Upload
simon-york
View
216
Download
0
Embed Size (px)
Citation preview
WSN05
Toulouse, France, 5-9 September 2005
Geostationary satellite-based methods for nowcasting convective initiation, total lightning flash rates, and
convective cloud properties
John R. Mecikalski1, Kristopher M. Bedka2
Simon J. Paech1, Todd A. Berendes1, Wayne M. Mackenzie1
1Atmospheric Science DepartmentUniversity of Alabama in Huntsville
2Cooperative Institute for Meteorological Satellite StudiesUniversity of Wisconsin-Madison
Supported by:Supported by:
NASA New Investigator Program (2002)NASA New Investigator Program (2002)NASA ASAP InitiativeNASA ASAP Initiative
WSN05
Toulouse, France, 5-9 September 2005
Outline
• Current capability: Overview• Progress toward improvements
Error assessmentsConfidence analysisAssessment of “interest field” importance; convective regimes
• Current & new initiativesNighttime convective initiationLightning (event) forecastingExamples with MSG data over Europe (MODIS)Demonstration with hyperspectral data
WSN05
Toulouse, France, 5-9 September 2005
How this began…
• Which cumulus will become a thunderstorm?• GEO satellite seems to be well-suited to address this question.• What methods are available?• What changes to current, globally-developed codes are needed?
Who can benefit from this research? What user groups are interested (e.g., 0-2 h
nowcasting)
••
WSN05
Toulouse, France, 5-9 September 2005
Where are we now …
• Applying CI algorithm over U.S., Central America & Caribbean• Validation & Confidence analysis• Satellite CI climatologies/CI Index: 1-6 h• Work with new instruments• Data assimilation possibilities
WSN05
Toulouse, France, 5-9 September 2005
• Build relationships between GOES and NWS WSR-88D imagery:
- Identified GOES IR TB and multi-spectral technique thresholds and time trends present before convective storms begin to precipitate
- Leveraged upon documented satellite studies of convection/cirrus clouds [Ackerman (1996), Schmetz et al. (1997), Roberts and Rutledge (2003)]
- After pre-CI signatures are established, test on other independent cases to assess algorithm performance
Use McIDAS to acquire data, generally NOT for processing:• GOES-12 1 km visible and 4-8 km infrared imagery every 15 minutes• UW-CIMSS visible/IR “Mesoscale” Atmospheric Motion Vectors (AMVs)• WSR-88D base reflectivity mosaic used for real-time validation• NWP model temperature data for AMV assignment to cumulus cloud pixels …
based on relationship between NWP temp profile and cumulus 10.7 m TB
Other non-McIDAS data:• UAH Convective Cloud Mask to identify locations of cumulus clouds
Input Datasets for Convective Nowcasts/Diagnoses
WSN05
Toulouse, France, 5-9 September 2005
Convective Cloud Mask
• Foundation of the CI nowcast algorithm: Calculate IR fields only where cumulus are present (only 10-30% of a domain on average)…greatly reduces CPU requirements
• Utilizes a multispectral region clustering technique for classifying all scene types (land, water, stratus/fog, cumulus, cirrus) in a GOES image
• Identifies 5 types of convectively-induced clouds: low cumulus, mid-level cumulus, deep cumulus, thick cirrus ice cloud/cumulonimbus tops, thin cirrus anvil ice cloud
WSN05
Toulouse, France, 5-9 September 2005
“Mesoscale” Atmospheric Motion Vector Algorithm
“Operational
Settings”
New Mesoscale
AMVs
(only 20% shown)
• We can combine mesoscale AMV’s with sequences of 10.7 m TB imagery to identify growing convective clouds, which represent a hazard to the aviation community
50% shown
WSN05
Toulouse, France, 5-9 September 2005
Oceanic Convective Cloud Growth Product
• Satellite AMVs are used to track clouds in sequential images and compute cloud-top cooling rates
• Rapid cloud-top cooling induced by convective cloud growth likely correlate well with vigorous updrafts and strong CIT
30 Minute
WSN05
