Thunderstorms are identified by cold cloud tops
from
the GOES-11/12 IR window channel (11 micron) imagery or from radar
echos. An algorithm originally
developed by Brian Mapes (19??) is used to automatically detect cloud
clusters
with temperatures less than a cloud top temperature threshold. Here,
the algorithm has been
adapted to identify and track areas where the radar reflectivity
exceeds
a threshold. The areas or clusters can be
followed in time by searching for partial overlap in their position
between
successive images. (This implies that for a given speed of movement and
image frequency, that the clusters must exceed a certain size threshold
to be sucessfully tracked. For example, the diameter must be greater
than
15 km for a movement of up to 60 km/hr and 15 min image frequency). A
time
series of the centroid position, surface area, mean, maximum and
mimimum
reflectivity are evaluated for each cluster.
When run on GOES data, the cloud tracks can be
displayed for different cloud top temperature thresholds ranging from
-43 to -73 deg. C. For WSR-88 data, the algorithm is run with
different reflectivity
thresholds (5 dbZ intervals from 30 to 60 dbZ). For each value, a
unique
set of precipitation clusters, tracks and reflectivity statistics are
obtained. The user can decide which threshold to use in displaying the
output.
Real-time movies include data from the past 3 hours
and and future projection of images from the K-means
algorithm developed by V. Lakshmanan of NSSL
and CIMMS University of Oklahoma.
Fig. 1. GOES movies
Past
3 hours
Including
60,120 minute nowcast
Fig. 2. WSR-88D movies
Past
3 hours
Including
30, 60 minute nowcast
Java based applications used for interactive animations were developed by Tom Whittaker of the Space Science and Engineering Center, University of Wisconsin-Madison. The animations can take a while to load, depending on network speed, computer systems, etc. Also, there can be a problem viewing these on certain machines (Macs).