Hybrid Classification Approach and NDVI for Land Cover Change Detection in a Newly Reclaimed Area of Egypt.

Land cover change detection using remotely-sensed data provides effective and accurate evaluation of human impact on the environment. Agriculture is a key element of the Egyptian social fabric and economy. The Egyptian government adopted policies aimed at extending cultivated lands near the fringes of the Nile delta. The Bustan 3 area, located in the desert region of the west delta in Egypt, was targeted for reclamation in the 1990s, and occupies ≈34,127 ha (341.27 km²). The main objective of this study was to explore the change in land cover in the Bustan 3 area using multi-temporal Landsat satellite images from 1984 to 2008. Temporal change detection was achieved using both a hybrid classification approach (composited supervised and unsupervised classification techniques) and normalized difference vegetation index (NDVI) from 1984, 1990, 1999, 2004, and 2008. The hybrid classification results showed that this area involves four land cover classes: urban or built-up land, agricultural land, water, and barren land. From 1984 to 1990, the study area was mainly barren land. However, change detection results showed that reclamation processes in the 1990s produced remarkable changes in land cover. NDVI results indicated that the area of vegetated land increased after reclamation efforts were initiated. However, NDVI results were not suitable in some cases (2004 and 2008). Accuracy assessment for the classified images produced from the hybrid approach ranged from 94.5% to 100%, while images that produced from NDVI varied from 77.5% to 100%. The rate of change was very high from 1999 to 2004 (around 62% of the study area changed according to cross tabulation matrix results for the thematic raster classified images). In summary, from 1984 to 2008, the Bustan 3 area of Egypt was changed from 100% barren land to ≈78.8% agricultural land, as a result of successful land reclamation efforts.

Keywords: Hybrid classification, NDVI, Landsat imagery, land cover change detection