<%BANNER%>

Improving Rip Current Forecasting Techniques for the East Coast of Florida

HIDE
 Title Page
 Dedication
 Acknowledgement
 List of Tables
 List of Figures
 Abstract
 Introduction
 Rip currents
 Forecasting
 Data
 Development of improved rip current...
 Use of index as forecasting...
 Summary and conclusions
 Appendices
 References
 Biographical sketch
 

PAGE 1

IMPROVING RIP CURRENT FORECASTING TECHNIQUES FOR THE EAST COAST OF FLORIDA By JASON ROLLAND CUMMINS A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2006

PAGE 2

Copyright 2006 by Jason Rolland Cummins

PAGE 3

To Arthur and Mildred Cummins.

PAGE 4

ACKNOWLEDGMENTS I would first like to thank all of my family, for their unconditional love and support throughout my college career. I thank Dr. Robert Thieke for giving me the opportunity to work on such an interesting project. His knowledge and guidance were always made available to me. I also thank Dr. Andrew Kennedy and Dr. Ashish Mehta, for their participation on my supervisory committee, as well as their insight into the research. I extend my appreciation to the Florida Sea Grant Program, for their financial support. I thank Oceanweather Inc., for liberally supplying the hindcast wind and wave conditions needed for the project. I also thank the Volusia County Beach Safety Division, for providing the lifeguard rescue information. iv

PAGE 5

TABLE OF CONTENTS page ACKNOWLEDGMENTS .................................................................................................iv LIST OF TABLES ............................................................................................................vii LIST OF FIGURES ...........................................................................................................ix ABSTRACT .......................................................................................................................xv CHAPTER 1 INTRODUCTION........................................................................................................1 2 RIP CURRENTS..........................................................................................................4 Formation of a Rip Current...........................................................................................4 Characteristics of a Rip.................................................................................................8 3 FORECASTING.........................................................................................................11 4 DATA.........................................................................................................................16 Site Description..........................................................................................................16 Rip Current Rescues...................................................................................................17 Hindcast Data..............................................................................................................18 Tides...........................................................................................................................19 WAVEWATCH III.....................................................................................................19 Lifeguard Observations..............................................................................................20 Completed ECFL LURCS Worksheets......................................................................20 Melbourne Beach and NOAA Data Buoys.................................................................20 5 DEVELOPEMENT OF IMPROVED RIP CURRENT INDEX................................23 Analysis......................................................................................................................23 Ocean Correlations..............................................................................................27 Wave height..................................................................................................30 Wave period.................................................................................................32 Wave direction.............................................................................................34 Low tide........................................................................................................37 v

PAGE 6

Testing.................................................................................................................39 Trial 1: Extraction of the wind factor...........................................................43 Trial 2: Inclusion of a wave direction factor................................................45 Trial 3: Modification of the swell period factor...........................................47 Trial 4: Redevelopment of the tidal factor...................................................48 High-risk examination..................................................................................50 Summary.....................................................................................................................53 6 USE OF INDEX AS FORECASTING TOOL...........................................................58 Analysis......................................................................................................................58 Summary.....................................................................................................................67 7 SUMMARY AND CONCLUSIONS.........................................................................70 APPENDIX A WAVE CONDITIONS AND RIP CURRENT RESCUES........................................75 B RIP CURRENT THREAT VALUES AND RESCUES.............................................83 C DAILY SUMMER THREAT VALUES, DAILY RESCUE TOTALS AND RESULTING PERFORMANCE STATISTICS........................................................91 D DAILY SUMMER THREAT VALUES, DAILY RIP TOTALS AND RESULTING PERFORMANCE STATISTICS........................................................98 LIST OF REFERENCES.................................................................................................105 BIOGRAPHICAL SKETCH...........................................................................................108 vi

PAGE 7

LIST OF TABLES Table page 5-1 Percentage of rip-related rescues occurring in the summer (defined as day 75250), 1998 to 2003...................................................................................................25 5-2 Performance results of the modified index after trial 1 (extraction of the wind factor), and the original ECFL LURCS averaged over the years from 1998 to 2003 (excluding 2002). A positive % change corresponds to an improvement over the ECFL LURCS............................................................................................44 5-3 Performance results of the modified index after trial 2 (inclusion of wave direction), and the original ECFL LURCS averaged over the years from 1998 to 2003 (excluding 2002). A positive % change corresponds to an improvement over the ECFL LURCS............................................................................................46 5-4 Performance results of the modified index after trial 3 (modification of the swell period factor), and the original ECFL LURCS averaged over the years from 1998 to 2003 (excluding 2002). A positive % change corresponds to an improvement over the ECFL LURCS......................................................................47 5-5 Performance results of the modified index after trial 4 (redevelopment of the tidal factor), and the original ECFL LURCS averaged over the years from 1998 to 2003 (excluding 2002). A positive % change corresponds to an improvement over the ECFL LURCS............................................................................................49 5-6 The percentage of the high rip current rescue and incident days for each year from 1998 to 2003. As well as the average of all the years weighted by the number of rips/rescues for each year........................................................................51 6-1 Performance results of the WAVEWATCH III forecasting the daily rip current threat levels for the summer of 2005........................................................................60 6-2 The WAVEWATCH III forecasted wave conditions and associated rip current threat levels, as well as the threat levels documented by the National Weather Service on the ECFL LURCS worksheets for the high-rescue days in the summer of 2005 (w corresponds to a weekend and subsequent decrease in the warning threshold)....................................................................................................63 vii

