4. Use of remote sensing and GIS in the site selection process
A land use/cover map of the Langkawi Island was produced from satellite data of 1995 and field data using a Maximum Likelihood Supervised classification (Baban and Wan Yusof, 2000). The land use/cover map was used to extract information and construct layers of information for constraints 1, 4 and 5, i.e. to identify urban areas, to protect agricultural and forest areas respectively. IDRISI, a raster based, GIS was employed to create information layers corresponding to each constraint. In implementing the criteria, these relevant information layers were combined using two different methods. First, the Boolean method which considered all the layers as being equally important and gave them an equal weight. Second, the Weighted Linear Combination (WLC) method which grouped the layers and graded them according to their perceive importance.

In the Boolean method, all the criteria are reduced to constraint Boolean images of areas which are suitable and not suitable. The constraint layers were subsequently overlayed consecutively, by using the OVERLAY multiply function to produce a single suitability Boolean image. Subsequently, the suitable areas in the last Boolean image were regrouped to determine the surface area. The output image was achieved by using the GROUP and AREA functions of IDRISI. The GROUP function was used to assign identifiers to unite groups of pixels and the AREA function was used to calculate the surface area for each group. Finally, the RECLASS function was used to select only those areas (groups) that can comply with the specified surface areas (Table 1) (Fig.2).

In the WLC technique, the criteria are standardised to a continuous scale of suitability from the least to the most suitable, thus giving a flexibility in the site selection (Eastman, 1997). The procedure in the WLC required that the principal eigenvector of the pairwise comparison matrix be computed to produce the best-fit set of weights. Subsequently, the acceptable best fit of the respective weights were used in the Multi Criteria Evaluation function (MCE) to calculate the weighted linear combination (WLC) using the factors (slope, height, land value, and erosion) and constraints (forest reserve, settlement zone, and geological foundation) images. Then using the GROUP, AREA and RECLASS functions again, potential sites with the highest suitability were produced (Fig.3).

5. Results and discussions
A field study to locate reservoir sites on the Langkawi Island based on topography, catchment areas, hydrology, land use/cover, accessibility and socio-economic factors has identified six possible reservoir sites (Fig.4) (Syed, 1992). These areas are the Limbong, Ulu Melaka, Upper Ayer Tawar, Lower Ayer Tawar, Batu Asah and Langkanah. This information has been used to evaluate the outcomes from the two methods employed in this study.

In the Boolean approach, the selected areas were absolute i.e. there were no trade-offs with other constraints criteria. Consequently, suitability in one criterion cannot compensate for a lack of suitability in any other image (Eastman, 1997). In terms of risk, this approach is very conservative. The final outcome of potential sites produced satisfied all the criteria: the reservoir sites selected were within the surface area required, were outside the settlement zone, located at an altitude between 25-90m, were on a gentle slope of 0 to 11^o, have a strong foundation on either granite or