Spatial prediction of susceptibility to gully erosion in Jainti River basin, Eastern India: a comparison of information value and logistic regression models

• Published in: Modeling earth systems and environment 2019-06, Vol.5 (2), p.689-708
• Main Authors: Hembram, Tusar kanti, Paul, Gopal Chandra, Saha, Sunil
• Format: Article
• Language: English
• Subjects: Agricultural practices; Agronomy; Catchment area; Chemistry and Earth Sciences; Computer Science; Correlation analysis; Earth and Environmental Science; Earth Sciences; Earth System Sciences; Economic benefits; Ecosystems; Elevation; Environment; Erosion; Geographic information systems; Geographical information systems; Geology; Gullies; Gully erosion; Land cover; Land use; Math. Appl. in Environmental Science; Mathematical Applications in the Physical Sciences; Moisture content; Original Article; Physics; Rain; Rainfall; Regression analysis; Regression models; Remote sensing; River basins; Rivers; Satellite navigation systems; Slopes; Soil; Soil types; Statistical analysis; Statistical methods; Statistics; Statistics for Engineering; Training
• ISSN: 2363-6203; 2363-6211
• DOI: 10.1007/s40808-018-0560-8

The research on effective methods for assessing areas susceptible to gully erosion has environmental, agricultural and economic benefits to society. This study aimed to identify the areas with high susceptibility to gully erosion using two statistical methods and compare the performances. Remote sensing and field data along with two statistical methods, information value (InfVal) and logistic regression, were applied in the analysis. Gullies, boundary demarcation and regulating erosion processes were surveyed and mapped in a GIS environment. Gully sites were split into training and validation set for modelling and validating susceptibility results. The geo-environmental variables selected, according to study area, were land use/land cover, slope steepness, slope aspect, elevation range, length-slope factor, topographical wetness and position index, stream power index, geology, soil type, rainfall erosivity index and altitude. Weights for each class of each variable were computed through InfVal method and probabilities of occurrence of 1,000,000 random locations were extracted to run the logistic regression operation. A weighted linear combination approach and an inverse distance weighted approach were applied for the InfVal and logistic regression methods, respectively, to generate the gully proneness maps. Both results were classified into five susceptibility classes and they depict 18.44% (Infval) and 12.67% (logistic regression) area of the catchment as high to very highly susceptible to gully erosion. Correlation analysis between the results revealed a coefficient value of 0.8379. For InfVal model, the success and prediction rate is 78% and 75% whereas for the logistic regression model, it is 73% and 72.9% accordingly.