PAGE 8

6-3 Performance results of the WAVEWATCH III forecasting the daily rip current threat levels for the summer of 2005, after the modification to the multiplicative direction factor.........................................................................................................65 6-4 The WAVEWATCH III forecasted wave conditions and associated rip current threat levels (after modification to the multiplicative direction factor), as well as the threat levels documented by the National Weather Service on the ECFL LURCS worksheets for the high-rescue days in the summer of 2005 (w corresponds to a weekend and subsequent decrease in the warning threshold).......66 viii

PAGE 9

LIST OF FIGURES Figure page 2-1 Rip current and its component parts (feeders, neck and head) as well as the commonly associated currents (represented by arrows) (from Shepard et al. 1941)...........................................................................................................................5 2-2 Profile of the mean water level and the envelope of the wave height for a typical experiment. Wave period, 1.14 sec; H o = 6.45cm; H b = 8.55cm; tan = 0.082 (from Bowen et al. 1968)...........................................................................................7 4-1 Map displaying the study site of Volusia and Brevard Counties as well as the locations of the tidal gauge, the AES40 grid point 3278, the NOAA buoy 41009, and the Melbourne Beach Wave Gauge...................................................................17 4-2 Lifeguard observation worksheet completed on August 15, 2005...........................21 4-3 A completed ECFL LURCS worksheet from September 2, 2005...........................22 5-1 Time-series plot of daily rescues for each individual year (top) and the total daily rescues (bottom) as logged by the Volusia County lifeguards (1998)..24 5-2 Time-series plots of wave height (ft), wave period (sec), wave direction (deg) and rip current incidents with associated rescues for 2001. The x marks correspond to days with more than 15 rescues.........................................................28 5-3 Time-series plots of wave height (ft), wave period (sec), wave direction (deg) and rip current incidents with associated rescues for 1998. The x marks correspond to days with more than 15 rescues.........................................................30 5-4 Correlation histogram-plot of offshore significant wave height (ft) along with rip current incidents and associated rescues for years 1998 to 2003. The blue (1 st ) bar represents the percentage of days that the wave height was within the respective range. The green (2 nd ) and red (3 rd ) bars represent the percentage of total rescues and the percentage of rip current incidents respectively, that occurred when the wave height was within that range. The numbers on the top of each plot correspond to the ratio of the blue and green bar magnitudes..............31 ix

PAGE 10

5-5 Correlation histogram-plot of peak wave period (sec) along with rip current incidents and associated rescues for years 1998 to 2003. The blue (1 st ) bar represents the percentage of days that the peak wave period was within the respective range. The green (2 nd ) and red (3 rd ) bars represent the percentage of total rescues and the percentage of rip current incidents respectively, that occurred when the peak wave period was within that range. The numbers on the top of each plot correspond to the ratio of the blue and green bar magnitudes.......34 5-6 Correlation histogram-plot of offshore wave direction (deg) along with rip current incidents and associated rescues for years 1998 to 2003. The blue (1 st ) bar represents the percentage of days that the wave direction was within the respective range. The green (2 nd ) and red (3 rd ) bars represent the percentage of total rescues and the percentage of rip current incidents respectively, that occurred when the wave direction was within that range. The numbers on the top of each plot correspond to the ratio of the blue and green bar magnitudes..............35 5-7 Correlation histogram-plot of the time of low tide (24hr) along with rip current incidents and associated rescues for each year from 1998 to 2003. The blue (1 st ) bar represents the percentage of days that low tide occurred within the respective time range. The green (2 nd ) and red (3 rd ) bars represent the percentage of total rescues and the percentage of rip current incidents respectively, that occurred when low tide occurred within that time range. The numbers on the top of each plot correspond to the ratio of the blue and green bar magnitudes..........................38 5-8 The daily-calculated threat values of both the ECFL LURCS and the modified index (top) along with the daily rip current incidents with associated rescues (bottom) for 1998. The horizontal lines represent the warning threshold for each respective index........................................................................................................40 5-9 Calculated rip current threat values for days with more than 15 rescues (19982003). The horizontal line represents the weighted average value..........................52 5-10 Calculated rip current threat values for days with more than 25 rescues (19982003). The horizontal line represents the weighted average value..........................52 5-12 Correlation histogram-plot of the offshore wave height (ft), peak wave period (sec), offshore wave direction (deg), and time of low tide (24hr) along with rip current incidents and associated rescues combined over the years from 1998 to 2003 (excluding 2002). The blue (1 st ) bar represents the percentage of days that each parameter occurred within the respective range. The green (2 nd ) and red (3 rd ) bars represent the percentage of total rescues and the percentage of rip current incidents respectively, that occurred when the parameter occurred within that range. The numbers on the top of each plot correspond to the ratio of the blue and green bar magnitudes.................................................................................54 5-13 ECFL LURCS daily rip current threat levels for the summer of 1998. The daily rip current rescue totals are indicated by the marker symbols.................................55 x

