m. ek - land surface in weather and climate models; "surface scheme"
TRANSCRIPT
1
Land Surface in Weather and Climate Models
“IV WorkEta: Esquema de superfície” Michael Ek
Environmental Modeling Center (EMC) National Centers for Environmental Prediction (NCEP)
NOAA Center for Weather and Climate Prediction (NCWCP) 5830 University Research Court
College Park, MD 20740
National Weather Service (NWS) National Oceanic and Atmospheric Administration (NOAA)
IV WorkEta – Workshop em Modelagem Numérica de Tempo e Clima em Mesoescala utilizando o Modelo Eta: Aspectos Físicos e Numéricos CPTEC/INPE, Cachoeira Paulista, São Paulo, 3-8 de Março de 2013
NOAA Center for Weather and Climate Prediction (NCWCP), College Park, Maryland, USA
2 World Weather Building
Development, upgrade, transition, maintain models
www.noaa.gov
www.nws.noaa.gov
www.ncep.noaa.gov
www.emc.ncep.noaa.gov
“Determine the predictability of climate and effect of human activities on climate.”
“Weather, Water, Climate”
“…improve accuracy, lead time and utilization of weather prediction.”
www.gewex.org
www.wmo.int
www.wcrp-climate.org
“www.wmo.int “WWRP”
5
• Role of Land Surface Models (LSMs) • Requirements: - atmospheric forcing, valid physics, land data sets/
parameters, initial land states, cycled land states • Applications for Weather & Climate • Validation • Improving LSMs • Expanding role of Land Modeling as part of an integrated Earth System
• Summary
Outline
6
• Role of Land Surface Models (LSMs) • Requirements: - atmospheric forcing, valid physics, land data sets/
parameters, initial land states, cycled land states • Applications for Weather & Climate • Validation • Improving LSMs • Expanding role of Land Modeling as part of an integrated Earth System
• Summary
Outline
• Traditionally, from a coupled (atmosphere-ocean-land-ice) Numerical Weather Prediction (NWP) and climate modeling perspective, a land-surface model provides quantities to a parent atmospheric model: - Surface sensible heat flux, - Surface latent heat flux (evapotranspiration) - Upward longwave radiation
(or skin temperature and surface emissivity), - Upward (reflected) shortwave radiation
(or surface albedo, including snow effects), - Surface momentum exchange.
Role of Land Models
7
8
Land modeling example:
• Products and models are integrated & consistent throughout time & space, as well as across forecast application & domain.
Static vegetation, e.g. climatology
or realtime observations
Dynamic vegetation, e.g. plant
growth Dynamic
ecosystems, e.g. changing
land cover
Weather & Climate a “Seamless Suite”
9
…to close the surface energy budget, & provide surface boundary conditions to NWP and climate models.
seasonal storage
Atmospheric Energy Budget
10
• Land models close the surface water budget, and provide surface boundary conditions to models.
Water Budget (Hydrological Cycle)
History of Land Modeling (e.g. at NCEP) – 1960s (6-Layer PE model): land surface ignored • Aside from terrain height and surface friction effects – 1970s (LFM): land surface ignored – Late 1980s (NGM): first simple land model introduced (Tuccillo) • Single layer soil slab (“force-restore” model: Deardorff) • No explicit vegetation treatment • Temporally fixed surface wetness factor • Diurnal cycle is treated (as is PBL) with diurnal surface radiation • Surface albedo, skin temperature, surface energy balance • Snow cover (but not depth) – Early1990s (Global Model): OSU land model (Mahrt& Pan, 1984) • Multi-layer soil column (2-layers) • Explicit annual cycle of vegetation effects • Snow pack physics (snowdepth, SWE) – Mid-90s (Meso Model): OSU/Noah LSM replaces Force-Restore – Mid 2000s (Global Model): Noah replaces OSU (Ek et al 2003) – Mid 2000s (Meso Model: WRF): Unified Noah LSM with NCAR – 2000s-10s: Noah “MP” with explicit canopy, ground water,CO2
INCOMING SOLAR AND LONGWAVE
SOIL HEATING
EVAPORATION
BOUNDARY-LAYER TURBULENT MIXING
CLOUDS
HEATING THE AIR
& OUTGOING LONGWAVE
MODEL PREDICTIONS: PROCESSES
LAND-SURFACE
ATMOSPHERIC BOUNDARY
LAYER
SOIL: TEMPERATURE MOISTURE
ATMOSPHERIC: TEMPERATURE HUMIDITY WINDS
MODEL PREDICTIONS: STATES FREE ATMOSPHERE
14
• Many land-surface/atmospheric processes. • Interactions & feedbacks between the land and atmosphere must be properly represented. • Global Energy and Water Exchanges (GEWEX) Program / Global Land/Atmosphere System Study (GLASS) project: Accurately understand, model, and predict the role of Local Land-Atmosphere Coupling “LoCo” in water and energy cycles in weather and climate models. • All land-atmosphere coupling is local… …initially.
Land-Atmosphere Interaction
15
LoCo: Local land-atmospheric modeling
15
• NEAR-SURFACE LOCAL COUPLING METRIC: Evaporative fraction change with changing soil moisture for both bare soil & vegetated surface = function of near-surface turbulence, canopy control, soil hydraulics & soil thermodynamics,
• Utilize GHP/CEOP reference site data sets (“clean” fluxnet),
• Extend to coupling with atmos. boundary-layer and entrainment processes,
• Link with approaches by Santanello et al.
From “GEWEX Imperatives: Plans for
2013 and Beyond”
16
Betts et al (1996)
Diurnal time-scales
Seasonal
Century
Land-Atmosphere Interaction
17
• Role of Land Surface Models (LSMs) • Requirements: - atmospheric forcing, valid physics, land data sets/
parameters, initial land states, cycled land states • Applications for Weather & Climate • Validation • Improving LSMs • Expanding role of Land Modeling as part of an integrated Earth System
• Summary
Outline
18
• To provide proper boundary conditions, land model must have: - Necessary atmospheric forcing to drive the land
model, - Appropriate physics to represent land-surface
processes (for relevant time/spatial scales), - Corresponding land data sets and associated
parameters, e.g. land use/land cover (vegetation type), soil type, surface albedo, snow cover, surface roughness, etc., and
- Proper initial land states, analogous to initial atmospheric conditions, though land states may carry more “memory” (e.g. especially in deep soil moisture), similar to ocean SSTs.
