Towards Passive Microwave Radiance Assimilation of Clouds and Precipitation

Ralf Bennartz1, Tom Greenwald2, Andrew Heidinger3,

Chris O’Dell1, Martin Stengel1, Kenneth Campana4, Peter Bauer5

1: Atmos. & Oceanic Sci.,University of Wisconsin2: CIMSS,University of Wisconsin3: NOAA/NESDIS4: NOAA/NCEP5: ECMWF

Outline

SOI model and applications

Cloud/precipitation overlap

3-year accomplishments

Further plans

SOI Radiative Transfer Model

• Hybrid multi-stream solution method (doubling plus iteration)

(Heidinger et al. 2006; O’Dell et al. 2006)

• Implementation

- Less scattering: 2-stream solution

- More scattering: 4-stream solution

• Forward, tangent linear and adjoint models integrated into CRTM

Accuracy of Results (Eddington and SOI versus Monte-

Carlo model)

Forward and adjoint simulation example

AMSR-E ObservationsGFS Simulations

Infrared Applications: MSG SEVIRI

MSG SEVIRI Comparison Results

Channel RMSE [K] Bias [K]

6.2 m 1.93 0.22

7.3 m 1.91 -1.25

8.7 m 1.54 1.17

10.8 m 1.38 0.73

12.0 m 1.37 0.64

13.4 m 1.37 1.06

Different cloud/precipitation overlap models

• Conventional approach uses cloud cover to subdivide NWP pixel in cloudy/precipitation

• New approach derives two or three optimal columns based on subscale distribution of precipitation columns with similar optical properties

• Numerically efficient (2-3 radiative transfer calculations per NWP grid point)

• Highly accurate against independent column/Max-Random-overlap reference

• Optimal approach reduces errors due to cloud overlap from maximum values of 5-10 K to values < 1K

Different cloud/precipitation overlap models

O’Dell, Bauer, Bennartz, JAS, 2006, submitted

Currently operational at

ECMWF

1Column 2Columns 3columns 3 optimal

Different cloud/precipitation overlap models

Currently operational at

ECMWF

New scheme with much better error

characteristics

O’Dell, Bauer, Bennartz, JAS, 2006, submitted

1Column 2Columns 3columns 3 optimal

3-year Accomplishments

• Fast forward RT model (SOI) developed, tested and integrated in CRTM (3 publications)

• Tangent linear and adjoint model developed, tested, and integrated in CRTM

• Bias GFS/SOI statistics for passive microwave and infrared

• Fast and accurate new cloud overlap scheme for use in NWP radiance assimilation (forward/adjoint); manuscript submitted

Further plans

• Monitor bias statistics over longer time period, especially:

– Fully include scattering (need more complete GFS input data)

– Biases in IR including scattering • Precipitation assimilation:

– Include cloud diagnostics to generate precipitation rate

– Further test and integrate cloud overlap with NCEP/GFS.