PAGE 11

5-14 The modified index daily rip current threat levels for the summer of 1998. The daily rip current rescue totals are indicated by the marker symbols........................56 5-15 RIPDEX (rip current predictive index) worksheet...................................................57 6-1 Daily rip current threat levels computed with the modified index using the WAVEWATCH III oceanographic output information for the summer of 2005. The daily rip current rescue totals are indicated by the marker symbols.................61 6-2 Daily rip current threat levels computed with the modified index using the WAVEWATCH III data for the summer of 2005 (after modification to the multiplicative direction factor). The daily rip current rescue totals are indicated by the marker symbols.............................................................................................67 6-3 The daily-calculated threat values of both the modified index using the WAVEWATCH III oceanographic output information (top) along with the daily rip current incidents with associated rescues (bottom) for the summer of 2005. The horizontal lines represent the warning threshold..............................................69 A-1 Time-series plots of wave height (ft), wave period (sec), wave direction (deg) and rip current incidents with associated rescues for 1997. The x marks correspond to days with more than 15 rescues.........................................................75 A-2 Time-series plots of wave height (ft), wave period (sec), wave direction (deg) and rip current incidents with associated rescues for 1998. The x marks correspond to days with more than 15 rescues.........................................................76 A-3 Time-series plots of wave height (ft), wave period (sec), wave direction (deg) and rip current incidents with associated rescues for 1999. The x marks correspond to days with more than 15 rescues.........................................................77 A-4 Time-series plots of wave height (ft), wave period (sec), wave direction (deg) and rip current incidents with associated rescues for 2000. The x marks correspond to days with more than 15 rescues.........................................................78 A-5 Time-series plots of wave height (ft), wave period (sec), wave direction (deg) and rip current incidents with associated rescues for 2001. The x marks correspond to days with more than 15 rescues.........................................................79 A-6 Time-series plots of wave height (ft), wave period (sec), wave direction (deg) and rip current incidents with associated rescues for 2002. The x marks correspond to days with more than 15 rescues.........................................................80 A-7 Time-series plots of wave height (ft), wave period (sec), wave direction (deg) and rip current incidents with associated rescues for 2003. The x marks correspond to days with more than 15 rescues.........................................................81 xi

PAGE 12

A-8 Time-series plots of wave height (ft), wave period (sec), wave direction (deg) and rip current incidents with associated rescues for 2004. The x marks correspond to days with more than 15 rescues.........................................................82 B-1 The daily-calculated threat values of both the ECFL LURCS and the modified index (top) along with the daily rip current incidents with associated rescues (bottom) for 1997. The horizontal lines represent the warning threshold for each respective index........................................................................................................83 B-2 The daily-calculated threat values of both the ECFL LURCS and the modified index (top) along with the daily rip current incidents with associated rescues (bottom) for 1998. The horizontal lines represent the warning threshold for each respective index........................................................................................................84 B-3 The daily-calculated threat values of both the ECFL LURCS and the modified index (top) along with the daily rip current incidents with associated rescues (bottom) for 1999. The horizontal lines represent the warning threshold for each respective index........................................................................................................85 B-4 The daily-calculated threat values of both the ECFL LURCS and the modified index (top) along with the daily rip current incidents with associated rescues (bottom) for 2000. The horizontal lines represent the warning threshold for each respective index........................................................................................................86 B-5 The daily-calculated threat values of both the ECFL LURCS and the modified index (top) along with the daily rip current incidents with associated rescues (bottom) for 2001. The horizontal lines represent the warning threshold for each respective index........................................................................................................87 B-6 The daily-calculated threat values of both the ECFL LURCS and the modified index (top) along with the daily rip current incidents with associated rescues (bottom) for 2002. The horizontal lines represent the warning threshold for each respective index........................................................................................................88 B-7 The daily-calculated threat values of both the ECFL LURCS and the modified index (top) along with the daily rip current incidents with associated rescues (bottom) for 2003. The horizontal lines represent the warning threshold for each respective index........................................................................................................89 B-8 The daily-calculated threat values of both the ECFL LURCS and the modified index (top) along with the daily rip current incidents with associated rescues (bottom) for 2004. The horizontal lines represent the warning threshold for each respective index........................................................................................................90 C-1 ECFL LURCS daily rip current threat levels for the summer of 1998. The daily rip current rescue totals are indicated by the marker symbols.................................92 xii

