The use of bands ratio derived from Sentinel-2 imagery to detect built-up area in the dry period (North-East Algeria)

Abstract

In this research, the band rationing technique was used to expect accurate detection of built-up in a dry period over El-Eulma city (North-East Algeria). In this context, the VNIR Sentinel-2 bands were examined statistically over the study area. Consequently, two bands ratio (BR) which are mainly the blue-near-infrared (B2/B8) and the green-near-infrared (B3/B8), were selected to be used singly as input data, for the binarization process via the use of Otsu method. To evaluate the approach and find the optimal bands ratio for built-up detection in the dry period, the accuracy assessment was done, using the high-resolution Google Earth images as a reference map. Also, the results obtained were compared to the both built-up mapping resulting from the support vector machine (SVM) classification and built-up area index (BAI). The findings showed that the BR (B2/B8) works approximately similar to the SVM classification result. In contrast, the BR (B2/B8) works better than the BR (B3/B8) and BAI index; the overall accuracy (OA) and kappa coefficient of the first BR (B2/B8) are about 92,33% and 80,81%, respectively. In contrast, the (OA) and kappa coefficient of the second BR (B3/B8) are about 90,97% and 76,72% respectively, Meanwhile, the (OA) of the BAI index is about 89.33% and its kappa coefficient is about 74,80%. Therefore, the results present BR (B2/B8) as a simple automatic technique that could be suitable for mapping cities accurately in a dry climate, for better land use planning.