Plymouth State University Meteorology
CCAFS/KSC Warm-Season Convective Wind Climatology
Background: This convective wind climatology was developed using 18 years (1995-2012) of warm-season (May-September) peak wind data
from 82 anemometers on 36 weather towers over a 30 km x 40 km area on and around the Cape Canaveral Air Force Station (CCAFS)/Kennedy Space Center (KSC)
complex. Other observational data were also used, such as the surface observations at the Space Shuttle Landing Facility (KTTS) at KSC and from the
Skid Strip (KXMR) at KSC. Additional surface observations for the area, NEXRAD radar (primarily from Melbourne (KMLB)), and GOES satellite data were
also used to better define convective periods. Warnings for strong convective winds account for the second highest number of weather warnings (after
lightning) issued by the Air Force 45th Weather Squadron (45WS) for the CCAFS/KSC area. The specific 45WS convective wind warning/advisory criteria are
listed here. This climatology helps in mission planning, forecaster training, and operational decision-making.
This climatology was started to update and expand a microburst climatology of warning level convective events, previously done by
Sanger (1999), which was based only on data for 1995-1998.
Loconto et al. (2006) updated the convective wind climatology by identifying and
examining warning-level convective events for the period of 1995-2003 and also examined various convective wind prediction techniques. The
next study (Cummings et al. 2007) increased the period of record by two additional
years (2004 and 2005) and looked at all convective periods rather than just warning-level microburst events. This page was later updated to include
fine tuning all of the previously identified convective periods based by using high resolution NEXRAD data and also added a radar climatology for all
warning level events from 1996-2005 (see Dinon et al. (2008) for details). The
climatology was again expanded by Ander et al. (2009) to add data for 2006 and 2007
and the radar climatology added these years and also went back to include all non-warning level convective events for all of the years. Data for 2008 through 2012
were then added and refined by Laro (2011; 2012) and Lupo (2013).
The introductory chapter of Loconto (2006) provides a nice literature review and summary of convective wind mechanisms.
Wind Tower Data: Five-minute wind data for 45 towers were initially examined, but nine of the towers did not meet the threshold requirement that
the wind sensors at a given site must have valid peak wind observations for at least 70% of the possible times over the period of record with no single
monthly availability value for any site dropping below 65% in order to be included in the climatology. Fortunately, most of the towers eliminated were
well to the west of the CCAFS/KSC complex. A study by Koermer and Roeder (2008) did look at
the data from these far western towers and concluded that besides their low data availability, they were also not very useful for anticipating convective
wind events, since strong convective winds to the west did not tend to develop and propagate eastward and impact the CCAFS/KSC complex. Tower 0112 was replaced and only slightly
relocated with Tower 0211 in 2001 and data for both were considered as from a single tower (denoted by 0211) for the purposes of this study. Figure 1 below
shows the individual tower sensor locations with their 4-digit numeric identification and the locations of KTTS and KXMR. Tower IDs in black numbers were used in this study and
towers in white numbers were not, since they did not meet the availability criteria.
Figure 1. Map of locations of CCAFS/KSC wind towers and KTTS and KXMR. Data from the black four-digit numeric tower identifiers were used in this study.
Most towers have wind sensors at multiple levels and six towers have dual sensors at the same levels on opposite sides of the tower. Peak winds
are not recorded at the 12-foot elevation on any towers and hence their data were not used. We also eliminated tower observations above 300-feet,
since this is above the level for any 45WS convective wind warning criteria. Applicable sensor elevations on various towers for providing data for this study included 30, 54,
60, 90, 145, 162, 204, and 295 feet. Sensor records were very consistent for most sites and the only small variation came from a few sensors at the
54-foot level, just west of the Indian River. Table 1 provides a summary of the towers and elevations of various sensors. Figure 2 graphically
depicts the number of sensors actually used in this study versus their elevation.
Table 1. Listing of tower identification numbers, location, and sensor elevations.
Figure 2. Height distribution of 82 sensors (on the 36 towers that were used in this wind climatology study).
Note: The 12-foot elevations sensors were not used in this study, since they did not provide peak wind data.
Data Quality Control: The initial tower dataset for 1995-2003 had been prepared by the NASA Applied Meteorology
Unit (AMU) and had undergone automated quality control (QC) screening procedures, described by
Lambert (2002). We first merged all of the
reports for a given month from the separate sensor files into a single chronological file and extracted all peak wind reports ≥
35 knots into single monthly files. This allowed us to quickly zero in on potential problems with warning-level reports. Based on a
manual review of these summaries, synoptic data, and radar observations, some bad observations were obviously missed in the initial QC process.
Most had common characteristics--they had high peak wind reports; they were the only observations with anywhere near these magnitudes; and they
only had 0 to 1 degrees of directional variability over the 5-minute period. A large sample of these observations were checked against other
observations and found to be inconsistent with any other wind reports from that tower or nearby towers. It was noted that even fairly weak peak
wind reports almost always had larger directional variability. As a result, another automatic QC program was developed to identify and eliminate
these unrealistic reports and manual QC checks were done to eliminate strong synoptic pressure gradient cases and/or those lacking nearby convective
activity.
For 2004-2012, raw tower 5-minute data were provided by CSR. Data were converted to "AMU" format for consistency and checked for low directional,
variability, but were not put through the AMU automated QC process. Data were merged as before and manually scrutinized for all identified convective
periods. This review only identified a handful of probable bad reports that were later flagged in the database. We did not see nearly
as many inconsistencies in more recent reports as we had seen in the data from earlier years.
