Numerical Enrivonmental Prediction, on the Way Towards More Integrated Forecasting of the Earth System Stéphane Bélair Meteorological Research Division Environment Canada WWOSC, Montreal, August 19th, 2014 Numerical Weather Prediction NWP NEP Numerical Environmetnal Prediction Numerical Weather Prediction NWP NEP Numerical Environmetnal Prediction ‘’Traditional’’ NWP… Plenty of Environmental Processes ATMOSPHERIC DYNAMICS / CIRCULATIONS CLOUDS ATMOSPHERIC RADIATION PRECIPITATION CITIES VEGETATION LAND SEA-ICE OCEANS LAKES GLACIERS SNOW ‘’Traditional’’ NWP… Characteristics “In-line” treatment Single code (most often) Same timestep Same spatial resolution Optimized for meteorology Incomplete The Larger and more Modular View of NEP AIR QUALITY MODELS FOREST FIRES ATMOSPHERIC DISPERSION SYSTEMS HYDROLOGY HYDRODYNAMICS WAVES LAKE MODELS (1D and 3D) WAVES OCEANS and SEA-ICE SYSTEMS SURFACE PREDICTION SYSTEM (land, vegetation, cities) The Larger and more Modular View of NEP AIR QUALITY MODELS FOREST FIRES ATMOSPHERIC DISPERSION SYSTEMS HYDROLOGY HYDRODYNAMICS WAVES LAKE MODELS (1D and 3D) WAVES OCEANS and SEA-ICE SYSTEMS SURFACE PREDICTION SYSTEM (land, vegetation, cities) Distinct systems Distinct timesteps Distinct codes Distinct spatial resolutions Coupled (one-way or twoway) Optimized for own applications Own assimilation system An Example: Land Surface Prediction Systems The Canadian Land Data Assimilation System (CaLDAS) CaLDAS IN Ancillary land surface data Orography, vegetation, soils, water fraction, ... LAND MODEL (SPS) Analyses of… xb ASSIMILATION EnKF + EnOI Atmospheric forcing OBS Observations Screen-level (T, Td) Surface stations snow depth L-band passive (SMOS, SMAP) MW passive (AMSR-E) *Optical / IR (MODIS, VIIRS) Combined products (GlobSnow) Surface Temperature Soil moisture y T, q, U, V, Pr, SW, LW OUT EnKF xa = xb+ K { y – H(xb) } Snow depth or SWE with K = BHT ( HBHT+R)-1 Carrera et al. 2014 (in revision) Vegetation* *) not done yet… Coupling CaLDAS with GEM 2.5-km model Upper-air assimilation system 4DVAR– (10 km regional) UA ICs and LBCs Atmospheric model (GEM 2.5 km) Land surface ICs Forcing and first guess Land data assimilation system (CaLDAS) GEM 2.5-km with and without CaLDAS : Dew point temp., Bias, summer, 00 UTC cases BC Prairies North Maritimes USA Que - Ont GEM 2.5-km with and without CaLDAS: Dew point temp., STDE, summer, 00 UTC cases BC Prairies North Maritimes USA Que - Ont CaLDAS-screen (Pan-Canada – 2.5 km) Near-Surface Soil Moisture (0-10 cm) Valid on June 25, 2011, at 1200 UTC Coming… For both global and regional suites Ensemble Kalman Filter (EnKF) Atmosphere ICs Ensemble Prediction System Land surface ICs CaLDAS Forcing and first guess Land surface ICs EnsembleVariational (EnVar) Atmosphere ICs Deterministic Prediction System Land surface prediction system (SPS) ATMOS MODEL LOW-RES 3D INTEGRATION ATMOSPHERIC FORCING at FIRST ATMOS. MODEL LEVEL (T, q, U, V) ATMOSPHERIC FORCING at SURFACE (RADIATION and PRECIPITATION) External Land Surface Model With horizontal resolution as high as that of surface databases (e.g., 100 m) HIGH-RES 2D INTEGRATION Computational cost of off-line surface modeling system is much less than an integration of the atmospheric model 100-m SPS for the 2010 Vancouver Games 100-m snow analyses Great decrease of T2m errors (bias shown here) (Bernier et al. 2011, 2012) (Thanks to Juan Sebastian Fontecilla) Urban Heat Island Modeling (Montreal) Comparison with MODIS MOD11A1 product Resolution: 1km (exactly 928 m) Atmospheric effects corrected Satellite View Angle : 15° • Radiative Surface Temperature (°C) July 6th 2008 (10:54 LST) Warm and Sunny Urban off-line modeling system Resolution: 120 m Z0h: Kanda (2007) (Leroyer et al., 2011) Two-way coupling GEM 2.5 km CaLDAS 2.5 km Lower BCs Nudging surface variables Forcing + first guess Surface Prediction System An ‘’horizontal’’ challenge SINGLE GEM (ATMOSPHERE) GRID AREA (LOW RES) 1 1 ______ KT w' ' t z z z ______ w' ' CT u* S f S NK _ atm S ______ w' ' NK _ atm S CT u* S CT u* f S Spatially averaged SPATIAL AVERAGE OF IMPLICIT LOWER BC FOR VERT. DIFFUSION WATER LAND / VEG (ISBA / SVS) URBAN (TEB) MULTIPLE SURFACE GRID AREAS (HIGH RES) Potential contribution of two-way coupling ~115 Wm-2 95% 75% 25% ~40 Wm-2 5% Subgrid-scale variability of turbulent fluxes for 25-km grid spacing model based on external 2.5-km land surface model (Provided by M. Rochoux, EC) ~115 Wm-2 ~40 Wm-2 INCREASED VERTICAL RESOLUTION A ‘’vertical’’ challenge SINGLE GEM (ATMOSPHERE) GRID AREA (LOW RES) SPATIAL AVERAGE of IMPLICIT LOWER BC for VERT. DIFFUSION (to be applied over atmospheric level just above canopy / soil water / ice) SPATIAL AVG of TENDENCIES for EACH INTERSECTING LEVEL WATER LAND / VEG (ISBA / SVS) URBAN (TEB) MULTIPLE SURFACE GRID AREAS (HIGH RES) Coupling Urban Canopy w/ Atmosphere CaM-TEB (Canadian Multilayer version of TEB) Several model levels intersect the buildings. Variable building heights exist within a grid cell. (Husain et al. 2013) To be tested with Pan Am and TOMACS Real-time 250-m GEM runs over the Toronto region in preparation of the Pan American Games. Here, precip rates and surface winds for 17 June 2014. Offline runs with SPS over Tokyo. Here, surface air temperature for 26 August 2011.