Comparative assessment of spectral indices for long-term flood monitoring in poljes using Sentinel-2 data: a case study from Muğla, Türkiye

Karst poljes are highly sensitive hydro-geomorphological systems where flooding poses persistent risks to agriculture and settlement. Their distinctive hydrological behavior and vulnerability demand systematic monitoring through remote sensing. However, flood mapping in poljes faces major challenges, including spectral confusion from reflective carbonate rocks, permanent water bodies, shadows, and the selection of appropriate thresholds. This study develops an improved approach by comparing seven spectral indices and designing a thresholding strategy based on representative zones for different flood types. Using Sentinel-2 data processed on GEE, we evaluated a benchmark flood event (14 January 2022) against 182 ground-truth points and PlanetScope imagery. The Modified Normalized Difference Water Index (MNDWI) achieved the highest performance (97.3% overall accuracy, Kappa 0.945). Applying this method to an 8-year series (2018–2025) enabled detection of 12 major floods, precisely delineating their onset and duration, and producing a flood frequency map of recurrently inundated zones. The results demonstrate how spectral index–threshold integration can transform flood monitoring from isolated events into continuous records of flood regimes. This work contributes to the relatively limited body of research on remote sensing–based flood monitoring in Turkish poljes and highlights its value for decision-making in flood risk mitigation and land-use planning.