This page contains links to access the entire raw quality controlled CCAFS/KSC tower data by month used in this study. These
data can also be queried by various search criteria from this page. Missing or bad data are indicated by "999(6-9)" entries.
Convective Period Identification: The purpose of this study was to examine convectively driven winds and hence there was the need to identify periods
of probable convection. In order to do this, we first reviewed all the available KTTS observations (download)
taken over the period of 1995-2011, which were provided by 14 WS in Asheville, NC. These observations are high quality, manually taken observations, with
very descriptive remarks and cloud identification information unlike most current ASOS/AWOS automated observations. After 2011, the manual KTTS
observations ceased for a period and then were replaced by automated observations, which lacked most of the detailed convective storm information that were
included in the manual observations. Additionally, we used surface observations from KXMR, Titusville (KTIX) and Patrick AFB (KCOF) for periods, as needed
for further clarification. We also used National Climatic Data Center (NCDC) radar archives, NOAA National Hurricane
Center (NHC) tropical cyclone archives and Plymouth State Weather Center (PSWC) archives that include some
NCEP/NCAR reanalysis data, surface data, upper air data, radar summary data, and satellite imagery. Melbourne (KMLB) National Weather Service (NWS) NEXRAD radar
data were extremely valuable in fine-tuning start and end times of these events and in distinguishing stratiform from convective precipitation events, using a
minimum 40 dBz threshold.
From these data, probable convective periods were identified as a period being defined as beginning on the top of the hour when convection in the area
first occurs and ending at the top of the hour after the last evidence of convection that is followed by a break in convective activity for a period of 6 hours or
more. All warning-level events were manually reviewed. Periods of strong synoptic pressure gradients, primarily often associated with frontal passages in May
or September or resulting because of the proximity to tropical systems (see summary), were also identified and these periods were
eliminated from the climatology, since the focus of the climatology is on primarily convectively driven winds.
These activities resulted in the identification of 1151 probable convective periods, which equates to an average of 64 periods per year during the warm season months of
May through September. These periods are listed along with some brief details about them in monthly summaries that can be accessed from this
page. Recall that tower winds were not available for two of the identified convective periods and had to be eliminated from the study.
The start and end times for the remaining 1149 periods were used for developing the statistical summaries and climatological results in the sections that follow.
Florida Flow Regimes: Lericos et al. (2002) showed how lightning activity
could be related to various synoptic patterns (or flow regimes) over the Florida Peninsula. The regimes were based upon the 1000-700 hPa average wind direction from Jacksonville
(KJAX), Tampa (KTBW), and Miami (KMFL, formerly KMIA) 12 UTC radiosonde soundings. Lambert (2007)
refined the technique to reduce the large number of "Other" and "Missing" classifications by using the Cape Canaveral Air Force Station (KXMR) 10 UTC sounding, when a regime was not fully
defined by the KJAX, KTBW, and KMFL soundings. The flow regimes used for this study are based upon this updated technique. This page
allows a user to view maps with the averaged 1000-700 hPa average wind barbs for radiosonde sites in the region and a gridded wind vectors from the analysis of the 1000-700 hPa winds. The site also provides the
flow regime designation from the NASA Applied Meteorology Unit from 1995 through 2009. The flow regimes can be summarized in the following table:
| Table 2. Flow regimes based upon the position of the subtropical ridge axis relative to Florida as
previously established by Lericos et al. (2002). Click on flow regime symbol to
obtain a chart depicting that specific flow regime. |
|
FLOW REGIME |
SUBTROPICAL RIDGE POSITION |
|
SW-1 |
Subtropical ridge south of Miami |
|
SW-2 |
Subtropical ridge between Miami and Tampa |
|
SE-1 |
Subtropical ridge between Tampa and Jacksonville |
|
SE-2 |
Subtropical ridge north of Jacksonville |
|
NW |
Subtropical far to south and extending far in Gulf of Mexico and stronger than normal |
|
NE |
Subtropical far to north and extending into SE US and much stronger than normal |
|
Other |
Subtropical ridge position not defined |
|
Missing |
Missing synoptic data to determine flow regime |
Statistical Summaries: After the tower observational data were quality controlled and convective periods were identified, several programs were written to
analyze the data and extract counts based on certain search criteria and output summary reports on the peak winds observed. The first summaries focused
on counting the peak wind speeds in 5-knot categories (0-4 kt, 5-9 kt, 10-14 kt, ...) for each individual convective event for each month/year. The second
summaries break these peak wind speed category counts further down by hours (UTC) of the day. The next set of summaries takes the categories and characterizes
the peak wind observations by tower sensor elevation. An additional set of summaries matched the peak wind speed categories by their directions. Another
analyzed peak wind speeds versus flow regimes for the region developed by
Lericos et al. (2002) and refined by
Lambert (2007). The final set of individual
event summaries includes individual tower sensors versus the peak wind speed categories. Each summary also indicates the start/stop times, the prevailing
synoptic flow regime, the times of the first and last peak wind ≥ 20 knots (if any), the times of the first and last peak winds ≥ 35 knots
(if any), the times of the first and last winds ≥ 50 knots (if any), and the maximum peak wind reported during the convective period along with the wind direction, tower
ID, sensor elevation, and time of the event. All of these summaries are set up in individual files by month/year and each file contains all of the individual
convective period data for that month.
