Système de modélisation et d assimilation de surface à

Système de modélisation et d’assimilation de surface à Environnement Canada Stéphane Bélair, Bernard Bilodeau, Marco Carrera, Dorothée Charpentier, Isabelle Doré, Vincent Fortin, Pablo Grunmann, Aude Lemonsu, Alexandre Leroux, Pierre Pellerin, Wei Yu, and Ayrton Zadra Recherche en prévision numérique (RPN) Division de la recherche en météorologie Environnement Canada

Impact of Surface Processes on NWP: Short-Range Regional Model (2001) Near surface soil moisture m 3 m-3 Precipitation bias (24 -48 h) Old OP model New surface scheme with soil moisture assimilation (OI) (h. Pa) Temp. Better soil moisture resulted in significant improvements for: (valid at 1200 UTC 22 October 2004) (K) Errors Low-level air Temp. Errors • Low-level air temp. and humidity RMS • Diurnal cycle of the PBL bias RMS • Precipitation biases NOTE: mostly in summer bias (K) (hour) (Bélair et al. ) 48 -h integrations

Impact of Surface Processes on NWP: Medium-Range Global Model (2006) Near surface soil moisture m 3 m-3 120 -h, Europe (valid at 0000 UTC 15 December 2001) Precipitation Threat Score (Day 4)- SHEF ISBA + soil moisture Control Has been implemented in the global forecasting system (31 October 2006). (Bélair et al. )

End-to-End Land Surface System (under construction) OBSERVATIONS LAND SURFACE DATABASES and HIGH-RES ANALYSES TRANSFER MODELS SURFACE FIELDS GENERATOR OFF-LINE LAND SURFACE MODELING SYSTEM (MEC) GRID PARAMETERS DOWNSCALING MODELS LAND SURFACE INITIAL CONDITIONS LAND SURFACE MODELS (in-line or off-line) 3 compontents • Fields generator • Assimilation and analyses • Models LAND SURFACE FORECASTS or BOUNDARY CONDITIONS FOR ATMOSPHERIC AND HYDROLOGY MODELS DOWNSCALING MODELS LAND SURFACE DATA ASSIMILATION SYSTEM (Ca. LDAS) OUTPUT PROCESSOR BEST ESTIMATES ATMOSPHERIC FORCING FORECASTS MODELS or CLIENTS

Land Surface Databases Soil texture STATSGO in the CONUS (~ 1 km) AG-Can in Canada (variable resolution) FAO global (~ 7 -8 km) Updated FAO global (~ 7 -8 km) Vegetation USGS global (~ 1 km) GLOBCOVER global (~ 300 m) SAFORAH Canada (~ 25 m forests) Topography USGS global (~ 1 km) US Navy global (~ 1 km) SRTM-DEM Canada (~ 30 m) SRTM-DEM global (~ 90 m) CDED Canada (~ 20 m) USGS-NED US (~ 10 m) Urban Satellite ASTER and LANDSAT in-house algorithm NTDB Canada (Vector data) Italics = not yet available in Gen-Gen software

Optimal Interpolation Analysis of LAI Using MODIS (Gu, Bélair, Mahfouf, Deblonde, 2006) αo = , and αc = Land cover databases do not provide information on LAI (usually specified using a look-up table). LAI is important for evapotranspiration. Using the LAI analysis from MODIS (or other instruments) could reduce an important source of errors.

Land Surface Databases Soil texture STATSGO in the CONUS (~ 1 km) AG-Can in Canada (variable resolution) FAO global (~ 7 -8 km) Updated FAO global (~ 7 -8 km) Vegetation USGS global (~ 1 km) GLOBCOVER global (~ 300 m) SAFORAH Canada (~ 25 m forests) Topography USGS global (~ 1 km) US Navy global (~ 1 km) SRTM-DEM Canada (~ 30 m) SRTM-DEM global (~ 90 m) CDED Canada (~ 20 m) USGS-NED US (~ 10 m) Urban Satellite ASTER and LANDSAT in-house algorithm NTDB Canada (Vector data) Italics = not yet available in Gen-Gen software

Orography and Urban Areas for the VO-2010 External Surface Modeling System Topography and the Vancouver urban area Topography from SRTM (30 m) Urban LULC from Lemonsu et al. (2007) Computational grid: 1400 x 1800 (100 m)

Example of Urban Covers: Montreal (Vector database, 44 classes)

RPN’s Surface Modeling System Water Sea ice Urban Soil and vegetation Glaciers Soils and vegetation ISBA, CLASS, Force-Restore Water Simple scheme with constant surface temperature (Lake model eventually) Urban covers TEB Glaciers Force-restore scheme (with snow), module from CLASS Sea ice 3 -layer model with snow on top Snow Very simple scheme over glaciers and sea ice; better in ISBA, CLASS and TEB, SNTHERM (IPY project)

