1. Upper tropospheric moisture assimilation using GOES observations
By William H. Raymond (CIMSS/UW-Madison) and
Gary S. Wade (NOAA/NESDIS/ORA ASPT)
The demonstrated assimilation procedure modifies only the moisture field in physical space (in contrast to radiance space). The numerical optimization of the differences for channel 3 brightness temperatures (water vapor at 6.7 um), between the forecast initial conditions (by a forward radiative model) and the observed GOES values, is incorporated directly into the CIMSS Regional Assimilation System (CRAS) forecast model. The partitioning of changes made to the model moisture field, in physical space, is directly proportional to the observational weighting functions.
1. These materials and information were actually presented by W. Raymond at a poster session (P3.61) at the conference on Tuesday afternoon 16 Oct 2001.
2. This PowerPoint file is named: gutmassim.ppt. Version is as of 16 Nov 2001.
3. Version of 26 Nov 2001 now includes notes for slide #2.
4. Version of 29 Nov 2001 now includes slides #4-5 showing image correlation improvement bar graphs.
As presented at the AMS 11th Conference on Satellite Meteorology and Oceanography (15-18 Oct 2001, Madison, WI)
UW-Madison

2. For modification of the model initial moisture field by GOES channel 3 (6.7 um)
Model initial field
Mixing ratio (q)
Relative humidity (RH)
Goal -
minimize difference
Forward radiative
transfer model
Model equivalent
brightness temperature
Satellite observed
brightness temperature
From the initial model fields (of temperature and moisture), a forward radiative transfer model is used to calculate the model equivalent brightness temperatures (specifically for GOES channel 3). These BTs are compared to those observed by the satellite. The resulting difference field for channel 3 is used to modify the weighted sum for the moisture field (in physical space, for RH or q). The partitioning of this modification made to the model moisture field is directly proportional to the observed (vertical) weighting functions. (No change is made to the temperature field.) When the channel 3 difference field has been made sufficiently small, the solution for the (final) model moisture field has been reached.
Difference field for
channel 3 brightness temperatures
Form weighted sum
for q or RH using
satellite weighting function
Modify sum using
numerical techniques
Redistribute perturbation
using weighting functions

3. 1. 2. Adjustment of the 6.7 um water vapor field in the model
(0000 UTC 15 Sep 2000)
Initial eta field in CRAS model
GOES observations (Ch 3) using recursive filter 1.
1. Note the following features which appear in the 6.7 um representation of the final model solution (lower right panel), which included assimilation of the GOES channel 3 brightness temperatures. Compare these features with what is evident in the “observed” GOES 6.7 um image (middle left panel).
- the very narrow moist axis that extends from E Pacific Ocean (20N 120W) to S ID.
- the comma shaped moist plume W and SW of the CA coast (27N 134W to 35N 127W) which extends even further N to just off the CA/OR border.
- the wave-like indentations along 130W on the east edge of the cloud mass (from 35 to 50N).
- two separate dry pockets (cen and W NC/SC border and Delmarva area) for the breaks in the cloud cover along the US East coast.
- expansion of cloud/moisture to W coast of FL and into S Bahamas.
- added structure to dry slot curving through W GlfMex.
- tendency for some dry extension into E GrtLks.
- more moisture/cloud over N cen US Plns.
2. Note that, where (thick) clouds persist (as from NE Mex to along the US GlfCst), the cloud structure remains the same for the final model solution (lower right panel) as in the initial field (upper right panel). 2.
Eta model field modified by assimilation of GOES observations

4. Inclusion of GOES channel 3 brightness temperature in the model assimilation improves correlation of model synthetic images compared to observed images
1. These correlations were done between:
initial eta field and - GOES observation field
as well as
the channel 3 brightness temperature
modified eta field - and - GOES observation field.
Referring back to slide 3, one can also see that these correlation pairs match accordingly with the images in this order:
upper right - vs - middle left
and
lower right - vs - middle left.
For these 00 UTC runs during the 10 day period, the average improvement in correlation was 12%.

5. Inclusion of GOES channel 3 brightness temperature in the model assimilation improves correlation of model synthetic images compared to observed images
1. Note that, although the average improvement was only 2.3%, this typical positive improvement did still exist AFTER 48 HOURS in the model.
2. The last case (21 Sep 2000) became unstable, and hence, the forecasts were poor. This case is still being re-evaluated.
Even through 48 hour forecasts, the correlation was generally improved (2.3%).