Using the individual period summaries, additional programs were written to summarize all of the individual periods by month for a given year and then summarize all
of the same months over the 18-year period of record. These monthly composite summaries are similar to the individual event summaries except one was not done for
peak speeds, since this information is included in all of the other monthly summaries. Each monthly summary file for a given breakout contains the summary data for
each month/year and at the end the combined data for all 18 months. All period and monthly summary files can be accessed here
In addition, monthly frequency data lists and maps showing individual tower sensor location for either a single elevation or all sensor elevations and for various wind
thresholds at or above 35 knots can be generated on this site.
KMLB Radar Data: During the summer of 2007, archived KMLB Radar data were obtained From NCDC covering the identified convective periods and for roughly 1-hour before through
1-hour after the period. The data primarily included base angle 0.5 radar reflectivity. KTBW data were also downloaded for those periods when KMLB data were
unavailable. This information was used to refine the previously identified convective periods and proved most beneficial during the earlier years of the study,
when high resolution radar data were not previously available. These radar data can be accessed here for all convective periods where KMLB
(or KTBW as backup) base reflectivity data were available can be obtained. Note that some other radar products may be available, but those products are not very complete.
Faster downloading of pre-built HAniS radar reflectivity loops with time matched peak tower winds covering each convective period (where data are available) can be
accessed here.
KXMR Sounding Data: Nearly all KXMR soundings have been obtained for all five warm season months from 1995 through 2012 and can be accessed in text or graphical form at this
site. These soundings were from the GTS transmissions. Until mid-November 2007, these data were encoded according to an old WMO
standard that required relative humidity values of 20% or less to be automatically assigned a dewpoint depression of 30°C. Since this could affect several important calculated
indices for predicting convective winds, it was decided to acquire all of the raw sounding data from CSR at CCAFS over the time of the old encoding. These raw data did not include
the WMO dewpoint depression convention and were converted to a pseudo standard format for plotting and display purposes. These data can be accessed at this
site. May-Sep 1995-2007 data from either the raw CSR data or the GTS can be displayed individually on thermodynamic diagrams or text from this
page. Dual soundings of these GTS and raw CSR based soundings for a selected date/time can be accessed at this site,
so that a user can see the differences between the two soundings through 2007. All daily KXMR soundings for all months from January 1995 through present in the Plymouth State archive
can be accessed here.
Climatological Analyses: This section provides the detailed climatological results. Based on the 18 years of warm-season wind data, data were stratified for
various time scales (yearly, monthly, and diurnally), by speed categories, by wind direction, by prevailing flow regime, and by tower elevation and location. We also
examined a number of potential aids based on climatology that may be useful for predicting warning-level wind events. These aids include the amount of time for ramp up from
20 knots to winds exceeding warning criteria, the relationship of cloud-to-ground lightning flash rate to observed peak wind speeds, and the relationship of GPS-derived
Integrated Precipitable Water (IPW) to convective wind speed.
 |
|
After classifying synoptically driven convective event days as non-convective days, this study found during the 153-day warm-season (number of days in May-September)
from 1995 through 2011 on average 59 days (~38.6%) are convective and 94 days (~61.4%) are non-convective during the warm-season. A day was classified as convective, if it had
at least one or more convective periods. Figure 3 shows that there was some annual variability. The year 2006 had the fewest convective days, while the year 2009 had
the most convective days. The year 2009 was also the only year where there were more convective (78) than non-convective days (75). |
| Figure 3. Annual distribution of the convective versus
non-convective days during the warm-season months (May-September) during the
18-year (1995-2012) span of the study. |
 |
|
Based on the classifications of convective and non-convective days, the warm-season (May-September) monthly averages reveal some notable trends, i.e. July on average has the most convective days versus non-convective days. May typically has the fewest convective days followed by September. |
| Figure 4. Annual distribution of the convective versus
non-convective days by warm-season month during the 18-year (1995-2012) span of the study. |
 |
|
Warm-season convection is definitely more likely than not to occur with a SW flow regime and less frequently with the other flow regimes. |
| Figure 5. Convective versus non-convective days categorized by flow regimes for all
warm-season days during the 18-year (1995-2012) span of the study. |
| Convective versus non-convective days by flow regime and month |
May |
Jun |
Jul |
Aug |
Sep |
 |
|
After eliminating synoptically driven pressure-gradient convective events, this study found a total of 1151 convective periods, which were defined in the "Convective
Period Identification" section above, during the warm-season (May-September) from 1995 through 2012 with an average of 64 convective periods per season, or about three events per
week. Figure 6 shows that there is some annual variability in the number of convective periods. The year 2006 had the fewest convective periods (34), while the year 2009
had the most convective periods (86) as would be expected from the convective/non-convective day results. |
| Figure 6. Distribution of 1151 convective periods from the warm-season months
(May-September) during the 18-year span of the study. |
 |
|
This figure provides a breakout of the convective periods in Figure 6 by batching them according to
the highest peak wind reported during the convective period during the warm-season (May-September) from
1995 through 2012. Years with fewer convective periods, tended to have a relative greater frequency of
warning-level events. Out of the warm season average of 64 convective periods, about 39 of them (61.4%) are
below warning thresholds and the remaining 25 (38.6%) had warning criteria winds. Of this latter
category, on average about 20 periods (31.6%) have peak winds in the range of 35-49 knots; and about 5
periods per year (7.4%) had peak winds in the ≥ 50 knot range.
