08-RR-TTT-##
Identifying and Tracking Severe Weather Precursor Signatures from High-resolution Satellite Data in Real-Time
FY 2008 Proposal to the NOAA HPCC Program
November, 15 2007
| Title Page | Proposed Project | Budget Page |
Principal Investigator:Robert Rabin
Line Organization:National Severe Storms Laboratory
Routing Code: R/NS
Address: 120 David L. Boren Blvd, Norman OK 73072
Phone:405-325-6336
Fax: 405-325-1774
E-mail Address:
Collaborators:
Valliappa Lakshmanan (CIMMS, University of Oklahoma): / Jaime Daniels (NOAA/NESDIS):Arnie Gruber (NOAA/CREST, City College of New York): / Kurt Hondl (NOAA/NSSL):
Steven Weiss (NOAA/SPC):
John Moses (NASA/Goddard, Geenbelt MD): / Wayne Feltz (CIMSS, University of Wisconsin-Madison):
Proposal Theme: Technologies for Modeling, Analysis or Visualization
Funding Summary:FY 2008 $ 37,000 (In-Kind $10,000)
______/ ______/ ______Dr. Robert Rabin / Dr. Jeff Kimpel / Linda Skaggs
Research Meteorologist / Director / Administrative Officer
NSSL / NSSL / NSSL
Identifying and Tracking Severe Weather Precursor Signatures from High-resolution Satellite Data in Real-Time
Proposal for FY 2008 HPCC Funding
Executive Summary:
A variety of severe weather detection, diagnosis and tracking algorithms for radar data are available. However, severe weather algorithms for satellite data have been hampered by many constraints – bandwidth, data resolution both in space and time and machine power.
Recently, we have demonstrated two key abilities: (1) the ability to receive, process and combine 1km visible data from both GOES satellites in real-time and (2) the ability to identify overshooting tops, which are precursors to thunderstorms at speeds capable of being implemented in real-time.
In this HPCC proposal, the first objective is to routinely, and in real-time, create products that identify and track overshooting tops from satellite visible images. We will disseminate these products in several ways: as a Java applet available over the internet, as a georeferenced data set available through publicly available viewers such as Google Earth and as netcdf files for easy incorporation into AWIPS and N-AWIPS.
The second objective is to provide output from this tool to operational units in NOAA. This includes testing at the Storm Prediction Center (SPC), transfer of imagery to AWIPS and N-AWIPS, and input to a nowcasting system for greater New York City (in collaboration with NOAA cooperative institute CREST). In addition, the output will be combined with output of other satellite-based algorithms being developed at NASA Goddard and the Cooperative Institute for Meteorological Satellite Studies (CIMSS) University of Wisconsin-Madison for testing usefulness in severe storm detection.
Problem Statement:
With significant HPCC funding, a suite of severe weather algorithms have been developed for radar data. These use information from all the radars in the CONUS, and combine with information from other sensors and numerical models to create derived products such as rotation tracks, probability of hail, storm cell tracking, etc.
A similar suite of severe weather algorithms for satellite data is missing. There is a convective initiation algorithm, and several promising attempts at MCS tracking. However, the satellite algorithms are not yet capable of providing warning guidance to a severe weather forecaster.
The key issues have been:
- Channel availability: The highest resolution satellite images are the visible channel images which are at 1km resolution. However, these are available only during daylight hours over the location being sensed.
- Spatial and temporal resolution: For availability, the best channels to use are the infrared channels. However, the spatial resolution of these channels is only 4km every 15 minutes. Cloud features associated with severe weather cannot always be fully resolved at such a data resolution.
- Volume: Unlike radar data which are mostly missing (the worst case scenario for a CONUS weather radar system is one where 1/3 of the country is experiencing weather), on satellite visible images, nearly all the data are always valid during daylight hours. Thus, a 1km-satellite image is similar in terms of processing requirements to a 0.5km resolution radar image covering the same geographic extent.
Hence, successful satellite algorithms have either been large scale (such as MCS tracking or cloud classification) or grid-point-based (such as precipitation estimation or convective initiation).