Toulouse, France, 5-9 September 2005
1000 900 800 700 600 500 400 300 200 100 mb
1/120th of vectors shown1/30th of vectors shown1/5th of vectors shown
Oceanic Satellite Atmospheric Motion Vectors
• Meso-scale satellite AMVs provide detailed depictions of flow near convective cloud features
• Validation of AMVs using ACARS and wind profiler data is a future ASAP effort
WSN05
Toulouse, France, 5-9 September 2005
CI Interest Field Critical Value
10.7 µm TB (1 score) < 0° C
10.7 µm TB Time Trend (2 scores)< -4° C/15 mins
∆TB/30 mins < ∆TB/15 mins
Timing of 10.7 µm TB drop below 0° C (1 score) Within prior 30 mins
6.5 - 10.7 µm difference (1 score) -35° C to -10° C
13.3 - 10.7 µm difference (1 score) -25° C to -5° C
6.5 - 10.7 µm Time Trend (1 score) > 3° C/15 mins
13.3 - 10.7 µm Time Trend (1 score) > 3° C/15 mins
CI Interest Fields for CI Nowcasting
from Roberts and Rutledge (2003)
WSN05
Toulouse, France, 5-9 September 2005
2000 UTC
2030 UTC
2100 UTC
CI Nowcast Algorithm: 4 May 2003
• Satellite-based CI indicators provided 30-45 min advanced notice of CI in E. and N. Cent. KS
CI Nowcast Pixels
These are forecasted CI locations!
WSN05
Toulouse, France, 5-9 September 2005ddddddddddddddddddddddd
1) Remap GOES data to 1 km gridded radar reflectivity data
• Correct for parallax effect by obtaining cloud height through matching the 10.7 m TB to standard atmospheric T profile
2) Identify radar/lightning pixels that have undergone CI/LI at t+30 mins
• Advect pixels forward using low-level satellite wind field to find their approximate location 30 mins later
3) Determine what has occurred between imagery at time t, t-15, & t-30 mins to force CI/LI to occur in the future (t+30 mins)
4) Collect database of IR interest fields (IFs) for these CI/LI pixels
5) Apply LDA: identify relativecontribution of each IF toward an accurate nowcast• Test LDA equation on
independent cases toassess skill of new method
ASAP CI/LI Linear Determinant Analysis (LDA)
WSN05
Toulouse, France, 5-9 September 2005
CI “Interest Fields”: 8 Total from GOESCI Interest Field Critical Value
10.7 µm TB (1 score) < 0° C
10.7 µm TB Time Trend (2 scores)< -4° C/15 mins
∆TB/30 mins < ∆TB/15 mins
Timing of 10.7 µm TB drop below 0° C (1 score) Within prior 30 mins
6.5 - 10.7 µm difference (1 score) -35° C to -10° C
13.3 - 10.7 µm difference (1 score) -25° C to -5° C
6.5 - 10.7 µm Time Trend (1 score) > 3° C/15 mins
13.3 - 10.7 µm Time Trend (1 score) > 3° C/15 mins
WSN05
Toulouse, France, 5-9 September 2005
“Interest Field” Importance: POD/FARCI Interest Field Critical Value
10.7 µm TB (1 score) < 0° C
10.7 µm TB Time Trend (2 scores)< -4° C/15 mins
∆TB/30 mins < ∆TB/15 mins
Timing of 10.7 µm TB drop below 0° C (1 score) Within prior 30 mins
6.5 - 10.7 µm difference (1 score) -35° C to -10° C
13.3 - 10.7 µm difference (1 score) -25° C to -5° C
6.5 - 10.7 µm Time Trend (1 score) > 3° C/15 mins
13.3 - 10.7 µm Time Trend (1 score) > 3° C/15 mins
• Instantaneous 13.3–10.7 um: Highest POD (82%)
• Time-trend 13.3–10.7 um: Lowest FAR (as low as 38%)
• Important for CI & Lightning Initiation
WSN05
Toulouse, France, 5-9 September 2005
Detecting Convective Initiation at Night
Detection of convective initiation at night must address severalunique issues:
a) Restricted to 4 km data (unless MODIS is relied upon)b) Visible data cannot be used to formulate cumulus maskc) Highly-dense, GOES visible winds are unavailable for trackingd) Forcing for convection often elevated and difficult to detect
(e.g., low-level jets, bores, elevated boundaries)
However, the advantages are:a) Ability to use ~3.9 m channel (near-infrared) data (!!!)b) More “interest fields” become available for assessing cumulus cloud
development
Therefore, new work is toward expanding CI detection fornocturnal conditions, and/or where lower resolution may bepreferred (i.e. over large oceanic regions).