|
| Figure 7. Distribution of the 1149 convective periods with maximum peak wind thresholds for
the warm-season months (May-September) during the 18-year study by year based on the highest peak wind reported
during the period. |
 |
|
The monthly climatology of convective periods forms a roughly bell-shape distribution of the
average number of convective events per month. On average, July and August had the greatest (and nearly
identical) number of convective periods and May had the fewest. The maximum number of convective periods
in any month in any individual year was 25 during July 2009, while the minimum number of convective periods in any
month was two in May 2007. The trends in the maximum and minimum number of convective periods from one month to
the next followed the average trend. |
| Figure 8. Distribution of 1149 convective periods from the 18-year wind
climatology (1995-2012) during the warm-season months (May-September). The
minimum column represents the year with the fewest number of convective periods
for a given month, while the maximum column represents the year with the
greatest number of convective periods for a given month. The average column
represents the average number of convective periods for the given month. |
 |
|
Warning-level periods roughly follow the average monthly distribution pattern with the maximum in July. However,
June had slightly more ≥ 50 knot events. |
| Figure 9. Average monthly distribution based on maximum peak wind categories based on the 1149
convective periods from the 18-year wind climatology (1995-2012) during the warm-season months (May-September). |
 |
|
This figure roughly shows how convective periods are distributed diurnally. The diurnal distribution of convective
wind observations spanning the convective periods for all warm-season months shows a single peak in the late afternoon
at 1900 UTC, which should be expected due to maximum daytime heating around that time and a relatively flat minimum
centered about 0900 UTC, associated with the coolest times of the diurnal cycle. By examining the individual months
the peak is an hour later in May and September and those months also show a flatter overall distribution with a
slightly higher percentage of nocturnal/early morning events, especially in September. The strong afternoon peak is
most evident for July, which is closely followed by August and then June. |
| |
| Figure 10. A 24-hour percentage distribution of all convective wind observations for all the convective
periods which occurred during all of the warm-season months in the 18-year study period. |
| Distribution of Maximum Convective Wind Observations by Month |
May |
Jun |
Jul |
Aug |
Sep |
 |
|
The data showed that the average maximum wind speed was 32.4 kt
with little diurnal speed variation (σ = 3 kt) for the convective
periods studied. While convective periods are certainly less frequent at night and in the morning,
they are usually only a few knots weaker on average than the afternoon convective periods. |
|
| Figure 11. The diurnal distribution of the average maximum peak
wind speeds for the 1149 convective periods observed during the warm-season months of the 18-year
study. The number of convective periods that had the maximum peak wind during the specified hour
is noted on the bottom of each bar. |
| The Diurnal Distribution of Averages Peak Wind Speeds by Month |
May |
Jun |
Jul |
Aug |
Sep |
 |
|
Approximately, 38.6% of the identified convective periods has at least one warning-level wind
report. Most of these observations ≥ 35 kt generally occurred during the afternoon with
a primary peak at 1900-2000 UTC. The result also reveal a much smaller secondary around 0100
UTC, which was primarily due to a single event on 25 September 2001, which had 134 warning-level wind observations,
which was one of the highest numbers for any single convective event over the entire warm season period of record. |
|
| Figure 12. Diurnal percentage distribution of peak wind observations ≥ 35 kt
during the 443 (out of 1149 - 38.6%) periods with warning-level winds for the warm-season months of the 18-year study period. |
|
| Diurnal Percentage Distribution of Peak Wind Observations > 35 kt |
May |
Jun |
Jul |
Aug |
Sep |
 |
|
The data show that the most strong convective wind events (≥ 50 kt) ramps up during the afternoon hours with a peak at 20 UTC.
Again, the secondary peak at 01 UTC is attributable to a single strong event on 25 September 2001, which had 33 observations ≥ 50 kt.
This was more than any other single convective event over the entire period of record.
|
|
| Figure 13. The diurnal distribution of the 370 observations for 85 convective periods (out of 1149 - 7.4%) where the maximum peak wind speed was 50 kt or greater over the 1995-2012 warm seasons. The time is based upon the hour when the highest wind was reported during the period within the tower network. |
 |
|
The distribution of the maximum wind speed, associated with each convective period in 5 kt increments, resulted
in a skewed bell-shape distribution with a peak in the 25-29 kt range. The Gumbel curve, shown in green,
represents a theoretical, best-fit probability distribution associated with the observational data.
The probability of meeting or exceeding any speed threshold, given that convective winds occur, can be found by
integrating the Gumbel curve from that speed threshold to infinity. This can be done can be done mathematically by
using the equation shown in Figure 14 and plugging in a threshold value for x (in knots). Setting x equal to the
threshold values of 35, 50, and 60 knots, yields the probability values shown.
|
| Figure 14. Frequency distribution of maximum convective peak wind observations by
5-knot increments, with a Gumbel probability curve fit to the observed data for the 1149 convective periods
for the warm-season (May-Sep) months in the 18-year (1995-2012) study period. |
 |
|
Weaker winds become more likely in August and even more so in September. There is not much difference in the
probability distributions for May through July.