PAGE 13

C-2 The modified index daily rip current threat levels for the summer of 1998. The daily rip current rescue totals are indicated by the marker symbols........................92 C-3 ECFL LURCS daily rip current threat levels for the summer of 1999. The daily rip current rescue totals are indicated by the marker symbols.................................93 C-4 The modified index daily rip current threat levels for the summer of 1999. The daily rip current rescue totals are indicated by the marker symbols........................93 C-5 ECFL LURCS daily rip current threat levels for the summer of 2000. The daily rip current rescue totals are indicated by the marker symbols.................................94 C-6 The modified index daily rip current threat levels for the summer of 2000. The daily rip current rescue totals are indicated by the marker symbols........................94 C-7 ECFL LURCS daily rip current threat levels for the summer of 2001. The daily rip current rescue totals are indicated by the marker symbols.................................95 C-8 The modified index daily rip current threat levels for the summer of 2001. The daily rip current rescue totals are indicated by the marker symbols........................95 C-9 ECFL LURCS daily rip current threat levels for the summer of 2002. The daily rip current rescue totals are indicated by the marker symbols.................................96 C-10 The modified index daily rip current threat levels for the summer of 2002. The daily rip current rescue totals are indicated by the marker symbols........................96 C-11 ECFL LURCS daily rip current threat levels for the summer of 2003. The daily rip current rescue totals are indicated by the marker symbols.................................97 C-12 The modified index daily rip current threat levels for the summer of 2003. The daily rip current rescue totals are indicated by the marker symbols........................97 D-1 ECFL LURCS daily rip current threat levels for the summer of 1998. The daily rip current incident totals are indicated by the marker symbols...............................99 D-2 The modified index daily rip current threat levels for the summer of 1998. The daily rip current incident totals are indicated by the marker symbols......................99 D-3 ECFL LURCS daily rip current threat levels for the summer of 1999. The daily rip current incident totals are indicated by the marker symbols.............................100 D-4 The modified index daily rip current threat levels for the summer of 1999. The daily rip current incident totals are indicated by the marker symbols....................100 D-5 ECFL LURCS daily rip current threat levels for the summer of 2000. The daily rip current incident totals are indicated by the marker symbols.............................101 xiii

PAGE 14

D-6 The modified index daily rip current threat levels for the summer of 2000. The daily rip current incident totals are indicated by the marker symbols....................101 D-7 ECFL LURCS daily rip current threat levels for the summer of 2001. The daily rip current incident totals are indicated by the marker symbols.............................102 D-8 The modified index daily rip current threat levels for the summer of 2001. The daily rip current incident totals are indicated by the marker symbols....................102 D-9 ECFL LURCS daily rip current threat levels for the summer of 2002. The daily rip current incident totals are indicated by the marker symbols.............................103 D-10 The modified index daily rip current threat levels for the summer of 2002. The daily rip current incident totals are indicated by the marker symbols....................103 D-11 ECFL LURCS daily rip current threat levels for the summer of 2003. The daily rip current incident totals are indicated by the marker symbols.............................104 D-12 The modified index daily rip current threat levels for the summer of 2003. The daily rip current incident totals are indicated by the marker symbols....................104 xiv

PAGE 15

Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science IMPROVING RIP CURRENT FORECASTING TECHNIQUES FOR THE EAST COAST OF FLORIDA By Jason Rolland Cummins December 2006 Chair: Robert J. Thieke Major: Coastal and Oceanographic Engineering This study documents the development of an improved rip current predictive index through the detailed examination of a long-term record of oceanographic and meteorological conditions with concurrent rip current events. Rip current rescue statistics are used as a proxy for actual in situ rip current measurements; the correlations of various meteorological and oceanographic parameters with the occurrence of rescues are used to establish the relative importance (and hence weighting) of these factors in the predictive scheme. The correlation analysis was conducted on a long-term data set consisting of rip-related rescues and hindcast wind and wave data in Volusia County, Florida, extending from 1997 to 2004. In addition to the established dependence on wave height and period (already incorporated in National Weather Service [NWS] predictions), the relative risk of daily rip current activity was found to increase during periods of shore-normal waves and when the occurrence of low tide coincided with times of peak beach attendance. The existing rip current forecasting methods practiced by the NWS were accordingly xv

PAGE 16

modified to include both a wave direction parameter and a low tide parameter. After these adjustments, the modified predictive index exhibited significant improvements. Currently, the NOAA (National Oceanic and Atmospheric Association) wave buoys off the east coast of Florida (which are used on a same-day basis for NWS rip current warnings at present) are not equipped to measure the wave direction. To overcome this obstacle, the wave direction was incorporated into the rip current predictive scheme by implementing a readily available forecast (directional) wave model called WAVEWATCH III (currently used by NWS in other connections). An investigation of the performance of the modified index using oceanographic output information from the wave model was conducted in blindfold fashion using the Volusia County rescue information for the summer of 2005. The results indicate that the modified index outperforms the index currently employed by the NWS. These improvements in performance, as well as the advantages of advanced notice, justify the incorporation of a forecast wave model (WAVEWATCH III) into the operational rip current prediction process. xvi

PAGE 17

CHAPTER 1 INTRODUCTION A rip current is a strong, channeled flow of water extending seaward from the shoreline. According to the United States Lifesaving Association rip currents are the cause of approximately 80% of their rescues nationwide. Since 1995, over 17,500 documented rescues in the U.S. were directly related to rip current activity. Rip current activity affects the safety of beach-goers visiting the coastal waters of this nation and others. As a result of these dangerous conditions, it is reported that in the U.S. alone over 100 deaths annually are attributed to rip currents. An investigation was conducted at the University of Florida to examine the existing rip current prediction methods implemented by the National Weather Service. Analysis of rip-related rescues correlated with oceanographic conditions assisted in improving the accuracy of the presently employed rip current index. The index is a scale used to calculate the level of risk for ocean-goers due to local rip current formation. Knowledge of these conditions aids the local governmental authorities in the issuance of warnings to the public. Public awareness, as well as the talent and dedication of the beach lifeguards, plays a vital role in the prevention of rip-related drownings. The development of a rip current prediction scheme began in south Florida with the Lushine Rip Current Scale (LURCS), which utilized wind speed and direction along with swell height and the time of low tide to assess the daily level of risk associated with rip currents (Lushine 1991). The LURCS prediction scheme was then later adapted for use on the central east coast of Florida (ECFL LURCS) with the inclusion of the swell period 1