Wayne Mackenzie, MS student
WSN05
Toulouse, France, 5-9 September 2005
Detecting Convective Initiation at Night
Nighttime CI: SoutheastOklahoma
Enhanced 10.7 m
SHV: 2:57 - 3:44 UTC
WSN05
Toulouse, France, 5-9 September 2005
Detecting Convective Initiation at Night
What we’ve learned so far:
10.7-3.9 m channel difference (Ellrod “fog” product)13.3-3.9 m6.5-3.9 m
Evaluation is being done in light of the forcing for the convection (e.g., low-level jets, QG).
WSN05
Toulouse, France, 5-9 September 2005 kkoooooooookkkkkkkkkkk
Satellite-Lightning Relationships• Current Work: Develop relationships between IR TB/TB trends and lightning source counts/flash densities toward nowcasting (0-2 hr) future lightning occurrence* Supported by the NASA New Investigator Program Award #:NAG5-12536
2047 UTC
2147 UTC
Northern Alabama LMA Lightning Source Counts
2040-2050 UTC
2140-2150 UTC
WSN05
Toulouse, France, 5-9 September 2005
The Meteosat Second Generation (MSG) satellite system could be used effectively for CI nowcasting within this algorithm:
• 12 spectral bands; ~3 km resolution,• 2 water vapor channels centered on two different central wavelengths.• 8.7 µm, 9.7 µm, ~1.5 µm channels• Plus many of the GOES capabilities.
Meteosat Second Generation (MSG)
Nighttime CI event over Italy
9.7 µm
8.7 µm
WSN05
Toulouse, France, 5-9 September 2005
Hyperspectral: GOES-R
LEFT
8.508-10.98 m Band
Difference: Red (’s >0)
= Ice
Comparison between the 10.98
m (right) and 11.00 m (far right) bands:
22 UTC 6.12.2002
Small wavenumber changeresults in significant changesin view:
Low-level water vapor Surface temperature Subtle cloud growth
& microphysical changes
WSN05
Toulouse, France, 5-9 September 2005
Contact Information/Publications
Contact Info:
Prof. John Mecikalski: [email protected]
Kristopher Bedka: [email protected]
UW-CIMSS ASAP Web Page:
nsstc.uah.edu/johnm/ci_home (biscayne.ssec.wisc.edu/~johnm/CI_home/)
http://www.ssec.wisc.edu/asap
Publications:
Mecikalski, J. R. and K. M. Bedka, 2005: Forecasting convective initiation by monitoring the evolution of moving cumulus in daytime GOES imagery. In Press. Mon. Wea. Rev. (IHOP Special Issue, ~Late 2005).
Bedka, K. M. and J. R. Mecikalski, 2005: Applications of satellite-derived atmospheric motion vectors for estimating mesoscale flows. In Press. J. Appl. Meteor.
Mecikalski, J. R., K. M. Bedka, and S. J. Paech, 2005: A statistical evaluation of GOES cloud-top properties for predicting convective initiation. In preparation. Mon. Wea. Rev.