Current Research

1
Current Research

• Roongroj (KIJ) Chokngamwong
• CEOSR/GMU
• Prof. Long Chiu
• NASA/GMU
• 01/04/2008

2
Outline

• GPCP-PSPDC
• Satellite Rainfall Estimation using
  Microwave-calibrated Infrared Split-window
  Technique (MIST)

3
Why We Estimate Rainfall from Space?

• Rain gauge networks are limited only to over land
  and remain sparse over most of the globe.
• Radar networks are limited to only a few areas
  and a few countries. Inter-radar calibration,
  mountain blockage
• Satellite sensors provide an excellent complement
  to continuous monitoring of rain event both
  spatially and temporally.
• Geostationary satellites VIR/IR observation,
  good coverage
• Low Earth orbiting satellites MW observation,
  provide less frequent observations but direct
  rainfall estimation

4
The Global Precipitation Climatology Project
(GPCP)

• To combine the precipitation information
  available from each of several sources into a
  final merged product by taking advantage of the
  strengths of each data type.
• The infrared (IR) precipitation estimates are
  computed primarily from geostationary satellites
  (United States, Europe, Japan)
• The Atmospheric Infrared Sounder (AIRS) data from
  Aqua
• Outgoing Longwave Radiation Precipitation Index
  (OPI) data from NOAA series satellites
• The gauge data are assembled and analyzed by the
  Global Precipitation Climatology Centre (GPCC)
• The microwave estimates are based on Special
  Sensor Microwave/Imager (SSM/I) data from the
  Defense Meteorological Satellite Program (DMSP,
  United States) satellites
• Adler et al. (2003) The Version 2 Global
  Precipitation Climatology Project (GPCP) Monthly
  Precipitation Analysis (1979-Present). J.
  Hydrometeor., 4,1147-1167.

5
The Polar Satellite Precipitation Data Centre
(PSPDC)

• Oceanic Rainfall derived from Special Sensor
  Microwave Imager (SSM/I) data
• IR can only provide cloud observations
• Microwave interacts directly with hydrometeors in
  the rain column (more physical approach).
• At microwave frequencies, the ocean is a highly
  reflective background and the atmosphere is
  highly transparent under most circumstances.

6
Oceanic Rainfall Retrieval Algorithm

• Developed by Wilheit, Chang and Chiu (1991)
• The rain rate, r, (in mm/hr) can be empirically
  related to the brightness temperature (Tb) via
  the following relationship
• where T0 is the average Tb for the non-raining
  portion of the Tb histogram, rc is 25/FL1.2 and
  FL is the freezing height (in kilometers) of the
  rain layer.
• Tb is the temperature that a blackbody would
  need to have in order to emit radiation of the
  observed intensity at a given wavelength.

7
(No Transcript)
8
Rainfall Estimation using Microwave Calibrated
Split-window Infrared Technique (MIST)
9
TRMM

• Launched in November 1997 (altitude of 350 km)
• Boosted to 405 km altitude in August 2001
• Extends satellite life to 2009 (current decision)
• TRMM rain instruments VIRS, TMI and PR
• The VIRS measures scene radiance in five spectral
  bands (0.6, 1.6, 3.8, 10.8 and 11.9 micron).
• The five TMI frequencies are 10.65, 19.35, 37 and
  85.5 GHz (vertical and horizontal polarization),
  and 21 GHz (only vertical polarization).
• The PR onboard TRMM, the first rain radar in
  space, is a cross-track scanning instrument
  operating at 13.8 GHz.

10
3B42 Version 5

• Coincident TCA and VIRS are analyzed to establish
  transfer coefficient relationship.
• Use the relationship to calibrate GOES IR rain
  estimates by adjusting the GPI to form 3B42
  (Adjusted GPI)

11
3B42 Version 6

• Put all available TCI-calibrated MW (TMI, SSMI,
  AMSR and AMSU) into 3 hourly 0.25 degree bin and
  fill missing bin with MW-calibrated IR rain rate
• The data are summed over a month to create
  monthly multi-satellite product (MS)
• The MS and gauge analysis are merged to create
  Gauge-Satellite (GS) product
• 3B42 is scaled as the ratio of MS/GS limited to
  0.2-2.

12
Rainfall Climatology Comparison
13
Daily Comparison
14
Skill Score

• How well the TRMM algorithms estimate rain events

• POD A/(AB)
• FAR C/(AC)
• CSI A/(ABC)

Observation
Algorithm

• A is (Algorithm rain, observation rain)
• B is (Algorithm no, observation rain)
• C is (Algorithm rain, observation no)
• For perfect algorithm, POD 1, FAR 0 and CSI
  1

15
Sensitivity of CSI
16
Comparison Summary

• The satellite-only product (V5) overestimates low
  rain events and underestimates heavy rain events
• The V6 TRMM algorithms show improvement over the
  V5, in terms of the mean RR, error statistics,
  correlation, rainfall CDF, and temporal and
  spatial autocorrelation structure
• The TRMM algorithms overestimate rain fraction
  and underestimates conditional rain rates
• No improvement of CSI from V5 to V6

17
Rainfall Information using Split Window Technique
18
Split Channel Technique

• Inoue (1987) has used a split window technique to
  delineate raining areas from NOAA-7 AVHRR
  imagery.
• The principle that underlies this technique is
  that while both cirrus and cumulonimbus clouds
  may be cold and bright, the spectral dependence
  of the emissivity of ice and water clouds differs
  in the infrared window.
• The two clouds can be discriminated by comparing
  the difference in the brightness temperatures at
  11 and 12 micron.
• The split window technique yields a better false
  alarm rate and also corrects the problem of the
  rainfall overestimation due to the error in
  rainfall area delineation when using only one
  channel IR.

