Are temporary stream observations useful for the calibration of a lumped hydrological model?

Since 2012 the Observatoire National des Étiages (ONDE) has collected observations about the state of headwater streams across France. Streams that occasionally dry up (also called temporary or intermittent streams) are visited at least five times a year by trained staff and the flow state is visually assessed (dry stream, standing water, flowing water). These data have been used to calculate regional probabilities of drying with statistical models. However, the usefulness of this kind of data for the calibration or validation of bucket-type hydrological models has so far not been assessed.

We, therefore, used the ONDE dataset for the calibration of the HBV model to evaluate the potential of temporary stream observations to improve model predictions in data-scarce regions. The model was calibrated for almost 90 catchments throughout France. We used the information on the flow state of temporary streams, either alone or in combination with limited observations of discharge or water level at the outlet of the catchment for model calibration and evaluated the simulations based on the measured discharge. Because the ONDE data set is large, we could do an extensive analysis of the value of temporary stream observations and the way to optimize their use in hydrological models. While the study focuses on catchments in France, visual observations of temporary streams can - with the help of citizen scientists - be collected at any place in the world. Thus, the method has the potential to be applied in truly data scarce regions.