Land Model Requirements
19
Precipitation Incoming solar
Incoming Longwave
Wind speed
Air temperature Specific humidity
+ Atmospheric Pressure Example from 18 UTC, 12 Feb 2011
Atmospheric Forcing to Land Model
• Four soil layers (10, 30, 60, 100 cm thick).
• Linearized (non-iterative) surface energy budget; numerically efficient.
• Soil hydraulics and parameters follow Cosby et al.
• Jarvis-Stewart “big-leaf” canopy cond.
• Direct soil evaporation. • Canopy interception. • Vegetation-reduced soil
thermal conductivity. • Patchy/fractional snow
cover effect on surface fluxes; coverage treated as function of snowdepth & veg type 20
Unified NCEP-NCAR Noah land model
• Freeze/thaw soil physics. • Snowpack density and snow water
equivalent. • Veg & soil classes parameters.
21
Soil Moisture (Θ):
• “Richard’s Equation”; DΘ (soil water diffusivity) and KΘ (hydraulic conductivity), are nonlinear functions of soil moisture and soil type (Cosby et al 1984); FΘ is a source/sink term for precipitation/evapotranspiration. Soil Temperature (T):
• CT (thermal heat capacity) and KT (soil thermal conductivity; Johansen 1975), non-linear functions of soil/type; soil ice = fct(soil type/temp./moisture). Canopy water (Cw):
• P (precipitation) increases Cw, while Ec (canopy water evaporation) decreases Cw.
Prognostic Equations
22
Surface Energy Budget
Rn = Net radiation = Sê - Sé + Lê - Lé Sê = incoming shortwave (provided by atmos. model) Sé = reflected shortwave (snow-free albedo (α) provided by atmos. model; α modified by Noah model over snow) Lê = downward longwave (provided by atmos. model) Lé = emitted longwave = εσTs
4 (ε=surface emissivity, σ=Stefan-Boltzmann const., Ts=surface skin temperature)
H = sensible heat flux LE = latent heat flux (surface evapotranspiration) G = ground heat flux (subsurface soil heat flux)
SPC = snow phase-change heat flux (melting snow)
• Noah model provides: α, Lé, H, LE, G and SPC.
23
Surface Water Budget
ΔS = change in land-surface water P = precipitation R = runoff E = evapotranspiration
P-R = infiltration of moisture into the soil
• ΔS includes changes in soil moisture, snowpack (cold season), and canopy water (dewfall/frostfall and intercepted precipitation, which are small).
• Evapotranspiration is a function of surface, soil and vegetation characteristics: canopy water, snow cover/ depth, vegetation type/cover/density & rooting depth/ density, soil type, soil water & ice, surface roughness.
• Noah model provides: ΔS, R and E.
24
Potential Evaporation
Δ = slope of saturation vapor pressure curve Rn-G = available energy"
ρ = air density cp = specific heat Ch = surface-layer turbulent exchange coefficient U = wind speed δe = atmos. vapor pressure deficit (humidity) γ = psychrometric constant, fct(pressure)"
open water surface
(Penman)
25
Surface Latent Heat Flux
• LEc is a function of canopy water % saturation. • LEt uses Jarvis (1976)-Stewart (1988) “big-leaf”
canopy conductance. • LEd is a function of near-surface soil % saturation. • LEc, LEt, and LEd are all a function of LEp.
soil
canopy canopy water
Transpiration (LEt)
Canopy Water Evap. (LEc)
Bare Soil Evaporation (LEd)
(Evapotranspiration)
26
Surface Latent Heat Flux (cont.) Canopy Water Evaporation (LEc):
• Cw, Cs are canopy water & canopy water saturation, respectively, a function of veg. type; nc is a coeff.
Transpiration (LEt):
• gc is canopy conductance, gcmax is maximum canopy conductance and gSê, gT, gδe, gΘ are solar, air temperature, humidity, and soil moisture availability factors, respectively, all functions of vegetation type. Bare Soil Evaporation (LEd):
• Θd, Θs are dry (minimum) & saturated soil moisture contents, =fct(soil type); nd is a coefficient (nom.=2).
max sê
27
Latent Heat Flux over Snow LE (shallow snow) LE (deep snow) <
• LEns = “non-snow” evaporation (evapotranspiration terms). • 100% snowcover a function of vegetation type, i.e. shallower for grass & crops, deeper for forests.
soil
snowpack
Shallow/Patchy Snow Snowcover<1
Deep snow Snowcover=1
LEsnow = LEp
LEsnow = LEp
LEns = 0
Sublimation (LEsnow)
LEns < LEp
28
Surface Sensible Heat Flux
ρ, cp = air density, specific heat Ch = surface-layer turbulent exchange coeff. U = wind speed
Tsfc-Tair = surface-air temperature difference
• “effective” Tsfc for canopy, bare soil, snowpack.
soil
canopy snowpack bare soil
(from canopy/soil snowpack surface)
29
Ground Heat Flux
KT = soil thermal conductivity (function of soil type: larger for moister soil, larger for clay soil; reduced through canopy, reduced through snowpack)
Δz = upper soil layer thickness Tsfc-Tsoil = surface-upper soil layer temp. difference
• “effective” Tsfc for canopy, bare soil, snowpack.
soil
canopy snowpack bare soil
(to canopy/soil/snowpack surface)
30
Land Data Sets (e.g. uncoupled Noah)
Soil Type (1-km, STATSGO-FAO)
Vegetation Type (1-km, USGS)
Max.-Snow Albedo (1-deg, Robinson)
Snow-Free Albedo (seasonal, 1-deg, Matthews)
Green Vegetation Fraction (monthly, 1/8-deg, NESDIS AVHRR)
July Jan
• Fixed climatologies or (near real-time) observations. • Some quantities predicted/assimilated (e.g. soil moist., snow,
vegetation).
summer winter
31
Land Data Sets (e.g. meso NAM)
Soil Type (1-km, STATSGO-FAO)
Vegetation Type (1-km, IGBP-MODIS)
Max.-Snow Albedo (1-km, UAz-MODIS)
Green Vegetation Fraction (weekly, 1/8-deg,
new NESDIS/AVHRR)
mid-July mid-Jan
Snow-Free Albedo (monthly, 1-km,
BU-MODIS)
July Jan
32 32
Land Data Sets (e.g. GFS/CFS, GLDAS)
Soil Type (1-deg, Zobler)
Vegetation Type (1-deg, UMD)
Green Vegetation Fraction (monthly, 1/8-deg, NESDIS/
AVHRR)
Max.-Snow Albedo (1-deg, Robinson)
Snow-Free Albedo (seasonal, 1-deg,
Matthews)
July July Jan Jan
33
Land surface model physics parameters (examples)
• Surface momentum roughness dependent on vegetation/land-use type.
