Title: Assimilation of AMSR-E soil moisture into a coupled land surface-mesoscale model in the Land Information System using an ensemble Kalman filter
1Assimilation of AMSR-E soil moisture into a
coupled land surface-mesoscale model in the Land
Information System using an ensemble Kalman
filter Clay Blankenship and Bill
Crosson USRA
2Overview
- Goal Improve predictions in a coupled
(land/atmosphere) weather model by assimilating
observations of soil moisture into a land surface
model. - LIS (Land Information System)
- Coupled mode--WRF and SHEELS LSM
- AMSR-E soil moisture observations
- Data Assimilation by Ensemble Kalman Filter
-
- Methodology
- Add SHEELS as a new land surface model in LIS.
- Add coupled-run and AMSR-E data assimilation
capability to SHEELS in LIS. - Run data assimilation experiments.
3Land Information System (LIS)
A modeling and data assimilation system with the
capability to run several different LSMs. It is
very customizable with the ability to swap out
LSMs, forcing datasets, etc. LSMS VIC,
Noah, CLM, Catchment,SiB2, Hyssib
Base Forcings ECMWF, GDAS, NLDAS...
Supplemental Forcings TRMM 3B42, Agrrad,
Cmap, Cmorph, Stg4... Parameters
Landcover, soils, greenness, albedo, LAI,
topography, tbot Data Assimilation
algorithm, observation, perturbation method
4SHEELS
SHEELS (Simulator for Hydrology and Energy
Exchange at the Land Surface) is a
spatially-distributed (grid cell) surface
flux-hydrology model that can be run as a
stand-alone model with meteorological input, or
coupled with a meteorological model. Uses of the
model include Provide areal soil moisture and
surface energy flux estimates. Validate
remotely-sensed moisture and temperature
measurements where observations are sparse or
absent. Provide surface boundary conditions for
mesoscale weather models. Heritage Based on
Biosphere-Atmosphere Transfer Scheme of
Dickinson, 1986. Second generation Ex-BATS of
Smith et al., 1993. (Bill Crosson) Third
generation SHEELS, developed to include major
modifications to soil layer structure,
1994. Added Kalman filter-based soil moisture
assimilation scheme, 1998. Added full soil
temperature diffusion scheme, 1999. Introduced
overland flow and stream routing, 2001. Current
version described in Martinez et al., 2001
Crosson et al., 2002.
5SHEELS Simulator for Hydrology and
Energy Exchange at the Land Surface
6SHEELS Input
- Time-dependent input
- Wind speed
- Air temperature
- Relative humidity
- Rainfall
- Atmospheric pressure
- Downwelling solar radiation
- Downwelling longwave radiation
- Required static variables
- Soils Landcover
- Saturated hydraulic conductivity canopy height
- Saturated matric potential fractional vegetation
cover - Soil wilting point minimum stomatal resistance
- Rooting depth leaf area index
- Soil porosity reflectance properties
- Topography
- Surface elevation or slope
7SHEELS Output
SHEELS estimates many time-dependent variables at
each grid point based on spatially-distributed
meteorological input as well as soil, vegetation
and topographic properties Surface latent and
sensible heat fluxes, Ground heat flux including
soil, canopy contributions Net radiation Vapor
mass fluxes from soil, canopy Reflected solar
radiation Solar and longwave radiation
absorbed Drag coefficients by canopy and
ground Soil surface and canopy temperatures Surfac
e temperature Soil temperature for each
layer Infiltration Soil moisture for each layer
Runoff Depth of water on canopy (dew or
rain) Ponded water
8SHEELS Output Examples Time/Depth Soil Moisture
Soil moisture is estimated for each layer. In
this example there are 15 layers in the 2 m soil
column. Note - Surface drying - Penetration
of wetting fronts - Transition at bottom of the
root zone (1 m)
9Adding SHEELS to LIS
- Added SHEELS as an LSM in LIS (rewrite code to
fit into LIS structure) - Enables use of LIS capabilities.
- Run-time selection of base forcing, supplemental
forcing, static data - Coupled WRF runs
- EnKF data assimilation
- Easy intercomparison with other models
- MPI enabled
- Allows subgrid variability in vegetation type
- Makes SHEELS available to LIS users.
10LSM Input Data
11Layer 1 Total Water (LiquidIce), hourly
12Layer 1 Soil Ice (hourly)
13Snow cover (daily)
14Model ResultsNebraskaJAN-JUL 2003
Fractional soil moisture (waterice)
Soil Temperature
15Model ResultsN. TexasJAN-JUL 2003
Fractional soil moisture (waterice)
Soil Temperature
16AMSR-E Soil Moisture Data
- Advanced Microwave Scanning Radiometer-EOS
- On board NASAs EOS Aqua satellite
(sun-synchronous polar orbiter) - Twice daily coverage
- Measures microwave radiance at 6 frequencies with
10km sampling - Soil moisture measured by 10.65 GHz channel with
resolution of 51x30 km - Measures volumetric liquid water concentration in
top 1 cm - Dataset AMSR-E Level 3 Daily land product (25 km
gridded)
17Ensemble Kalman Filter
- Ensemble Kalman filter (EnKF)
- The EnKF is initialized by creating an ensemble
of N initial condition - fields around a mean model state at t0, with an
assumed covariance. - The spread of an ensemble of N model
trajectories is used to estimate the error
covariances. The full non-linear dynamic
equations are used to propagate each ensemble
member forward in time, thus determining the
trajectories. This is in contrast to the
traditional Kalman filter in which linearized
model dynamics are used to propagate error
covariances. - The mean or median of the N ensemble states is
used to define the state vector estimate. - When observations are available, each ensemble
member is updated based on the difference between
the observation and the model state, weighted by
the Kalman gain (as in the EKF). Random error is
added to the observation based on assumed noise
characteristics this ensures that the variance
of the updated ensemble matches the true
estimation error covariances (Burgers et al.,
1998, Mon. Wea. Rev.) - Propagation of error covariance matrix is more
stable than in the traditional Kalman filter,
especially if there are strong non-linearities in
the model. - Assumptions/issues in EnKF
- Gaussion error distributions are assumed this
may be violated in some applications. - A large ensemble may be needed for accurate
determination of the error covariances.
18Settings for Data Assimilation
In an EnKF, the relative strength of the
background and observations is controlled by the
observation error (specified) and the background
error (derived from the ensemble, which depends
on the state perturbations). Observation AMSR-E
Surface Soil Moisture (g/cm3) Range .01 to
.52 Error .05 State SHEELS 14-layer water
fraction (0 to 1) Perturbation type normal Stand
ard deviation .005 at surface, less
below Correlation (vertical) 50 at 2.3m
19DA ResultsSoil Moisture Layer 1 Hourly Change
09Z
08Z
20Assimilation Results
Soil Moisture Layer 1
21Assimilation ResultsSoil Moisture Layer 1 (Case
2)
22(No Transcript)