PROJECT TITLE:



Senior Capstone: Forecast Verification of Tropical Cyclones in the

North Atlantic Ocean (provisionary title)

Nicole Light

School of Meteorology, University of Oklahoma, Norman, Oklahoma

Stephanie Malone

School of Meteorology, University of Oklahoma, Norman, Oklahoma

Max Perron

School of Meteorology, University of Oklahoma, Norman, Oklahoma

14 February 2008

Abstract

(Coming soon…)

1. Introduction

Forecasting tropical cyclones has been a part of meteorology for over 100 years. With the exponential increase in coastal populations, the challenge of predicting a tropical cyclone’s projected path and intensity has taken a proportionally important role. Now, several decades after the first weather satellites were launched, some facets of this field have advanced far less than others. A November 2003 NCAR news release mentioned that though track forecast accuracy has steadily increased by 1-2% since the 1960s, almost no progress has been done in the areas of tropical cyclone intensity and precipitation forecasts (Hosansky and Gillis 2003). This project will attempt to determine what meteorological factors lead tropical forecasters and computer models astray. We will focus on tropical cyclone intensity rather than track or precipitation.

Since the first encounters with hurricanes in the Caribbean, it was hypothesized that these storms could only survive in the tropics. By the 1950s, meteorologists had determined that tropical cyclones were born over areas with warm sea surface temperatures (SSTs) and that this warm water was a source of energy for the storm. Edwin Fisher (1958) observed that tropical cyclones left behind cold pools of water, leading to the dissipation of extremely slow-moving or stationary ones, and that they seemed to follow areas of high SSTs in the absence of other forcings (now shown to be false.) More recently, it was shown that warm core rings shed by the Loop Current in the Gulf of Mexico could significantly strengthen tropical cyclones (Shay et. al 2000). As such, the depth of warm water is also quite important when forecasting the intensity.

Another important factor that limits the intensity of a tropical cyclone is the nature of the atmospheric flow in and around it. Previous research in vertical wind shear patterns inside tropical cyclones suggest that a high amount will tilt the storm vortex horizontally and tend to make the circulation quite disorganized, reducing the intensity of the storm. Wind shear also acts to disturb heat and moisture patterns at the mid-levels, which in turn tend to inhibit convective movement inside the tropical cyclone (DeMaria 1996). Merrill (1988) has also suggested that upper-level wind patterns in the vicinity of a tropical cyclone could have an effect on its intensity.

There are certain processes that affect the intensity of a tropical cyclone sinusoidally. One is the eyewall replacement cycle, caused when turbulence in the eye of a tropical cyclone causes it to collapse while a larger eye forms around it (Black and Willoughby 1992). Another is the solar diurnal cycle, in which radiation hitting the storm’s cloud tops during the day results in less intense convection (Hobgood 1986). Both of these processes only affect the intensity of a storm on a daily basis and have little effect on long-term forecasts.

Meteorologists use a variety of tools in order to accurately forecast the progress of tropical cyclones. An analysis of the performance of several models during Hurricane Hugo showed that no one model consistently performed better than the others during the duration of the study. However, it was found that models perform worse when Hugo was located in regions where atmospheric and oceanic measurements were scarce. The models performed best as Hugo approached the baroclinic zone over the United States (Ward 1990). This shows us that the amount of measurement sites in the tropical Atlantic and our knowledge of barotropic environments are crucial to improving future forecasts. A summary of parameters used in the NCEP Global Model in 1995 is found in Surgi et al.’s 1998 paper.

2. Data Sources

We received our sea surface temperature data from the Unisys Weather[1] website, the NOAA AMOL Heat Content[2] website, and from the Remote Sensing Systems website.[3]

The Unisys Weather website includes an archive of sea surface temperatures averaged over a week of time for the world. The NOAA AMOL Heat Content website includes a tropical cyclone heat potential loop, a sea height anomaly, the depth of the 26 degree Celsius isotherm, and the sea surface temperature loop. Finally, the Remote Sensing Systems website includes daily sea surface temperatures around the world. These measurements were taken at 8am local time every day.

We analyzed synoptic charts from the National Center for Environmental Prediction, Hydrometeorological Prediction Center’s website[4], which allowed us to look at and compare the synoptic set-ups over the United States and the Gulf of Mexico during the time periods of all four hurricanes. Information was also taken from the National Hurricane Center’s Tropical Cyclone Reports[5], which are completed for every tropical storm and hurricane that occur in a season. These reports included a synoptic summary of each event and a forecast and warning critique, where forecast errors where compared to average errors.