• Stomatal control dependent on vegetation type, direct effect on transpiration.
• Depth of snow (snow water equivalent, or s.w.e.) for deep snow and assumption of maximum snow albedo is a function of vegetation type.
• Heat transfer through vegetation and into the soil a function of green vegetation fraction (coverage) and leaf area index (density).
• Soil thermal and hydraulic processes highly dependent on soil type (vary by orders of magnitude).
34
• Valid land state initial conditions are necessary for NWP and climate models, & must be consistent with the land model used in a given weather or climate model, i.e. from same cycling land model.
• Land states spun up in a given NWP or climate model cannot be used directly to initialize another model without a rescaling procedure because of differing land model soil moisture climatologies.
May Soil Moisture Climatology from 30-year
NCEP Climate Forecast System Reanalysis (CFSR), spun up from Noah land model coupled with CFS.
Initial Land States
Initial Land States (cont.)
Air Force Weather Agency snow cover & depth
• In addition to soil moisture: the land model provides surface skin temperature, soil temperature, soil ice, canopy water, and snow depth & snow water equivalent.
National Ice Center snow cover
35
Initial Land States (cont.) • In seasonal (and longer) climate simulations, land states are “cycled” so that there is an evolution in land states in response to atmospheric forcing and land model physics.
• Land data set quantities may be observed and/or simulated, e.g. green vegetation fraction & leaf area index, and even land-use type (evolving ecosystems).
36
• Better initial conditions of land states for numerical weather prediction (NWP) and seasonal climate model forecasts.
• Assess model (physics) performance and make improvements by assimilating real land data (sets).
• Application to regional and global drought monitoring and seasonal hydrological prediction.
Land Data Assimilation: Motivation
37
• Observations incorporated into weather, seasonal climate and hydrological models in analysis cycles.
• In each analysis cycle, observations of current (and possibly past) state of a system are combined with results from a weather/climate/hydro model.
• Analysis step is considered “best” estimate of the current state of the weather/climate/hydro system.
• Analysis step balances uncertainty in data and in forecast. The model is then advanced in time & its results becomes the forecast in next analysis cycle.
Land Data Assimilation (LDA)
38
• Minimization of a "cost function” has the effect of making sure that the analysis does not drift too far away from observations and forecasts that are known to usually be reliable.
Land Data Assimilation (LDA)
39
Figure 4: Changes in annual-average terrestrial water storage (the sum of groundwater, soil water,
surface water, snow, and ice, as an equivalent height of water in cm) between 2009 and 2010, based on
GRACE satellite observations. Future observations will be provided by GRACE-II.
Figure 5: Current lakes and reservoirs monitored by OSTM/Jason-2. Shown are current height variations
relative to 10-year average levels. Future observations will be provided by SWOT.
Figure 2: Annual average precipitation from 1998 to 2009 based on TRMM satellite observations. Future
observations will be provided by GPM.
Figure 1: Snow water equivalent (SWE) based on Terra/MODIS and Aqua/AMSR-E.
Future observations will be provided by JPSS/VIIRS and DWSS/MIS.
Figure 3: Daily soil moisture based on Aqua/AMSR-E. Future observations will be
provided by SMAP.
Developing Land Data Assimilation Capabilities
40 C. Peters-Lidard et al (NASA)
41
NASA Land Information System
Topography, Soils
(Static)
Land Cover, Leaf Area Index
(MODIS, AMSR, TRMM, SRTM)
Meteorology: Modeled & Observed (NLDAS, DMIP II,
GFS,GLDAS,TRMM GOES, Station)
Observed States (Snow
Soil Moisture Temperature)
Land Surface Models SAC-HT/SNOW17
Noah, CLM, VIC, Catchment
Data Assimilation Modules (DI, EKF, EnKF)
Surface Energy Fluxes (Qh,Qle)
Soil Moisture Evaporation
Surface Water Fluxes
(e.g., Runoff)
Surface States
(Snowpack)
Inputs Outputs Physics
Water
Supply & Demand
Agriculture
Hydro- Electric Power, Water
Quality
Improved Short Term
& Long Term Predictions
Applications
41 Christa Peters-Lidard et al., NASA/GSFC/HSB
Figure 3: Daily soil moisture based on Aqua/AMSR-E. Future observations will be
provided by SMAP.
Soil Moisture Data Assimilation
Data Assimilated: • AMSR-E LPRM soil moisture • AMSR-E NASA soil moisture
Variables Analyzed: • Soil Moisture • Evapotranspiration • Steamflow
Experimental Setup: • Domain: CONUS, NLDAS • Resolution: 0.125 deg. • Period: 2002-01 to 2010-01 • Forcing: NLDASII • LSM: Noah 2.7.1,3.2
Peters-Lidard, C.D, S.V. Kumar, D.M. Mocko, Y. Tian, 2011: Estimating evapotranspiration with land data assimilation systems, Hydrological Processes, 25(26), 3979--3992, DOI: 10.1002/hyp.8387
42
LDA: Ensemble Kalman Filter (EnKF)
yk
Nonlinearly propagates ensemble of model trajectories. Can account for wide range of model errors (incl. non-additive). Approx.: Ensemble size.
