ABSTRACT

Rate and distribution of erosion are important to know in management of watershed, because it is related to priority of rehabilitation and conservation of soil. Integration of digital image analysis and Geographic Information System (GIS) is one method to assess rate of soil erosion that is able to reduce the cost and time of research. Nevertheless, capability and accuracy of the method need tobe examined before being used widely. The study relating to erosion assessment that was carried out in the downstream of Ayung watershed, Badung Regency, Province of Bali. The objectives of the research are : (1) to know ability and accuracy of Landsat TM digital image analysis and GIS in presenting the erosion estimation variables, and (2) to estimate and to compare the erosion assessment result with direct measurement method of erosion variables in the field, in previous research at the same watershed.

Calculation of erosion rate used Universal Soil Loss Equation (ULSE). Erosion components of the USLE formula was analyzed in digital form using ILWIS (Integrated Land and Water Information System) software version 1.4 that was able be used to digital image processing and GIS based analysis. Rainfall erosivity index (R) was obtained through computation of the R-value at each rainfall station in the study area and its surrounding, and gridding process. Soil erodibility (K) was calculated from soil properties at each soil map unit and then it was linked with the graphic data. Length and slope factor (LS) was derived from topographic map by way of digitizing, rasterization, and interpolation of contour lines and elevation points. Plant and land management factor (CP) were based on landuse map that was gotten from multispectral classification of Landsat TM image. Erosion rate in ton/ha/year was obtained from calculation of the fourth erosion component maps. Accuracy test was applied for landuse, slope, and soil maps by means of compare analysis result with actually condition in the field. Comparisons of soil erosion estimation resulted from the method and measurement in the field was done with statistic test (t-test).

The result shows that 60.15% of erosion rate in study area is classified as very slight to slight (less than 29.3 ton/ha/year), 14.58 % is included rather heavy to very heavy (more than 47 ton/ha/year), and 25.27% is grouped as moderate erosion level (29.4 ? 46.9 ton/ha/year). The accuracy level of landuse, slope, and soil maps are 83.56%, 87.67%, and 86.30% respectively. It means that the accuracies are higher than the minimum limit of accuracy level acceptable. Comparison of soil erosion prediction result with early research through statistic test showed that t-calculation value is 1.02 less than t-table 5% value (1.98). Thus, integration of digital image analysis and GIS could be used to predict rate of erosion with acceptable capability and accuracy.