External Land Surface System: Refining CMC’s Forecasts at the Surface Global (33 km) Urban (200 m) MODEL OUTPUT Local (2. 5 km) 2 days 1 day MODEL OUTPUT Regional (15 km) MODEL OUTPUT Grid size 10 days ATMOSPHERIC FORCING External Surface Model With horizontal resolution as high as that of surface databases (e. g. , 200 m) 10 days Cost of the external surface modeling system is much less than an integration of the full atmospheric model

External Surface Modeling System ATMOSPHERIC FORCING from GEM (forecasts) INITIAL SURFACE CONDITIONS Temperatures Soil water content Soil ice content Snow characteristics Urban surfaces wetness Near-surface air characteristics (temperature, humidity, winds) Surface pressure Incident radiation (solar and infrared) Precipitation (rain and snow) Low res forcing DOWNSCALING MODELS High res forcing EXTERNAL SURFACE MODEL LAND SURFACE CHARACTERISTICS Topography Roughness Land/water fractions Soil texture Natural cover types Urban cover types Glaciers OUTPUTS All surface prognostic variables Low-level air temperature and humidity Low-level winds (from adaptation+roughness) Surface fluxes (coupling with atmos models) Surface runoff and drainage (coupling with hydro)

Snow in Mountains: The Effect of Downscaling Air Temperature Snow Water Equivalent Orography (GEM) 15 km ALB ALB 1 km (MEC) (Valid 0000 UTC 1 December 2006) (kg/m 2)

Urban Environment (UHI): The Effect of High-Res Databases OKC urban area Near-surface air temperature (K) 27 -hr 250 -m forecasts of near-surface air temperature over Oklahoma City using the Town Energy Balance scheme, MEC offline, valid at 0000 local time, 18 July 2003 (Lemonsu et al. 2007, in preparation)

Canadian Land Surface Data Assimilation System (Ca. LDAS) – under construction ATMOSPHERIC FORCING ANCILLARY DATA Ancillary data Vegetation types (USGS, GLOBCOVER) Soil texture (STATSGO, FAO) Water bodies Cities (NLCD, CMC) Topography (USGS, SRTM, CDED, USGSNED) – – – Meteorological Analyses - low-level winds, temps, and humidity Precipitation analyses - meteorological forecasts - Canadian Precipitation Analysis (Ca. PA) Radiation analyses Best estimates from models and assimilation cycles Downscaling Land schemes ISBA, CLASS Regional - North America (1 -2 km) OFF-LINE SURFACE MODELING SYSTEM Global (5 -10 km) Integrated on CMC's supercomputer Transfer models (emission, backscatter, surface layer…) Surface temperature From remote sensing: GOES, … (IR) OBSERVATIONS Vegetation From remote sensing: MODIS, AVHRR (NDVI) Snow on ground From surface obs and remote-sensing: SSM/I, MODIS (visible, MW) Freeze/thaw From remote-sensing: Quik. Scat (Ku-band) Hydros (L-band) SSM/I (Microwave) Soil Moisture low-level air characteristics IR heating rates C-band radiances L-band Tb and o (Hydros) -

Variational Assimilation of Soil Moisture and Surface Temperatures (Simplified 2 D-Var) Cost function 15 July 2005 Linear hypothesis In this technique, the linear observation operator H is evaluated using a finite difference approach, from two perturbed model integrations. Also, the minimum of J(x) is directly obtained from 1 October 2005 The analyzed state xa is thus given by: where K is the gain matrix: This formulation of the variational problem could be easily converted to an Extended Kalman Filter (Balsamo, Mahfouf, Bélair, Deblonde, 2006 a, b)

Satellite Observations of Soil Moisture ~40 -50 km spatial resolution hourly L-band Tb C-band Tb 6 -hourly IR Ts T/H 2 m

CMC’s Snow Analysis Global Analysis (33 km) Valid 5 December 2006 Strategy similar to Ca. PA (statistical interpolation) Background snow depth is given by a simple off-line snow model Reports from SYNOP and METAR are used (Brasnett) When done on a 2. 5 -km grid, with adaptation of temperature forcing for the high-resolution orography, the analysis seems more realistic in mountainous regions

Surface Externalisée et la Modélisation du Climat Régional • La surface externalisée, à haute résolution, pourrait être d’intérêt pour la modélisation du climat régional • Plusieurs éléments météorologiques d’intérêts sont associés à la surface (température de l’air, vent, phase de la précipitation, couverture de neige, …) • Première version du système de prévision sera avec couplage 1 -way, mais couplage 2 -way est aussi possible (déjà fait avec les modèles de glace marine) • Caractéristiques de surface peuvent évoluer d’une manière pronostique (e. g. , modèle de végétation du CCCma) ou de manière prescrite (prévision d’urbanisation de la région Montréalaise).

