Evaluation of the effects of uncertainty on the predictions of landslide occurrences using the Shannon entropy theory and Dempster–Shafer theory

Abstract

One of the requirements for planning and decision-making to develop the infrastructures is to prepare the landslide occurrence hazard map. For this purpose, in this article, the Shannon entropy and Dempster–Shafer intuition theory methods were used to prepare the hazard map and applying the data uncertainty in the Tutkabon region, Guilan Province. In this study, parameters of the slope, elevation, geomorphological conditions, the curvature of the earth, proximity to the river and proximity to faults were used as the affecting factors on the landslide occurrence. By using these parameters, the map of the landslide occurrence hazard was prepared using the entropy index; besides, the belief values were calculated by the Dempster–Shafer method. To investigate the uncertainty, the disbelief and uncertainty values were calculated by the Dempster–Shafer method. Besides, in the Shannon entropy method the maps were compared before and after applying the entropy. Finally, by evaluating the results using comparing the landslide occurrence places in the study area and modeled hazard map, the value of 0.69 was obtained for the area under the prediction rate curve (as the parameter of prediction total precision) in the status with entropy and the value of 0.54 was obtained for the status without entropy. Similarly, the evaluation of the hazard belief map by the Dempster–Shafer method indicates that 65% of the landslide occurrence places occur in the classes of high and very high hazard and the value of 0.74 was obtained for the area under the prediction rate curve in the belief map.