Flood extent mapping using optical remote sensing indices: a case study of the upper Medjerda River basin (Northwestern Tunisia)

Abstract

Floods represent one of the most frequent and damaging natural hazards in northwestern Tunisia, particularly along the upper Medjerda River valley, which has experienced several major flood events in recent decades. Accurate and timely mapping of flood extent is therefore essential for effective risk management and decision-making. This study investigates the capability of optical remote sensing data for flood extent mapping using Landsat-8 imagery acquired before and after the February 2015 flood event. Several widely used spectral water indices, including the Normalized Difference Water Index (NDWI), Modified Normalized Difference Water Index (MNDWI), Automated Water Extraction Index (AWEI), Normalized Difference Moisture Index (NDMI), and Water Ratio Index (WRI), were applied and compared to assess their effectiveness in discriminating flooded areas and to identify the most suitable index under semi-arid Mediterranean conditions. The results indicate that indices incorporating the short-wave infrared (SWIR) band, particularly the MNDWI, provide improved delineation of flooded surfaces with sharper boundaries and reduced confusion with irrigated agricultural land and built-up areas. In addition, a rule-based decision-tree approach combining MNDWI, the soil adjusted vegetation Index (SAVI), and the normalized difference built-up index (NDBI) was implemented to enhance flood extent extraction by minimizing misclassification related to vegetation and urban surfaces. To ensure the reliability of the results, Sentinel-1 SAR data were utilized as an independent validation reference. The quantitative analysis demonstrates high accuracy for the proposed methodology, with overall accuracy (OA) values ranging from 73.4% to 81.1% and a Kappa coefficient reaching up to 0.83. The findings demonstrate the relevance of freely available Landsat-8 data and hybrid spectral index-based decision-tree approach for rapid flood mapping in semi-arid environments. This methodology can be transferred to similar flood-prone regions and provides a practical tool for supporting flood risk assessment and management strategies.