Linearized update.
xki state vector (eg soil moisture)
Pk state error covariance Rk observation error covariance
Propagation tk-1 to tk:
xki+ = f(xk-1
i-) + wki
w = model error
Update at tk: xk
i+ = xki- + Kk(yk
i - xki- )
for each ensemble member i=1…N Kk = Pk (Pk + Rk)-1
with Pk computed from ensemble spread
From Rolf Reichle (2008)
Soil moisture Assimilation -> soil moisture (Evaluation vs SCAN)
Anomaly correlation
OL NASA-DA LPRM-DA
Surface soil moisture (10cm)
0.55 +/- 0.01
0.49 +/- 0.01
0.56 +/- 0.01
Root zone soil moisture (1m)
0.17 +/- 0.01
0.13 +/- 0.01
0.19 +/- 0.01
ALL available stations (179)
(21) Stations listed in
Reichle et al. (2007)
Anomaly correlation
OL NASA-DA LPRM-DA
Surface soil moisture (10cm)
0.62 +/- 0.05
0.53 +/- 0.05
0.62 +/- 0.05
Root zone soil moisture (1m)
0.16 +/- 0.05
0.13 +/- 0.05
0.19 +/- 0.05 44
Latent Heat Flux (Qle) Estimates over US (“Observed” vs. “Modeled Open Loop” (OL))
Pg. 45
Peters-Lidard, C.D, S.V. Kumar, D.M. Mocko, Y. Tian, 2011: Estimating evapotranspiration with land data assimilation systems, Hydrological Processes, 25(26), 3979--3992, DOI: 10.1002/hyp.8387
DJF MAM JJA SON
FLUXNET
MOD16
GLDAS
NLDAS
(Noah 2.7.1)
NLDAS
(Noah 3.2)
1
Soil Moisture Assimilation -> Evapotranspiration (Qle)
Pg. 46
Peters-Lidard, C.D, S.V. Kumar, D.M. Mocko, Y. Tian, 2011: Estimating evapotranspiration with land data assimilation systems, Hydrological Processes, 25(26), 3979--3992, DOI: 10.1002/hyp.8387
FLUXNET MOD16
RMSE
Bias
1
Where Does Soil Moisture Assimilation Help Improve Qle (i.e. Reduce RMSE) ?
Pg. 47
Peters-Lidard, Christa D., Sujay V. Kumar, David M. Mocko and Yudong Tian, (2011), Estimating Evapotranspiration with Land Data Assimilation Systems, In press, Hyd. Proc.
FLUXNET MOD16
DJF
MAM
JJA
SON
1
FLUXNET MOD16
DJF
MAM
JJA
SON
1
LPRM-DA NASA-DA
Where Does Soil Moisture Assimilation Help Improve Qle (i.e. Reduce RMSE) ?
Pg. 48
Qle RMSE % Difference
(DA-OL)
FLUXNET MOD16
Landcover type NASA-DA
(Wm-2)
LPRM-DA
(Wm-2)
NASA-DA
(Wm-2)
LPRM-DA
(Wm-2)
Evergreen needleleaf forest 17.6 7.9 10.5 -3.6
Deciduous broadleaf forest 3.2 12.7 0.3 0.7
Mixed forest 1.8 8.0 -0.7 -0.9
Woodlands 16.4 18.9 11.5 -5.9
Wooded grassland 8.8 -0.5 9.6 0.3
Closed shrubland 7.3 3.4 2.5 8.9
Open shrubland 9.0 7.4 3.6 12.1
Grassland 23.9 7.1 32.9 46.4
Cropland 12.3 34.7 30.9 40.8
Bare soil -0.1 0.6 -0.8 1.4
Urban -0.1 -0.1 -0.2 -0.3
Soil Moisture Assimilation -> Streamflow Evaluation vs. USGS gauges – by major basins
49
Soil Moisture Assimilation -> Streamflow (average seasonal cycles of RMSE– using Xia et al. (2011) stations)
0
10
20
30
40
50
60
70
80
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
RMSE
(m3/
s)
NEW ENGLAND
OLLPRM-DA
10
15
20
25
30
35
40
45
50
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
RMSE
(m3/
s)
MID ATLANTIC
OLLPRM-DA
12 14 16 18 20 22 24 26 28 30 32
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
RMSE
(m3/
s)
SOUTH ATLANTIC
OLLPRM-DA
6 8
10 12 14 16 18 20 22 24 26
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
RMSE
(m3/
s)
GREAT LAKES
OLLPRM-DA
40
60
80
100
120
140
160
180
200
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
RMSE
(m3/
s)
OHIO
OLLPRM-DA
6
8
10
12
14
16
18
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
RMSE
(m3/
s)
TENNESSE
OLLPRM-DA
30 40 50 60 70 80 90
100 110 120
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
RMSE
(m3/
s)
UPPER MISSISSIPPI
OLLPRM-DA
5
10
15
20
25
30
35
40
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
RMSE
(m3/
s)
LOWER MISSISSIPPI
OLLPRM-DA
0
5
10
15
20
25
30
35
40
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
RMSE
(m3/
s)
SOURIS RED RAINY
OLLPRM-DA
50
0 10 20 30 40 50 60 70 80 90
100 110
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
RMSE
(m3/
s)
MISSOURI
OLLPRM-DA
8 10 12 14 16 18 20 22 24 26
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
RMSE
(m3/
s)
ARKANSAS-WHITE-RED
OLLPRM-DA
1 2 3 4 5 6 7 8 9
10
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
RMSE
(m3/
s)
UPPER COLORADO
OLLPRM-DA
0 1 2 3 4 5 6 7 8 9
10
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
RMSE
(m3/
s)
LOWER COLORADO
OLLPRM-DA
0 1 2 3 4 5 6 7 8 9
10
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
RMSE
(m3/
s)
GREAT BASIN
OLLPRM-DA
10 20 30 40 50 60 70 80 90
100
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
RMSE
(m3/
s)
PACIFIC NORTHWEST
OLLPRM-DA
0
10
20
30
40
50
60
70
80
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
RMSE
(m3/
s)
CALIFORNIA
OLLPRM-DA
Soil Moisture Assimilation -> Streamflow (average seasonal cycles of RMSE– using Xia et al. (2011) stations)
51
SMOS (ESA satellite) soil moisture assimilation tests in the NCEP Global
Forecast Model
52
" The simplified ensemble Kalman Filter (EnKF) was embedded in the GFS latest version to assimilate soil moisture observation
" Case: 00Z July 6, 2011. (GFS free forecast)
" Experiments:
CTL: Control run EnKF: Sensitivity run
Weizhong Zheng and Xiwu Zhan (NESDIS/STAR)
SMOS CTL
EnKF
53
Comparison of Soil Moisture between GFS & SMOS
Comparison of Soil Moisture between GFS & SMOS SMOS CTL
EnKF
54
CTL
OBS EnKF-CTL
EnKF
Improved Precipitation in EnKF vs CTL
55
Snow Data Assimilation
Comparison of snow cover fraction between the MODIS (blue circles), the open loop simulation (black line) and the assimilation simulation (green line).
