Land use/Land Cover Change Analysis and Prediction in the Northwestern Coastal Desert of Egypt
Halmy, Marwa1; Gessler, Paul2; Hicke, Jeff2; Salem, Boshra1
1Alexandria University, EGYPT; 2University of Idaho, UNITED STATES
Land use/land cover (LULC) classes were mapped for three stages of modification of a landscape located in the northwestern coastal desert of Egypt. Landsat TM data and ancillary data were classified using random forest (RF) ensemble classification approach. The use of random forest (RF) to classify the Landsat TM images for the years 1988, 1999, and 2011 produced LULC classification with an overall accuracy over 90% and kappa index of agreement of more than 0.9. This indicates the merit of using this technique in mapping LULC in similar arid and semi-arid ecosystems. The analysis of the distribution of LULC classes over the three different stages revealed that the landscape has been subject to dynamic changes in land use/land cover over the last three decades. The change occurring in the area is a reflection of the policy by the Egyptian government towards expanding outside the heavily populated Nile valley and Delta through developing desert fringes. The strategy that has been taken for the northwestern desert is to develop it through tourism and agriculture reclamation projects. The summer resort development on the coastal dunes and the establishment of irrigation system in the area are likely the main causes of changes in the area as revealed in the current study. The analysis of the change in the spatial distribution of the LULC in the landscape reveals that the landscape has become patchier and less naturalized. The natural land cover (rangelands, coastal dunes, mixed barren lands, and wet & marsh areas) continues to suffer from dissection and attrition, while the artificial and semi-artificial land covers (resorts, built-up areas, quarries and croplands) are undergoing creation and aggregation. This trend is expected to continue in the future as is revealed through the simulation of the potential situation in the landscape by 2023. The use of the Markov Chain-Cellular automata model was successful in providing a prediction of the LULC distribution for the year 2011 and was comparable to the actual for 2011. The simulated potential distribution of the LULC classes in the landscape for the year 2023 shows predicted expansion in the croplands westward and northward and an expected increase in quarries and growth in the residential centers in the area. The changes that the landscape has experienced might continue and will likely impact the distribution of species in the area.