The National Hurricane Center Forecast Verification site[6] was used as well, to find the values of the official intensity errors for each hurricane.

As for wind data, we found 850 mb and 200 mb streamline analyses for the tropical Atlantic from Colorado State University’s archives.[7] We also have surface wind field data from the NOAA’s Hurricane Research Division.[8]

3. Methodology

To begin, we decided to analyze tropical cyclones that affected the United States after the year 2000. After going through the National Hurricane Center’s verification reports, we picked out three hurricanes that were poorly forecasted (especially in intensity) and one that was remarkably well forecasted. It turned out that our control storm was Hurricane Claudette, which made landfalls in the Yucatan Peninsula and Texas in August 2003. Our three tropical cyclones of interest were Alex, which flirted the Carolina coast in 2004, Katrina, the famous New Orleans disaster in 2005, and Ernesto, which affected the eastern seaboard in 2006.

After going through each forecast for a particular tropical storm, we found the ones that were wrong and searched for unusual patterns in each of the data sets mentioned earlier (sea surface temperatures, wind shear, synoptic features, etc.)

We looked at the sea surface temperatures, the heat potential, and the depth of the 26 degree Celsius isobar. We then looked at the path that the hurricane or tropical storm took and saw the conditions it was in when it either increased or decreased in intensity. For instance, we looked at whether the storm was in an area of high heat potential when it increased in intensity, or if the storm continued to increase in strength despite traveling in relatively cool water.

We then looked at the 850 mb and 200 mb streamline data for every twelve-hour period during the tropical cyclone’s life. Of particular interest were areas of substantial shear surrounding the storm and times when the upper level organization of the storm was favorable for strengthening.

Analyzing the synoptic conditions was done by looking at the synoptic charts from each hurricane. Charts were available for each hurricane as it entered into the Gulf of Mexico and approached the coast of the United States. We used 500-mb and surface maps to look at the locations of high and low pressure systems, which helped us to verify the storm tracks. Verifying the track data helped us to see why some of the intensity forecasts were made, and why some of them were under or over-forecasted.

Finally, we looked at satellite imagery to find signs of dry air entrainment, eyewall replacement cycles, upper-level divergence, and overall storm organization.

4. Results

a. Hurricane Claudette

Tropical Storm Claudette formed from a tropical wave on 8 July 2003 south of the Dominican Republic. During the first few days, it traveled westward toward the Yucatan Peninsula. Although situated over very warm waters, Claudette struggled to become organized because of the high-shear environment it was located in. Just before making landfall in the Yucatan, Claudette suddenly became very disorganized and its center became hard to locate. As it arrived into the Gulf of Mexico, Claudette featured constant 45 to 50 knot winds until 14 July. The storm entered an area of more favorable wind shear and became a category 1 hurricane just before hitting Texas late on 15 July.

On 11 July 2003 Hurricane Claudette’s primary circulation was overtaken by a secondary circulation lying west of the cyclone. This was a result of dry air entrainment and wind shear. Hurricane Claudette was a hurricane that was well forecasted throughout the entirety of its life. We looked at Claudette first, and used it as a control for all of the other hurricanes that we researched.

b. Hurricane Alex

Tropical Depression Alex began 31 July 2004 off the coast of South Carolina. It increased intensity to a tropical storm on 1 August, where it slowly crept up the coast of South Carolina. It then became a category 2 hurricane on 3 August as it traveled north and came close to making landfall on North Carolina. After that, Alex began to veer more into the ocean as it traveled northward. Hurricane Alex was the first major hurricane to form north of 35 degrees north, and became a category 3 hurricane on 5 August. Alex continued its hurricane status until 6 August when it traveled over cold water and reduced strength to a extratropical cyclone.