Comparison of snow water equivalent between the open loop simulation (green), the assimilation simulation (red) and the in-situ measurement (black) averaged over all SNOTEL sites in the study region.
Snow Cover Fraction
Snow Water Equivalent
56
57
• Role of Land Surface Models (LSMs) • Requirements: - atmospheric forcing, valid physics, land data sets/
parameters, initial land states, cycled land states • Applications for Weather & Climate • Validation • Improving LSMs • Expanding role of Land Modeling as part of an integrated Earth System
• Summary
Outline
APPLICATIONS: e.g. Noah land model in NCEP Weather and Climate models
58 NLDAS
(drought)
Air Quality
WRF NMM/ARW Workstation WRF
WRF: ARW, NMM ETA, RSM
Satellites 99.9%
Regional NAM WRF NMM
(including NARR)
Hurricane GFDL HWRF
Global Forecast System
Dispersion ARL/HYSPLIT
Forecast
Severe Weather
Rapid Update for Aviation (ARW-based)
Climate CFS
1.7B Obs/Day
Short-Range Ensemble Forecast
MOM 2-Way Coupled Oceans
HYCOM
WaveWatch III
NAM/CMAQ
Regional Data Assimilation
Global Data Assimilation
North American Ensemble Forecast System
GFS, Canadian Global Model
NOAH Land Surface Model
59
• Surface energy (linearized) & water budgets; 4 soil layers.
• Forcing: downward radiation, precip., temp., humidity, pressure, wind.
• Land states: Tsfc, Tsoil*, soil water* and soil ice, canopy water*, snow depth and snow density. *prognostic
• Land data sets: veg. type, green vegetation fraction, soil type, snow-free albedo & maximum snow albedo.
NCEP-NCAR unified Noah land model
• Noah coupled with NCEP models: North American Mesoscale model (NAM; short-range), Global Forecast System (GFS; medium-range), Climate Forecast System (CFS; seasonal), and uncoupled NLDAS (drought) & GLDAS (climate & drought).
60
Global Land Data Assimilation System (GLDAS)
• GLDAS (run Noah LSM under NASA/Land Information System) forced with CFSv2/GDAS atmos. data assimil. output & blended precip. • “Blended precipitation”: satellite (CMAP; heaviest weight in tropics where gauges sparse), surface gauge (heaviest in middle latitudes) and GDAS (modeled; high latitudes). • CFSv2/GLDAS ingested into CFSv2/GDAS every 24-hours (semi –coupled) with adjustment to soil states (e.g. soil moisture). • Snow cycled in CFSv2/GLDAS if model within 0.5x to 2.0x observed value (IMS snow cover, and AFWA snow depth products), else adjusted to 0.5 or 2.0 of observed value.
IMS snow cover AFWA snow depth GDAS-CMAP precip Gauge locations
Climate Forecast Syst. & 30-yr Reanalysis
61
• Global Land Data Assimilation System (GLDAS): Daily update of land states via semi-coupled Noah model run in NASA Land Information System.
• Driven by CFS assimilation cycle atmos. forcing & “blended” precipitation obs (satellite, gauge, model).
• Reasonable land climatology: energy/water budgets.
January (top), July (bottom) Climatology from 30-year NCEP CFS Reanalysis (Precip, Evap, Runoff [mm/day]; Soil Moisture, Snow [mm])
Runoff Precipitation Evaporation Soil Moisture Snow
62 62
Land Modeling and Drought (NLDAS) • North Amererican Land Data Assimilation System • Noah (& 3 other land models) run in an uncoupled
mode driven by obs. atmos. forcing to generate surface fluxes, land/soil states, runoff & streamflow.
• Land model runs provide 30-year climatology. • Anomalies used for drought monitoring, and
seasonal hydrological prediction (using climate model downscaled forcing); supports the National Integrated Drought Information System (NIDIS).
NLDAS four-model ensemble soil moisture monthly anomaly
July 30-year climatology July 1988 (drought year) July 1993 (flood year)
• NLDAS is a multi-model land modeling & data assimil. system… • …run in uncoupled mode driven by atmospheric forcing
(using surface meteorology data sets)… • …with “long-term” retrospective and near real-time output of
land-surface water and energy budgets.
NLDAS Configuration: Land models
NLDAS Data Sets and Setup
www.emc.ncep.noaa.gov/mmb/nldas ldas.gsfc.nasa.gov/nldas
NLDAS Drought Monitor
NLDAS Drought Prediction
Anomaly and percentile for six variables and three time scales: • Soil moisture, snow water, runoff, streamflow, evaporation, precipitation • Current, Weekly, Monthly
NLDAS website
June 1988: Drought Year Monthly total soil moisture anonamlies for NLDAS land models and ensemble mean
June 1993: Flood Year Monthly total soil moisture anonamlies for NLDAS land models and ensemble mean
• System uses VIC land model and includes three forecasting approaches for necessary downscaled/ensemble temperature and precipitation forcing data sets: (1) NCEP Climate Forecast System (CFS), (2) traditional ESP, and (3) CPC. • System jointly developed by Princeton University and University of Washington. • Transitioned to NCEP/EMC local system in November 2009 as an experimental seasonal forecast system. • Run at the beginning of each month with forecast products staged on NLDAS website by mid-month. • For CFS forcing, transitioning from use of CFSv1 to now-operational CFSv2.
NLDAS Seasonal Hydrological Forecast System
Weekly Drought Forecast System Using CFS forcing
Drought Monitoring (top) & Corresponding Monthly Fcst (bottom)
Courtesy Dr. L. Luo at Michigan State U.
