CHAPTER 2 LITERATURE REVIEW
2.6 The Efforts on Reducing Landslides Hazards
2.6.2.1 LHA in Cameron Highlands using Qualitative Methods
Omar, et al. [9] applied a qualitative method for constructing LHZ map for Pos Selim to Cameron Highlands areas using GIS and remote sensing data The study utilized factor maps, i.e. land use land cover (derived from Landsat TM5 image), slope gradient, slope aspect, and height/elevation. The last three maps were derived from DEM (Digital Elevation Model) of topographic map with a scale of 1:50,000. Multi temporal factors were not included in this work. All factors were assigned weight values before proceeding to GIS overlay process for constructing a LHZ map. The authors adopted the weight values from the works of others outside Malaysia rather than extracting from their own study area. For example, Equation 2.3, the probability of high risk model developed by Gao and Lao [81], was used to produce a height risk map. The authors used expert opinion from DeGraff and Romesburg [62] for producing aslope aspect risk map and made their own assumptions for creating a slope risk map. The final LHS showed three level of landslide risk: low risk, medium risk and high risk. About 6.21% of the area was categorized as high risk area while 83.93 % was low risk area. The accuracy of the developed LHZ map has not been know so far because this works did not include process of map validation.
Another investigation on LHZ that used a qualitative method, LHEF rating system, was carried out by Ramli, et al. [93]. The authors utilized GRASS (Geographical Resources Analysis Support System) which is open source GIS software for assessment of LHZ in Tanah Rata, part of Cameron Highlands district.
This investigation involved five landslide contributing factors namely lithology, slope angle, structure (lineament), relative relieve, hydrological conditions (water features), and land use land cover. Anbalagan [2] LHEF rating system was modified to suit with the available factors and respective classes. The final LHZ map was divided into five categories namely very low hazard, low hazard, moderate hazard, high hazard, and very high hazard. About 93 percent of the study area fell under low hazard category.
Meanwhile, about 15 percent and 0.2 percent of the study area were categorized as moderate hazard and high hazard respectively. None of the area was categorized as very high hazard area. Some notes regarding this work are that expert opinion in form of LHEF rating system was used showing a typical of a qualitative method; secondly,
this work did not take into account the temporal factors; and lastly, this work did not apply map validation, the stage which is important for measuring the accuracy of the developed LHZ map.
2.6.2.2 LHA in Cameron Highlands using Quantitative Methods
Gahgah, et al. [43] applied another qualitative method which is a combination of heuristic method that is, using expert opinion of Van Westen [111], and index overlay as GIS analysis method for investigation of landslide hazard in Cameron Highlands – Gua Musang road. The final landslide showed five categories of hazard namely very low hazard, low, moderate, high, and very high hazard. Slope and elevations were found to be the most affecting factors for landslide occurrence. This work has the same shortcomings as the previously mentioned work.
Talib [8] applied a bivariate statistical analysis method that is Information Value Method for investigation of slope instability and hazard zonation in Cameron Highlands. The study involved digital elevation data from topographic map, geology map, land use land cover map, distance to fault, drainage, and road map. Weight values were derived from pixel-based pair wise comparison between landslide map and factor maps. The final LHZ map divided the area into three categories: low, medium and high hazard. The author highlighted several places prone to landslide such as, sloping areas at the road and a gardening area in Bertam, Ringlet; some areas near the edge of urban areas in Tanah Rata where settlements were built on sloping hills; road slopes at Berinchang (north of Tanah Rata), Ringlet (near the main lake of CH), and slopes parallel to the Bharat the plantation. All these works have offered the objectivity of defining the relationship between landslide occurrences and causative factors which is implemented in defining a weighting system. However, temporal environmental factors have not been involved in the modeling LHZ and the verification test of the LHZ map was not mentioned in this work.
Back-propagation neural network model has been used to study landslide susceptibility in Cameron Highlands by Pradhan and Lee [12]. Ten landslide contributing factors were involved in this study namely, slope, slope aspect,
topographic curvature, lithology, soil type, rainfall, vegetation index (derived from SPOT 5 satellite image), distance from drainage and lineament. Using an advanced neural network model for analyzing all factors, landslide susceptibility map was produced. This investigation found that that the topographic slope has been the most influencing landslide factor followed by the distance to drainage, and lithology. This was indicated by their weight values resulted from back-propagation training method in succession 0.205, 0.141 and 0.117. The final landslide susceptibility map showed a good agreement, 83% accuracy, with landslide data after validation process. This work focused on the introduction of another quantitative method proven to be more satisfactory that the previous method in term of the accuracy of prediction of susceptible areas. The subjectivity of the expert was removed as this method is considered as data-driven modeling. This work did not take into account the temporal environmental factors.