The intensity of Hurricane Alex was under forecasted throughout most of its life. On 3 August 2004, Hurricane Alex was forecasted to weaken due to its movement over cooler sea surface temperatures and increased wind shear. However, the low level trough to the west of Alex dug, advecting warm air into the low to mid levels. This warmed the column, which in turn amplified the upper level ridge. This in turn created difluence aloft in the southeast quadrant of the storm. Hurricane Alex did not weaken in intensity as expected due to this difluence aloft. From 4 August into 5 August, Alex was still forecasted to decrease in intensity due to its movement over cooler sea surface temperatures. Yet, Alex was embedded in the westerlies and its shear decreased. It was also following a warmer than usual Gulf Stream, so it did not weaken as predicted. Consequently, Hurricane Alex was the strongest hurricane ever that formed over 35 north latitude. On 6 August models believed that Alex would deintensify slowly. However, Alex moved over extremely cold water and died of rather quickly.

The initial long-term forecasts expected wind shear to completely destroy Hurricane Alex, yet it survived. Then, after the first half of its life, long-term forecasts expected relatively cooler sea surface temperatures to destroy Alex, yet Alex survived over sea surface temperatures in the lower twenties. It wasn’t until Alex moved over waters at 20 degrees Celsius that it was killed off.

c. Hurricane Katrina

Hurricane Katrina began as Tropical Depression 12 near the Bahamas on 23 August 2005. The depression traveled northwestward towards Florida, becoming a tropical storm on 24 August, and a category 1 hurricane on 25 August, shortly before making landfall on the south Florida coast. As Katrina tracked across the southern tip of Florida, it weakened to tropical storm strength, but satellite shows that the eye became better defined. The storm was over land, mostly the moist Florida Everglades, for only a total of six hours, and quickly re-intensified into a hurricane once it re-entered the warm waters of the Gulf of Mexico. As Hurricane Katrina steered west-southwest, it intensified to a category 3 on 27 August. On 28 August, Katrina underwent rapid intensification to a category 5 hurricane as it continued to track across the Gulf towards the southeast coast of the United States. A few hours before making landfall on the Mississippi Delta on 29 August, Katrina underwent rapid de-intensification from a category 5 hurricane to a category 3, and made landfall at that strength. It made a second landfall on the coast of Mississippi. After landfall, the storm rapidly weakened to a category 1 by 1800 Z, and was further reduced to tropical storm status as it moved northeast over the Tennessee Valley. The storm was classified as a tropical depression by 1200 Z on the 30th of August, and underwent extra-tropical transition by 00 Z on August 31st.

The intensity of Hurricane Katrina was under forecasted the majority of its life. It was not until it hit land that it was over forecasted. From 26 August to 27 August, Katrina had been forecasted to move straight west into Florida. However, Katrina dipped southwest into the Florida Everglades and stayed inland for only a few hours. This unexpected path was due to the strengthening ridge over the southeastern portion of the U.S. As Katrina exited Florida, upper level difluence became significant, which resulted in Katrina’s intense strengthening. On 28 August the loop current intensified the storm by 50 knots in 12 hours, which was not forecasted. From 29 August to 30 August, the intensity of the storm was over forecasted. Dry air entrainment and a disruption of the eyewall replacement cycle resulted in weakening of Katrina.

Long-term forecasts of Katrina were incorrect because Katrina was forecasted to continue straight through Florida, and then once making its way into the Gulf of Mexico steer north and hit the Florida panhandle.

d. Hurricane Ernesto

Ernesto began as a tropical wave just west of the Lesser Antilles on 24 August 2006. By the end of the next day, it was upgraded to a Tropical Storm, as it was located in the center of the Caribbean Sea. It slowly strengthened over the next 48 hours and had sustained winds of 65 knots at 12 Z 27 August. Several hours later, it traveled over the mountainous terrain of Haiti’s southwestern tip and subsequently moved over much of Cuba’s length. This brought Ernesto back to tropical storm status as it exited into the Straights of Florida and made its third landfall in South Florida. Eventually, it traveled over the Gulf Stream and regained some of its strength, but made landfall yet again on 1 September in North Carolina.

The intensity of Hurricane Ernesto was briefly under forecasted in the beginning of its life, but then for the majority of its life was over forecasted. On 27 August, upper level outflow became better organized, resulting in a quick intensification that was not forecasted. From 18 Z 27 August to 12 Z 28 August Ernesto was weakened by mountains in Haiti, which was not forecasted. On 31 August Ernesto was forecasted to remain consistent with its intensity due to the forecaster’s caution and desires to remain consistent in their forecasts. Forecast discussions mentioned the possibility for strengthening, yet wanted to take a cautious approach. But due to warm waters and low wind shear, Hurricane Ernesto strengthened on 31 August.

