|
Titel |
Multiscale verification of water fluxes and states over Pan European river basins |
VerfasserIn |
Luis Samaniego, Oldrich Rakovec, David Schaefer, Rohini Kumar, Matthias Cuntz, Juliane Mai, John Craven |
Konferenz |
EGU General Assembly 2014
|
Medientyp |
Artikel
|
Sprache |
Englisch
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 16 (2014) |
Datensatznummer |
250088812
|
Publikation (Nr.) |
EGU/EGU2014-2969.pdf |
|
|
|
Zusammenfassung |
Developing the ability to predict the movement of water at regional scales with a spatial
resolution from 1 to 5 km is one of grand challenges in land surface modelling. Coping with
this grand challenge implies that land surface models (LSM) should be able to make reliable
predictions across locations and/or scales other than those used for parameter estimation.
Validating LSM only against integral basin response such as streamflow is a necessary but not
a sufficient condition to warranty the appropriate partitioning of incoming precipitation and
radiation into different water budget components. Extensive in-situ observations of state
variables (e.g., soil moisture), on the contrary, are not feasible at regional scales.
Remote sensing has been considered as the solution for this dilemma because they
constitute a cost-effective source of information and provide a valuable insight about the
spatio-temporal patterns of state variables. Their main disadvantage is their large
uncertainty.
The mesoscale hydrologic model (mHM 5.0 http://www.ufz.de/index.php?en=31389) is
used in this study to estimate uncalibrated water fluxes and states and then to investigate
which are the effects of conditioning this model with freely available multiple-scale data sets.
The main characteristic of mHM is the treatment of the sub-grid variability of input
variables and model parameters which clearly distinguishes this model from existing
precipitation-runoff models or land surface models. It uses a Multiscale Parameter
Regionalization (MPR) to account for the sub-grid variability and to avoid systematic
re-calibration. Another key characteristic of mHM is that it can simultaneously estimate
fluxes in nested-scales and/or in multiple basins keeping its global parameters (i.e.,
regionalization coefficients) unaltered across scales and basins. These key characteristics
of the model would allow to assimilate disparate sources of information such as
satellite data, streamflow gauging stations, and eddy covariance data at their native
resolutions.
To address these objectives, mHM was set up over more than 280 Pan-European river
basins. This model was forced with the gridded EOBS data set (25x25 km2) obtained from
the European Climate Assessment & Dataset projec. The required morphological data was
derived from the FAO soil map (1:5,000,000), the SRTM DEM (500 m) and three CORINE
land cover scenes (500 m). MODIS LAI (NASA) was used to estimate a dynamic LAI model
for every land cover class. mHM simulations were obtained at 25 km spatial resolution for the
period 1950-2012. The multi-scale verification of simulated water fluxes was carried out
using observation data sets such as: latent heat flux obtained from more than 150 eddy
flux stations (FLUXNET), streamflow in more than 250 gauging stations (GRDC),
and the remotely sensed Earth’s gravity field anomalies retrieved by the Gravity
Recovery and Climate Experiment (GRACE) release 05 (Landerer and Swenson, 2012,
WRR). The former are used a proxy of the total water storage anomalies in mHM. In
Germany, over 1000 weakly groundwater stage stations were used to evaluate and/or
condition groundwater level anomalies. mHM water storage anomalies simulated
over Europe from 2003 to 2012 at monthly time step were compared with those of
GRACE. Results lead to the conclusion that mHM water fluxes are robust since less
than 25% of river basins exhibit Nash-Sutcliffe efficiencies (NSE) of 0.5 or less.
Likewise, the soil moisture and groundwater anomalies, specially in severe drought
years such as 2003, exhibit a large spatial correlation with those obtained from
remotely sensed products. Comparison against observed latent heat indicates that
the dynamics and magnitude of the simulated values were well captured by the
model at most locations. In general, deficient model performance (NSE |
|
|
|
|
|