Computed probabilities for an individual month or the entire warm-season for various convective wind speed thresholds
(in knots) in 5-knot increments are shown in Table 3 and for 1-knot increments in
Table 4. |
| Figure 15. The Gumbel probability distribution for maximum convective peak winds during a
convective period in 5-knot increments, with the probability curve fit to the observed monthly (May-Sep) data for
the 1149 convective periods in the 18-year (1995-2012) study period. The more detailed breakouts by individual month
(like Figure 14) can be accessed from the links in the table below. |
 |
|
The frequency and average intensity of peak convective wind speeds varied with the associated direction of the wind.
The largest percentage of convective periods had a maximum peak wind with a westerly direction, followed by southwest and
south directions. Convective periods that had a maximum wind with a northeasterly component were the least frequent.
Maximum peak convective wind speeds were the strongest on average when they had westerly component directions, while
convective winds associated with easterly component directions were generally the weakest. |
| Figure 16. Distribution of the maximum peak wind directions and speeds associated with the
1149 Convective periods over the eight cardinal directions for the 1995-2012 warm seasons. The average maximum peak wind speed for
each direction is noted at the top of each bar. |
 |
|
This figure shows a very strong correlation of warning-level peak wind events having westerly components and
weaker events more frequently associated with easterly components. This can be attributed to prevailing westerly flow,
which is most often associated with stronger convective development. |
| Figure 17. Distribution of the maximum peak wind directions by speed categories associated with the
1020 convective periods over the eight cardinal directions for the 1995-2012 warm seasons. A more detailed look by
individual categories can be obtained from the selections below. |
 |
|
Convective periods with a SW‑1 flow regime had the strongest average maximum convective wind, while the SW‑2 flow regime
contained the greatest number of convective periods. Flow regimes with higher averaged maximum convective wind speeds had a west
wind component, while flow regimes with an east wind component had lower averaged maximum convective winds.
The data also suggest that the wind components of the synoptic scale flow regime were often reflected in the typical direction of the maximum
convective wind. For example, maximum convective winds observed in a SE‑2 flow regime tended to have a southeasterly direction. |
| Figure 18. Distribution of 1149 maximum convective wind speeds in the eight flow regime categories during the
warm-season months of the 18-year study. The average maximum convective wind speed for each flow regime in each flow regime category is noted
under the bars. |
 |
|
Clearly the southwesterly (SW-1 and SW-2) flow regimes dominate the warning periods. SW-2 also occurs with the largest
number of non-warning events, but SE-1 edges out SW-1 for the next largest for non-warning events. |
| Figure 19. Frequency of occurrences of flow regimes associated with identified convective periods broken down by whether they were warning level or non-warning level events for the entire 18-year period of record. |
 |
|
Convective wind periods meeting various threshold speed criteria for the various sensor heights was analyzed and normalized by the
number of sensors that report winds at each elevation. The normalization was required since the number of sensors varied widely
by height (See Figure 2) with more observations at 54 ft than at any of the other elevations. The normalized data show that
peak wind events for criteria ≥ 20 kt generally increased with increasing elevation, with the highest at 295 ft.
As elevation increases there are increasingly stronger convective wind reports due to decreasing friction (Holton 2004) and
increasing kinetic energy with increasing size of the turbulent cells. When plotted in proportion to the height, these data
suggest a turbulent mixed layer with an average depth of about 150 ft with a log-wind law beginning to apply above that depth. |
| Figure 20. Distribution of normalized convective wind periods by sensor height for < 20 kt, 20-34 kt, 35-49 kt, and ≥ 50 kt threshholds over the
18-year period of record. |
 |
|
The Coastal/Causeway towers overall had the highest number of maximum peak wind reports that correspond to the convective periods,
when compared to other areas of the study region. Higher elevation sensors did fairly well as expected, but towers 19, 22, 300, and
1007, all reporting at 54 ft, were more often reporting the strongest peak wind. |
|
| Figure 21. Distribution of maximum peak winds for a convective period by Coastal/Causeway towers included
in the 18-year study during the warm-season months. The sensor elevations of the various towers are noted by the legend. |
 |
|
Peak winds most frequently show westerly components at most of the towers in the Coastal/Causeway area. |
| Figure 22. Wind directions reported with the maximum peak wind events by tower for the convective period in
Figure 20. |
 |
|
The CCAFS/Merritt Island tower sensors reported the maximum peak wind speeds for convective periods compared to the
Coastal/Causeway towers. Higher elevation sensors for 3131 and 3132 did well, but tower 421 at 54 ft and one of the most
northern towers had a significantly higher number of reports than the other CCAFS/Merritt Island site. This northern area
seems to be more convectively active than most other areas in the study area. |
| Figure 23. Same as Figure 21, except for the CCAFS/Merritt Island towers. |
 |
|
Peak winds still generally have westerly components at most towers. |
| Figure 24. Same as Figure 22, except for the CCAFS/Merritt Island towers. |
 |
|
The Mainland tower sensors had the smallest number of observations of maximum wind speed reported for the convective periods
of the study. Tower 1612, the most westerly site, had the highest number of events recorded. This area often gets events which
move to the northeast and don't affect the KSC/CCAFS areas. Tower 1612 also has one of the best records with little missing data
when compared to the other towers in this region. Note that all these Mainland sensors are located at the 54 ft elevation. |
| Figure 25. Same as Figure 21, except for the Mainland towers. |
 |
|
Although most towers still show westerly component for the small number of periods, where they reported the maximum peak wind,
Tower 1612 shows more of a mix of directions. |
| Figure 26. Same as Figure 22, except for the Mainland towers. |
Radar Studies: During the third year of the project,
Dinon et al. (2008) obtained and analyzed high resolution
NEXRAD data from the National Climatic Data Center (NCDC) archive. These data consisted mainly of Melbourne (KMLB) 0.5°
reflectivity which was used to determine cell strength, cell initiation, cell structure, cell group movement, individual cell
movement and location of maximum peak wind with respect to the cell (all six features as well as their respective potential
categorizations are shown in the table below). However, this analysis was only done for all warning level events for the eleven
warm seasons between 1995 and 2005. During the fourth year of the project,
Ander et al. (2009) obtained mainly Melbourne (KMLB) 0.5° reflectivity data for the years 2006 and 2007 increasing the number
of years in the data set to 13. To expand upon the radar study, they also analyzed all non-warning level cases between 1995 and 2005
and all cases for 2006 and 2007. Laro (2011 and 2012) expanded the period of record out through 2011 and Lupo (2013) extended it to include 2012. The results of the study are shown in the series of six graphs below. In order to simplify the analysis, radar data were time-matched and overlaid with the corresponding peak wind tower data as well as the user's choice of wind barbs or streamlines.
