soil moisture and numerical weather prediction

Soil Moisture and Numerical Weather PredictionSoil Moisture and Numerical Weather Prediction

SMAP Applications Workshop, Silver Spring, September 2009

Stephane BelairScience and Technology Branch,

Environment Canada

LL

MOTIVATION for BETTER SOIL MOISTURE in NWP MODELS

1 2

3

Low-level air Temp. Errors

bias

RMS

(K)

(hour)

(hPa)

(K)

Temp. Errors

Soil moisture + assimilation

Control

Precipitation Threat Score (Day 4)- SHEF (US)

NEAR-SURFACE AIR CONDITIONSBOUNDARY-LAYER andMIXING

CLOUDS and PRECIPITATION

RMSE

BIAS

Control

With improvedsoil moisture

RMSE

BIAS

Improvedsoil moist.

Ctrl

1 and 2: Implementation of short-range regional system in 2001. 48 cases (summertime)3. Implementation of global medium-range in 2006. 116 cases (summertime)

POSSIBLE IMPROVEMENT at MEDIUM-RANGE (GLOBAL MODEL)

m3m-3

(valid at 0000 UTC 15 December 2001)

Near surface soil moisture

m3m-3



Global

Medium-Range prediction system

116 cases

120-h, Northern Hem.

red = improved soil moisture

Zonal wind Wind speed

Temperature Geopotential height

BiasRMSE


Global, Medium-Range prediction system

116 cases, Northern Hemisphere


RMSE, Geopotential Height

500 hPa


?

RMSE, Geopotential Height

500 hPa


Global, Medium-Range prediction system

116 cases, Northern Hemisphere


OTHER ASPECTS BECOMING MORE IMPORTANTSOIL MOISTURE and AIR QUALITY / DISPERSION

Dry Soil Dry SoilUrban Center

Humid Soil Humid SoilUrban Center

Solar

Solar

MEAN FLOW

GRADUAL DIMINUTION OF TURBULENCETRANSITION TO FREE ATMOSPHERETRANSPORT POLLUTANT AWAY FROM CITY

MORE SHALLOW BOUNDARY LAYERBUT A SMALLER PORTION OF POLLUTANTS IS INTRODUCED IN THE DOWNWIND WELL-MIXED LAYER

BOUNDARY LAYER IS DEEPERAND “MIXES” MOST OF THE POLLUTANTS PRODUCED BY THE CITY

SMAP DATA in the ENSEMBLE KALMAN FILTER of theCANADIAN LAND DATA ASSIMILATION SYSTEM (CaLDAS)

LAND SURFACEMODEL (M)

OBS OPERATOR (H)

GEO DATA

FORCING DATA

(perturbed)

OBS DATA

(perturbed) Background error covariances (Bt+1)

Obs error covariances (Rt+1)

Background(first guess)

Gain Matrix (Kt+1)

Increments and updates

New analyses

( )−

+1tx ( )+

+1tx

( )i1ty +

Previousanalyses

( )+tx

SMAP DATA in the ENSEMBLE KALMAN FILTER of theCANADIAN LAND DATA ASSIMILATION SYSTEM (CaLDAS)

LAND SURFACEMODEL (M)

OBS OPERATOR (H)

GEO DATA

FORCING DATA

(perturbed)

OBS DATA

(perturbed) Background error covariances (Bt+1)

Obs error covariances (Rt+1)

Background(first guess)

Gain Matrix (Kt+1)

Increments and updates

New analyses

( )−

+1tx ( )+

+1tx

( )i1ty +

Previousanalyses

( )+tx

SMAP

ANCILLARY DATA for the ASSIMILATION of SMAP DATA

FORCING DATA

(perturbed)GEO DATA

Analyses or best estimates of:

Precipitation (quantity and phase)

Radiation incident at surface (LW and SW –direct and diffuse)

Low-level air temperature

Low-level air humidity

Low-level winds

Information on:

Orography (DEM)

Land / water coverage fractions (databases)

Soil texture (databases)

Vegetation type (databases / remote sensing)

Vegetation conditions, i.e., LAI (analysis based on remote sensing)

Urban cover types (databases)

Snow conditions, i.e., coverage at least (analyses)

Surface temperature (analyses)

RESEARCH and DEVELOPMENT ISSUES

Assimilation of SMAP data

Impact of SMAP data

Ancillary data

Specification of background and observations error covariances

Role and impact of other data (i.e., screen-level obs)

Impact on boundary layer, clouds, and precipitation is expected

Impact on large-scale upper-air features?