The track of Ernesto was influenced by a shortwave trough moving into the southeast portion of the U.S., allowing Ernesto to move north into Florida and then into the Atlantic. The low pressure system moving east-southeast confused models on the path of Ernesto. With the projected path being over water, and not moving into Florida, the result was an over forecast of Ernesto’s intensity.

5. Conclusions

In conclusion, there are several aspects of forecasting hurricanes that need to be improved upon. When there abrupt changes in the sea surface temperatures or any variable of the environment, the models do not forecast these changes well. Also, potential heat and the depth of warm water were a major factor in the abrupt intensification of Hurricane Katrina over the Gulf of Mexico. The potential heat and depth of the warm water should be better coded into the models.

Also, many long-term forecasts are extremely inaccurate, and depend a lot upon the track that the hurricane takes. Many cases of intensifications in hurricanes that go unexplained are a result of upper level patterns. As in Hurricane Ernesto, upper level patterns steered the track of the storm, and because of this error in track forecast, the intensification was also poorly forecasted.

As seen in Hurricane Alex, hurricanes can thrive even when they are embedded in the westerlies if tropical conditions are present.

Lastly, water vapor satellite images are a key tool to utilize when forecasting the intensity of a hurricane. The presence of dry air is quite important. As seen in Hurricane Claudette, dry air entrainment will kill off a hurricane quite quickly.

References

Black, M. L., and H. E. Willoughby, 1992: The Concentric Eyewall Cycle of Hurricane Gilbert. Monthly Weather Review, 120, 947-957.

DeMaria, M., 1996: The Effect of Vertical Shear on Tropical Cyclone Intensity Change. Journal of the Atmospheric Sciences, 53, 2076-2088.

Fisher, E. L., 1958: Hurricanes and the Sea-Surface Temperature Field. Journal of the Atmospheric Sciences, 15, 328-333.

Hobgood, J. S., 1986: A Possible Mechanism for the Diurnal Oscillations of Tropical Cyclones. Journal of the Atmospheric Sciences, 43, 2901-2922.

Hosansky, D., and C. Gillis, 2003: 2003 Hurricane Season: USWRP Research Led to More Accurate Track Forecasts. NCEP News Release 2003-48. (Available online at .)

Merrill, R. T., 1988: Environmental Influences on Hurricane Intensification. Journal of the Atmospheric Sciences, 45, 1678-1687.

Shay, L. K., G. J. Goni, and P. G. Black, 2000: Effects of a Warm Oceanic Feature on Hurricane Opal. Monthly Weather Review, 128, 1366-1383.

Surgi, N., H.L. Pan, and S. J. Lord, 1998: Improvement of the NCEP Global Model over the Tropics: An Evaluation of Model Performance during the 1995 Hurricane Season. Monthly Weather Review, 126, 1287–1305.

Ward, J. H., 1990: A Review of Numerical Forecast Guidance for Hurricane Hugo. Weather and Forecasting, 5, 416–432.

-----------------------

[1]

[2]

[3]

[4]

[5]

[6]

[7]

[8]

-----------------------

Fig. 1. Infrared images of the tropical Atlantic Ocean at 2215 Z 10 July 03 (left) and 1015 Z 11 July (left). Tropical Storm Claudette is located just east of the Yucatan Peninsula.

Fig. 4. Three-day Ernesto track forecasts from 2 pm EDT 26 August 2006 (top), 2 pm EDT 27 August 2006 (middle), and 2 pm EDT 28 August 2006 (bottom).

Fig. 2. Sea surface temperatures of the tropical Atlantic Ocean on 31 July 2004. The track of Hurricane Alex is overlaid in black.

Fig. 3. Depth of the 26-degree Celsius isotherm (left) and tropical cyclone heat potential (right) for the Gulf of Mexico on 28 August 2005. The path of Hurricane Katrina is overlaid in black.

Table 1. Maximum winds and official National Hurricane Center forecasts for (a) Claudette, (b) Alex, (c) Katrina, and (d) Ernesto (in knots.) The mean absolute error for each forecast period is calculated on the right side. The color shadings represent the error of each forecast, where blue is an underforecast and red is an overforecast, and progressively darker shades indicate a larger error. Green text represents good forecasts (within 5 knots), with bolded forecasts being perfect.

................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download