Using streamflow and baseflow separation to characterize spells of low and high flows and their transitions

Extreme hydrometeorological events such as floods and droughts can lead to severe socio-economic and environmental impacts, which can be amplified through temporally compound events, when successive hazards occur before the system can recover from the first event. This situation may arise not only from repeated occurrences of the same hazard, but also from shifts between contrasting hydrometeorological hazards. In this study, we propose a framework for consistently identifying and characterizing high-flow spells (HFS) and low-flow spells (LFS), and the transitions from one type of spell to another that might be of particular interest to stakeholders. We use baseflow as a proxy to determine catchment recovery between spells, and a mixed threshold approach to identify the spells in long discharge time series. We apply the methodology to 643 catchments of the CAMELS-FR dataset in France, with at least 30 complete hydrological years of data over the 1970-2021 period. The spells were characterized in terms of duration and severity. We further analyzed the spatiotemporal characteristics of consecutive spells of the same type and the transitions between spells, investigating their frequency and transition times. The application of the framework allowed us to identify over 140,000 spells across all catchments, with 74% classified as HFS and 26% as LFS. HFS of short duration (less than 3 days) and high severity (above 99th percentile) occur more often in catchments located in mountainous areas, while LFS of long duration (over 90 days) and high severity (below 5th percentile) occur more often in Northern France. Our results also indicate that consecutive short-duration HFS occur more often in the dataset studied than consecutive long duration LFS. Rapid transitions (less than 14 days) from LFS to HFS mainly occur in the Mediterranean part of France in the beginning of the winter season. The framework developed to identify spatiotemporal patterns of high and low flow spells can be potentially useful to new generation early warning systems and support first responders in flood disaster and drought management.

This work is funded by Horizon Europe under grant agreement No. 101074075 (MedEWSa project).