Abstract. Citizen science
and crowdsourcing are gaining increasing attention among hydrologists. In a
recent contribution, Mazzoleni et al. (2017) investigated the integration
of crowdsourced data (CSD) into hydrological models to improve the accuracy
of real-time flood forecasts. The authors used synthetic CSD
(i.e. not actually measured),
because real CSD were not available at the time of the study. In their work,
which is a proof-of-concept study, Mazzoleni et al. (2017) showed that
assimilation of CSD improves the overall model performance; the impact of
irregular frequency of available CSD, and that of data uncertainty, were also
deeply assessed. However, the use of synthetic CSD in conjunction with
(semi-)distributed hydrological models deserves further discussion. As a
result of equifinality, poor model identifiability, and deficiencies in model
structure, internal states of (semi-)distributed models can hardly
mimic the actual states of complex systems away from calibration points.
Accordingly, the use of synthetic CSD that are drawn from model internal
states under best-fit conditions can lead to overestimation of the
effectiveness of CSD assimilation in improving flood prediction. Operational
flood forecasting, which results in decisions of high societal value,
requires robust knowledge of the model behaviour and an in-depth assessment
of both model structure and forcing data. Additional guidelines are given
that are useful for the a priori evaluation of CSD for real-time flood
forecasting and, hopefully, for planning apt design strategies for both model
calibration and collection of CSD.