These data can be found in the form of HAniS loops here.
| Table 6. List and definitions of the various cell features used for the radar portion of the study. |
|
|
|
|
|
|
Sea Breeze Front (SBF) |
linear |
weak/broken (<45 DBZ) |
16 cardinal wind directions |
16 cardinal wind directions |
behind |
Outflow Boundary (OFB) |
individual |
moderate (45-55 DBZ) |
variable/stationary |
variable/stationary |
overhead |
SBF & OFB |
cluster |
Strong (>55 DBZ) |
|
|
ahead |
No SBF or OFB |
|
|
|
|
|
 |
|
The Cell Initiation (%) graph shows that a majority of the warning level cases were produced by thunderstorms which
initiated as a result of some form of mesoscale boundary interaction (SBF & OFB, OFB Only and SBF Only). On the other
hand, non-warning level winds were produced by thunderstorms that initiated without the help of any mesoscale boundary
interaction nearly 69% of the time. This shows the importance of mesoscale boundary interactions in producing stronger
thunderstorms. However, about 45% of warning level events could not be associated with boundary interactions. While the
addition of the wind barbs to the java loops did help to identify SBFs and OFBs, there is a possibility that some of these
boundaries were overlooked as they were not always picked up in the 0.5° base reflectivity data obtained from KMLB. |
| Figure 27. Most likely sources (see the first column of Table 6) of cell initiation for the convection
responsible for the maximum peak wind as indicated by KMLB radar and tower wind observations, for warning versus non-warning
periods over the warm season months of the 18-year study period. |
 |
|
Cell Structure indicates that storms that produce warning level wind gusts are better "organized" than storms that
produce non-warning level gusts. Nearly 80% of warning level gusts were produced by either a linear or cluster grouping
of cells. Considering that only about 7% of non-warning level gusts were associated with linear storms, this is fairly
significant. The majority of the non-warning level gusts were produced by storms that were somewhat less "organized",
such as a cluster of storms or an individual cell (pulse or airmass). What may help to contribute to these results is
the fact that in general, the linear storms already have a velocity associated with the direction that they are moving
whereas cluster and individual cell type storms are less likely to have a discernible direction of motion. |
| Figure 28. Most likely type of cell structure (see the second column of Table 6) associated with the
convection responsible for the maximum peak wind as indicated by KMLB radar and tower wind observations for warning versus
non-warning periods over the warm season months of the 18-year study period. |
 |
|
This shows that most of the time both warning level and non warning level events are produced by moderate strength
cells, which more commonly occur. As might be expected, a large majority of strong cell events do produce warning level
winds, whereas the opposite was true for weak/broken cells. Warning level events in the weak/broken and moderate
categories might be due to collapsing storm cloud cores generating the downdraft with the strongest winds at the surface
occurring slightly after the drop in radar reflectivity, so that the reflectivity at the time of the peak wind is less
than the reflectivity that drove the downburst. To further reinforce this theory, it was often observed that a cell would
approach a tower maintaining a moderate or strong strength and become significantly weaker just prior to the peak wind gust.
This idea was suggested by Roeder and forecasters from the 45 WS (personal communication, 2008). |
| Figure 29. Summary of observed cell strength (see column three of Table 6) of the cell most likely responsible
the maximum peak wind, as indicated by KMLB radar and tower wind observations for warning versus non-warning periods over the warm
season months of the 18-year study period. |
 |
|
The Cell Group Movement graph indicates that warning level winds are generally associated by storms that are traveling
towards the east. On the other hand, non-warning level events are more often associated with westward movement. The maximum
for the non-warning level cases occurs in the variable/stationary category. This is understandable as many of the weaker,
non-warning level cases were pulse or airmass type thunderstorms that would form and dissipate in the same general area.
According to this graph the most dominant movements can be tied to SW-1 and SW-2 flow regimes, show in Figures 18-19.
