Title: Bob.Rabin@noaa.gov
1Nowcasting of thunderstorms from GOES Infrared
and Visible Imagery
- Valliappa.Lakshmanan_at_noaa.gov
- Bob.Rabin_at_noaa.gov
- National Severe Storms Laboratory University of
Oklahoma - http//cimms.ou.edu/lakshman/
2Nowcasting Thunderstorms From Infrared and
Visible Imagery
- Tracking Storms Existing Techniques
- Overview of Method
- Identifying Storms at Multiple Scales
- Motion Estimation and Forecast
3Methods for estimating movement
- Linear extrapolation involves
- Estimating movement
- Extrapolating based on movement
- Techniques
- Object identification and tracking
- Find cells and track them
- Optical flow techniques
- Find optimal motion between rectangular subgrids
at different times - Hybrid technique
- Find cells and find optimal motion between cell
and previous image
4Some object-based methods
- Storm cell identification and tracking (SCIT)
- Developed at NSSL, now operational on NEXRAD
- Allows trends of thunderstorm properties
- Johnson J. T., P. L. MacKeen, A. Witt, E. D.
Mitchell, G. J. Stumpf, M. D. Eilts, and K. W.
Thomas, 1998 The Storm Cell Identification and
Tracking Algorithm An enhanced WSR-88D
algorithm. Weather Forecasting, 13, 263276. - Multi-radar version part of WDSS-II
- Thunderstorm Identification, Tracking, Analysis,
and Nowcasting (TITAN) - Developed at NCAR, part of Autonowcaster
- Dixon M. J., and G. Weiner, 1993 TITAN
Thunderstorm Identification, Tracking, Analysis,
and NowcastingA radar-based methodology. J.
Atmos. Oceanic Technol., 10, 785797 - Optimization procedure to associate cells from
successive time periods - Satellite-based MCS-tracking methods
- Association is based on overlap between MCS at
different times - Morel C. and S. Senesi, 2002 A climatology of
mesoscale convective systems over Europe using
satellite infrared imagery. I Methodology. Q. J.
Royal Meteo. Soc., 128, 1953-1971 - http//www.ssec.wisc.edu/rabin/hpcc/storm_tracker
.html - MCSs are large, so overlap-based methods work well
5Some optical flow methods
- TREC
- Minimize mean square error within subgrids
between images - No global motion vector, so can be used in
hurricane tracking - Results in a very chaotic wind field in other
situations - Tuttle, J., and R. Gall, 1999 A single-radar
technique for estimating the winds in tropical
cyclones. Bull. Amer. Meteor. Soc., 80, 653-668 - Large-scale growth and decay tracker
- MIT/Lincoln Lab, used in airport weather tracking
- Smooth the images with large elliptical filter,
limit deviation from global vector - Not usable at small scales or for hurricanes
- Wolfson, M. M., Forman, B. E., Hallowell, R. G.,
and M. P. Moore (1999) The Growth and Decay
Storm Tracker, 8th Conference on Aviation, Range,
and Aerospace Meteorology, Dallas, TX, p58-62 - McGill Algorithm of Precipitation by Lagrangian
Extrapolation (MAPLE) - Variational optimization instead of a global
motion vector - Tracking for large scales only, but permits
hurricanes and smooth fields - Germann, U. and I. Zawadski, 2002
Scale-dependence of the predictability of
precipitation from continental radar images. Part
I Description of methodology. Mon. Wea. Rev.,
130, 2859-2873
6Need for hybrid technique
- Need an algorithm that is capable of
- Tracking multiple scales from storm cells to
squall lines - Storm cells possible with SCIT (object-identificat
ion method) - Squall lines possible with LL tracker (elliptical
filters optical flow) - Providing trend information
- Surveys indicate most useful guidance
information provided by SCIT - Estimating movement accurately
- Like MAPLE
- How?
7Nowcasting Thunderstorms From Infrared and
Visible Imagery
- Tracking Storms Existing Techniques
- Overview of Method
- Identifying Storms at Multiple Scales
- Motion Estimation and Forecast
8Technique Stages
- Clustering, tracking, interpolation in space
(Barnes) and time (Kalman)
Courtesy Yang et. al (2006)
9Technique Details
- Identify storm cells based on reflectivity and
its texture - Merge storm cells into larger scale entities
- Estimate storm motion for each entity by
comparing the entity with the previous images
pixels - Interpolate spatially between the entities
- Smooth motion estimates in time
- Use motion vectors to make forecasts
Courtesy Yang et. al (2006)
10Why it works
- Hierarchical clustering sidesteps problems
inherent in object-identification and
optical-flow based methods
11Advantages of technique
- Identify storms at multiple scales
- Hierarchical texture segmentation using K-Means
clustering - Yields nested partitions (storm cells inside
squall lines) - No storm-cell association errors
- Use optical flow to estimate motion
- Increased accuracy
- Instead of rectangular sub-grids, minimize error
within storm cell - Single movement for each cell
- Chaotic windfields avoided
- No global vector
- Cressman interpolation between cells to fill out
areas spatially - Kalman filter at each pixel to smooth out
estimates temporally
12Nowcasting Thunderstorms From Infrared and
Visible Imagery
- Tracking Storms Existing Techniques
- Overview of Method
- Identifying Storms at Multiple Scales
- Motion Estimation and Forecast
13K-Means Clustering
- Contiguity-enhanced K-Means clustering
- Takes pixel value, texture and spatial proximity
into account - A vector segmentation problem
- Hierarchical segmentation
- Relax intercluster distances
- Prune regions based on size
14Example hurricane on radar (Sep. 18, 2003)
Image
Scale1
Eastward
s.ward
15Satellite Data
- Technique developed for radar modified for
satellite - Funding from NASA and GOES-R programs
- Data from Oct. 12, 2001 over Texas
- Visible
- IR Band 2
- Because technique expects higher values to be
more significant, the IR temperatures were
transformed as - Termed CloudCover
- Would have been better to use ground temperature
instead of 273K - Values above 40 were assumed to be convective
complexes worth tracking - Effectively cloud top temperatures below 233K
C 273 - IRTemperature
16Segmentation of infrared imagery
Not just a simple thresholding scheme
Coarsest scale was used because 1-3 hr forecasts
desired.
17Nowcasting Thunderstorms From Infrared and
Visible Imagery
- Tracking Storms Existing Techniques
- Overview of Method
- Identifying Storms at Multiple Scales
- Motion Estimation and Forecast
18Motion Estimation
- Use identified storms in current image as
template - Move template around earlier image and find best
match - Match is where the absolute error of difference
is minimized - Not root mean square error MAE is more
noise-tolerant - Minimize field by weighting pixel on difference
from absolute minimum - Find centroid of this minimum region
- Interpolate motion vectors between storms
19Processing
Clustering, Motion estimation
IR to CloudCover
Motion estimate applied to IR and Visible
20Forecast Method
- The forecast is done in three steps
- Forward project data forward in time to a
spatial location given by the motion estimate at
their current location and the elapsed time. - Define a background (global) motion estimate
given by the mean storm motion. - Reverse obtain data at a spatial point in the
future based on the current wind direction at
that spot and current spatial distribution of
data.
21Forecast Example (IR, 1hr, 2hr, 3hr)
22Forecast Example (Visible, 1hr, 2hr, 3hr)
Varying intensity levels are a problem
23Skill compared to persistence
24Conclusions
- Advection forecast beats persistence when storms
are organized - Does poorly when storms are evolving
- IR forecasts are skilful
- Visible channel forecasts are not