xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID EG9B02SAL_4VJYL8 INGEST_TIME 2018-05-18T18:47:57Z PACKAGE UF00091523_00853
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
University of Florida | Journal of Undergraduate Research | Volume 1 9 Issue 2 | Spring 201 8 1 Analysis of the Relationship between Flooding Potential of Landfalling Tropical Cyclones and their Size and Forward Speed Alex DesRosiers and Corene Matyas College of Liberal Arts and Sciences, Department of Geography, University of Florida Tropical cyclones striking land often bring heavy rainfall which can last for a long duration and result in flooding events along the path of the storm. The size and forward speed of the storm has been shown to affect the amount of flooding caused. This study focuses on rainfall and related flood events caused by hurricanes Charley and Frances from the 2004 season. Both storms made landfall ov er Florida, but were of different intensities and sizes and took different trajectories aft er landfall. Instances of flooding events from the National Centers for Environmental Information Storm Events Database and flood warnings iss u ed by the National Weather Service are used to identify possible flood locations. The locations are analyzed with actual rainfall totals and timing obtained from a database that combines radar estimates with rain gauge observations. was expected to cause more flooding. The results show that Frances had more flood events and warni ngs. Frances also had a significantly greater magnitude and duration of flooding in the tested areas. However, rainfall rates were not greater in flood warning areas during Frances likely due to the tested events proximity intensity. The results of this study encourage future work to identify the relationship of tropical cyclone size and speed on flooding potential after landfall INTRODUCTION ropical cyclones are a grandiose force of nature that occur annually during summer and fall in the North Atlantic basin. Although they form and intensify over the ocean, they can also cross over land and cause much destruction through fast winds, storm surge, and heavy rainfall. D uring these destructive events that the preservatio n of life and property are of the utmost concern. A determining factor in the evacuation behaviors during hurricanes in the US is the intensity of the storm on the Saffir Simpson Hurricane wind scale (Whitehead et al. 2001). This scale uses wind speed to c ategorize hurricanes based on destructive potential (Lakshmi 2006). The categories range from a tropical depression up to a C ategory 5 hurricane. Since this scale only incorporates wind speed, it fails to evaluate all factors that lead to the destructive p otential of a landfalling hurricane event. The potential for heavy rainfall Table 1. Measurements of Charley and Frances Factors such as size of the storm and its forward speed can play an important role in the potential for damange of these storms as they move over land (Rego and Chunyan 2009). Larger tropical cyclones can bring higher storm surge (Irish, Resio and Ratcliff 2003 2013) and more rainfall to a wider area (Matyas 2014). A tropical cyclone that moves quickly can produce a higher amount of wind related damage and storm surge, but rainfall totals tend to be lower. Although rainfall from tropical cyclones is respons ible for a high amount of deaths in the U.S. (Rappaport 2014), a study done in the aftermath of Hurricane Rita found that the relationship between risk perception of inland flooding and evacuation action was insignificant (Stein, Dueas Osorio and Subrama nian 2010). Rather, evacuation potential has been found to be related to storm intensity (Whitehead et al. 2001). In this study, the relationship among storm intensity, forward speed, size, and rainfall amounts will be examined using two storms from the 20 04 hurricane season, Charley and Frances. Charley was a fast moving storm with Category 4 intensity and a smaller circulation. Frances was double the size of Charley and crossed the Florida coastline at Category 2 intensity moving more slowly over land. T h ese differences are shown below. This study will look at locations where flooding did occur, according to reports after the storm or was forecast to occur according to the National Weather Service. Rainfall events will be analyzed in terms of their durat ion, magnitude, and T
A LEX D ES R OSIERS AND C ORENE M ATYAS University of Florida | Journal of Undergraduate Research | Volume 19 Issue 2 | Spring 2018 2 rainfall rate. The differences in size and speed between these storms should result in Frances having produced more significant flooding events over land. DATA AND METHODS The first step was to determine where flood events occurred,. P olygons representing flood warnings issued by the National Weather Service were obtained from the Iowa Mesonet (IEM 2017) and used to identify the regions that received heavy rainfall. Since these polygons cover the entire country, the first task was to de termine which had been triggered by the hurricanes. Polygons that bounded the regions receiving at least 12.5 mm of rainfall from each hurricane were used (Zhou and Matyas 2017). Both the flood warning polygons and hurricane rainfall polygons were imported into a Geographic Information System (GIS) and any flood warnings that fell outside of the rainfall boundary were removed from the study. To further confirm that the flood warnings occurred while rainfall was being produced by the hurricane, a visual ins pection was performed by examining the Wunderground during the time when the warnings were occurring.Overall, 19 flood warnings were issued during the passage of Hurricane Charley, and 180 flood warn ings were issued during the passage of Hurricane Frances. Regardless of whether a flood warning was issued, it was necessary to document where and when actual flood events occurred. The NCEI Storm Events Database (NCEI 2017) houses information on inclemen t weather events taking place across the U.S. Spanning multiple decades and containing key information such as when, where, and what kind of event took place, this dataset was useful for identifying confirmed flood events. Using the timing and location of track, satellite imagery from the Wunderground hurricane archive, and descriptions of the events within the database, we identified 108 total flooding events that occurred due to each storm. Using the geographic coordinates representing the cen troid of each event, all events were imported into the GIS and represented as points. To be sure that these events occurred within the region of rainfall produced by each storm, these points were also compared to the rainfall polygons developed by Zhou and Matyas (6). Eight flood events caused by rainfall during Hurricane Charley and 100 during Hurricane Frances were identified The main limitation of the Storm Events Database was that no reports were available for some areas which were known to have sustained damage from these storms. For example, in the immediate area around Punta Gorda, Florida landfall, there were no event s of any kind recorded, including those for wind damage. However, Punta Gorda is the location where the eye made landfall and the location of several damage reports from the storms including hospitals going out of service and roofs being torn off (Orlando Sentinel 2004). Thus, it was important to consider both the flood warnings and storm event reports since some flood reports are likely missing, and some flood events may not have received a warning. Stage IV data (EOL 2017) were used to determine the rain rates that occurred in association with each storm. This dataset consists of gridded hourly rain rate values that were derived from ground based weather radars and corrected using rain gauge observations. One advantage of using this dataset over other rai nfall datasets is that it benefits from manual quality control measures implemented by river forecast centers before it becomes publicly available. The Stage IV data for the storms in this study were summed from the 12 hours prior and after each 6 hourly s torm center location. The Analysis Process The next part of the analysis determined the timing of the flood warnings and events relative to the passage of the circulation center of each storm. The time of the 6 hourly track position was taken as a tempor al center point. All warning and events from the storm were then assigned to a 6 hourly center track position. If the starting time of the event or warning fell within the three hours prior or the three hours after of a center point, they were grouped tog ether irrespective of their location relative to the storm center. This allowed each 24 hour set of Stage IV data that were centered around the 6 hourly center track positions to then be used to analyze events that began around that time. Once storm event s and warnings were assigned to their respective center points, the amount of rainfall per each hour in the 24 hour period surrounding the center location time to which they belonged was calculated. Within the GIS, each storm event point was assigned to th e value of the rainfall grid cell in which it resided. The warning polygons were slightly more complex because they are an area instead of one single point in space that only overlapped a single pixel value. The average value of all pixels falling within t he polygon as a whole was calculated to represent rainfall within the warning polygon. To facilitate a faster analysis, a python script was written to compile the data and export values to a spreadsheet. A model of this process is summarized by the flow ch art in F igure 1. Statistical Analysis To determine the duration of each rainfall event, the number of hours that rainfall occurred out of the 24 hour s analyzed was recorded. Rainfall totals for each event were created by adding all hourly values over the 24 hour period. The average rainfall rate was computed by averaging the rainfall experienced per hour over the 24 hour s. Once these numbers had been created for all warnings and events, they were grouped based on whether they were caused by Charley or Fran ces.
A NALYSIS OF THE R ELATIONSHIP BETWEEN F LOODING P OTENTIAL OF L ANDFALLING T ROPICAL C YCLONES AND THEIR S IZE AND F ORWARD S PEED University of Florida | Journal of Undergradua te Research | Volume 1 9 Issue 2 | Spring 2018 3 These groups of data were then ready for statistical analysis. The magnitude, or rainfall totals, for all flood warnings grouped from each storm were tested against each other using a Wilcox Rank Sum test. The significance level alpha was set to 0.05 and the means were tested for a statistically significant difference by setting the null hypothesis for the mean differences between the two groups to be zero. This same test was d one for the duration and rainfall rate of flood warning areas. The statistical tests were then repeated for the NCEI events. RESULTS AND DISCUSSION The statistical analysis done on flood magnitude, duration, and precipitation rates show ed a significant d ifference between the two hurric anes in all comparisons (Table 2 ). In the parameters of flood duration and magnitude, Frances had significantly more hours of reported rainfall in a 24 hour period and greater rainfall accumulation totals. This observed difference is in support of the hypothesis that Frances as the slower and larger storm would have a higher potential for flooding. Frances also had a greater rainfall rate at NCEI flood event locations. However, Charley had greater rainfall rates in the flood warning areas which was likely due to the fact that Charley was a more intense storm than Frances. Precipitation rates have been linked to tropical cyclone intensity, especially in the core region of the storm (Knutson and Tuleya 2004). As shown in F were located closer to the center of the storm than those of Frances, possibly causing this difference in rainfall rates. Table 2 Results of Statistical Comparisons Frances has a radius of outermost close d isobar at landfall that is double that Charley. Frances is also close to half the average land speed of Charley. A larger and more slowly moving tropical cyclone should produce more rainfall over a larger area and the results show that Frances produced l onger durations of rainfall and higher totals in both the warned areas and at the event locations. Frances also Figure 1 Schematic of the Analysis Process
