Satellite Data Assimilation

1
Satellite Data Assimilation Activities at CIMSS
for FY2003 Robert M. Aune Advanced Satellite
Products Team NOAA/NESDIS/ORA/ARAD Cooperative
Institute for Meteorological Satellite
Studies Madison, Wisconsin Research /
Applications Collaborations. Concerns /
Shortfalls.
2
Research and Applications.. Tools CIMSS Regional
Assimilation System (CRAS) First real time
prediction system to use cloud and moisture
products from the GOES sounders (1996) New for
FY2003 Real time CRAS now at 61 km resolution
with 20 km nest using single field-of-view
(10km) GOES retrievals to initialize moisture and
clouds (Raymond) Polar stereographic version of
CRAS running over Antarctica Multi-layer cloud
initialization using GOES/MODIS Pacific CRAS for
THORPEX
3

• Retrieved Cloud Top Pressure/Effective Cloud
  Amount

600 hPa
300 hPa
This information can be used to initialize 3D
cloud fields in numerical prediction models
50
98
4
Cloud top pressure from GOES-8 Sounder
GOES-8 IR image
CRAS 15 hr forecast
CRAS is currently running in real time at 20 km
horizontal grid spacing to study how cloud
physics software responds to finer horizontal
resolution. Above (center) is a CRAS 15 hour
forecast IR image valid 15UTC 06Aug01. Also
shown are images of cloud top pressure (left) and
IR window from GOES at the validating time. CRAS
cloud forecasts are monitored for shape and
position.
5
Forecast Bust in Madison, WI On April 18,
2000 the official predicted high temperature for
Madison was 65 F. Persistent cloud cover kept
temperatures from rising. The CIMSS Regional
Assimilation System (CRAS) was the only model
that predicted overcast skies which resulted in a
cooler forecast. Figure 1 shows the forecast
surface temperature from the CRAS and from the
Aviation, Eta, and NGM models of the National
Center for Environmental Prediction (NCEP),
initialized at 12UTC, April 17. The observed
temperatures at Madison are plotted with a solid
line. Figure 2 shows the predicted cloud-top
pressure from CRAS and the observed cloud-top
pressure from GOES-8. The GOES-8 sounder detected
higher clouds moving in after 12UTC on the 18th
while the low clouds persisted The CRAS runs at
80km resolution and uses information from
the GOES sounders to initialize water vapor
and clouds. Daily CRAS forecasts can be viewed
at http//cimss.ssec.wisc.edu/model/daily.htm
l
Figure 1
Figure 2.
6
Using MODIS Moisture Products to Initialize
Forecasts for Antarctica
Cloud-top pressure and total
precipitable water from the MODerate resolution
Imaging Spectroradiometer (MODIS) are currently
being evaluated in the polar version of the CIMSS
Regional Assimilation (PCRAS). The data are
being used to adjust mixing ratio and cloud water
in the forecast model during a 24 hour forecast
initialization period. Approximately 800,000
observations at 5 km resolution were used from
each of the 14 satellite passes during the
period. Initial and boundary conditions are
provided by the NCEP Aviation model. The
figure at right shows the differences in the
total precipitable water (MODIS minus no MODIS)
at the end of the 24 hour initialization period.
For this case, valid Dec 7, 2000, the MODIS data
has reduced the amount of water vapor in the
coastal regions (blue areas).
7
24 hour CRAS spinnup forecast with mixing ratio
and cloud adjustments using cloud-top pressure
and total precipitable water retrieved from MODIS
valid 07 Dec 2000.
Composite IR window channel image from AVHRR
valid 07 Dec 2000.
8
3h 20km RUC cloud-top fcst w/ GOES cloud
assimilation
Verification Cloud-top pressure based on NESDIS
product
3h 40km RUC cloud-top fcst No GOES cloud
assimilation
Effect of GOES data on 3-h RUC cloud-top
fcsts Valid 1200 UTC 9 Dec 2001
9
40km RUC
20km RUC w/ cloud analysis
3h 20km RUC cloud-top fcst w/ GOES cloud
assimilation
Verification Cloud-top pressure based on NESDIS
product
Cloud-top forecast verification - correlation
coefficient between forecast and
NESDIS cloud-top product - much improved cloud
forecasts even at 12h 28 Sept 2 Oct 2001
10
Nowcasting Analysis System Bob Aune, NESDIS/ORA
and Ralph Petersen, NWS/EMC HYPOTHESIS Can we
design a fast objective analysis system that
places emphasis on observations, not forecast
model requirements? The influence of
observations must be optimized in space and time.
Previous observations are projected forward in
time to augment spatial and temporal coverage
(GOES). Such an analysis system would not be
influenced by model initialization constraints or
model forecast errors A real time implementation
would provide forecasters with fast 3D snapshots
of the atmosphere.
11
Typical GOES-8 sounder retrieval coverage at 12
UTC.
Coverage after forward trajectories are computed
for three levels (300, 500, and 700 hPa).
Information is spread in different directions due
to wind shear.
12
Statistics comparing trajectory observations to
actual equivalent potential temperature
(K) retrievals from GOES-8. Observation
count for GOES-8 drops due to increasing
clouds. Statistics indicate trajectory
observations retain useful information for four
to five hours. Low-level trajectory
observations degrade faster.
13
700 hPa mixing ratios from GOES-8 valid 12 UTC
projected forward in time and compared with
corresponding GOES retrievals.
initial
3 hours
Fits remain good through 3 hours. Biases
increase after that, especially at low levels
where diurnal effects are large.
6 hours
14
The CIMSS Regional Assimilation System (CRAS) is
being used to study the predictability of fog in
Wisconsin. Time series plots from the 20 km
CRAS are being produced for instrumented sites
(blue squares) from the Wisconsin Department of
Transportation surface mesonet. Plotted
parameters are Temperature (Fahrenheit) Dew
point temperature (Fahrenheit) Mean sea-level
pressure (millibars) Total cloud water
(millimeters 1000) Wind barbs (speed (knots)
and direction) Accumulated precipitation
(inches100)
FOG
GOES-12 Visible
15
GOES Sounder Radiances in the Eta?
Total precipitable water differences, GOES 3x3
(CIMSS) minus EDAS analysis for 12UTC 14Mar02
At present, the Eta Data Assimilation System
(EDAS) assimilates GOES sounder radiances over
water, and 3-layer precipitable water (PW) over
land. The difference between GOES 3x3 total PW
from CIMSS and the total PW from the Eta analysis
is plotted above. Red indicates Eta is too dry.
16
Validating EDAS Radiance Assimilation against
GOES Total Precipitable Water Retrievals
PW differences (mm) at the end of
the assimilation cycle (tm00) using
two iterations in the RTE calculation for each
insert time.
PW differences (mm) at the end of
the assimilation cycle (tm00) using one iteration
in the RTE calculation for each insert time.
Iterations used to solve RTE EDAS 1 GDAS
2 NESDIS Retrieval 3