03 May 2012 28 June 2012 02 August 2012
Global Land Data Assimilation System (GLDAS) for drought
� GLDAS running Noah LSM part of NCEP Climate Forecast System (CFSv2) Reanalysis (CFSR) 1979-2010; similar to NLDAS. � GLDAS/Noah uses atmospheric forcing from Climate Data Assimilation System (CDAS) from CFSR and observed precipitation. � GLDAS part of the now operational CFSv2 (January 2011). � GLDAS being developed for Global Drought Information System (GDIS); connections with World Climate Research Programme.
71
CFSR/GLDAS Soil Moisture Climatology
71
72 72
GLDAS CFSR Soil Moisture Anomaly
May 1988 (US drought)
May 1993 (US floods)
July 2012 (US drought)
GLDAS CFSR Soil Moisture Anomaly
74
• Role of Land Surface Models (LSMs) • Requirements: - atmospheric forcing, valid physics, land data sets/
parameters, initial land states, cycled land states • Applications for Weather & Climate • Validation • Improving LSMs • Expanding role of Land Modeling as part of an integrated Earth System
• Summary
Outline
75
Land Model Validation • Land model validation uses (near-) surface
observations, e.g. routine weather observations of air temperature, dew point and relative humidity, 10-meter wind, along with upper-air validation, precipitation scores, etc.
• To more fully validate land models, surface fluxes and soil states (soil moisture, etc) are also used.
• Comparing monthly diurnal composites useful to assess systematic model biases (averaging out transient atmospheric conditions).
• Such validations suggest land model physics upgrades.
Near-surface verification regions
Western US! Eastern US!
76
Western US Eastern US
tem
p
RH
1C daytime warm bias slight nighttime warm bias!1C daytime warm bias!
slight daytime dry bias nighttime dry bias
daytime dry bias (increasing dry trend)
slight nighttime dry bias
77
July 2008 diurnal monthly mean NAM (dashed line) vs obs (solid)
00-84hr, 00z cycle, 2-m temp & RH
78
Validation with flux site data (e.g. via “Fluxnet”, “GEWEX/
CEOP reference sites).
SURFX (Surface Flux) Network!NOAA/ATDD, Tilden Meyers et al!
Compare monthly diurnal composites of model!output versus observations from flux sites to!
assess systematic model biases. !
79
January!
Ft. Peck,!Montana!
(grassland)!
80
Ft. Peck,!Montana!
(grassland)! July!
81
January!
NOAA/ATDD Surface Flux Network!
Audubon!Grassland,!
Arizona!
82
July !
NOAA/ATDD Surface Flux Network!
Audubon!Grassland,!
Arizona!
83
January!
NOAA/ATDD Surface Flux Network!
Ozarks,!Missouri!
(mixed forest)!
84
July !
NOAA/ATDD Surface Flux Network!
Ozarks,!Missouri!
(mixed forest)!
85
Walker Branch,!Tennessee!
(deciduous forest)!
NOAA/ATDD Surface Flux Network!
January!
86
July !
NOAA/ATDD Surface Flux Network!
Walker Branch,!Tennessee!
(deciduous forest)!
87
SURFACE ENERGY BUDGET TERMS
January 2008 monthly averages:
model vs obs
Ft. Peck, Montana (grassland)
Downward Solar
Downward Longwave Reflected Solar
Upward Longwave
Ground Heat Flux Latent Heat Flux Sensible Heat Flux 88
Ft. Peck, Montana (grassland)
Sensible heat flux monthly averages:
model vs obs January 2008
April 2008 July 2008
October 2008
July 2007
better surface-layer physics in
NAM vs GFS
89
July 2005, Sensible heat flux monthly averages: NAM
(North American Mesoscale) model vs obs
Ozarks, MO (deciduous)
Brookings, SD (grassland)
Black Hills, SD (conifer)
Walker Branch, TN (deciduous)
Bondville, IL (cropland)
Ft. Peck, MT (grassland)
obs
model
low bias
high bias
90
91
Land Model Benchmarking • Comprehensive evaluation of land models that goes beyond traditional validation, with an established level of a priori performance for a number of metrics, e.g. modeled surface fluxes within ±ε [W/m2] of the observations, PDFs of observed & modeled surface fluxes overlap by at least X%, an annual carbon budget within Y%, etc. • Metrics depend on community, e.g. climate, weather, carbon, hydrology. • Empirical approaches: e.g. statistical, neural network, other machine learning; relationship between training and testing data sets manipulated to see how well a model utilizes available information. • Land model (physics) should be able to e.g. out-perform simple empirical models, persistence (for NWP models), climatology (for climate models).
© Crown copyright Met Office
Benchmarking vs Evaluation (Martin Best et al, GEWEX/GLASS)
• Evaluation
Ø Run model
Ø Compare output with observations and ask:
q How good is the model?
Model A Model B
Site / Variable E
rror
Benchmark
• Benchmarking Ø Decide how good model needs to be
Ø Run model and ask:
ü Does model reach the level required?
PALS Land sUrface Model Benchmarking Evaluation pRoject (PLUMBER)
93 93
• GLASS tools, i.e. Protocol for the Analysis of Land Surface models (PALS): www.pals.unsw.edu.au.
• GEWEX Hydroclimatology Panel (GHP) provides reference site/model output data sets for different regions, seasons, &variables: evaluate energy, water & carbon budgets; coupled, uncoupled model output.