PAGE 18

2 and a modification to the tidal parameter (Lascody 1998). The ECFL LURCS was later modified by Engle (2003). Engle (2003) eliminated the wind parameters and also incorporated the incident wave angle, directional spreading and the tidal level. The intention of this study was to further justify the incorporation of a swell direction parameter and establish a way to integrate it into the prediction scheme, with the ultimate goal of using the index as an operational forecasting tool. An analysis using Volusia County rescue information extending from 1997 to 2004 was completed to establish the importance and respective range of each oceanographic parameter used in the rip current prediction scheme. The modifications were then individually tested against the performance of the existing ECFL LURCS method for the same time period. The incorporation of wave direction improved the accuracy of the rip current predictions. However, the NOAA (National Oceanic and Atmospheric Association) data buoys on the east coast of Florida are presently not capable of measuring wave direction; this represented a significant stumbling block in the implementation of the modified index as an operational tool. The application of the WAVEWATCH III model (Tolman 1997, 1999a) was introduced into the prediction scheme in order to resolve this quandary. The modified prediction index developed here used the oceanographic output information from the wave model to calculate the daily rip current threat levels for the summer of 2005. The results were then compared with the Volusia County rescue information and the completed ECFL LURCS worksheets to examine if the model could be used to accurately predict rip current conditions. Previous research concerning the driving forces and theoretical underpinnings of rip current formation, as well as general characteristics of a rip are reviewed in Chapter 2.

PAGE 19

3 An overview of the LURCS, ECFL LURCS and the Modified ECFL LURCS is presented in Chapter 3, along with motivation for improvements to the existing rip current prediction scheme. Chapter 4 discusses the data sources used during the analysis process of this study. The correlation between rescues and specific oceanographic parameters, and the subsequent modifications to the existing ECFL LURCS is presented in Chapter 5. A comparison between the resulting modified index implementing the WAVEWATCH III forecast model and the documented ECFL LURCS worksheets is presented Chapter 6. Finally, the summarized results of the study and overall conclusions are then presented in Chapter 7.

PAGE 20

CHAPTER 2 RIP CURRENTS Formation of a Rip Current In 1936 the term Rip Current was suggested by F. P. Shepard to describe the phenomena observed by lifeguards and others on the coast of California. Originally, the phenomenon of seaward flow was referred to as a Rip Tide. Yet the occurrence has little connection with the tidal flow itself, hence the recommendation by Shepard to rename the process. Rip currents are defined as narrow, seaward-directed currents that extend from the inner surf zone out through the line of breaking waves (Haller and Dalrymple 2001). Five years later F. P. Shepard, along with K. O. Emery and E. C. Lafond (1941), reported various qualitative observations regarding rip currents. They divided the rip current, also referred to as a rip, into three specific parts: 1) the feeder currents, 2) the neck and 3) the head (Figure 2.1). The flow was perceived to be strongest in the nearshore through the surf zone, or neck, and then the observed speeds reduced as the current traveled further offshore into the rip head. Another important observation made by Shepard et al. was the association of rips with certain meteorological conditions. They noticed that an increase in rip current intensity was directly associated with an increase in wave height. In a later study performed by Shepard and Inman (1950), it was found that rip currents are an integral component of a larger nearshore circulation system. They hypothesized the driving force of these circulation cells was the convergence and 4

PAGE 21

5 divergence of the incoming wave field due to the effect of the refraction, producing a longshore variation in breaking wave conditions. The resulting wave set-up field creates conditions where water is driven away from regions of the larger waves towards areas of smaller waves in the form of a longshore current. These currents eventually converge and turn seaward in the form of a rip current. The physics of the forcing mechanism was not completely understood until Longuet-Higgins and Stewart (1964) introduced the concept of radiation stress, and linked it analytically to the wave set-up. Radiation stress is defined as the excess momentum flux conveyed by a progressive wave. It is a function of the wave energy and therefore proportional to the square of the waves height. Figure 2-1. Rip current and its component parts (feeders, neck and head) as well as the commonly associated currents (represented by arrows) (from Shepard et al. 1941).

