May 15, 2002 MURI Hyperspectral Worksh op 1 Cloud and Aerosol Products From GIFTS/IOMI Gary Jedlovec and Sundar Christopher NASA Global Hydrology and Climate Center University of Alabama - Huntsville Research goals (year 1): • Identifying cloud and surface characteristics in high spectral resolution data that best delineate clouds, aerosols, and surface characteristics from one another, and leads to a superior cloud product. • Refine the Tracking Error Lower Limit (TELL) parameter to include instrument characteristics and observing requirements of the GIFTS/IOMI. Presentation centers on capabilities to satisfy these goals
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
May 15, 2002 MURI Hyperspectral Workshop 1
Cloud and Aerosol Products From GIFTS/IOMI
Gary Jedlovec and Sundar ChristopherNASA Global Hydrology and Climate Center
University of Alabama - Huntsville
Research goals (year 1):
• Identifying cloud and surface characteristics in high spectral resolution data that best delineate clouds, aerosols, and surface characteristics from one another, and leads to a superior cloud product.
• Refine the Tracking Error Lower Limit (TELL) parameter to include instrument characteristics and observing requirements of the GIFTS/IOMI.
Presentation centers on capabilities to satisfy these goals
May 15, 2002 MURI Hyperspectral Workshop 2
Cloud DetectionCurrent geostationary cloud property retrieval technique at GHCC –
detect clouds and retrieve cloud information, mask for atmospheric & surface parameter retrieval (http://wwwghcc.msfc.nasa.gov/goesprod)
Cloud detection
• Bi-spectral THreshold (BTH) method (Jedlovec and Laws 2001)
Used operationally at the GHCC (24h a day)
GOES Imager or Sounder
Single pixel resolution (4 or 10 km)
• 3.7 - 11 micrometer difference provides key cloud signature• Three (3) tests applied to difference image
1. ~2.8K spatial pixel deviation (edge detection)
2. ~2.1K adjacent pixel (element direction) change (fills in clouds)
3. Historical (20 day) minimum difference image check for each time (detects low clouds/fog, and incorporates synoptic influences)
Performance documented against NESDIS products (Jedlovec and Laws 2001) link
May 15, 2002 MURI Hyperspectral Workshop 3
Cloud and Aerosol Products
Parameter retrieval
• Cloud height (CTP) – infrared look-up with model guess for GOES imager and opaque cloudso Easy to implement, uses model T(p) as a reference
o Highly accurate for opaque clouds
o CO2 slicing H2O intercept possible with Sounder (currently not implemented)
• Cloud phase – water or ice, mixed – reflective information at 3.7 micrometers (under development)
• Aerosol optical thickness (AOT) – visible channel approach to retrieve AOT in cloud-free regions (Zhang and Christopher, 2001)
o DISORT model (Ricchiazzi et al. 1998) used to generate look up tables describing radiance, AOT, ,
o Correlation with sun photometer data as high as 0.97 chart
May 15, 2002 MURI Hyperspectral Workshop 4
Cloud Product ComparisonsGOES-8 CTP – 16:45 UTC 18 April 2002
NESDIS Sounder CTPGHCC BTH Imager CTP
May 15, 2002 MURI Hyperspectral Workshop 5
Cloud Product ComparisonsGOES-8 vs MODIS - 18 April 2002
MODIS CTP (1635 UTC)GHCC BTH - GOES Imager CTP (1645 UTC)
May 15, 2002 MURI Hyperspectral Workshop 6
Cloud Research Focus
Examine spectral signature of clouds, aerosols , and dust for unique features
• Use AIRS radiance data for selected periods
• Begin to adapt the Bi-spectral Threshold method for for high spectral measurements for the retrieval of cloud products
May 15, 2002 MURI Hyperspectral Workshop 7
Sources of wind tracking errors• When clouds and wv features are non-conservative tracers of wind• Changes in cloud shape (often result of too large of image separation)• Improper height assignment• Mis-identification of targets (dependent on tracking algorithm)• Incorrect image displacements (navigation and registration inaccuracies)
The effect of incorrect image displacements on the cloud-tracked wind is a function of image registration, image separation time, and image resolution.
