TRMM Ground Validation Some Lessons and Results

1
TRMM Ground ValidationSome Lessons and Results

• David B. Wolff, David Marks, David Silberstein
  Richard Lawrence
• TRMM Satellite Validation Office
• NASA/GSFC

2
Summary

• TRMM GV Data Processing is up-to-date
• Major issues are gauge data problems and
  (especially at KWAJ) radar calibration
  uncertainties.
• GSFC is developing an automated method of
  correcting relative calibration of the KWAJ radar
  to salvage historical KPOL data.
• Comparisons of GV estimates to TRMM (3G68) show
  good agreement, when radar calibration is not
  major issue.
• Comparisons of GV estimates to other datasets
  (e.g. MPA) also show good agreement and bode well
  for GPM era sampling expectations

3
Importance of Stable Calibration

• Calibration is a Major Issue for TRMM GV
• Houze et al. (2004) cite the error associated
  with a 2 dB calibration error is 30, in
  rain rate, respectively.
• Actual calibration uncertainty at KWAJ frequently
  well over 2 dB and sometimes gt 10 dB!
• Absolutely critical that proper calibration
  procedures be in place for GPM.
• Calibration uncertainty is the single largest
  source of error in Kwajalein rainfall estimates.
• While absolute calibration is not always possible
  using conventional radars, stable calibration is
  essential.
• GSFC is developing a method to correct historical
  calibration uncertainties using probability
  distributions of clutter area reflectivity.

4
GV v5 Kwajalein Monthly Radar/Gauge Rainfall
Statistics
5
Detrimental Effects of Unstable Radar Calibration
Waveguide change occurred on or around July 1,
2003
6
Statistical/Automated Correction of Calibration
TRMM GV uses a clutter-map in quality control
step at KWAJ.

• This map provides a lookup
• table for areas (r, ?) that have a
• high probability of echo when
• no precipitation is present.
• Given KPOL resolution, there are 4000 points
  per VOS.
• Calculate daily PDF of Clutter Area Reflectivity
  (CAR).
• Fond that upper percentiles (e.g. 95th) are
  remarkably stable when the radar calibration is
  stable, even when precipitation is present.

7
Statistical/Automated Correction of Calibration
Calibration changes are often traceable to
engineering changes to the radar system here a
7 dB gain was applied.
8
Statistical/Automated Correction of Calibration
95th Percentile Reflectivity in Clutter Areas is
Stable
Silberstein et al. 2005 (32nd Radar Conference)
9
RCA vs. Engineering Changes

• Relative Calibration Adjustment (RCA)

http//trmm-fc.gsfc.nasa.gov/trmm_gv/gv_products/G
Vproducts.html
10
Statistical/Automated Correction of Calibration
UW estimates based on comparison to TRMM
PR (Houze et al. 2004)
11
Quantifying RCA Effects vs. Z-R Changes
V5 2002 WPMM, no RCA V6 Seasonal WPMM, RCA
Marks et al. 2005 (32nd Radar Conference)
12
V5 Biases - KWAJ 1999-2004 GV vs. TRMM
13
V5 Monthly Means - KWAJ 1999-2004
14
V6r Biases - KWAJ 1999-2004
15
V6r Monthly Means - KWAJ 1999-2004
16
Comparing GV to Other Rain Estimates (MPA)

• A similar comparison was done using the
  Multi-Satellite Precipitation Analysis (3B42)
  Rain Estimate (Huffman et al. 2005)
• 0.25 x 0.25 gridded product, available every
  3-hours
• GV gridded similarly
• Compared the 3-hour accumulations from both GV
  and MPA over these pixels.

17
KWAJ Comparisons to Other Datasets (MPA)
18
MELB Comparisons to Other Datasets (MPA)
19

• Summary
• TRMM GV data processing is current for all sites
• Calibration issues at Kwajalein a major source of
  error
• Relative Calibration Adjustment (RCA) shows
  promise but is still a work in development.
• Comparisons to TRMM over MELB generally within
  10 on year-to-year basis
• Comparisons to other estimates (MPA) also show
  good agreement in MELB and KWAJ (during period
  when radar calibration is mitigated).
• TRMM GV is providing GPM GV a significant number
  of lessons learned, which are being applied in
  GPM GV development

20
Backup Slides
21
Statistical/Automated Correction of Calibration
Diurnal Differences in 95th Percentile Clutter
Reflectivity
22
Comparison of Rain Intensity Distributions
Probability distributions of rain rate were
derived from TRMM (PR, TMI and COM) and GV
estimates TRMM --gt global (land or ocean) for
Feb 1998 (ITE110 - official V6, thanks to S.
Yang) GV --gt KWAJ (Ocean) and MELB (Land) for
period 07-12/1999 Over ocean all PDFs nearly
log-normal with some exceptions. Over land,
there remains work to be done, especially for
2A12 (TMI).
23
TRMM GV Processing Status
24
Statistical/Automated Correction of Calibration

• Benefits of clutter map approach
• 95th percentile clutter-area reflectivity is
  remarkably stable even in the presence of
  precipitation. Little or no diurnal effect
• Easily calculated for near-real-time monitoring
  of the current state of the relative radar
  calibration
• Current effort determine the best means of
  applying these relative calibration changes to
  historical data to develop improved GV products
• Signal is strong, amplitude is in question. Under
  further investigation.

25
Version 6 GV Site Surface Masks
26
3G68 GV Land/Coast/Ocean Masks
27
Comparing GV and TRMM Data

• Notes on Bias Calculations
• Two references (GV and Satellite)
• GV_Ref Mean 0.5 GV rain intensity
• Sat_Ref (PR TMI COM)/3
• Bias (Reference - Measurement) / Measurement

28
Comparing TRMM and GV Rain Intensities

• TRMM 3G68 (gridded PR, TMI and COM) rain
  intensity estimates compared to similarly gridded
  TRMM GV estimates. All TRMM estimates are Version
  6!
• 0.5 x 0.5 over land, coast and ocean
• Analysis of 1999-2004 TRMM-v6 has been completed