A LEX D ES R OSIERS AND C ORENE M ATYAS University of Florida | Journal of Undergraduate Research | Volume 19 Issue 2 | Spring 2018 4 produced more flood warnings and events than Charley Distribution. Thus, the findings of this study support the hypothesized relationship and previous research on rainfall distributions in landfalling tropical cylones. These findings not only support the need for future research on other tropical cyclones, but can be used for emergency preparedness. A previous study showed inland flooding risk perception and evacuation behavior to not be strongly correlated. However, the researchers did fi nd a positive correlation between local media and evacuation behavior (Stein, Dueas Osorio and Subramanian 2010). According to the results of the study, local media should work especially hard to warn people of the dangerous flooding that can occur as a l arge and slow moving tropical cyclone moves inland. CONCLUSION This study sought to demonstrate the relationship between size and average forward speed of a landfalling hurricane to its flooding potential. A spatial analysis of rainfall, flood warnings, and reported flood events for hurricanes Charley and Frances were conducted using a GIS. Frances was a large and slow moving storm that tracked inland. It produced more flood events and events had a longer duration and higher rainfall totals. However, aver age rain rates were higher for the events that occurred during Hurriane Charley, which could be due to the stronger winds in this storm converging more moisture to fall as precipitation. There were limitations to the data being used in this study, the most notable being the omission of several areas from the S torm E vent s D ataba se that should have experienced inclement weather events as a result of these storms. The averaging of rainfall within a warning polygon did not allow for the highest rain rates to be explored, which may have contributed most to the flood event. The fact that storm center positions are only available every six hours means start or issuance of the warning cannot be determi ned. The relationships between size, forward speed, and flooding potential should continue to be examined given the results of the current study. A large enough samp le of storms exhibit ing these relationships could be useful information to disseminate to t he public with future storms which could result in increasing public perception of hurricane flood risk ACKNOWLEDGMENTS Thank you to the National Science Foundation for funding this research with grant B CS 1053864. Thank you to the University Scholars Program at the University of Florida for funding me in this project. Also, a special thanks to my advisor Dr. Corene Matyas. REFERENCES Demuth, J., M. DeMaria, and J.A. Knaff ( 2006 ). Improvement of advance d microwave sounder unit tropical cyclone intensity and size estimation algorithms. J. Appl. Meteor, 45 1573 1581. EOL data archive GCIP/EOP s urface: Precipitation NCEP/EMC 4KM gridded data (GRIB) stage IV d ata ( 2016 ). Retrieved from Data.eol.ucar.edu Hurricane And Tropical Cyclone Activity ( 2016 ) Weather Underground Figure 2 A few of the 6 hourly center locations and their related watches and warnings of Charley (left) and Frances (right )
A NALYSIS OF THE R ELATIONSHIP BETWEEN F LOODING P OTENTIAL OF L ANDFALLING T ROPICAL C YCLONES AND THEIR S IZE AND F ORWARD S PEED University of Florida | Journal of Undergradua te Research | Volume 1 9 Issue 2 | Spring 2018 5 IEM : NWS watch warning archive download ( 2016. ) Iowa Environmental Mesonet, 2017. Retrieved from Mesonet.agron.iastate.edu Irish, Jennifer L., Donald T. Resio, and Jay J. Ratcliff. (2008) The influence of storm size on hurricane surge Journal of Physical Oceanography 38 ( 9 ) 2003 2013. Kantha, L. (2006 ). Time to replace the Saffir Simpson hurricane scale ?" Eos, Transactions American Geophysical Union 87 ( 1 ) 3. Knutson, T. R., and Tuleya R E (2004) Impact of CO2 induced warming on simulated hurricane intensity and precipitation: sensitivity to the choice of climate model and convective parameterization. Journal of Climate 1718 3477 3495. Matyas, C. J. ( 2010 ). Associations between the size of hurricane rain fields at landfall and their surrounding environments. Meteorol. Atmos. Phys. 106, 135 148. Orlando Sentinel ( 2004 ). List of known damage caused by Hurricane Charley Rappaport, E. N. ( 2014 ). Fatalities in the United States from Atlantic tropical cyclones new data and interpretation. Bull. Amer. Meteor. Soc. 95 341 346. Rego, J .L. and Li C ( 2009) On the importance of the forward speed of hurricanes in storm surge forecasting : A numerical stud y Geophysical Research Letters 36 ( 7 ). Stein, R M., Dueas Osorio, L., and Subramanian D (2010) Who evacuates when hurricanes approach? T he role of risk, information, and location Social Science Quarterly 91 ( 3 ), 816 834. Storm Events Database ( 2014 ) National C enters for Environmental Information. Whitehead, J C. et al. (2001) Heading for higher ground : Factors affecting real and hypothetical hurricane evacuation behavior Environmental Hazards 2 (4) 133 142. Zhou, Y. and Matyas, C. J. ( 2017. Spatial char acteristics of storm total rainfall swaths associated with tropical cyclones over the eastern United States. International Journal of Climatology, DOI : 10.1002/joc.5021.