Work

Identification and Classification of Non-Perennial Rivers Using Integrated Remote Sensing and Hydrological Modeling Approaches

Year: 2025

Type: conference-paper

Abstract: Mapping and understanding the hydrological dynamics of non-perennial rivers (NPRs) is critical for their conservation and management, yet current methodologies often fail to capture the spatial and temporal complexities of flow intermittency and dry riverbed extension. Leveraging advancements in satellite remote sensing, the RIVERTEMP Erasmus+ project develops an innovative web platform that analyzes Sentinel-2 multispectral imagery to classify NPR hydrological conditions. These data allow for the analysis of rivers with an active channel width larger than 40 m and not covered by vegetation. Through the visual interpretation of false-color images, derived from the SWIR, NIR, and RED bands, the platform allows users to identify the hydrological condition of the river, distinguishing between three main hydrological condition a NPR can present: the ‘Flowing (F)’ condition, characterized by continuous surface flow; the ‘Ponding (P)’ condition, characterized by ponds of water that are not connected to each other; and the Dry (D)’ condition, in which the river is completely without surface water. the ‘Flowing (F)’ condition, characterized by continuous surface flow; the ‘Ponding (P)’ condition, characterized by ponds of water that are not connected to each other; and the Dry (D)’ condition, in which the river is completely without surface water. This methodology makes it possible to generate time series that measure the frequency and duration of each condition, allowing a monitoring of rivers with an average time resolution of approximately 3-5 days in the absence of cloud cover. The tool, which is available free of charge, also includes user manuals and training material for university students. The project further integrates hydrological modeling with satellite observations, using the karst SWAT hydrological model in the Keritis River basin (Chania, Greece) as a case study. The hydrological model fills in gaps in the satellite data, providing information on flow conditions on days without imagery or when cloud cover exists. Satellite imagery classified approximately 60% of the river as "flowing" and around 40% as "ponding," with a small percentage of observations from 2023 identifying "dry" conditions. By combining modelled flow rates and satellite-derived classifications, a flow duration curve for 2019–2023 reveals significant overlap between the three hydrological conditions (Fig. 1), underscoring strong interannual variability in the relationship between modelled flow rate and the hydrological condition. Therefore, an annual basis approach was preferred to determine the threshold flow values representing the transition from one hydrological condition to another. Time-series analyses highlight notable annual variability in hydrological behavior, further supported by ternary plots showing hydrotype classifications. Both hydrological modeling and satellite observations consistently place the river reaches within the intermittent-fluent area, demonstrating strong agreement between methods. The analysis reveales a clear temporal pattern transitioning from intermittent-fluent hydrotype to intermittent-stagnant one, indicating significant changes in the river's hydrological regime (Fig. 2). This integration demonstrates that satellite-derived classifications effectively validate hydrological model predictions, while models can extrapolate hydrological conditions in the absence of satellite data. The research highlights the potential of combining remote sensing and hydrological modeling to enhance the monitoring and management of NPRs, offering robust tools to address existing data limitations.

Source: ARPHA Conference Abstracts

DOI: 10.3897/aca.8.e150535