Tracking Error Lower Limit (Tell) is the theoretical lower limit error in wind tracking algorithms due to image resolution (), time separation (), and image stability or registration accuracy () uncertainties. TELL = ( ) /
GOES• infrared pixel resolution () is 4km• image-to image registration accuracy () is typically about 2km (~0.5 pixel)For 15 minute images ( = 15), TELL = 2.22 ms-1
This means that GOES derived winds under these conditions will typically have a 2 ms-1 error component due to these image uncertainties alone!
Satellite-derived Wind Errors
May 15, 2002 MURI Hyperspectral Workshop 8
Science Requirement: Accurate mesoscale winds for diagnostic and modeling studies (<2.0 ms-1)• use small time intervals• high resolution imagery• accurate image-to-image registration
Imaging Requirement:Resolution trades/constraints:• as image separation () is decreased (point 1 to 2), the registration accuracy (R) must improved to maintain quality of wind data • if image resolution () is improved, registration accuracy can be relaxed (point 2 to 3) for an equivalent image separation interval ()
Imaging Requirements for Cloud-drift Winds
20001000
500250
12563
0.5
1.0
2.0
4.0
8.0
16.0
0
5
10
15
20
25
30
35
40
45
50
55
60
Image Interval, Resolution, and Registration Accuracy Constraints
1
23
TELL =(R*)/
GOES-R
= 15 min
R = 0.125
= 4km
TELL = 0.55
TELL Surface of 0.55
-
Im
age
Sep
arat
ion
Tim
e (m
in)
R - Image Registration Accuracy (m) - Im
age R
esol
utio
n (km
)
May 15, 2002 MURI Hyperspectral Workshop 9
Wind Tracking Error Emphasis
Refine Tracking Error Lower Limit (TELL) for GIFTS
• Instrument characteristics
• Observing scenarios
May 15, 2002 MURI Hyperspectral Workshop 10
Summary / Deliverables
Focus of researchExamine spectral signature of clouds, aerosols , and dust for unique features
• Use AIRS radiance data for selected periods
• Begin to adapt the Bi-spectral Threshold method for for high spectral measurements for the retrieval of cloud products
Refine Tracking Error Lower Limit (TELL) for GIFTS
• Instrument characteristics
• Observing scenarios
DeliverablesKey spectral signatures and wavelengths for the detection of clouds and aerosols
Insight on how these characteristics can be included in a cloud product algorithm
Estimates of the lower limit on satellite derived wind errors from GIFTS
May 15, 2002 MURI Hyperspectral Workshop 11
Backup Charts
Cloud and Aerosol Products From GIFTS/IOMI
Gary Jedlovec and Sundar ChristopherNASA Global Hydrology and Climate Center
University of Alabama - Huntsville
May 15, 2002 MURI Hyperspectral Workshop 12
15 points (locations on the image to right) used each hour to validate cloud detection schemes
subjective determination of clouds (man in the loop)visible, multiple channel IRany pixel cloudy in 32x32km area, then all cloudy
• Statistical performance at hourly intervals - 2 times below• Results are for:
CLC = ground truth clear – retrieval scheme correctCLI = ground truth clear – retrieval scheme incorrectCDC = ground truth cloudy – retrieval scheme correctCDI = ground truth cloudy – retrieval scheme incorrect
NESDIS = NESDIS operational algorithm (Hayden et al. 1996)BSC = Bi-spectral Spatial Coherence method (Guillory et al. 1998) used operationally at GHCCBTH = Bi-spectral Threshold algorithm – under development
Daytime: 1845 Statistics
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Ra
tio
CLC
CLI
CLR
CLD CDC
CDI
BTH
CLC
CLI
CDC
CDI
CLC
CLI
CDC
CDI
BSC NESDIS
Night: 0645 Statistics
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Ra
tio
CLC
CLI
CLRCLD
CDC
CDI
BTH
CLC
CLI
CDC
CDI
CLC
CLI
CDC
CDI
BSC NESDIS
Cloud Detection ValidationCase Study: September 11 – October 8, 2001
back
May 15, 2002 MURI Hyperspectral Workshop 13
MODIS IR C31 – 16:35 UTC 18 April 2002
May 15, 2002 MURI Hyperspectral Workshop 14
Related Documents
