|
Titel |
A High-Resolution Dataset of Water Fluxes and States for Germany accounting for Uncertainties in the Parameter Estimation |
VerfasserIn |
Matthias Zink, Rohini Kumar, Matthias Cuntz, Luis Samaniego |
Konferenz |
EGU General Assembly 2015
|
Medientyp |
Artikel
|
Sprache |
Englisch
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 17 (2015) |
Datensatznummer |
250109937
|
Publikation (Nr.) |
EGU/EGU2015-9891.pdf |
|
|
|
Zusammenfassung |
Long term, high-resolution data of hydrologic fluxes and states are needed for many
hydrological applications such as i) impact assessment studies (e.g. drought, flood or climate
change analyses), ii) studies that need the state or variability of hydrometeorological or
hydrologic variables (e.g. downscaling of climate model outputs), iii) modeling studies
that need hydrologic variables as input or boundary conditions (e.g. recharge for
groundwater modeling). Since long-term, large-scale observations of such fluxes and states
are not feasible, hydrological or land surface models are applied to derive them.
Usually such datasets are provided as single model realization without accounting
for input, model structural or uncertainty caused by equifinal model parameter
sets.
This study aims to analyze and provide a high resolution dataset of hydrological fluxes
and states accounting for uncertainties caused by the estimation of model parameters.
Furthermore, the spatiotemporal distribution of uncertainties in various hydrological variables
as well as the superposition of uncertainties through different model compartments is
investigated.
The hydrological variables of interest are evapotranspiration, soil moisture, recharge, and
generated discharge. They are estimated for entire Germany in the period 1950 - 2010
employing the mesoscale hydrological model mHM (www.ufz.de/mhm). The spatial
resolution is 4 km and the temporal resolution is 1 day. The ensemble of 100 model
realization is based on 700 parameter sets which are derived from 100 calibration runs in the
seven, major German river basins. These 700 parameter sets are filtered for those exceeding a
Nash-Sutcliffe efficiency (NSE) of 0.65 in each of the seven catchments, which leads to the
final 100 parameter sets.
The model is evaluated against observed runoff in 222 additional catchments. In this
catchments the mean and the standard deviation are for daily discharge 0.68 and 0.09 and for
monthly discharge 0.81 and 0.09, respectively. Modeled evapotranspiration is evaluated
against eddy covariance stations. The estimated evapotranspiration exhibits the largest error
in spring (RMSE = 0.39 mm d-1) during the onset of the vegetation period compared to the
other seasons, which have an average RMSE value of 0.07 mm d-1. This may be explained
by the absence of a dynamic vegetation model within mHM. The uncertainty of the fluxes and
states is assessed by the coefficient of variation of the 100 ensemble members. The
lowest uncertainty is observed for evapotranspiration with an average coefficient of
variation of 0.02, while the highest uncertainty is observed for recharge with an
average coefficient of variation of 0.2. The uncertainty of the hydrologic variables
varies throughout the course of a year with exception of evapotranspiration, which
stays almost constant. For example, the generated discharge exhibits its largest
uncertainty in the end of summer and beginning of autumn. At this time the amount of
water in the soil and groundwater reservoir is lowest and thus the slow interflow and
baseflow is lowest. Furthermore, we found that the magnitudes of uncertainty of
evapotranspiration, soil moisture or recharge do not superpose to the modeled discharge. |
|
|
|
|
|