Atmospheric forcing (e.g., precipitation) of crucial importance

Geophysical data (soil texture not that good)

Vegetation analysis (combination of satellite and model products)

THANK YOU for your ATTENTION

SOIL MOISTURE at ENVIRONMENT CANADAHISTORICAL PERSPECTIVE

1986

1995

2001

2006

2011?

Force-restore scheme with a time-evolving “soil wet ness” variable. Initial conditions provided by a “climat ology”

Simple assimilation of soil moisture based on scree n-level air temperature and humidity

Land surface scheme with a soil moisture variable. Improved method for the initial conditions, but sti ll based on screen-level data (regional, short-range system)

Same but for the global, medium-range system

Assimilation of SMOS passive data

2014? Assimilation of SMAP passive and active data

ASSIMILATION of SPACE-SPACE REMOTE SENSING DATA

Local

Regional

Global

Land Surface

Surface fluxes (H, LE, momentum)

Boundary-layer mixing

Near-surface air conditions

Clouds

Precipitation

Convective systems

Large-scale systems

Spa

tial a

nd T

ime

scal

es

PROCESSES MODELS

• Determine exchanges of heat, water, and momentum between the surface and the atmosphere

• Impact on near-surface conditions, boundary-layer mixing, clouds, precipitation, and maybe even large-scale systems

• Important for local, continental, and global models, … at short, medium, and long ranges

• Land surface is even more important for new environmental systems now in development at EC, e.g., external land surface system, hydrology.

SOIL MOISTURE and NUMERICAL WEATHER PREDICTION

IMPACT on NEAR SURFACE and BOUNDARY LAYER


(valid at 1200 UTC 22 October 2004)

m3m-3(hPa)

(K)

Temp. Errors

48-h integrations

(Bélair et al. 2003)

CONTROLIMPROVED SURFACE

RMSEBIAS


bias

RMS

(K)

(hour)

IMPACT on NEAR SURFACE and BOUNDARY LAYER


(valid at 1200 UTC 22 October 2004)

m3m-3


bias

RMS

(K)

(hour)

(hPa)

(K)

Temp. Errors

48-h integrations

(Bélair et al. 2003)

CONTROLIMPROVED SURFACE

RMSEBIAS

IMPACT on PRECIPITATION

Precipitation bias (24-48h)

Control

Soilmoisture assimilation (OI)

Soil moisture + assimilation

Control

Precipitation Threat Score (Day 4)- SHEF (US)

REGIONAL FORECASTING SYSTEM (2001)

48 CASES

GLOBAL FORECASTING SYSTEM (2006)

116 CASES

THERMODYNAMIC versus MOMENTUM SURFACE FLUXES... Based on our experience at EC

� H and LE surface fluxes

� Surface temperatures

� Soil moisture

� Water / land fractions

H LE

Significant impact on boundary layer, clouds, and precipitation. But not clear yet how it influences large-scale upper-air features

� Resolved orography

� Blocking effect (subgrid-scale)

� Turbulent mixing (subgrid-scale)

Significant impact on large-scale features, but not on clouds and precipitation

THERMO MOMENTUM

m3m-3



Global


116 cases

120-h, Europe




BiasRMSE


m3m-3



Significant impact for this region... but not always like this. Impact on medium-range upper-air prediction still TBD



BiasRMSE


Global


116 cases

120-h, Europe


soil moisture and numerical weather prediction

Documents