GOES12 Sounder SFOV PW vs' GPS PW

1
GOES-12 Sounder SFOV PW vs. GPS PW

• NOAA/FSL (Dan Birkenheuer. Seth Gutman) have
  been assessing the NESDIS GOES Sounder SFOV PW
  product (from ORA/OPDB) against collocated (20km,
  30mins) the GPS PW in real time since late March
  2005 (See http//gpsmet.noaa.gov/cgi-bin/sat/goe
  s.cgi )
• These comparisons have highlighted several
  problems with our GOES SFOV PW products
• Cloud mask
• We generated retrievals where it was clearly
  cloudy (low cloud in particular) during the day
  and night time hours (See Slide 2 for example)
• Moist Bias
• GOES-12 PW retrievals significantly more moist
  than collocated GPS PW observations (See Slide 3
  for example)
• Periodic in nature with minimums at/near 00Z and
  12Z and maximums at 06Z and 18Z

2
Blue dots indicate where a GOES-12/GPS
collocation exists Clearly should not have done
a retrieval over this area where significant
cloud exists. GOES GPS PW diffs in the 5-10mm
range!!
3
D. Birkenheuer/S. Gutman (NOAA/FSL)

• PW bias time series is periodic and very
  similar looking to FSLs findings during IHOP
  (Dan Birkenheuer)
• Bias ramps up significantly as you move from
  00Z and 12Z with a max at 18Z and another
  secondary max at 06Z.
• 00Z Mean Bias 0.40mm 12Z Mean Bias 0.93mm
  18Z Mean Bias 3.51mm
• Possible Causes
• - Model Guess preprocessing (We checked and
  ruled this out)
• - Radiance Bias Correction (Significant
  Contributor more later)
• - Cloud mask Make retrievals where it is
  really cloudy (usually low cloud)

4
Our Response

• Address Cloud Mask Problems (in cldclr)
• 1) More emphasis on SWIR-LWIR difference check to
  identify clouds at night (BT17-BT08 lt -2.0 vs
  -6.0)
• 2) More emphasis on using Visible data to better
  identify low clouds (Cu) (VIZCLD10 instead of
  VIZCLD25)
• 3) Increase window channel threshold over land
  (DT5.5 instead of 20.) to better identify low
  clouds at night
• Tighten Up Quality Control
• Constrain a bit more, how far retrieved PW can
  move from guess (GFS) PW
• Revisit Radiance Bias Correction
• Reinstitute hourly model-based (GFS) bias
  correction
• Setup GOES/GPS PW Validation System
• Automated, routine (hourly) access of GPS
  observations from FSL reformat into McIDAS MD
  files
• Collocate (50km 30mins) GOES PW with GPS PW
  Observations
• Generate validation statistics for every hour of
  the day in order to characterize GOES-GPS PW
  differences as a function of time

5
ResultsRadiance Bias CorrectionRaob-Based vs
Model-Based
6
Radiance Bias CorrectionRaob-Based vs Model-Based
GOES PW GPS PW (mm)
Time (Z)
Routine System Raob-Based radiance bias
correction used Parallel System Model-based
radiance bias correction used
7
ResultsChanges to Cloud Mask (cldlcr)
Radiance Bias Correction
8
Summary of Changes Made

• Cloud Masking (Cldclr)
• DT5.5 over land (was 20.)
• VIZCLD 10.0 (was 25.0)
• SPW (moisture attentuation estimate) limited to
  2.5K (unlimited before)
• Main Retrieval Program (Sounder1)
• Model-based radiance bias corrections computed
• (TPWret TPWguess ) difference threshold changed
  
• Cloud Height (CO2 Slicing) Routine (clouds4)
• Model-based radiance bias corrections computed

9
Routine System Parallel System
GOES PW GPS PW (mm)
Time (Z)
14Z June 10, 2005 12Z June 13, 2005

• Significant reduction in mean PW bias!!
• Moist bias still evident at night time (May
  warrant further tightening up of SWIR-LWIR
  threshold experimenting with)

10
13Z Hour prior to us switching to parallel system
and sending FSL PWs from this system Note all
the BLUE dots
11
14Z First hour after we switched to parallel
system and sending FSL PWs from this system Note
that most of the BLUE dots are gone!!
12
15Z Second hour after we switched to parallel
system and sending FSL PWs from this system Note
that most of the BLUE dots are gone!!
13
CLOSE-UP VIEW
13Z Hour prior to us switching to parallel system
and sending FSL PWs from this system Note the
moist (relative to GPS) GOES PW retrievals (GREEN
BLUE dots) over Texas where lots of low clouds
(Cu) are evident
14
CLOSE-UP VIEW
14Z First hour after we switched to parallel
system and sending FSL PWs from this system No
GOES PW retrievals over Texas now!!
15
15Z Second hour after we switched to parallel
system and sending FSL PWs from this
system Still no retrievals over Texas... A good
thing!!
16
13Z
1 hour prior to switch
14Z
First hour after switch
Second hour after switch
15Z
17
GOES-GPS
18
RUC 20-GPS
19
RUC 13-GPS
20
Now What?

• Can we improve upon cloud detection at night over
  land by further tightening up SWIR-LWIR
  threshold?
• IF (ABS(BT17-BT8) .GT. 1.5) Call pixel cloudy?
• Continue to monitor performance of GOES cloud
  mask
• Continue to monitor GOES/GPS PW differences
• Impact of more aggressive cloud mask
• Better clear-sky radiance PW products
• Good for NWP
• DPI Products
• Cloudier DPI Cloud Product
• Less dense PW, LI, etc DPI Products (Is this
  OK??? if not, what do we do???)
• Suggestions?