Forum geografic. Studii și cercetări de geografie și protecția mediului Volume XXIV, Issue 1 (June 2025), pp. 50-59; DOI: 10.5775/fg.2025.1.3688 © 2025 The Author(s). Published by Forum geografic. Open Access article. Creative Commons Attribution-NonCommercial-NoDerivatives License. 50 Applying impervious indices using Sentinel-2 data in semi-arid land (North-East Algeria) Khaled ROUIBAH 1,* 1 École Normale Supérieure Messaoud Zeghar-Sétif, Teacher Education College of Setif Messaoud Zeghar, B.P.N 556 El Eulma 19600, Setif, Algeria * Corresponding author: k.rouibah@ens-setif.dz Received on 28-01-2025, reviewed on 22-03-2025, accepted on 21-06-2025 Abstract The urban area estimation in dry and semi-dry climate, is still a difficult. Therefore, identify a reliable impervious index is critical. For that reason, Ain Azel city that located in semi-arid land of North-East Algeria, was the area of test of three impervious indices, namely: the Built-up Area Index (BAI), the Normalized Impervious Surface Index (NISI) and the Urban Area Index (UAI). These indices were derived from Sentinel-2 data and subjected to the Support Vector Machine (SVM) classification method, to extract built-up area class. The accuracy assessment results showed that the Overall accuracy (Oa) of the NISI index achieved 92,67 %, which is a little less compared to the (Oa) of UAI index which is about 93.67 %. Based on this, the both indices provided satisfactory result, although the UAI index, relatively, overestimated the built-up area. However, the BAI index that use the Blue and Near-Infrared bands is sufficiently discriminative between highly similar of buildings materials and dry soil; the BAI index produced the accurate built-up mapping with well detection of road networks with (Oa) achieved 95.67% and the highest kappa coefficient which is about 86.54 %. This is due to BAI band’s degree of sensitivity and their high spatial resolution of 10 m. Consequently, the SVM segmentation-based BAI index worked well and the result is promising for accurate modeling of cities with same environment condition, for its sustainable development. However, the performance level evaluation of indices applied in this study, should be retested over different regions in dry land. Keywords: Sentinel-2, BAI Index, NISI Index, UAI Index, Support Vector Machine-SVM, semi-arid land Introduction The remote sensing technology is useful in exploring the spatial distribution of built-up areas and its extension (Kaur & Pandey, 2022). In terms of impervious detection, there are different methods (Lu et al, 2014; Wang & Li, 2019). For instance, the Support Vector Machine (SVM) as supervised classification method (Vapnik, 1995), the k- means as unsupervised classification method (MacQueen, 1967). However, according to the literature review, usually, the fast detection of the urban area is performed widely using Index-Based Methods. In fact, there are vast list of various built-up indices developed (Javed et al., 2021; Kaur & Pandey, 2022; Pirowski & Szypuła, 2024). However, their capability to differentiate urban areas from bare soil is challenging (Kaya & Dervisoglu, 2023). Often, some pixels of bare lands were misclassified as built -up areas and vice versa, these errors occurs when the spectral signatures of urban features and bare land can be very similar (Martin, 2024; Sun et al., 2016; Waqar et al., 2012), this is what leads to confusion between them as referred to previously (Deng & Wu, 2012). So, the similarity reflectance poses a problem which is a real challenge concerning the detection of impervious surface, as it influences certainly the effectiveness level of most impervious indices. Consequently, they producing lower accuracy mapping of urban area (Kaur & Pandey, 2022). This issue is increasing especially in the dry period, where the difficulty become very clearly (Valdiviezo et al., 2018), because bare soil with high albedo in summer, causes the main disturbance in urban impervious surface (Wang & Li, 2019). The semi-arid land with strong resemblance in reflectance, are knowing same problem as well (Dib et al., 2022; Rasul et al., 2018; Rouibah & Belabbas, 2020). In same point, it is reported that the spectral confusion between the dry soil and bright pixels of urban area affected the built-up indices that become more sensitive (Javed et al., 2021). Based on the previous literatures review and analysis, it is stated that the dry soil and the barren, lights highly in dry climate, that leads to be confused with bright building and asphalt. Therefore, the impervious indices application, leads to non-correctly classify the built-up areas; i.e. higher misclassification rates and low accurate built-up mapping. The suitable impervious algorithms are critical steps in the general procedure of mapping. However, the appropriate method for a specific study area, is still poorly understood (Lu et al., 2014). In this context, recently, several researches worked to evaluate the performance of numerous spectral indices for impervious surface extraction, with making comparatives (Chen et al., 2020; Kebede et al., 2022; Li et al., 2021; Mshelia, 2022; Valdiviezo et al., 2018; Xi et al., 2019). However, so far, similar studies that carried on semi-arid