PAGE 22

6 Originally the development of nearshore currents was attributed to the shoreward mass-transport of water by waves and the subsequent localized changes in sea level. In a qualitative field experiment, McKenzie (1958) observed that the water brought in toward the shore by breakers and translatory waves tends to cause longshore currents close to the beach. At variable intervals the longshore currents turn seaward and form outgoing rip currents. Longuet-Higgins and Stewart (1964) adopted a different approach by using the concept of radiation stress to analyze the conservation of momentum flux and observed changes in sea level. According to the theory, any change in the cross-shore radiation stress is balanced by a hydrostatic pressure gradient, or change in water level. In their study they theoretically predicted a decrease in the water level, know as set-down, when the waves approach the breaking point. Adhering to the continuity of momentum flux, the decrease in the water level is due to the increase in energy from the shoaling of a wave. They also predicted an increase in water level, known as set-up, shoreward of the breaking zone. The increase in water level is attributed to the energy dissipation, and subsequent decrease in radiation stress, of the wave during breaking. Figure 2.2 displays measurements from a lab experiment performed by Bowen et al. (1968) and their calculated results applying the theory developed by Longuet-Higgins and Stewart (1964). Bowen (1969) investigated the observed nearshore circulation system using the concept of radiation stress developed by Longuet-Higgins and Stewart (1964). As mentioned earlier, cross-shore variations of the radiation stress is the cause of set-up and set-down. Since the radiation stress is proportional to the wave height, a longshore variation in the incident wave height will result in a longshore variation of set-up and set-down. The variation in set-up induces a pressure field, driving a flow of water in the surf

PAGE 23

7 zone away from regions of high waves toward the regions of low waves (Bowen 1969). When two of these flows converge on the same location exhibiting low wave energy, they turn offshore and exit the surf zone as a confined rip current (MacMahan 2006). Figure 2-2. Profile of the mean water level and the envelope of the wave height for a typical experiment. Wave period, 1.14 sec; H o = 6.45cm; H b = 8.55cm; tan = 0.082 (from Bowen et al. 1968). In review, the generation of a rip current begins with longshore variations of the incoming wave field. These variations in wave height can be derived from incident and edge wave interactions (Bowen and Inman 1969), the convergence or divergence of wave rays over offshore bottom topography, and/or the induced variability in wave height

PAGE 24

8 caused by coastal structures (Dalrymple 1978). Independent of cause, the forcing disparity in the incident wave field drives nearshore circulation cells. These cells exhibit wide regions of shoreward flow separated by narrow regions of offshore flow. If the narrow regions are strong enough they will appear as rip currents (Haller and Dalrymple 2001). Characteristics of a Rip Rip currents are not confined to a specific type of beach and have been observed on the east and west coasts of Florida (Sonu 1972, Engle 2003), the coast of California (Shepard et al. 1941 & 1950, Bowen and Inman 1969, Cook 1970, MacMahan 2004) and on the coasts of Australia (McKenzie 1958, Short 1985, Brander 1999 & 2001, Haas 2002). Each location contains a different offshore topography and incoming wave field, yet the same phenomenon was seen in all. Rip currents have also been noted to occur around man-made structures, such as jetties, groins and piers. In this study the concentration is on the straight, typically barred beaches seen on the east coast of Florida. Rips have been observed in such locations and have been established in a similar controlled lab environment. Experiments conducted by Haller et al. (1997) at the University of Delaware demonstrated the occurrence of cell circulation on an alternating bar and channel configured shoreline, similar to the bathymetry on Floridas east coast. On a barred beach the rip current is typically associated with a rip channel, or a gap in the bar where the out-flowing current is located. Since the rip itself can be unstable in its location, the associated channel can wander as well. However, if there is structural influence, the channel will tend to remain stationary. Generally, rip currents are not constant features, they can flow intermittently, the head swings back and forth, and their channels may migrate (Cook 1970). The migration of the channel can be the result of

PAGE 25

9 changing wave conditions. When the incident wave field increases in intensity, the rips tend to transform from many small rips to a few large rips (McKenzie 1958). Another mechanism for migration is the variability in strength of the feeder currents, which are rarely the same length and intensity. Uneven feeder currents can orient the rip obliquely to the shore, and cause the head to become unstable, moving from side to side. Rip currents have been known to form both orthogonally and diagonally across the surf zone (McKenzie 1958). Due to the many variable factors influencing the nature of a rip current, the flow of the rip can also be highly unsteady, constantly varying in flow intensity. The changing speed or pulsation of the rip current is an important factor when evaluating the safety of beachgoers. A swimmer can be situated in a channel during a lull period and remain in control, yet when the rip pulsates and increases its flow the swimmer suddenly becomes in danger. Such unsteadiness has been observed on several occasions. Shepard and Inman (1950) noted that rip currents tend to register all variations in wave strength with a short time lag. Typical average rip current velocities are O(1.0)ft/s (0.3m/s), but on shorter time scales velocities can reach a max of 6.6ft/s (2.0m/s). Sonu (1972) observed pulsations at high tide corresponding to variance of the incoming swell, and at low tide corresponding with surf beat frequencies. Others have observed the latter phenomenon and generally associate the pulsations with wave groups at the infragravity level (0.004-0.04Hz) (Shepard et al. 1941, Shepard and Inman 1950, Brander and Short 2001, MacMahan et al. 2004). Field experiments performed by MacMahan et al. (2004) confirmed the pulsations were driven by infragravity cross-shore standing waves, also known as surf beat.

