Extraction of Urban Areas Using Spectral Indices Combination and Google Earth Engine in Algerian Highlands (Case Study: Cities of Djelfa, Messaad, Ain Oussera)

The fundamental difficulty in mapping urban areas, especially in semi-arid and arid environments, is the separation of built-up areas from bare lands, owing to their similar spectral characteristics. Accordingly, this study aims to identify the suitable spectral index that can provide high differentiation, between urban areas and bare lands, in semi-arid areas of three cities of the province of Djelfa, namely, Djelfa, Messaad, and Ain Oussera (Algerian central highlands), through a selection of four spectral indices including Urban Index (BUI), Band ratio for built-up area (BRBA), Normalized Difference Tillage Index (NDTI) and Dry Bare-soil Index (DBSI). In order to increase the mapping accuracy of the built-up in studied areas, a multi-index approach has been applied focusing on identifying an adequate combination of spectral indices of remote sensing that provides the highest performance compared to the images of sentinel 2A. The multi-index approach was developed using three spectral indices combinations and was created using a layer stack process. For forming bare land layer stacking data, both NDTI and DBSI indices were used, while the built-up area layer stacking data was made with both BUI and BRBA indices. The main process was carried out on the Cloud Computing Platform based on geospatial data of Google Earth Engine (GEE) and using machine learning classification by the Support Vector Machine (SVM) algorithm, based on imagery from sentinel 2A acquired during the dry season. The results indicated that the thresholds of the built-up areas are difficult to delineate and distinguish from bare land efficiently with a single index. The obtained results also revealed that the use of multi-index including BUI index provided the best results as they showed the highest effects with NDTI index and DBSI index compared to BRBA index, where the overall accuracies of the multi-index (DBSI/ NDTI/ BUI) were 98.7% in Djelfa, 96.5% in Messaad, and 97.87 % in Ain Oussera, and the kappa coefficients were 97.3%, 85.4%, and 95.3% respectively. These results show that this multi-index is effective and reliable and can be considered for use in other areas with similar characteristics.