So, far, we have been limited by our bandwidth and CPU to addressing severe weather applications on satellite. When NSSL scaled up the WDSS-II hardware system to handle 0.5-km resolution radar data, we finally got machines and bandwidth capable of receiving and processing 1km-resolution satellite visible data and performing sophisticated image processing on the dataset.
Solution:
Overshooting tops when seen on satellite images are precursors to thunderstorms on the ground. Thus, an automated ability to detect, diagnose and track overshooting tops is the first step towards building severe weather algorithms which combine satellite and radar information. These can complement the radar-only algorithm suite as they may increase lead time for severe weather detection during daylight hours, especially in areas not in close proximity to radars.
Figure 1: The output of the overshooting tops algorithm (red contours) shown superimposed over the visible image from which the overshooting tops were automatically identified.
We have developed a McIDAS script to retrieve satellite data from both GOES images, fuse them, remap them into a Platte Carre (equidistant cylindrical) projection and convert the data into netcdf images. The background image of Figure 1 demonstrates the 1km visible data received and processed using our software.
We have also developed an overshooting tops algorithm that does extremely well in detection. The red contours on the visible imagery in Figure 1 demonstrate the output of the completely automated overshooting tops identification algorithm. The overshooting tops can be collocated with severe weather signatures identified from IR imagery. Detection of IR signatures has been funded by NASA in a joint project with the Cooperative Institute for Meteorological Satellite Studies (CIMSS, University of Wisconsin-Madison) and NASA Goddard. The tracking software will be built as part of the HPCC effort.
Besides the web-based tool for viewing data, data will be imported on an experimental basis to N-AWIPS in collaboration at the SPC, and in AWIPS in collaboration with NESDIS and MDL. Output will include image loops containing observed and projected storm tracks, which is currently under development as part of the "System for Storm Analysis Using Multiple Data Sets" project funded by the HPCC.
Analysis:
This project provides a cost effective means to further develop tools for research and operations which utilize frequent, high resolution satellite data from NOAA. The project will leverage tasks from other existing projects in NOAA: 1) NASA funded research for incorporating feature detection from satellite with a sensor-web approach to forecasting, 2) a collaborative effort between NESDIS (Jaime Daniels) and MDL. In that project, an effort is underway to import various GOES satellite products into AWIPS (in parallel to SCAN which uses radar data), 3) a collaborative effort between NSSL and MDL which has successfully pushed LMA and radar analysis products to AWIPS workstations at four different WFO's in 2005, 4) a collaborative effort with NESDIS and CREST (NOAA Cooperative Institute) to develop a nowcasting system which uses satellite data. Our proposed HPCC efforts are of interest to these programs and we will benefit from working together on the import of data into AWIPS and implementation and testing of the tracking of overshooting cloud tops.
There are four parts to this proposal. Risk reduction efforts have been carried out on all the subcomponents, as indicated in the following table:
Activity / Status- Receive, fuse and remap visible 1km satellite data from both GOES satellites over the CONUS.
- Detect overshooting tops from the visible images.
- Track overshooting tops over time and attach radar and satellite based parameters with these overshooting tops so that forecasters can study the time evolution.
- Display algorithm output (real-time) in Java applet
Since all the components of the proposed solution have been tested (and timed) offline, there is little risk that this proposal will not be able to meet its goals.
Performance Measures
- Overshooting algorithm output available over the web in a Java applet.
Milestones
Receive award notification (tentative) – 15 March 2008
Order and receive hardware – 15 April 2008
Have overshooting tops algorithm running in real-time – 15 May 2008
Deliverables
Overshooting algorithm output will be available over the web in a Java applet.
Budget Summary:
Category / Detailed Description / Amount / In-Kind AmountPersonnel Compensation / Project Management PI (NOAA FTE)
Scientist – JI Salary (CIMMS) / $0
$15K / $ 10K (NSSL)
Contracts or Services
Rent, Communications, Utilities
Capital Expenses / Real-time Algorithm Server / $15K
Supplies and Materials
Training/Travel / HPCC Review
IIPS Conference or NOAA Tech / $ 3K
$ 4K
Total Requested: / $ 37K / $ 10K
Administrative Officer:Linda Skaggs
Phone: 405-325-6910
E-mail Address:
FMC Number: 50-26