PAGE 26

10 Although short-term pulsations can adversely affect ocean-goers, the overall strength and severity of a rip current is the primary determinant of ocean safety. It is almost uniformly agreed that an increase in wave height is directly correlated with an increase in rip current flow. An experiment conducted by Shepard et al. (1941) showed an increase of rip intensity with every period of larger waves. When conditions of larger swell occur, the surf zone increases in width and a system of larger and more active rips can establish itself (McKenzie 1958). Another important factor contributing to the intensity of a rip current is the tidal stage. During low tide there are two responses: (1) the breaking on the bar intensifies and (2) the flow concentrates in the rip channel. Both of these reactions contribute to the increased intensity of the rip current. Shepard (1941) first observed the concentration of seaward flow in the rip channel during low tide, preserving form even in less than ideal conditions. Additional field experiments conducted by Sonu (1972) helped validate the association of rip current strength with tidal stage, as well as a correlation with swell direction. During an experiment in Australia, Brander (1999) measured the greatest flow velocities at low tide. It has been noted and observed by many researchers, that there is a definite correlation between rip current intensity and wave height, as well as tidal stage. Therefore, both the wave height and tidal stage become crucial when attempting to interpret the severity of the rip current conditions, with an eventual goal of providing appropriate warnings to the public.

PAGE 27

CHAPTER 3 FORECASTING Researchers have long been examining rip currents to identify their forcing mechanisms and the conditions associated with their occurrence in nature. This information was incorporated into more recent works in an attempt to correlate rip currents with specific oceanographic parameters. With such knowledge one could accurately assess the rip current related hazards in the surf zone and then inform beach-goers. There are more rip-related deaths in Florida each year than hurricanes, tropical storms, tornados, severe thunderstorms and lighting combined (Lascody 1998). With a more accurate prediction system and adequate warning methods hopefully the number of rip current victims can be decreased. In one of the first attempts at rip current predictions, Lushine (1991) examined the reported rip-related drownings and rescues in southeast Florida and the concurrent oceanographic and meteorological conditions. Since the availability of long-term records on rip current incidence is scarce, rip-related rescues proved to be a useful indicator of rip current events. Through his work, an experimental scale (Lushine Rip Current Scale, LURCS) was developed to calculate the risk level of the surf zone due to rip currents. The scale ranges from zero to five, zero corresponding to no weather-related rip current danger and five meaning high danger for all swimmers. In the development of the LURCS, Lushine (1991) found a strong correlation between rip current rescues and wind conditions. In southeast Florida wind is the primary source for wave generation, because the islands of the Bahamas intercept most of the distant swell. Wind then became the 11

PAGE 28

12 primary foundation of the scale, although it also includes swell height and small factors for tidal stage and persistence. An increase in wind and/or swell height will raise the scales value respectively. The tidal stage factor is only added to the scales value if the time is between two hours before and four hours after low tide. The persistence factor accounts for continuing rip current conditions. Confirmation of the ability to accurately predict rip currents with the LURCS was achieved through testing on an independent data sample. Three parameters were used to interpret the results 1) Probability of Detection (POD), 2) False Alarm Ration (FAR), and 3) Critical Success Index (CSI). The results were convincing, and the LURCS was recognized as a beneficial approach to forecasting the occurrence of rip currents. The scale developed by Lushine (1991) was intended for use in southeast Florida, an area dominated by locally generated wind waves. This is evident in the LURCS, because of the emphasis placed on wind conditions. The southeast division of the National Weather Service (NWS) implemented the forecasting technique, and warnings were issued when rip current activity was calculated high enough to pose a threat. Other sectors of the NWS noticed the advantage of the LURCS and a modified version was prepared for use along the central east coast of Florida. The ECFL (East Central Florida) LURCS, developed by Lascody (1998), was derived from the original LURCS with the addition of a swell period factor and small changes to other parameters. Lascody (1998) realized the limitations of the LURCS application because of the dependence on wind waves, where east central Florida is more disposed to long period swell conditions. Less emphasis was placed on the wind conditions, and the inclusion of wave period into the scale assured dependence on swell

PAGE 29

13 conditions. The modified scale predicts well and is used today, yet a high false alarm ratio in testing showed the ECFL LURCS still needed some improvement (Lascody 1998). A later investigation by Engle (2003) examined the ECFL LURCS and looked for such methods of improvement. Engle (2003) proposed two parameters, wave direction and tidal stage, to be particularly influential when attempting to assess rip current activity. The modified scale was based on the ECFL LURCS. The first change was removing the wind speed and direction parameters completely. Secondly, wave direction and tidal level were incorporated into the index. Tests were completed to compare the performance of the ECFL LURCS with the modified version, using rescue data and associated weather conditions. The modified scale showed improvements over the ECFL LURCS when using the POD and FAR for comparison. Alarm Ratio (AR) was another statistical parameter introduced by Engle (2003) to develop balance between the two scales. AR is the percentage of days the scale is predicting rip currents (Engle 2003). Engles (2003) modified index showed promising improvements over the original ECFL LURCS. Another analysis was executed by Schrader (2004) to reinforce the accuracy of the modifications to the ECFL LURCS. This analysis applied a smaller independent data set, but still demonstrated the validity of the improvements claimed in earlier work by Engle (2003). These more recent studies established the need to include wave direction in a predictive index, however there was no ready way to implement the new directional parameters. The existing rip current index applies information obtained from the NOAA weather buoys located off the east coast of Florida. These buoys measure the wave height

