Remotely sensed data capacities to assess soil degradation

This research has tried to take advantage of the two-field based methodology in order to assess remote sensing data capacities for modeling soil degradation. Based on the findings of our investigation, preprocessing analysis types have not shown significant effects on the accuracy of the model. Conversely, type of indicators and indices of the used field based model has a large impact on the accuracy of the model. In addition, using some remote sensed indices such as iron oxide index and ferrous minerals index can help to improve modeling accuracy of some field indices of soil condition assessment. According to the results, the model capacities can significantly be improved by using time-series remotely sensed data compared with using single date data. In addition, if artificial neural networks are used on single remotely sensed data instead of multivariate linear regression, accuracy of the model can be increased dramatically because it helps the model to take the nonlinear form. However, if time series of remotely sensed data are used, the accuracy of the artificial neural network modeling is not much different from the accuracy of the regression model. It turned out to be contrary to what is thought, but according to our results, increasing the number of inputs to artificial neural network modeling in practice reduces the actual accuracy of the model.