Soil erosion prediction based on ANFIS and RUSLE model / Md. Rabiul Islam

• Main Author: Md. Rabiul, Islam
• Format: Thesis
• Published: 2018
• Subjects: TA Engineering (General). Civil engineering (General)
• Online Access: http://studentsrepo.um.edu.my/9093/2/MD_RABIUL_ISLAM.bmp; http://studentsrepo.um.edu.my/9093/11/rabiul.pdf; http://studentsrepo.um.edu.my/9093/
• Institution: Universiti Malaya

This study investigated erosion that have occurred at the slopes of Guthrie Corridor Expressway (GCE), Malaysia, by setting up eight experimental plots with existing vegetation (N) and re-vegetated (P) of size 8 m × 8 m. The treatment of vegetation densities were bare (B), less dense (LD- 50% coverage), and dense (D- 100 % coverage). The commercialized microbe fertilizer was then applied to few plots and known as the microbe treated (M) and non-microbe treated (NM- control) plots. This study was carried out on the basis of three important objectives. Firstly, derivation of vegetation cover and management factors (C) using a digital image followed by soil erosion prediction by the Revised Universal Soil Loss Equation (RUSLE) model in the Geographic Information System (GIS) environment. The actual density was computed as the C value using several established formulas the prediction of erosion rate determined from the proposed method and from the Malaysia guideline were compared. Result shows that C value from the proposed method were ranged from 0.016 to 0.125 and therefore, the erosion susceptibility was predicted low with the rate of 0.372 t ha-1 y- 1 to 0.842 t ha-1 y-1, respectively. As for the second objective, the effect of rainfall throughout the northeast monsoon (November till March 2014 to 2015) on erosion rate was investigated. Result shows that the maximum runoff and soil loss occurred in Natural Bare Non Microbes (NBNM) plot followed by the Planted Dense Microbes (PDM) plot whilst Natural Dense Microbes (NDM) plot was the lowest. In PDM plot, the average runoff and soil loss were found to be 7.84 mm m-2 and 1173.32 gm m-2 respectively, which were 92.16% and 35.99% less than that of NBNM plot. It can be concluded that the soil detachment, transportation and deposition rate are greatly influenced by the vegetation cover and topography. The third objective was to develop an erosion model using an Adaptive Neuro-Fuzzy Interface System (ANFIS). The novelty of this study was the introduction of a new parameter called microbes (Mf) in the RUSLE model. The model was developed based on the monthly erosion data for the year of 2015. The result shows that ANFIS model capable of accurately predicting the erosion with the correlation value (R2) of 0.83 between predicted and observed erosion. It was found that if the entire erosion conservation practices considered unchanged for a given period, the soil erosion would be inversely proportional to the value of Mf. Thereby application of microbes could be an alternative or additional erosion conservation technique for slope protection.