Seamless prediction of water fluxes across scales
Luis Samaniego1, Oldrich Rakovec
2, Rohini Kumar
1, Juliane Mai
3, David Schaefer
3, Matthias Cuntz
3, Martin Schrön
3, Stephan Thober
3, Matthias Zink
3 and Sabine Attinger
3, (1)Helmholtz Centre for Environmental Research - UFZ, Computational Hydrosystems, Leipzig, Germany, (2)Helmholtz Centre for Environmental Research - UFZ, Leipzig, Germany, (3)Helmholtz Centre for Environmental Research-UFZ, Leipzig, Germany
Contact First Author: Luis Samaniego; luis.samaniego@ufz.de
Previously Published Material: Some results have been presented at previous AGU and EGU meetings 2014. mHM/MPR related issues were published in Samaniego et al. 2010 WRR, Kumar et al. 2013ab WRR, as well as, in Samaniego et al. 2011 JHM. Manuscript in preparation for WRR.
English Abstract:
Developing the ability to seamlessly predict streamflow and other state variables like soil moisture at catchment, regional, continental or global scales with spatial resolutions varying from hundreds to thousands of meters is fundamental for improving our understanding of the water balance at scales relevant for decision making as well as for improving our understanding of the potential impacts of climate change on water resources. Hydrologic observations, however, are hardly available at the scale at which the predictions are needed. Streamflow, for example is a quite reliable signal but it represents the integral response over the whole basin. In situ soil moisture observations have a very small control volume and are hardly available at a regional scale. Remote sensing products such as the total water storage anomaly or soil moisture, on the other hand, are spatially explicit but observed at resolutions much larger than those required. It is therefore necessary to develop a framework that incorporates observations at their native resolutions into a hydrologic model without the need of using ad hoc up/downscaling techniques to match observations.
Here, we show the capability of the Multiscale Parameter Regionalization (MPR) technique within the mesoscale Hydrologic Model (mHM) to perform this task. MPR is an effective method to find quasi scale invariant parameter sets tested over 250 Pan-EU river basins varying from 100 - 500,000 km2. The model is forced using the E-OBS data set available at a (0.25×0.25)° resolution from ECA&D during the period 1951-2012. The effective parameters obtained with simulations at this scale can effectively reproduce the total water storage anomalies retrieved by GRACE (NASA) at the spatial resolution (1 × 1)°, the gridded evapotranspiration estimated by LandFlux (ETH) (0.5×0.5)° as well as at tens of eddy flux stations (FLUXNET) whose footprint is about 1 hm2. Cross-validation experiments lead to the conclusion that mHM estimated water fluxes are robust since less than 25% of river basins exhibit NSE of 0.4 or less for daily discharge. Likewise, the TWS anomalies, exhibit a large spatial correlation with those obtained from GRACE. Comparison against observed latent heat indicates that the dynamics and magnitude of the simulated values are well captured by the model at most locations.
