Modelling the intermittence of watercourses in the small French Mediterranean catchments of the Maures massif (Réal Collobrier ) with the SMASH platform (Spatially distributed Modelling and ASsimilation for Hydrology)

Intermittent rivers and ephemeral streams represent half of the global river network and span all climates. The intermittent rivers and ephemeral streams is a short-hand term for all flowing water that ceases to flow or that dries up completely at some point in time and/or space They are more frequent in arid and semi-arid areas but are also present in temperate, tropical humid, boreal, and alpine areas, where they are mainly located in headwaters. Their abundance is increasing due to climate change and water withdrawals for human activities.

The objective of this study is to represent the spatio-temporal dynamics of flow intermittence at the reach level in river of the seven sub-catchments of the Maures massif (between 1.5 km² and 70 km²).

First, two hydrological continuous models of varying complexities are performed: GR6J (lumped and conceptual), and SMASH (spatially distributed and conceptual) in terms of temporal calibration/validation, by dissociating dry and wet years, to asses the models’ability to simulate observed drying event over time. The metrics are based on daily flow records observed in the 7 catchments since 1968 to 2023.

In the second part, a regionalization method is tested on the spatially distributed model (SMASH). The HDA-PR approach (Hybrid Data Assimilation and Parameter Regionalization) incorporating learnable regionalization mappings, based on multivariate regressions is used. This approach consist to search for a transfer function that quantitatively relates physical descriptors to conceptual model parameters from multi-gauge discharge in order to produce a regional model.

Flow condition observed from multiple data sources (daily flow time series from gauging stations, phototrap installed along the river network taking daily pictures from 2021-04-01 to 2023-31-12, daily conductivity measurements series from 2021-01-01) are used to validate the ability of the regional model to simulate flow intermittence (prediction of dry events) at river section level.

The distributed modelling approach, with a high-resolution conceptual hydrological modeling at 0.250 km² and coupled with Hybrid Data Assimilation and Parameter Regionalization descriptors shows results highlight the effectiveness of HDA-PR surpassing the performance of a uniform regionalization method with lumped model parameters. However, the results on smallest catchments area are lowest.

The study shows the interest of using daily photos which are a good indication of the hydrogical state of the streams to obtain intermittence data and increasing the spatial coverage of observations in order to validate regional model.