Understanding temporal transitions and spatial connections of high and low flow spells at local and regional scales

When floods and droughts occur successively before the river catchment is able to recover, their socio-economic and environmental impacts are amplified. While the drivers of individual extremes are well understood, the spatial connections and timescales of transitions between occurrences of high and low flow spells remain understudied, especially regarding their implications for operational risk management. Here we detect and characterize high and low flow spells at the catchment (local) scale as well as at the regional (decision-making) scale of operational flood forecasting centers. We use daily streamflow data from 643 catchments of the CAMELS-FR dataset in France over the 1970-2021 period, and investigate the occurrences of such spells within the 17 regional centers of the national forecasting service (hereafter called ‘SPC’, for ‘Service de Prévision des Crues’ in French). We initially use a mixed threshold approach combined with baseflow estimation as an indicator for catchment recovery to detect the spells and then analyze their frequency, duration, temporal transition and spatial connections (spatially compounding events). Our results show that consecutive occurrences of the same spell type are more predominant, with consecutive high flow spells being more common. Transitions occurring in less than a month from low to high flows show distinct spatial variability, with the shortest transition durations concentrated in the regions of the Rhone-Mediterranean and Rhine-Meuse river basins. These transitions mainly occur in autumn and early winter. On the other hand, transitions from high to low flows are typically slow, developing over more than 90 days. In addition, it is identified that the SPC Alpes du Nord (located in the Rhone-Mediterranean region) shows transition frequencies above the national mean for all transition types, while the SPC Bassin du Nord (in northern France) has lower frequencies in all transition types. We also applied a synchronization approach to investigate spatially compounding events by identifying pairs of catchments with concurrent high flow spells, low flow spells, and opposing spells (e.g. when one catchment is experiencing a high flow spell, while another is experiencing a low flow spell). The analysis allowed us to detect regions with high spatial connectedness, as well as patterns of inter- and intra-annual variability of spatially connected spells. Finally, perspectives on the application of the developed methodology at the European scale are discussed.

This work is funded by EU Horizon Europe project MedEWSa (Mediterranean and pan-European forecast and Early Warning System against natural hazards) under Grant Agreement 101121192.