Spatiotemporal Flood Characterization and Early Warning System Development in the Rokan Watershed Using Sentinel-1 SAR and Water Level Data Rijaldi R.M.(1*), Sidik R.F.(1), Liyantono.(1), Setiawan Y.(1) and Faskayana Y.S.(2)
1) Center for Environmental Research, IPB University, Indonesia
2) The Graduate School of Agriculture, Tokyo University of Agriculture and Technology, Japan
*rizkirijaldi44[at]gmail.com
Abstract
The Rokan watershed in Sumatra, frequently experiences flood inundation with the most extensive event recorded between late 2023 and early 2024 that last for over three months. Despite this, the long-term spatiotemporal characteristic of floods in this watershed remain unstudied and the early warning system has not been developed yet to mitigate its impacts. This study aims to characterize flood dynamics over a 10-year period (2014-2024) and provide a foundation for a flood early warning system in the Rokan Watershed. Flood extent was identified using Sentinel-1 synthetic aperture radar (SAR) data, applying a -19 dB threshold to VH-polarized imagery. Unlike optical sensors, Sentinel-1 SAR can penetrate cloud cover and is not affected by sunlight enabling reliable and consistent flood mapping under all weather conditions. The threshold-based classification was implemented using Google Earth Engine through Google Colab, enabling rapid and consistent mapping across multiple time periods. Validation using 439 ground reference points within 100-meter buffer yielded an overall detection accuracy of 83.8%. Detection results revealed the extensive inundation that occured in 2014, 2018, 2019, and especially during the 2023-2024 period, indicating an increasing trend in flood severity and spatial extent. The results were compared with water level observation data from BWSS III monitoring stations to explore temporal relationships between hydrological conditions and flood onset. The comparison highlights potential time lag between upstream water levels and flood inundation. Based on data from Lubuk Bendahara Station, when the water level reaches 220 cm during the rainy season at the end of the year, increased vigilance is necessary within 2 to 5 days due to the potential risk of flooding in Bonai Village. These findings can enhance flood preparedness and support early warning system development in the Rokan Watershed.
Keywords: flood dynamics- Sentinel-1 SAR- early warning systems- Google Earth Engine
