Spatial Prediction of Soil Organic Matter with Soft and Axillary Data Using Bayesian Maximum Entropy Method

Soil organic matter (SOM) is one of the important soil quality factors and knowledge of its condition in soil, is one of the most important steps in the management of land resources and controlling soil losses. As SOM monitoring is an expensive and time-consuming task, any method which can produce high quality maps of SOM with available axillary soil data and less samples, would be a step forward in reaching the goals of sustainable agriculture. The aim of this research is to predict SOM using soft data, auxiliary data and Bayesian maximum entropy method (BME). Soil samples were gathered from the Bonab-Miandoab plain, and almost 122 samples were collected from 0-20 cm depth of surface soil. SOM and some other soil properties including soil texture, aggregate stability, and calcium carbonate equivalent were measured. Later spatial prediction of SOM was done using SOM soft data, auxiliary data and generalized linear model (GLM) using BME method. Results showed that the highest R, lowest RMSE and nRMSE with values of 0.97, 0.07 and 0.12 respectively, belonged to spatial prediction of SOM with soft data and error. Results also revealed that the developed GLM model with calculated error, resulted in better R, RMSE and nRMSE in comparison to predictions with GLM model without error (R, RMSE and nRMSE improved from 0.65, 0.58 and 0.55 to 0.85, 0.31 and 0.29 respectively). As a conclusion, BME method has provided the possibility of merging error resulted from the use of soft data, in spatial prediction equations and through that, has helped to improve spatial prediction of SOM.