Joint GEWEX/GLASS-GHP project: Land Model Benchmarking
GHP “CEOP” reference sites Fluxnet
94
050
150
250
DJF hour of day
Late
nt h
eat f
lux
W/ m
2
0 6 12 18 23
Obs: BondvFluxnet.1.2 Model: Cab1.4bBondv
Observed
Modelled
Total score: 0.46
(NME)
Score: 1.7
(NME)
050
150
250
MAM hour of day
Late
nt h
eat f
lux
W/ m
2
0 6 12 18 23
Obs: BondvFluxnet.1.2 Model: Cab1.4bBondv
Score: 0.57
(NME)
050
150
250
JJA hour of day
Late
nt h
eat f
lux
W/ m
2
0 6 12 18 23
Obs: BondvFluxnet.1.2 Model: Cab1.4bBondv
Score: 0.37
(NME)
050
150
250
SON hour of day
Late
nt h
eat f
lux
W/ m
2
0 6 12 18 23
Obs: BondvFluxnet.1.2 Model: Cab1.4bBondv
Score: 0.28
(NME)
LE
Rn
020
040
060
0
DJF hour of day
Net
radi
atio
n W
/ m2
0 6 12 18 23
Obs: BondvFluxnet.1.2 Model: Cab1.4bBondv
Observed
Modelled
Total score: 0.3
(NME)
Score: 0.59
(NME)
020
040
060
0
MAM hour of day
Net
radi
atio
n W
/ m2
0 6 12 18 23
Obs: BondvFluxnet.1.2 Model: Cab1.4bBondv
Score: 0.19
(NME)
020
040
060
0
JJA hour of day
Net
radi
atio
n W
/ m2
0 6 12 18 23
Obs: BondvFluxnet.1.2 Model: Cab1.4bBondv
Score: 0.37
(NME)
020
040
060
0
SON hour of day
Net
radi
atio
n W
/ m2
0 6 12 18 23
Obs: BondvFluxnet.1.2 Model: Cab1.4bBondv
Score: 0.22
(NME)
G
−50
050
100
150
DJF hour of day
Gro
und
heat
flux
W/ m
2
0 6 12 18 23
Obs: BondvFluxnet.1.2 Model: Cab1.4bBondv
Observed
Modelled
Total score: 1.6
(NME)
Score: 3.6
(NME)
−50
050
100
150
MAM hour of day
Gro
und
heat
flux
W/ m
2
0 6 12 18 23
Obs: BondvFluxnet.1.2 Model: Cab1.4bBondv
Score: 1.2
(NME)
−50
050
100
150
JJA hour of day
Gro
und
heat
flux
W/ m
2
0 6 12 18 23
Obs: BondvFluxnet.1.2 Model: Cab1.4bBondv
Score: 1.5
(NME)
−50
050
100
150
SON hour of day
Gro
und
heat
flux
W/ m
2
0 6 12 18 23
Obs: BondvFluxnet.1.2 Model: Cab1.4bBondv
Score: 1.8
(NME)
010
020
030
040
0
DJF hour of day
Sen
sibl
e he
at fl
ux W
/ m2
0 6 12 18 23
Obs: BondvFluxnet.1.2 Model: Cab1.4bBondv
Observed
Modelled
Total score: 1.1
(NME)
Score: 0.91
(NME)
010
020
030
040
0
MAM hour of day
Sen
sibl
e he
at fl
ux W
/ m2
0 6 12 18 23
Obs: BondvFluxnet.1.2 Model: Cab1.4bBondv
Score: 0.51
(NME)
010
020
030
040
0
JJA hour of day
Sen
sibl
e he
at fl
ux W
/ m2
0 6 12 18 23
Obs: BondvFluxnet.1.2 Model: Cab1.4bBondv
Score: 2.9
(NME)
010
020
030
040
0
SON hour of day
Sen
sibl
e he
at fl
ux W
/ m2
0 6 12 18 23
Obs: BondvFluxnet.1.2 Model: Cab1.4bBondv
Score: 0.55
(NME)
H spring
summer
autumn
winter
model obs
• PALS example: CABLE (BOM/Australia) land model, Bondville, IL, USA (crolandp), 1997-2006, average diurnal cycles
.
Joint GEWEX/GLASS-GHP project: Land Model Benchmarking
95
• Role of Land Surface Models (LSMs) • Requirements: - atmospheric forcing, valid physics, land data sets/
parameters, initial land states, cycled land states • Applications for Weather & Climate • Validation • Improving LSMs • Expanding role of Land Modeling as part of an integrated Earth System
• Summary
Outline
Land Model Improvement
• Urban-canopy & land-use/land cover changes • Refined surface layer turbulence (esp. stable cond.) • Refined evapotranspiration (empirical Jarvis-Stewart) • Surface/subsurface lateral flow (hydrology) • Dynamic (growing) veg. with multi-layer canopy* • CO2-based photosynthesis* • Groundwater/water table* • Multi-layer snowpack*, refined frozen soil processes
Surface flow
Saturated subsurface flow
Snowpack & frozen soil
Urban-canopy model
*Noah-MP upgrades
96
� Use surface fluxes (e.g. latent and sensible) to evaluate land-surface physics formulations and parameters, e.g. invert transpiration formulation to infer canopy conductance. (Similarly for aero. cond.)
97
Land Model Improvement: Transpiration
98
Noah-MP: a land component in Mesoscale Weather Research &
Forecasting (WRF) model • Noah-MP developed by Niu et al (2011, JGR), Univ.
Texas-Austin, NCAR, NCEP, Univ. Arizona • Explicit canopy/subcanopy layers, dynamic (growing)
vegetation, and canopy conductance based on CO2-photosynthesis
• Ground water (connection with hydrology models) • Multi-layer snowpack • Implicit solution of surface energy budget (SEB, for
better SEB closure) • Recent NCAR Improvements to Noah-MP version 3.5
99
SWdn
shaded fraction
What are the units of solar absorbed radiation in the canopy? - W/m2
grid : original two stream makes this assumption - W/m2
veg : Noah-MP calculates absorption for the grid but then applies it to the vegetated fraction
Original Noah-MP uses both fractional vegetation cover (turbulent fluxes) and vegetation stem density (radiation fluxes). In the original code, these were inconsistent. For Evergreen Needleleaf Forest, the prescribed stem density gives a constant horizontal coverage of 72% while turbulent flux fractional vegetation can be any value from 0% to 100% These inconsistencies were fixed in v3.4.1 by making the stem density a function of fveg.
Canopy Absorbed Solar Radiation in Noah-MP
100
Before (V3.4.1) After (Pre V3.5) • Reduced low-level (2-meter) air temperature bias
Noah-MP vs. Noah: Improved canopy radiation effect and surface-layer changes
in coupled mesoscale WRF v3.5
Tiled or Separate Land Model Grid � A land model grid may comprise sub-atmospheric-
grid-scale “tiles” of e.g. grass/crop/shrubland, forest, water, etc, of O(1-4km, or even less), with coarser-resolution atmospheric forcing, and an aggregate flux fed to the “parent” atmospheric.