Physically, it is understandable that synoptic scale low level flow from the SW would cause storms to move in an E to NE
direction. |
| Figure 30. Direction of movement of the group of cells (see column 4 of Table 6) containing the cell
most likely responsible for producing the maximum peak wind. It is important to note that the direction is the direction
"towards" which the cell group is moving. Results are based on KMLB radar and tower wind observations for warning
versus non-warning periods over the warm season months of the 18-year study period. |
 |
|
The results that were gathered as a result of the Individual Cell Movement category were very similar to the results
that were seen in the Cell Group Movement category for the same physical reasons. For warning-level events, the movement
was just slightly more to the north than for group movement. |
| Figure 31. Same as Figure 30, but for the movement of the individual cell (see column 5 of Table 6)
in a group. |
 |
|
An overwhelming majority (84%) of warning level wind gusts occurred when the cell was located nearly directly overhead or
extremely close to the tower that recorded it. This is somewhat counterintuitive at first as downburst winds are generally thought
to propagate down-shear of the cell. However, low to mid-level winds (1000-700 hPa) are often quite weak during the Florida warm
season, which tends to lessen the horizontal propagation of winds from individual cells. This result could potentially be more
evidence supporting the collapsing core hypothesis. Frictional effects within the generally deep summertime boundary layer
likely slow winds significantly. This implies that, although a downburst may have been above warning criteria at the time of
generation, it could have slowed down to below 35 kt by the time it reached a distant (5-10 km) tower due to frictional effects. |
| Figure 32. Location of the tower reporting the maximum peak wind with respect to (WRT) the cell (see column
6 of Table 6) most likely responsible for producing the wind. Results are based on KMLB radar and tower wind observations for
warning versus non-warning periods over the warm season months of the 18-year study period. |
Limited Related Studies: During the second year of this effort, several limited studies were completed to relate convective winds with lightning density
and integrated precipitable water (IPW) from the U.S. Coast Guard Station at Port Canaveral. Another aspect studies was looking into the "lead time" between when
convective winds first started to accelerate and then reach their maximum peak speed to see whether this could be used as a tool for short-term predictions. This
section summarizes the results of those limited studies.
 |
|
Based on CG lightning data, this scatter plot shows all 837 convective periods and the reported cloud-to-ground lightning associated with them.
There were 282 of these periods with no cloud-to-ground lightning reported. There was a slight positive trend of increased lightning with stronger convective
periods with a weak correlation of r2=.19. |
| Figure 32. This scatter-plot shows the number of cloud-to-ground lightning strikes in the CCAFS/KSC area versus maximum wind speed
for all convective periods with CG data from 1995-2005. Scatter plot includes convective periods with no reported lightning. |
 |
|
The number of lightning strikes per convective period was determined for all 837 convective periods from 1995-2005. The data
clearly show that on average the stronger the convective period, the greater are the number of
lightning strikes associated with it. The < 20 kt cases, 20-34 kt cases, 35-49 kt
cases and ≥ 50 kt cases had an average of 43, 146, 438 and 679 strikes, respectively. |
| Figure 33. This graph represents the average number of
cloud-to-ground lightning strikes for each maximum wind speed category of the
837 convective periods in the 1995-2005 portion of the climatology. The sample size of convective
periods for each speed category is noted at the bottom of the bar and includes
convective periods with no reported lightning. |
 |
|
A "lead time" here is defined as the time between the first reported 20 kt wind and the first reported 35 kt wind.
It was determined for all cases with a maximum convective wind speed between 35-49 kt. Nearly 67% of the time for these 35-49 kt cases,
the "lead time" was greater than 30 minutes. Less than 18% of the time, the ramp up from 20 kt to ≥ 35 kt took 15 minutes or less. |
| Figure 34. "Lead times" for convective periods that had gusts of 35-49 knots during the warm seasons of 1995-2005. |
 |
|
For this chart, a "lead time" was defined at the time between the first reported 20 kt wind and the first reported maximum convective wind
speed ≥ 50 kt. Some 56% of the cases had "lead times" > 30 minutes and 22% of the time it was ≤ 15 minutes, leading to the
conclusion that the strongest convective wind events on average ramp up more quickly.
|
| Figure 35. "Lead times" for convective periods that had gusts of ≥ 50 knots during the warm seasons of 1995-2005.
|
 |
|
The month of May had much lower IPW values compared to the other warm season months. Excluding May, the averaged IPW
values for the < 20 kt cases, 20-34 kt cases, 35-49 kt cases and ≥ 50 kt cases
were 1.97 inches, 2.01 inches, 1.88 inches, and 1.92 inches, respectively. The variation in IPW values with the maximum wind speed
of convective periods appears somewhat helpful in distinguishing between convective cases < 35 kt and ≥ 35 kt,
which have higher and lower IPW values, respectively. A very high IPW value (i.e. ≥ 2.0 inches) appears to indicate a saturated
atmosphere that does not allow sufficient evaporational cooling to help to generate stronger winds. |
| Figure 36. This graph shows the integrated precipitable water values (averaged from the observations that were valid
approximately 3-hours prior to the initiation convection) for 95 convective periods during the warm-season months (May-September). The GPS-based
IPW data from 2000-2004 was obtained from the Coast Guard Station at nearby Port Canaveral. |
Preliminary Case Studies Using 1-minute Tower Data: Two convective cases were selected to study in more detail using 1-minute tower data.
