Ardèche catchment (France) with a
Mediterranean climate characterized by flash
floods and long dry periods Focus on four catchments.
Streamflow computed using BaRatin to have an
estimation of the associated uncertainty (Le Coz et al., 2014)
using data from OHM-CV (Nord et al., 2017)
Forcing data: SAFRAN reanalysis data (Vidal et al. 2010)
J2K process-based distributed hydrological model (Krause et al. 2006)
→ long term water variability and availability
→ daily time step
Default parameter values:
no calibration, results of
prior studies; mainly based on available spatial data,
expert judgement
Diagnostic-evaluation of a distributed model using hydrological signatures
Horner I.*(1), Branger F.(1), Vannier O.(2), Braud I.(1)
(1) IRSTEA Lyon, Villeurbanne, France
(2) CNR, Compagnie National du Rhône, Lyon, France
Classical performance metrics
→ poor diagnostic on the model (dys)functioning Hydrological signatures
→ extract information from hydrological data
→ insights into the underlying hydrological processes Comparing observed and simulated hydrological
signatures
→ diagnostic of the model (structures and parameters)
Gupta et al. (2008)
Good performance given that the model wasn’t calibrated
But the streamflow time series show important discrepancies
How can these differences be “objectively” captured using hydrological signatures?
Main results:
→ Overestimation of evapotranspiration to the expense of streamflow
→ Overestimation of slow flow contributions
How can these results be translated into clear diagnostics on the model?
A sensitivity analysis could help disentangle the links between the hydrological signatures and the model
parameters.
Introduction Model and data
General results
Conclusion
Only precipitation P and streamflow Q are used
d𝑄
d𝑡 = 𝑎𝑄𝑏
Too much baseflow ? 1/ Streamflow response too dampened
2/ Overestimation of low flows?
1/ Change from fast to slow flow contribution too quick
2/ Too slow late recessions Too much
evapotranspiration to the expanse of
streamflow
(or the other way around)
Seasonal streamflow
response change too weak and too late
Input precipitation partitioning issues:
over/under estimation of Leaf Area Index, Crop coefficients, Sealing coefficients ?
Overestimation of slow flow component:
water infiltration in soil, soil drainage, percolation, groundwater characteristics ?
Labeaume Meyras
Claduègne
Goulette
Seasonal catchment change of state badly reproduced by the model:
over/under estimation of soil and groundwater reservoirs?
Runoff coefficient Flow duration curve Baseflow index Streamflow recessions Seasonal dynamic
Hydrological signatures analysis & model diagnostic
References:
Gupta et al. (2008) Hydrological Processes - 10.1002/hyp.6989
Le Coz et al. (2014) Journal of Hydrology - 10.1016/j.jhydrol.2013.11.016
Krause et al. (2006) Advances in Geosciences - 10.5194/adgeo-9-53-2006
Vidal et al. (2010) International Journal of Climatology - 10.1002/joc.2003
Nord et al. (2017) Earth System Science Data - 10.5194/essd-9-221-2017
We would like to acknowledge the support of CNR and OZCAR for funding as well as OHM-CV and French hydrometric services (Banque Hydro and SPC Grand Delta) for providing the data used herein.
AGU Fall Meeting, Washington DC, USA, 10-14 décembre 2018
