Abstract
A soil erosion risk map was developed for Langkawi Island, Malaysia using the Universal Soil Loss Equation (USLE), remote sensing and GIS. Spatially modelling soil erosion in the GIS required generating representative raster layers based on secondary data for the following parameters; rainfall erosivity, slope length/gradient, soil erodibility and conservation practices. A Landsat TM imagery dated March 1995 was utilised to produce a land use/cover map of the Island based on the Maximum Likelihood Classification method. This map was then, used to generate the conservation practice factor in the USLE. The analysis was performed using IDRISI, a raster based GIS software. Upon comparison, the produced erosion map showed significant similarities with an erosion risk map of the Island produced by conventional means in 1995. The majority of high erosion risk areas seem to be confined to the highlands. This study demonstrates the effectiveness of remote sensing and GIS in generating soil risk maps. The produced erosion risk map is a valuable resource for planners to minimise soil erosion problems caused by future and ongoing development projects on the Island.

Introduction
Land degradation has always been associated with the failure to identify areas that are prone to soil erosion. Moreover, land surveying using conventional method is expensive and time consuming. In contrast, mapping soil erosion using the integration of remote sensing and GIS could identify areas that are at potential risk of extensive soil erosion, provide information on the estimated value of soil loss at various locations. In addition it can provide answers to spatial queries; for instance whether the erosion is associated with specific factors such as the loss of continuos vegetation cover (Kertesz, 1993). This information is very useful in the decision making context to avoid land acquisition in the erosion risk areas or alternatively to recommend soil conservation measures to reduce soil loss if developments were to continue. The Universal Soil Loss Equation (USLE), developed by Wischmeier and Smith (1978), has been used within a Geographical Information System (GIS) frame work to calculate the total erosion loss. The spatially distributed parameters involved in the equation such as topography and land use, could be generated by remote sensing techniques (Moore and Wilson, 1992). These can then be converted into raster layers to input into a GIS to be analysed and, to produce a soil erosion risk map (Lal et. al., 1990; Perez et. al., 1993). In Malaysia, several soil erosion studies have been conducted using this approach. These include the soil erosion study of the Bakun Dam project (Samad and Abdul Patah, 1997) and soil erosion risk assessment for Genting Highlands (Jusoff and Chew, 1998). This study aims to develop and evaluate a preliminary soil erosion risk map for the Langkawi Island, Malaysia using the USLE, remote sensing and GIS.

The study area
Langkawi Island is located west of Peninsular Malaysia between longitude 6°10’ to 6°33’ and latitude 99°35’ to 100°0’(Fig.1). It has an area of 362 sq. kilometres. Langkawi Island is mainly covered by forest, mangroves, shrub, agricultural land, sand beaches and newly expanding developed urban areas. The topography varies from flat coastal plains, hilly areas to rugged mountains. Langkawi Island was selected for this study due to the rapid physical development in the Island since its declaration as a duty free port in 1987. This development has changed the existing land use/cover tremendously and without proper planning and management land degradation will inevitably occur.

Fig. 1 Study area
Methodology

Generating raster layers for USLE parameters from secondary data.
The USLE expresses the rate of soil erosion as soil loss (Wischmeier and Smith, 1978) and can be written as:
A = R*K*L*S*C*P              (1)
Where A is the computed annual soil loss (kg m^{- 2} year^{- 1}), R is rainfall erosivity factor (dimensionless), K is soil erodibility factor (dimensionless), L is slope length factor (dimensionless), S is slope steepness factor (dimensionless), C is cropping and management factor (dimensionless), P is conservation practices factor (dimensionless).

In this study, R factor was derived from the mean annual rainfall data collected from different rainfall stations in the Island (Shaaban and Sahat, 1995), K values were estimated from the digitised soil map of a 1984 semi-detailed soil map of Langkawi of scale 1:100 000. This was then rasterised and converted into a K value map by reclassing each soil polygon into its corresponding K value according to Morgan (1986). The soil K erodibility value was determined using a small portable rainfall simulator (Sulaiman et. al., 1995). LS values were computed from equation (2) (Morgan, 1986) :

where, L is slope length in m and S is slope steepness in percent (%).

A Digital Elevation Map (DEM) for Langkawi Island scale 1: 100 000 developed by Taher (1996) was utilized to produce the LS factor. A series of maps OVERLAYING and SCALAR operation in IDRISI for slope length (L) and slope steepness (S) were performed on the DEM to produce the LS map. The C factor is defined as the ratio of soil loss from cropped land which reflects the combined influences of vegetation cover, crop types, tillage and management techniques. C values were obtained from Morgan (1986). P factor takes into account the effectiveness of erosion control practices, such as contouring, terracing and strip-cropping (Jusoff and Chew, 1998).

Producing land use/cover map using remote sensing
A Landsat Thematic Mapper (TM) image with a spatial resolution of 30m for Langkawi Island taken in March 1995 was used to produce a land use/cover map. The image was geometrically and atmospherically corrected. A stratified supervised classification using Maximum Likelihood Method and ground referenced data was carried out to generate the land use/ cover map for the Island (Fig.2). This map layer was converted to the C factor layer in USLE through reclassification of each cover type into its corresponding C value according to Morgan (1986).

Fig. 2 Land use/cover map of the Langkawi Island (after Baban and Kamaruzaman, 1999)