Andreas Hartmann1,2, Luisa Hopp3, Jean-Lionel Payeur-Poirier3
University of Freiburg
1Andreas Hartmann, Assistant Professor
Chair of Hydrological Modeling and Water Resources
Estimating the value of hydrodynamic and hydrochemical signatures
to identify subsurface storm flow
2 Department of Civil Engineering
University of Bristol, UK
3 Department of Hydrology, University of Bayreuth (BayCEER), Bayreuth, Germany
At many catchments, subsurface storm ow SSF is an important runo generating mechanism. Although well explored and conceptualized at the hillslope scale, the importance of subsurface stormow
at the catchment scale is still uncertain due to a lack of adequate monitoring and analysis schemes.
This study explores the value of hydrodynamic and hydrochemical signatures to identify the contribution of SSF to streamow at the catchment scale. We hypothesize that signatures that include
information about subsurface stormow can be used to improve the simulation of subsurface storm ow dynamics with a simplied version of the process-based HBV model.
Furthermore, we propose that the mutual use of hydrochemical signatures and modelling provides a promising direction to nally explore the dynamics of subsurface storm ow at the catchment scale.
SIMPLIFIED VERSION OF THE HBV MODEL
The HBV model is a conceptual semi-distributed model that was developed for
the simulation of humid temperate catchments. In order to reduce the risk of
over-parameterization, we simplied its structure from originally 14 to 6
parameters, neglecting the simulation of the snow storage, deep groundwater
percolation, and channel routing. We run the model on daily resolution.
SSF assumed to be represented by shallow GW storage.
For more information on the HBV model see Seibert, J., Vis, M.J.P., 2012. Teaching hydrological modeling with a user-friendly
catchment-runo-model software package. Hydrol. Earth Syst. Sci. 16, 3315–3325. doi:10.5194/hess-16-3315-2012
IDENTIFIABILITY OF SUBSURFACE STORMFLOW
Stream ow increased by more than two orders of magnitude during wet-up of the catchment, coinciding with a pronounced
dilution of hydrochemical parameters in stream ow. Hydrograph separation and mixing analyses suggested that hillslope water,
i.e. SSF, contributed up to 50% to streamow during and after the
major period of the monsoon season (shaded area indicates the
entire duration of the summer monsoon season 2013, with DOY
190 separating the wet-up from the major period).
Day of the year Day of the year
Our results show that hydrodynamic and hydrochemical signatures have dierent inuence
on the parameter estimation of the HBV model. Hydrodynamic information helps to identify
the slow groundwater processes, while the explicit information on the fractions of subsurface
stormow (hydrochemical signature) allowed to distinguish realistic and unrealistic
simulations, which were -only by discharge- not distinguishable. Hence, the information on
subsurface stormow resulted in more realitic simulations
The changes of the distribution of the model parameters that control SSF processes within the
HBV model corroborate this nding. However, even with both hydrodynamic and
hydrochemical signatures included, 3 of the 6 parameters of the simplied HBV model are still
not identiable. Next steps will include the usage of a longer and continuous time series as
well as the use of other hydrodynamic and hydrochemical signatures.
Acknowldegments This research is a contribution to the research network „Subsurface Stormow: A
well-recognized but still challenging process in catchment hydrology research“ funded by the German
Research Foundation DFG (project number 299961754).
STUDY AREA & EXPERIMENTAL SETUP
Area: ~ 16 ha
Elevation range: 368 – 682 m a.s.l.
Mean slope angle: 24°
Mean soil depth: 0.6 m
Cambisol with loamy texture
61% deciduous, 39% coniferous
Climatic forcing, discharge at the outlet and hydrochemical parameters in stream ow and
dierent end members were measured in a small forested headwater catchment in South
Korea during the 2013 monsoon season.
HYDRODYNAMIC AND HYDROCHEMICAL SIGNATURES
As hydrodynamic signature, we use the Kling-Gupta
eciency (KGE), which indicates best model perfor-
mance with values close to 1 and worst model perfor-
mance with values close to 0.
As hydrochemical signature, we use the average
fraction of discharge provided from the shallow
groundwater storage of the model, i.e. simulated
SSF, which we compare to the SSF fraction that was
obtained from end-member mixing analysis.
To quantify the information content of the two
signatures, we run the HBV model 25,000 times and
calculate both hydrodynamic and hydrochemical
In a rst step, we reduce the sample by discarding all si-
mulations that perform badly in terms of the hydrodyna-
mic signature (KGE ≥ 0.75). We obtain a signicant reduc-
tion of the sample by ~50%.
In a second step, we reduce the sample by
discarding simulations that do not provide the
expected range of SSF fractions in stream
According to the ndings of the end-member
mixing analysis, SSF contributed up to ~50%
to the streamow. Accounting for uncertainty
in the mixing analysis, we chose an SSF
fraction of 30-60% as most realistic.
Discarding all simulations outside that range
resulted in 141 remaining simulations.
In order to check for equinality, we also
picked the ranges of 0-30% and 30-100%, and
received remaining simulations of 4,080 and
For more information on the the quantication of information content see Hartmann, A., Barberá, J.A., Andreo, B., 2017. On the value of water quality data and informative ow states in karst modelling. Hydrol.
Earth Syst. Sci. 21, 5971–5985. doi:10.5194/hess-2017-230
Changes of the distribution of a HBV model parameter indicate that the calibration data (hydrodynamic or hydrochemical signature) contains
information about the model process that is controlled by this parameter. For instance, the distribution parameter K1 that controls the slow
groundwater outow, is changed by the hydrodynamic signature.
Using the information on the SSF fractions (30-60%) results in changes of the K0 and UZL, both controlling the
dynamics of the shallow groundwater, i.e. SSF. Similar results are found for the rather unrealistic SSF fractions of 60-100%
and, less pronounced, 0-30%. The distributions of the other parameters (Beta, FC, LP) remain with minor changes.
Applying only the hydrodynamic signature, the simulations already
envelop most of the observations. Using the hydrochemical signature
with the most realistic SSF fractions, just a slight decrease of uncer-
tainty is obtained. Hence, realistic or unrealistic SSF fractions, the
simulated discharge dynamics are almost the same.
Normalized parameter ranges