SUPPORT PRACTICE CLASSIFICATION MODEL BASE ON SELF ORGANIZING MAP IN RUSLE EROSION ESTIMATION MODEL (Studi Area: Cidadap Sub-District and Lembang Sub-District - Bandung Basin)

• Main Author: YUSUP SAPTARI (NIM : 35110002); Pembimbing : Dr. Ir. S. Hendriatiningsih., MS; Dr. Ir. Agung Bu, ASEP
• Format: Dissertations
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/19935
• Institution: Institut Teknologi Bandung

Support Practice or P factor is one of the crucial factors in estimating erosion rate by using Revised Universal Soil Loss Equation (RUSLE). RUSLE is an erosion model designed to estimate annual erosions. P factor in the RUSLE erosion estimating model is computed when determining the types of acts on land that may <br /> decrease or increase erosion rate. The determination of Self Organizing Map (SOM)-based P factor value is a method of applied to find the P factor value of a land area automatically by utilizing the capacity of SOM Artificial Neural Network, so as to expedite a policy making on determining the extent of human act effect on <br /> land that may result in an accelerated erosion rate. The classification of the values of P factor produced is then used to find out the spread of erosion rate in an area by using RUSLE model. The problem thus far is that the observation of P factor should be performed directly in the field, and thus it will take a long time if mass <br /> observations are to be made in case the area observed is very wide and there are a large variety of P factor types to observe. This is because P factor must be classified systematically to be eligible as a factor in estimating RUSLE erosion automatically. <br /> In the present research, a P factor classification model that is fast, accurate, and may be performed massively by analyzing the data of land surface of Digital Terrain <br /> Model (DTM) data was created. A morphometric parameterization process was carried out to produce morphometric parameters such as: slope, maximum curvature, minimum curvature, and cross sectional curvature (4 dimensions) by using DTM data. By using an artificial neural network method, SOM was conducted to reduce morphometric parameters into 2 dimensions by a morphometric parameter training, steadily keeping their <br /> topological relations. SOM training was applied in data quantization stage to produce neuron as Best Matching Unit (BMU), a stage well known as self organize. <br /> BMU is a winning neuron in a concept of competitive learning used as a reference of other neurons in establishing a classification of morphometric parameter. In a neuron classification stage known as mapping, the result of morphometric c classification was transformed into P factor through an analysis of Wischmeier and <br /> Smith's P factor table. Data of rain fall, soil type, and slope length and gradient, and data of vegetation cover were derived from NDVI obtained from landsat satellite image data, implemented into a RUSLE erosion estimation model in Cidadap and Lembang Sub-districts.