SERVIR:
Central America & Caribbean- CI Nowcasting
Satellite Derived
Offset Technique
(click image to enlarge)
 |
Satellite-derived
Barnes Wind Analysis |
In order to account for cloud motion in the time interval between
satellite images, this study incorporates the UW-CIMSS satellite-derived
cloud-motion VIS and IR AMV algorithm (Velden et al. 1997, 1998)
toward the formation of satellite-derived offset vectors (SOVs)
for evaluating cloud-top TB and multi-spectral band difference
trends. A SOV is a measure of the distance that a cumulus cloud
pixel has traveled (in pixels) between successive satellite images.
Initialization settings associated with AMV algorithm were manipulated
in a way that allows for the identification of both synoptic-
and meso-scale flows (Bedka and Mecikalski 2005). The AMV algorithm
used in operational settings preferentially identifies synoptic-scale
flow in geostrophic balance. Adjustments were made to 1) reduce
the impact of the NWP model background, 2) increase the number
of cloud features tracked, 3) change the AMV editing methodology
for optimal meso-scale AMV detection. These changes all serve to
enhance detection of meso-scale cloud motions associated with boundary
layer (upper-tropospheric) convergence (divergence) that can be
used to more accurately obtain cloud-top trend information.
Once the AMVs are obtained, a Barnes objective analysis (Barnes
1964) is performed to produce a satellite wind analysis at 1-km
resolution for three atmospheric layers, 100-70 kPa, 70-40 kPa,
and 40-10 kPa. Convectively-induced cloud pixels are assigned an
AMV from one of the three layers depending on a comparison of the
pixel 10.7 µm TB and a NWP model temperature profile. Cloud-top
trends are calculated if a satellite wind is present at the appropriate
height within the near vicinity of a cumulus cloud (~0.25º radius).
The wind speed, direction, and the time interval between images
are used to obtain the SOV and thus the approximate pixel/cloud
location in previous images.
After applying the SOV, a check is performed to ensure that the
past pixel location does in fact represent a convectively-induced
cloud. Assuming a pixel passes both checks (wind availability and
past cloud presence) for both the 15- and 30-min time lags, cloud-top
cooling/multi-spectral band difference trends are calculated. The
passage of these checks therefore indicates that a convectively
induced cloud is being tracked, back to a reasonable prior location,
across successive images.
Despite use of the above checks, and every effort to obtain accurate
SOVs (in terms of direction and magnitude), errors in estimating cloud
TB trends using SOVs are inevitably present in the analysis below.
These errors have been determined to range from ~1-2 km at 15 mins,
to as large as 5 km by 30-45 mins. In fact, by the 45 min time lag
(and beyond), use of the SOV technique to track clouds is not reasonable,
except in conditions of uniform flow and nearly unidirectional vertical
shear. Nonetheless, SOVs obtained from satellite winds are perhaps
the only means of tracking clouds with reasonable accuracy over large
geographical regions in real-time, thereby offering a substantial
improvement over simple per-pixel differencing techniques. Other more
sophisticated methods exist for tracking clouds in satellite data
(e.g., Papin et al. 2000; Carvalho and Jones 2001), and these will
be tested as part of future improvements to SOV calculation for this
project. (back to top-->)
6.5-10.7
um Temporal Trends
(click image to enlarge)

30
minute trend in the 6.5-10.7 micron difference technique
|
As
described in the "Infrared Products Description" section,
the 6.5-10.7 um band difference is useful for inferring cloud top
height. The images above provide examples of temporal trends in
this band subtraction technique. Values of this band subtraction
technique typically range from around -50 to +3 K. The highest (postive)
values of this technique indicate the presence of high clouds at
or above ("over-shooting" cumulus tops) the tropopause
whereas the lowest (negative) values are indicative of clear sky.
Therefore, the highest temporal change in this technique can be
+/- 53 K. A highly negative (-20 or greater) temporal change indicates
one of two things: 1) high cloud movement away from a location (currently
negative - formerly postive = negative value) or 2) rapid high cloud
decay (currently highly negative - formerly slightly negative (positive)
= negative value). A highly positive value can also mean one of
two things: 1) high cloud movement into a location (i.e. cirrus
expansion, currently positive - formerly negative = positive value)
or 2) cumulus growth. Cumulus growth is what we are most interested
in determining. Trend values of 3 K/15 mins were found to precede
CI in several CI events syudied in the development of the CI nowcast
algorithm. Hence, this value is used as an interest field for CI
nowcasting. (back to top-->)
13.3-10.7
um Temporal Trends
(click image to enlarge)

30
minute trend in the 13.3-10.7 micron difference technique |
As
described in the "Infrared Products Description" section,
the 13.3-10.7 um band difference is useful for inferring cloud height
and cloud type. The image above provides an example of the 30 min
temporal trend in this band subtraction technique. Values of this
band subtraction technique typically range from around -30 to +3
K. The highest (postive) values of this technique indicate the presence
of high cirrus or mature cumulus clouds whereas the lowest (negative)
values are indicative of low clouds such as immature cumulus, stratus,
and fog. Therefore, the highest temporal change in this technique
can be +/- 33 K. A highly negative temporal change indicates cirrus
expansion (currently negative - formerly postive (or 0)= negative
value). A positive value indicates cumulus growth. (currently positive
- formerly negative (or 0) = positive value) Trend values of 3 K/15
mins were also found to precede CI in several CI events syudied
in the development of the CI nowcast algorithm. Hence, this value
is used as an interest field for CI nowcasting. (back
to top-->)
10.7
um Temporal Trends
(click image to enlarge)

GOES 10.7 micron: 5/04/03, 1930-2000 UTC |

15 minute change of 10.7 micron TB < -4 K
|
This
technique, documented in Roberts and Rutledge (2003), involves examination
of time trends in 10.7 um TB (cloud top temp.). In this work, the
authors noted that "the onset of vigorous cloud growth leading
to storm development was characterized by cloud tops that reached
sub-freezing temperatures and exhibited large cooling rates at cloud
top 15 min. prior to the first detection of 10 dBZ radar echoes
aloft and 30 min before 35 dBZ. The rate of cloud top temperature
change was found to be important for descriminating between weakly
precipitating storms (<35 dBZ) and vigorous convective storms
(>35 dBZ)." The cloud top cooling rate for "weak, limited
growth" was 4 K/15 mins and was 8 K/15 mins for "vigorous"
growth. The product above illustrates locations exhibitng "weak,
limited" growth" (i.e. the 4 K/15 mins threshold). We
utilized the "satellite-derived offset vector" technique
(described above) to perform the cloud top trend estimates. We have
also calculated the cloud top cooling rate over a 30 minute interval,
which was not performed in the forementioned paper. This allows
us to identify clouds exhibiting sustained growth. (back
to top-->) |