Scope of the Project Direct contributors: Bélair, Bilodeau, Charpentier, Fortin, Zadra, Chamberland, Yu, Doré, Carrera, Lemay, Grunmann, Barscz, plus several PDFs soon coming on external funding. Collaborators: Climate Research Division, National Water Research Institute, Canadian Meteorological Centre – Ouranos ? ? External funding: CRTI (CBRN Research and Technology Initiative), EPi. CC (Environmental Prediction in Canadian Cities), NAESI (National Agri-Environmental Standards Initiative), GRIP (Government-Related Initiatives Program), IPY (International Polar Year, TAWEPI and CFL), VO 2010 (Vancouver Olympic Games of 2010) Active projects: Surface geophysical fields, Canadian Precipitation Analysis (Ca. PA), Canadian Land Data Assimilation System (Ca. LDAS), Canadian Land Surface Scheme (CLASS), Atmospheric forcing and downscaling, Town Energy Balance, Momentum fluxes, blowing snow, surface-atmosphere coupling, near-surface wind forecasting, operational transfer of updates to ISBA

Objectives of the Land Surface Group/Project • Improve the performance of existing and new environmental prediction systems (atmosphere, hydrology) by providing better forcing data to them • Improve analyses/forecasts of the land surface state (soil wetness, snow on the ground) and of near-surface atmospheric conditions • Develop new products, such as snow conditions (and avalanche? ), conditions in cities, blowing snow, low-level winds

Met-Surface: Statistical Adaptation for Local Prediction INITIALIZATION / ASSIMILATION FORECAST LAM 1 km for day 1 Reg for day 2 Glb after GEM OBS FORCING Precip Tair, qair Wind Cloud cover OBS SURFACE VARIABLES SCRIBE Tsnow Snow depth Soil moisture Tsurf LAND SURFACE MODEL + VAR ASSIMILATION Assimilation window UMOS Radiation is calculated from cloud cover and Tair Precip Tair, qair Wind Cloud cover LAND SURFACE MODEL Assimilation window Forecast Initial Conditions

Land Surface Models, Analyses, and Assimilation in CMC’s Operational System ANALYSES TS TS, ES, TP Gaussian 1080 x 540 TP TM TS, ES Gaussian 1080 x 540 For snow anal Gaussian 1080 x 540 18 UTC Reg-576 x 641 TS, ES ASSIMILATION SEQ. ASSIMILATION SNOW SD Global 800 x 600 Gaussian 1080 x 540 SD ISBA fields SEQ. ASSIMILATION SD 6 -h forecasts Regional 576 x 641 SD ISBA fields SD 18 -h forecasts PR MODELS ENSEMBLES GLOBAL REGIONAL LOCAL GEM and SEF (ISBA, FR, glaciers, water) GEM 800 x 600 uniform (ISBA, glaciers, water) GEM 576 x 641 variable (ISBA, glaciers, water) GEM-LAM East and West (ISBA, glaciers, water) GENESIS DATABASES Soil texture, orography, vegetation, lakes, and glaciers ISBA fields: Tsurf(1, 2), Wsoil(1, 2), wice, snow albedo, snow density, wsliq, wlveg ISBA and snow fields

Long-Term Vision for CMC’s Land Surface Models, Analyses, and Assimilation ANALYSES TS, ES, TP TM-Lakes GL-Lakes External system’s Grid (high-res) TP Vegetation OI Satellite High-res grid Satellite Ca. PA Model, surface and satellite data TS, ES ASSIMILATION Canadian Land surface Data Assimilation System (Ca. LDAS) 2 D-Var for soil moisture and surface temperature (screen-level + sat) Snow mass and coverage (surface data + sat) High-res global grids (same as external system) Initial conditions for land surface schemes LAND SURFACE MODELS MESH External Land Surface Modeling System (MEC – Environmental Community Model) High-resolution grid over Canada (1 km or less) – Lower resolution grid over world (5 km or less) (CLASS or ISBA, TEB, WATER - possibly LAKES, SNOW, GLACIERS, EOLE, blowing snow) Forcings 2 -way coupling OTHER MODELS ENSEMBLES GLOBAL REGIONAL (FR, ISBA, CLASS, WATER, GLACIERS) DATABASES Gen-Gen Soil texture, orography, vegetation, water bodies, glaciers, and cities LOCAL HYDROLOGY