Continental-scale high resolution terrestrial hydrologic modeling - PowerPoint PPT Presentation

(The Energy oriented Centre of Excellence in computing applications) Continental-scale high resolution terrestrial hydrologic modeling over Europe using COSMO- REA6 reanalysis dataset. Bibi S. Naz 1,2 , Stefan Kollet 1,2 , Anne Springer 5 , Harrie-Jan Hendricks Franssen 1,2 , Carsten Montzka 1 , Klaus Goergen 1,2 , Carina Furusho 1,2 1 Research Centre Jülich, Institute of Bio- and Geosciences: Agrosphere (IBG-3), Jülich 52425, Germany 2 Centre for High-Performance Scientific Computing in Terrestrial Systems, Geoverbund ABC/J, Jülich 52425, Germany 3 Research Centre Jülich, Jülich Supercomputing Centre, Jülich 52425, Germany 4 Laboratory of Hydrology and Water Management, Ghent University, Ghent 9000, Belgium 5 Institute of Geodesy and Geoinformation, Bonn University, Nussallee 17, Bonn 53115, Germany International Symposium on Regional Reanalysis, July 17 - 19, 2018

Background • Water for Energy Hydropower production simulations (Regional to site- • specific) – Refine spatial resolution for hydrologic simulation (~3km) – Routing of streamflow at locations of interest (e.g., hydropower plants in Alpine Region) • How to quantify the uncertainty of hydrological predications? – Ensemble-based data assimilation • Need to have reasonable model performance for hydrologic states and fluxes Reservoirs, hydropower plants and gauging across different scales stations in the Italian Alpine Region.

Community Land Model (CLM-PDAF) CLM 3.5 Parallel Data Assimilation CLM-PDAF Framework (PDAF) Nerger et al. (2005) TerrSysMP-PDAF Interface: Kurtz et al. (2016) Oleson et al. 2008 CLM3.5 for the European CORDEX CCI-SM Domain Naz et al. 2018, HESSD Satellite soil moisture

Model setup Domain Extent and resolution: Ø EU-CORDEX at 0.03° x 0.03°(~ 3km) Land surface inputs: Ø Topography (GMTED2000 elevation) Ø Soil characteristics (FAO global dataset) Ø Vegetation LAI (global LAI product) Ø Land surface classification (MODIS)

Meteorological forcings • COSMO-REA6 reanalysis at 6km resolution (Bollmeyer et al., 2015), • Temperature, precipitation, Wind speed, humidity, short and longwave radiation, pressure • Compare precipitation with DWD station data (166 stations for 1997 – 2006 time period),

Model Validation GRDC stations (4105) Ø The Global Runoff Data Centre (GRDC) gauge stations Streamflow (2001 – 2015) Ø MODIS Snow Cover (2001 – 2015) Ø ESACCI satellite soil moisture observation (2000 – 2006) at 0.25 o resolution, E-RUN v1 gridded monthly runoff Ø E-RUN v1 gridded monthly runoff for 1950 - 2015 (Gudmundsson et al., 2016), Ø Total Water Storage from GRACE satellite (2003 – 2006)

CLM River Transport Model Continental scale high resolution river discharge simulations at 3Km resolution

CLM Streamflow comparison with GRDC stations Monthly flow (2001 – 2015) Gave Dòssau, France (NSE: 0.8) Nash–Sutcliffe efficiency for 884 stations Rhine River (NSE: 0.6) Danube River (NSE: -0.5) An NSE of 1 corresponds to a perfect match of Red: Observation modeled discharge to the observed data. Blue: Simulated

Data assimilation of soil moisture Soil moisture is a key component to control the water and energy exchange between the atmospheres and the land surface. Large scale estimates of soil moisture are available from: Land surface model forced w/ ESA CCI surface soil moisture observed meteorology. Complete Upper few cm, ~25km, ~daily. space/time coverage, incl. root zone. Weights based on uncertainties Model Soil RS Soil Moisture Moisture Assimilation (subject to (subject to data gaps) error) Optimal Soil Moisture/other hydrologic predictions

Data assimilation of soil moisture Ø ESACCI satellite soil moisture observation (2000 – 2006) at 0.25 o resolution. Ø Data coverage of observation is low in northern Europe and in winter. Ø For data assimilation 100 grid cells were randomly selected. Ø CCI-SM soil moisture: cross- validation over grid cells that were not used in the data assimilations Simulation period: Simulation Scenarios: Ø 1 January 2000 – 31 December Ø Open-loop (no data assimilation) 2006. Ø State update (soil moisture) Ensemble generation: Ø 12 realization of perturbed precipitation and soil texture.

