Impact of erroneous data in streamflow time series in large-sample-hydrology modelling experiments - INRAE - Institut national de recherche pour l’agriculture, l’alimentation et l’environnement Accéder directement au contenu
Communication Dans Un Congrès Année : 2022

Impact of erroneous data in streamflow time series in large-sample-hydrology modelling experiments

Résumé

Although often neglected, many sources of error in data, such as precipitation data, estimates of evapotranspiration or flow measurements, can affect the results of hydrological models. Various studies have investigated the impact of errors related to climatic inputs (typically precipitation, evapotranspiration and temperature data). Comparatively, the impacts of errors in streamflow time series has been less studied, and was mainly limited to stage-discharge relationship issues. However, it is well known that several other types of errors may also exist in observed flow series and may ultimately affect modelling experiments. The automatic detection of these errors is often difficult and time series have to be checked by expert judgement, which may be tedious in case of large samples of catchments. Therefore, there is a need to better quantify the actual impact of such errors. The aim of this work is to answer the following question: do erroneous data commonly found in streamflow time series have a significant impact on the performance and parameterization of hydrological models? To answer this question, 15 French catchments were randomly drawn from a large database of 147 catchments. Precipitation, evapotranspiration and flow data were gathered at the hourly time step over the 1998-2018 period. We used the lumped-conceptual GR5H hydrological model. The streamflow time series were inspected by two experts who identify possible errors in flow observations, which were sorted into four types. Sensitivity of modelling results to these errors in calibration and evaluation was analysed by leaving or removing these errors in observed series. As expected, the more erroneous a time series, the more the model is affected, whether in terms of parameterization or performance. However, these variations remain small, highlighting the stability of the GR5H model despite the presence of erroneous data. The model is also more sensitive when the errors affect high water, even if, on average on the sub-sample, the majority of erroneous time steps occur during low-flow periods. Finally, erroneous data in streamflow time series may be accepted as long as they do not reach outlier values. The implications of these findings in the use of large sample of catchments are discussed.
Fichier non déposé

Dates et versions

hal-03697738 , version 1 (17-06-2022)

Identifiants

  • HAL Id : hal-03697738 , version 1

Citer

Cyril Thébault, Charles Perrin, Vazken Andréassian, Guillaume Thirel, Sébastien Legrand. Impact of erroneous data in streamflow time series in large-sample-hydrology modelling experiments. IAHS-AISH Scientific Assembly 2022, May 2022, Montpellier, France. ⟨hal-03697738⟩
40 Consultations
0 Téléchargements

Partager

Gmail Facebook X LinkedIn More