Uncertainty Assessment of the Soil Salinity Boundary Prediction in Urmia Plane Using Bayesian Maximum Entropy Method.

In recent years, decrease in depth of Uromia Lake, has resulted in higher increase of salinity threat in agricultural lands around the Lake. The aims of this study were 1- Investigation of the spatial changes in soil salinity using the Bayesian Maximum Entropy method (BME); 2- prediction of the boundary between saline and agricultural lands; and 3- Assessment of the uncertainty involved with salinity boundary prediction in South Uromia Plane, North West of Iran. ME and MSE criteria were used for comparison of the results. Prediction error variance resulted from BME equations, was used for assessing the uncertainty of the soil salinity boundary prediction. Cross-validation results showed that BME method with ME and MSE equal to 0.42 and 0.33 for autumn 2009, and 0.2 and 0.64 for spring 2010 respectively, had a high accuracy in spatial prediction of soil salinity, considering that only probabilistic type soft data was used. A sharp boundary was detected between salty and non-salty lands in the area and the BME method had a high ability in predicting the uncertainty involved with boundary predictions. Monitoring soil property variation such as salinity is sometimes limited because it is costly and requires time to gather necessary the hard data. BME method has shown potential for using soft data in cases that hard data are not readily available.