Evaluation of climate change impacts on future gully erosion using deep learning and soft computational approaches

Abstract

The impact of climate change and its associated extreme rainfall event is the major threats of land resources in subtropical monsoon dominated region. In this study, we have considered the Support Vector Machine (SVM), Analytical Neural Network (ANN) and Deep Learning Neural Network (DLNN) for the estimation of gully erosion susceptibility in Dwarkeswar River basin. The ensemble Global circulation Model (GCM) has been considered for simulating the rainfall scenario in the projected period, i.e. 2100s. All selected parameters' VIF (variance inflation factor) and TOL (tolerance) ranges are 1.05 to 11.56 and 0.20 to 0.95, respectively. In the training dataset, AUC values of SVM, ANN and DLNN are 0.95, 0.95 and 0.96, respectively. In the case of validation, AUC values are .86, 0.87 and 0.90, respectively. Here, the DLNN is the most optimal model in terms of the predictive capacity. The AUC (Area Under Curve) values from ROC (Receiver operating characteristics Curve) of the DLNN model for training and validation datasets are 0.96 and 0.90, respectively. The values of sensitivity, specificity, PPV and NPV in the case of validation datasets in DLNN are 0.88, 0.76, 0.80 and 0.79, respectively. There is an increasing tendency of rainfall and gully erosion susceptibility in the projected period. This type of information is helpful to the decision maker in this respective region to implement the long-term planning for escaping this type of situation.

Keywords:

• Climate change
• deep learning neural network
• global circulation model
• gully erosion susceptibility

Disclosure statement

There is no conflict of interest.

Data availability statement

The data that support this study are available upon request from the authors.

Additional information

Funding

The authors acknowledge the NRDMS, DST for providing financial support (NRDMS/01/143/016) to carry out this research.