Impacts on seasonal surface soil moisture CLM-OL CLM-DA CCI-SM Ø CLM-OL has higher soil Winter water content (upper two soil layers) in all seasons over most part of Europe compared to the CLM-DA simulations Spring and CCI-SM. Ø The SWC in the summer and autumn is better reproduced in the CLM-DA simulations Summer than in the CLM-OL when compared with the CCI-SM Autumn

Daily surface soil moisture evaluation over Prudence regions

Evaluation of monthly runoff over PRUDENCE regions

Impacts on Total Water Storage Ø Total water storage anomaly from GRACE satellite was compared with CLM TWS for 2003 – 2006 time period. Ø Total water storage from CLM was calculated through vertical aggregation of: Ø Snow water, Ø Canopy water, Ø Soil ice, Ø Soil water, Ø Aquifer water. Ø Horizontal aggregation to ½ degree to compare with GRACE data

Summary and Outlook 1) Using COSMO-REA6 data, CLM 3km model realistically simulates hydrological states and fluxes 2) Assimilating daily satellite SM improved near-surface soil moisture simulations over most parts of Europe relative to open-loop simulations. 3) CLM-DA underestimated runoff in the summer and autumn seasons particularly in the mid and south Europe. 4) Assimilating SM data showed slightly reduced correlation with GRACE TWS compared to open-loop estimates. Future work will focus: Ø Calibrate model parameters/ ensemble size and parameter updates Ø Joint assimilation of SM and GRACE data Interface EU 3km RS Data model

Thank you Acknowledgements: Ø The Energy oriented Centre of Excellence in computing applications (EoCoE) (Horizon 2020 programme of the European Commission) Ø The Jülich Supercomputing Centre Naz, B. S., Kurtz, W., Montzka, C., Sharples, W., Goergen, K., Keune, J., Gao, H., Springer, A., Hendricks Franssen, H.-J., and Kollet, S.: Improving soil moisture and runoff simulations over Europe using a high-resolution data-assimilation modeling framework, Hydrol. Earth Syst. Sci. Discuss., https://doi.org/10.5194/hess-2018-24, in review, 2018.

Extra slides

CLM Snow Cover comparison with MODIS Ø Snow cover fraction from MODIS data was compared with CLM snow cover for 2001 – 2015 time period. Ø Snow cover fraction from CLM was extracted for area where MODIS data is available. Ø Horizontal aggregation to 0.05 o degree to compare with MODIS data

Improvements in monthly runoff estimates • Improvements in terms of RMSE are more pronounced in the SC,AL and EA regions and in the spring season • Negative bias error in CLM-DA

Impacts on seasonal runoff Ø CLM-OL estimated Winter greater magnitude of runoff over most parts of Europe compared to CLM-DA in all seasons. Ø The overestimation in Spring the CLM-OL runs is more pronounced in the summer and spring seasons. Ø CLM-DA underestimated Summer runoff in the summer and autumn seasons particularly in the mid and south Europe. Autumn

Improvements in Soil Moisture • Improvements in terms of RMSE are more pronounced in the SC,AL,MD and EA regions • Negative bias error in summer in CLM- DA

Evaluation of monthly runoff over PRUDENCE regions

Soil texture parameter update impacts on soil moisture and runoff Mean annual soil water content (2000 – 2006)

Soil texture parameter update impacts on soil moisture and runoff Mean annual Runoff (2000 – 2006)

CCI soil moisture data interpolation (25 km vs. 3km)

CLM-PDAF: The Global Ensemble Kalman Filter Ensemble generation: ensemble of Ø 12 realization of perturbed precipitation state vectors and soil texture. Ø Soil moisture updates was set to ensemble of 1 day. updated state vectors observation ensemble of Kalman Gain operator perturbed matrix observations ‘weight matrix’ Ø Determines by model and observation covariance matrices

CLM Snow cover fraction CLM SCF SCF (fraction of a grid covered by snow) is calculated as a function of snow density and snow depth CLM 3km FSNO is aggregated to 5km resolution.