Each of the cases were chosen because of the fairly large number of 5-minute peak wind reports greater than or equal to 35 knots. The objective
was to explore the onset, proliferation, and demise of these convective wind events. Preliminary results on the individual cases were reported
separately by Cummings (2007) and Dupont (2007) as part of their senior research projects. They present some interesting results that warrant
further investigation with more additional cases.
References:
Ander C. J., A. J. Frumkin, J. P. Koermer and W. P. Roeder, 2009: Study of sea-breeze interactions
which can produce strong warm-season convective winds in the Cape Canaveral area.
16th Conf. on Air-Sea Interaction/8th Conf. on Coastal Atmospheric and Oceanic Prediction
and Processes, January, Phoenix, AZ. DOWNLOAD
Cummings, K. A., 2007. Warm-season convective wind event for the Florida Space Coast: A case study -
23 July 2005. Senior Research paper, Dept. of Chemical, Earth, Atmospheric and Physical
Sciences, Plymouth State University, Plymouth, NH. DOWNLOAD
Cummings, K. A., E. J. Dupont, A. N. Loconto, J. P. Koermer and W. P. Roeder, 2007. An updated warm-season
convective wind climatology for the Florida Space Coastr. Preprint CD-ROM, 16th Conf. of Applied
Climatology, January, San Antonio, TX. DOWNLOAD
Dinon, H. A., M. J. Morin, J. P. Koermer, and W. P. Roeder, 2008. Convective winds at the Florida Spaceport:
year-3 of Plymouth State research. 13th Conf. on Aviation, Range, and Aerospace Meteorology,
January, New Orleans, LA. DOWNLOAD
Dupont, E. J., 2007. A case study of a warm-season convective wind period on the Florida Space Coast.
Senior Research paper, Dept. of Chemical, Earth, Atmospheric and Physical Sciences,
Plymouth State University, Plymouth, NH. DOWNLOAD
Holton, J. R., 2004. An Introduction to Dynamic Meteorology, 4th Ed.. Elsevier Press, 535pp.
Koermer, J. P., and W. P. Roeder, 2008. Assessment of the importance of certain wind towers in the
Cape Canaveral AFS/Kennedy Space Center mesonet for predicting convective winds. 13th Conf. on
Aviation, Range, and Aerospace Meteorology, January, New Orleans, LA. DOWNLOAD
Lambert, W. C., 2002. Statistical short-range guidance for peak wind speed forecasts on Kennedy Space
Center/Cape Canaveral Air Force Station: Phase I results. NASA Contractor Report CR-2002-211180. 39 pp.
DOWNLOAD
Lambert, W. C., 2007. Objective Lightning Probability Forecasting for Kennedy Space Center and Cape
Canaveral Air Force Station, Phase II. NASA Contractor Report CR-2007-214732. 59 pp.
DOWNLOAD
Laro, K. L., 2011. Updating the KSC/CCAFS Warm-Season Convective Wind Climatology. 10th Student Conference,
American Meteorological Society, January, Seattle, WA. DOWNLOAD
Laro, K. L., 2012. Warm season convective wind climatology for the CCAFS/KSC area. 2012 Posters on the Hill,
Council on Undergraduate Research, April, Washington, DC. DOWNLOAD
Lericos, T. P., H. E. Fuelburg, A. I. Watson, and R. I. Holle, 2002. Warm season lightning distributions
over the Florida Peninsula as related to synoptic patterns. Weather and Forecasting, 17, 83-98.
DOWNLOAD
Loconto, A. N., J. P. Koermer, and W. P. Roeder, 2006. An updated warm-season convective wind climatology for Cape
Canaveral Air Force Station/Kennedy Space Center. 12th Conf. on Aviation, Range, and Aerospace
Meteorology, January, DOWNLOAD
Loconto, A. N., 2006. Improvements of warm-season convective wind forecasts at the Kennedy Space Center
and Cape Canaveral Air Force Station. M.S. Thesis, Dept. of Chemical, Earth, Atmospheric and Physical
Sciences, Plymouth State University, Plymouth, NH. DOWNLOAD
Lupo, Kevin M., 2013. An Update to the Convective Wind Climatology of Kennedy Space Center/Cape Canaveral Air Force Station.
12th Annual Student Conference, American Meteorological Society, Austin, TX, 5-6 Jan., Paper S49.
DOWNLOAD
McCue, M. H., J. P. Koermer, T. R. Boucher, and W. P. Roeder, 2010. Validations and development of existing and
new RAOB-based warm-season convective wind forecasting tools for Cape Canaveral Air Force Station and
Kennedy Space Center. 22nd Conference on Climate Variability and Change, Atlanta, GA, 18-21 Jan., Paper P2.16.
DOWNLOAD
McCue, M. H., 2010. Validations and development of existing and warm-season convective wind forecasting tools
for Cape Canaveral Air Force Station and Kennedy Space Center. M.S. Thesis, Dept. of Atmospheric Science and
Chemistry, Plymouth State University, Plymouth, NH. DOWNLOAD
Rennie, J. J., J. P. Koermer, T. R. Boucher, and W. P. Roeder, 2010. Evaluation of WSR-88D methods to predict
warm-season convective wind events at Cape Canaveral Air Force Station and Kennedy Space Center.
22nd Conf. Climate Variability and Change, Amer. Met. Soc., Atlanta, GA, 18-21 Jan., Paper P2.17.
DOWNLOAD
