Organic Carbon (TOC) were considered. The optimal sampling scheme was designed using Spatial Simulated Annealing (SSA) with the minimized kriging variance as a criterion. This resulted in 96 locations in the first area and 7 locations in the second area. It was explored how spatial information from one area could be used in a second, topographically similar area. The study focused on pH as the key variable for soil health. Regression kriging performed best for all the soil variables at the surface and sub-surface levels. Bayesian kriging allows one to use prior information and hence was used to transfer from the first to the second area. A mean error of 0.15, a root mean square error of 0.28 and a residual variance equal to 0.73 respectively were observed. We conclude that with modern interpolation methods important information on soil health can be collected even with sparse amounts of data.