5-50km
ABL “blending” height
surface tiles with different fluxes
• Boundary-layer issues, e.g. when blending height > BL depth?
• Land-surface tiling run under NASA Land Information System (LIS); other LIS tools.
• NASA/LIS to be part of NOAA Environmental Modeling System (NEMS).
Aggregated surface flux
101
102
• Change to surface-layer turbulence physics (thermal roughness change) and surface microwave emissivity MW model… •…yields a reduction of large bias in calculated brightness temperatures was found for infrared and microwave satellite sensors for surface channels… •…so many more satellite measurements can be utilized in GSI data assimilation system.
Upgraded land/surface-layer physics: Improvement of Satellite Data Utilization over
Desert & Arid Regions in NCEP Operational NWP Modeling and Data Assimilation Systems
103
LST [K] Verification with GOES and SURFRAD
3-Day Mean: July 1-3, 2007
GFS-GOES: CTR
GFS-GOES: New Zot
Large cold bias
GFS-GOES: New Zot
Improved significantly during daytime!
Ch LST
(a) (b)
(d)
(c)
Aerodynamic conductance: CTR vs Zot
103
104
• Role of Land Surface Models (LSMs) • Requirements: - valid physics, land data sets/parameters, initial
land states, atmospheric forcing, cycled land states • Applications for Weather & Climate • Validation • Improving LSMs • Expanding role of Land Modeling as part of an integrated Earth System
• Summary
Outline
105
Expanding Role of Land Models � In more fully-coupled Earth System models, role is
changing, with Weather & Climate connections to: Hydrology (soil moisture & ground water/water
tables, irrigation and groundwater extraction, water quality, streamflow and river discharge to oceans, drought/flood, lakes, and reservoirs with management, etc),
Biogeochemical cycles (application to ecosystems --terrestrial & marine, dynamic vegetation and biomass, carbon budgets, etc.)
Air Quality (interaction with boundary-layer, biogenic emissions, VOC, dust/aerosols, etc.)
� More constraints, i.e. we must close the energy and water budgets, and those related to air quality and biogeochemical cycles.
106
NCAR Common Land Model (CLM)
107 107
Hurricanes and Inland Flooding • Physical-based Noah model included in (mesoscale)
Hurricane Weather Research & Forecasting model, with little degradation in track & intensity & precip.
• Inland flood forecasting (right) using Noah runoff & streamflow model.
• Extend to global/climate models: river discharge to oceans.
USGS gauge station
observations
GFS Soil Moisture
Initial Condictions (IC; default)
NLDAS IC
NAM IC
ops NAM
Hurricane Katrina
108
Freshwater nearshore
Hydro-Coastal Linkage � Impact of adding freshwater input to Gulf of Mexico (implemented RTOFS June 2007).
Sea Nettle Forecasting
Salinity
SST Likelihood of Chrysaora
Habitat Model
NCEP, with NESDIS, NOS
109
110
• Thousands of lakes in N. America on the scale of 4km (NAM target), not resolved by SST analysis.
• Influence of previously unresolved lakes may be felt on this scale and can no longer just be “filled in”.
Lake modeling
111 111
• Freshwater lake “FLake” model (Dmitrii Mironov, DWD). • In use in regional COSMO, HIRLAM (European
models), UKMO, ECMWF (global).
- two-layer - temperature &
energy budget - mixed-layer &
thermocline - snow-ice module - atmospheric
forcing inputs - specified depth
& turbidity
Lake modeling
112
• Role of Land Surface Models (LSMs) • Requirements: - atmospheric forcing, valid physics, land data sets/
parameters, initial land states, cycled land states • Applications for Weather & Climate • Validation • Improving LSMs • Expanding role of Land Modeling as part of an integrated Earth System
• Summary
Outline
113 113
Land Surface: Summary • Provide surface boundary conditions for weather &
climate models (e.g. NCEP models), proper representation of interactions with atmosphere.
• Valid physics, land data sets & parameters, initial land states, atmospheric forcing, cycled land states.
• Land model validation using (near-) surface observations, e.g. air temperature, relative humidity, wind, soil moisture, surface fluxes, etc …suggests model physics improvements.
• Expanding role of land models for weather & climate in more fully-coupled Earth System models with connections between Weather & Climate and Hydrology, Biogeochemical cycles (e.g. carbon), and Air Quality communities and models.
114 114
Land-Hydrology Partners/Projects • Noah land model development: NCAR, U.
Ariz., U. Texas/Austin, U. Wash., WRF land & PBL working groups, other national/internat’l partners; freshwater “FLake” community),
• NLDAS/GLDAS drought: Princeton, Univ. Wash., NASA, NWS Hydro., Climate Prediction Center),
• Remote sensing/land data sets/data assimil. of soil moisture, vegetation, surface temp, snow: JCSDA, NESDIS, NIC, NASA, AFWA,
• NOAA research: NLDAS, GLDAS support,
• WCRP/GEWEX: GLASS/land-hydrology, GABLS/boundary-layer, GHP/hydroclimate projects, UKMO, ECMWF, Meteo-France, etc.
115
Thank you! Obrigado! ¡Gracias!
116
Michael Ek ([email protected])
Environmental Modeling Center (EMC) National Centers for Environmental Prediction (NCEP)
NOAA Center for Weather and Climate Prediction (NCWCP) 5830 University Research Court
College Park, MD 20740
National Weather Service (NWS) National Oceanic and Atmospheric Administration (NOAA)
relative humidity temperature
sensible heat flux
moisture flux
soil heat flux soil temperature soil moisture
1
2
3
1
2 3
4
4 5
6 6
7
7
8
8 downward longwave
cloud cover
5
boundary-layer growth
canopy conductance
Local Land-Atmosphere Interactions
+positive feedback for C3 & C4 plants, negative feedback for CAM plants *negative feedback above optimal temperature
*
incoming solar
surface layer & ABL land-surface processes radiation
positive feedback negative feedback
+
entrainment
above-ABL dryness
above-ABL stability
precipitation
reflected solar albedo
emitted longwave
surface temperature
wind
turbulence
117