|
|
:
|
Forecasting the occurrence of gully erosion using different models and preparing risk zonation maps is the most suitable solution for land management planning in watersheds. The aim of this study is too zonate the risk of ditch erosion in Qara Toureh village of Bijar city in Kurdistan province, which has been achieved using frequency ratio methods, logistic regression and random forest algorithm, and a comparison between these methods Was performed. The village of Qara Toureh has shown great sensitivity to the danger of ditch erosion and damage has been done to this area and it has led to the production of sediment and its threats in the future. In order to reduce the damages of this phenomenon, scientific evaluation of these areas sensitive to the occurrence of Gully erosion is necessary. The location of the ditches around the studied village with an area of 353.61 hectares was recorded with the help of satellite images and control and visit of these points during field operations. Fourteen conditional factors in the occurrence of gully erosion including layers: geology, Aspect, degree of slope, Land use, Stream density, Stream Distance, Stream power index, topographic Witness index, soil depth and permeability, fault density, Fault Distance, the height above the sea level and the annual rainfall have been identified as effective factors in the occurrence of gully erosion in the study area and were used to prepare a sensitivity map to gully erosion in the study area. The spatial coordinates of 157 ditch points were recorded in the studied village, and about 30% of the recorded ditches (42 ditch points) were completely randomly isolated as experimental ditches, and the remaining 115 Gully erosion points were designated as ditches. Education was included in the analysis. In the method, the frequency ratio of assigned weights and values to each of the factors was done using statistical calculations in Excel software, and after algebraic summing of all layers with the help of raster calculations; The susceptibility map to gully erosion was prepared using the abundance ratio method. In the logistic regression method, after performing progressive step-by-step regression analysis, five effective factors in the occurrence of Gully Erosion among fourteen factors include: Land use, Aspect, Stream Distance, Stream density and soil hydrology group due to its meaning. The presence was statistically selected. The program related to the random forest algorithm was written under a main program in the R software environment, and after the execution of the obtained weights model, it was entered into the Arc GS 10.8 environment and a trench erosion risk zonation map was prepared. The decreasing average of precision and the average of Gini reduction were used to determine the priority of the influence of each effective factor and it was found that it includes the following factors: lithology, slope, Land use, Stream density and soil hydrology group in the order of greatest influence and the factors of soil depth, rainfall, The Fault Distance and the Fault density had the least effect on the erosion of the ditch in the studied area. The prediction accuracy of the different models used in this study was measured using the ROC curve. The area under the ROC curve in the frequency ratio (FR) method was equal to 0.91 percent and obtained the highest value. And the logistic regression method with the area under the ROC curve equal to 0.87 Percent showed very acceptable estimates in the zonation of ditches in the studied area. The area under the curve for the random forest algorithm model was also equal to 0.84 percent and it shows acceptable accuracy. The models used in this research in combination with GIS for the zonation of trench erosion areas are simple, fast and correct and will greatly help the planners and decision makers of issues related to environmental hazards.
|