NOAA-MDL Seminar 7 May 2008 Bob Rabin NOAA/National Severe Storms Lab Norman. OK CIMSS University of Wisconsin-Madison Challenges in Remote Sensing to.

NOAA-MDL Seminar7 May 2008

Bob RabinNOAA/National Severe Storms Lab

Norman. OK

CIMSS University of Wisconsin-Madison

Challenges in Remote Sensing to Improve Severe Weather Forecasting

Current Applications

A Web-based tool for monitoring MCSStorm Analysis Using Multiple Data Sets.

Robert Rabin, Tom Whittaker 2004

Advances in Visual Computing,

G. Bebis, R. Boyle, D. Koracin, B. Parvin, Ed(s)., Springer, 571-578.

Identify and track MCS

- Cold cloud tops

- Radar reflectivity

- Adjustable thresholdsTime trends of MCS characteristics

- Size

- Cloud top temperature stats

- Radar reflectivity stats

- Lightning

- Storm environment from RUC,...Real-time and archived data on-lineData access from NOMADS/THREDDS catalog

Data Flow

GOES

Radar

RUC model analysis

THREDDS

LightningTrackingAlgorithm Web Server

Example session:

Mesoscale Convective Complex: Mature Stage

Time Series: Mature Stage

Mesoscale Convective Complex: Decaying Stage

Time Series: Decaying Stage

Tornadic Storm Track

Time Series: Tornadic Storm Track

Real-time and archive data:

http://tracker.nssl.noaa.gov

Motion Estimation

Uses K-Means clustering and Kalman filters

Forecast dBZActual dBZ

30 min30 min

Need for new approach

Traditional centroid tracking Accurate at small scales, but not at large scales Inaccurate when storms merge or split Possible to extract trends from the information

Flow-based tracking Cross-correlation, Lagrangian methods, etc. Are accurate at large scales, but not at small scales Not useful in decision support because trends of storm

properties can not be extracted

K-Means clustering

K-Means clustering is a hybrid approach Cluster the input data to find clusters

Like centroid-based tracking methodsBut at different scales.

Track the clusters using flow-based methods (minimization of cost-functions)Like flow-based methodsDoes not involve cluster matching (e.g: Titan)

Example clusters

Two different scales shown

Both scales are

tracked

Extrapolation

Smooth the motion estimates spatially using OBAN

techniques (Gaussian kernel)

temporally using a Kalman filter (assuming constant velocity)

Repeat at different scales and choose scale appropriate to extrapolation time period.

Nowcasting Infrared Temperature

How good is the advection

technique

What is the quality of cloud cover nowcasts?

Effectively the quality of forecasting IR temperature < 233K

Blocks represent how well

persistence would do

The lines indicate how well the motion estimation technique does

1,2,3-hr nowcasts shown

Real-time loops (WSR-88D and GOES):

http://www.nssl.noaa.gov/~rabin/tracks

Detecting Overshooting Tops

Looked for high textural variability in visible

images

These are the thunderstorms to be identified and forecast

Shown outlined in red

Detection algorithm now running in real-time

at NSSL

http://www.nssl.noaa.gov/users/rabin/public_html/vis_1km/

Couplets

Another technique to identify thunderstorms developed by John Moses of NASA Looks for couplets of high

and low temperatures Data from 2200 UTC from the same

Oct. 12 case The pink tails indicate the

past position of these detections

As with our overshooting tops technique, persistence of detection is a problem

No. 17 jumps all over the place

No. 36’s direction is wrong

No. 39, 40, 41 have no real history

No. 37 is being tracked well

Real-time visible loops (comparison with

radar, upper-level divergence):

http://www.nssl.noaa.gov/~rabin/vis_1km

Mesoscale Wind Analysis from Water Vapor Imagery

Detecting Winds Aloft from Water Vapour Satellite Imagery in the Vicinity of Storms

Rabin, R.M., Corfidi, S.F., Brunner, J.C., Hane, C.E. Weather, 59, 251-257

GOES-8 Water Vapour Imagerydivergence (yellow, 10-5s-1), and absolute vorticity (red, 10-5s-1)

2300 UTC 19 July 1995

0445 UTC 20 July 1995

Upper air winds at 0000 UTC on 03 June 2003

300 mb rawinsonde analysis

Upper air winds at 0000 UTC on 03 June 2003 from satellite

black: 100-250, cyan: 251-350, yellow: 351-500 hPa

Surface weather map at 2300 UTC, 11 June 2003

Severe weather reports

wind damage (blue)

large hail (green)

tornadoes (red)

GOES-12 divergence at 300 hPa

11 June 2003 at 1945 UTC

12 June 2003 at 0045 UTC

First Guess (NOGAPS) divergence at 300 hPa

11 June 2003 at 1945 UTC

12 June 2003 at 0045 UTC

Surface weather map at 2300 UTC, 12 June 2003.

GOES-12 divergence at 300 hPa on 12 June 20032145 UTC

derived from satellite-winds

first guess model (NOGAPS)

VIS0115 UTC

WV0215 UTC

Greensburg, KSStorm

05 May 2007

Real-time and archive:

http://www.nssl.noaa.gov/~rabin/winds

http://cimss.ssec.wisc.edu/mesoscale_winds