Assessment of forest land change in southeast Pahang, Malaysia using remote sensing techniques

Tropical forest has been recognized as mostly disturbed natural land cover which contribute to deprived ecological balance. Malaysia is one of the primary exports of palm oil and timber in Southeast Asia. Hence, the objectives of this research mainly to evaluate the deforestation activity and predict the future extent of forest land in the highly deforested area in Pahang. These could be achieved through; (1) assessing the deforestation dynamics in Southeast Pahang using remote sensing, and (2) predicting the future spatio-temporal deforestation using CA-Markov model for year 2025 and 2035. The most critical issue is that forest land are usually inaccessible, therefore, remote sensing (RS) and Geographical Information System (GIS) are the important tools and recent approach for forest cover monitoring. The imageries from Landsat 5TM and 8 OLI satellite platforms were retrieved and categorized into four types of land cover including forest, nonforest vegetation, openland/ built-up, and waterbodies using Maximum Likelihood Classification (MLC). In addition, contour data was utilised to produce the elevation and slope maps, while Euclidean Distance analysis were applied to city centre, and Permanent Reserved Forest (PRF) for proximity maps. The overall accuracies are ranging from 83.7%, 80.9%, 85.6%, and 84.4% with Kappa value of 0.65, 0.65, 0.74, and 0.74 for 1990, 2000, 2010, and 2017 respectively. The results from land change analysis show that state land had changed from 42% of total forest area in 1990 to 21% in 2017 with a steady negative change over time. Dipterocarp reserved forest are consistently exploited for timber extraction, but the forest covers are able to be regenerated after more than 20 years from sustainable management practices. The distance to population centre has a positive relationship with deforestation, and the protected area have a clear restriction on deforestation in the area because forest loss inside protection region only happen after 20 years of study period compared to the outside of protected region. Furthermore, elevation and slope have similar effects on deforestation where the increase in their value will reduce the risk for deforestation. For land cover prediction, Markov chain integrated with cellular automata model were used for future forest land cover forecasting. The model calibrations achieved up to 62% accuracy for land cover prediction. The CA-Markov model prediction for year 2025 and 2035 suggests that the forest land cover will continuously reduce with 13 to 24 km2/year rate. Generally, state lands provide the highest level of deforestation in Rompin and Pekan district in both dipterocarp and peat swamp type, conversely the reserved forest in peat area are more protected compare to dipterocarp type. The comparison between using multiple and binary land cover as input suggest that the traditional CA-Markov model can simulate better when dealing with binary land cover. Other than that, the deforestation might be more than what were predicted in this study based on the standard error of the model.

Text MOHAMAD AL-EKHWAN - IR.pdf Download (987kB)

Actions (login required)

View Item