Spatial Prediction of Soil Micronutrients using Supervised Self-Organizing Maps

Abstract

Supplementing optimal quantities of nutrients in soil is considered to be one of the challenging tasks faced by farmers mainly because of the lack of information on the status of soil nutrients. Quantitative estimation of soil elements across agricultural farms is a hard problem to address. Although there are several studies proposed to spatially predict the soil nutrients employing geostatistical, computational, or by using AI techniques, most of these methods either do not have sufficient accuracy or perform well only with datasets similar to the model building dataset. In this study, we propose supervised Self-Organizing Maps (xyf-SOM) for the first time to quantitatively and spatially predict soil micronutrients viz. Boron, Iron, Manganese, Copper and Zinc. Soil nutrient data (2594 samples) pertaining to Alappuzha District, Kerala, India, collected during 2019–20 was used for the study. Geo-environmental predictors generated from remote sensing data such as topography, vegetation, land surface temperature, and precipitation were used as explanatory variables. The prediction accuracy was compared with Regression Kriging and Random Forest based spatial prediction. The results showed that supervised Self-organizing Maps predictions resulted significantly high and consistent prediction accuracy for all micronutrients when compared with Geostatistical and random forest predictions. The models were validated with test data set as well as with an independent dataset. The prediction model was applied to a data grid with a 200 x 200-m spatial interval, and the prediction results were converted and visualized in a geospatial framework.

Keywords

• Spatial Prediction
• Self-Organizing Maps
• Regression Kriging
• Soil Nutrients
• Soil Boron