19
Case 1 Mixed Clouds

• Blue refers to PR rain pixels
• Red refers to PR non-rain pixels

20
Case 2 Warm Clouds

• Blue refers to PR rain pixels
• Red refers to PR non-rain pixels

21
Case 3 Cold Clouds

• Blue refers to PR rain pixels
• Red refers to PR non-rain pixels

22
(No Transcript)
23
Summary

• The single threshold algorithm does not perform
  well for rain/no-rain classification for all
  cases
• The use of the split window technique is
  effective in eliminating non-raining pixels

24
Rainfall Estimation using Microwave-calibrated
Infrared Split-window Technique (MIST)
25
Operational Satellite Algorithms
26
Comparison of CSI over Australia
27
Goal of MIST

• Produce rainfall estimates by using only
  satellite product without merging with gauge
  measurements as an input
• Produce rain rate at the IR pixel resolution
• Improve over AGPI

28
Data Used

• Microwave Data (3B40RT)
• TMI, SSM/I, AMSR, AMSU
• IR data from GMS-5
• GMS-5 (0.75, 6.9, 10.8 and 11.5 µm)

29

• Reference Hsu et al. (1997) Figure 2 (b).

30
(No Transcript)
31
July2002
32
February2002
33
Comparison over Thailand
34
Skill Comparison
35
February 2002Thailand
36
July 2002Thailand
37
Application of MIST to GOES data

• Microwave Data (3B40RT)
• IR data from GOES-12

38
Oklahoma Comparison
39
Skill Comparisons
40
July 2006Oklahoma
41
February 2006 Oklahoma
42
Conclusion

• Develop Microwave-calibrated Infrared
  Split-window Technique (MIST)
• CSI can be improved by incorporating the split
  window technique
• The MIST algorithm provides reasonable
  performance and is comparable to the V6 3B42
• No direct rain gauge input
• Easily portable to other sensors and platforms

43
Previous and Ongoing Research

• Variation of Vegetation and Rainfall in Thailand
• Variability of aerosol optical depth and aerosol
  forcing over India
• Trends in Oceanic Rainfall Derived from Microwave
  Brightness Temperature Histograms
• Effect of TRMM boost on oceanic rain rate
  estimates based on microwave brightness
  temperature histograms
• Variation of Oceanic Rain Rate Parameters from
  SSM/I Mode of Brightness Temperature Histogram
• "Trends" and variations of global oceanic
  evaporation datasets from remote sensing
• Thailand Daily Rainfall and Comparison with TRMM
  Products
• Development of the Microwave calibrated Infrared
  Split-window Technique (MIST) for rainfall
  estimation

44
Next things to do

• V6 SSM/I PSPDC
• Reprocess data for 20 years for all satellites
  (F8, F10, F11, F13, F14, F15)
• Continue working on MIST
• Test the applicability of this technique with
  other IR sensors and channels
• Apply this technique to other Satellite
  Platforms, such as FY and Meteosat
• Incorporate the radar data which may improve the
  accuracy since the radar has the ability to
  detect fairly light rain
• Aerosol-Precipitation Interaction

45

• Thank you

46
Back up slides
47
Tropical Rainfall Diurnal Cycle

• http//daac.gsfc.nasa.gov/precipitation/trmm_apps/
  TRMM3G68_animation.shtml

48

• Importance of Rainfall in Thailand
• Effects on agriculture
• Thailand is the worlds largest rice exporter.
  Annual exports are 7.5, 7.2, 7.6, 10.13 (record
  high) million tons in 2001, 2002, 2003, and 2004,
  respectively (Source USDA)

49
POD, FAR and CSI
50
Sensitivity of CSI
Similar results are found by Katsanos et al.
(2004) over the central and eastern Mediterranean
51
Temporal Autocorrelation
52
February 2006
Cumulative Distribution Function
July 2006
53
(No Transcript)
54
Background (Cont.)

• Flood in August 2002 in Northern Region causing
  many lives and property damages. The flood caused
  about 32,000,000 damage (Source Dartmouth Flood
  Observatory). Floods are associated with heavy
  rain.

55
Typhoon Xangsane (Elephant)
56
(No Transcript)
57
(No Transcript)
58
TRMM Satellite Algorithm Data Flow
59
Principle of VIS/IR rainfall estimation

• The cloud information will be a good
  discriminator of rain/no rain classification
• No cloud, no rain
• Thick and cold clouds tend to rain

60
Comparison with Version 5
61
Comparison with Version 6
62
TRMM V5 and V6 difference
63
(No Transcript)
64
The assumptions

• In the model, a Marshall-Palmer distribution of
  raindrops as a function of rain rate is assumed
  to exist from the ocean surface to the freezing
  level in the atmosphere.
• A nonprecipitating cloud containing 0.25 kg m-2
  of integrated liquid water is assumed in the 0.5
  km below the freezing level.
• A constant lapse rate of 6.5 C/km is assumed.
• The relative humidity is assumed to increase
  linearly with height from 80 at the ocean
  surface to 100 at the freezing level and above.

65
July 2006Oklahoma