PAGE 30

14 and period, but provide no directional data. The aim of the present study is to overcome this shortcoming in an effort to make the predictive index operational. Through correspondence with the scientists operating the ECFL LURCS at the National Weather Service, the author determined what information was readily accessible to them during the forecasting process. One such asset is the output of the WAVEWATCH III model managed by the National Oceanographic and Atmospheric Association (NOAA). WAVEWATCH III is a global wave model used to predict height, period and direction at each of its grid points. It is executed every six hours (ex. 12am, 6am, 12pm, and 6pm) and the output is given on one-hour intervals for the following seven days. The resulting information could prove valuable by improving the index not only with the inclusion of wave direction but also through the extension of the forecast. Currently the rip current index calculation is prepared in the morning, and the rip current threat is ascertained for the same day. The conditions (and warnings if necessary) are distributed to the public through different avenues of the media, such as NOAA weather radio. Knowledge of the rip current threat a day in advance would help in the education and awareness of the public as well as establishing additional staffing needs of the local beach lifeguards. This study built upon the previous work of Engle (2004), by the inclusion of a wave direction parameter and a modified tidal stage parameter. The modified index created here, and each of its parameters, was refined through testing on a long-term data set. This data consisted of rip-related rescues and hindcast oceanographic conditions. After finalization of the modified index, an investigation was conducted implementing the WAVEWATCH III forecast wave data (which includes wave direction) into the

PAGE 31

15 prediction methods. The WAVEWATCH III examination was completed to assess the forecasting capabilities of a modified prediction scheme.

PAGE 32

CHAPTER 4 DATA Site Description The study site is separated into two sections, both of which are located on the central east coast of Florida (Figure 4-1). The northern section is comprised of the Volusia County beaches ranging from the North Ormond region south to New Smyrna. The southern section consists of the Brevard County beaches extending from Cape Canaveral south to Melbourne beach. The two sections are divided by the property of NASAs Kennedy Space Center. The coastline of Volusia and Brevard County consists of sandy beaches with a mean sediment diameter of about 0.23mm and 0.33mm respectively (Charles et al. 1994). The nearshore bathymetry of both locations typically includes a single shore-parallel bar and trough configuration. The approximate azimuth of Volusia County is 62 East of North. The shoreline angle of Brevard County shifts due to a coastline perturbation at the location of the Kennedy Space Center. The azimuth therefore migrates gradually from 100 in Cape Canaveral to approximately 65 in Melbourne Beach (Figure 4-1). The continental shelf in the Volusia County region extends out approximately 80km from shore and the contours are relatively shore-parallel (Engle 2003). The continental shelf narrows further south on the Florida coastline, and therefore the width decreases to approximately 60km in Brevard County. However, the bottom contours remain relatively shore-parallel. The tides in Volusia and Brevard County are semidiurnal and have a maximum range of approximately 6ft (2m). 16

PAGE 33

17 Figure 4-1. Map displaying the study site of Volusia and Brevard Counties as well as the locations of the tidal gauge, the AES40 grid point 3278, the NOAA buoy 41009, and the Melbourne Beach Wave Gauge. Rip Current Rescues Archived rescue logs were obtained from the Volusia County Beach Safety Division extending from 1997 through 2005. Each lifeguard reports their daily rescue activity including the date, location, type of rescue, and number of victims involved. This study utilizes both the number of daily rip related incidences and the number of victims rescued in each rip current incident. The reports are divided into six zones: (1) North Ormond to Flagler County line (2) Ormond Beach (3) Daytona Beach (4) Daytona Beach

PAGE 34

18 Shores (5) Dunlawton Avenue south to Ponce Inlet jetty and (6) New Smyrna. The rip current predictive index has been formulated for application on the entire central east coast of Florida. In order to limit the possible spurious correlations due to localized effects, the present study does not account for the zones separately but compiles the rescue information from each zone into one comprehensive record. Hindcast Data Historic oceanographic (wave height, period, and direction) and meteorological (wind speed and direction) conditions were provided by the AES40 North Atlantic Wind and Wave Climatology hindcast model called OWI 3-G. The OWI 3-G model is a direct spectral type based from the WAM model (WAMDI Group 1988), and was originally developed by Oceanweather Inc. during a project for the Meteorological Service of Canada. The model grid spans from the equator in the south to the 75.625 latitude in the north, with the North American coastline representing the western boundary and the 20 longitude as the eastern boundary. The grid is spaced at 0.833 increments in the longitude and 0.625 increments in the latitude, therefore consisting of 9023 wet grid points. The information in this study was extracted from grid point #3278, which is located at 28.75 N latitude and 80 W longitude (Figure 4-1). The OWI 3-G model has been tested to validate the precision of the models deepwater wind and wave output. A quantile-quantile evaluation performed by Swail et al. (2000) compared the model hindcast with recorded satellite and buoy information. A quantile-quantile assessment is used to determine if two data sets are comprised of a common distribution. The study exhibited a good correlation in the 1 st to the 99 th percentile between the model output and the documented real-time data.

PAGE 35

19 Tides The tidal data used in analyzing the ECFL LURCS and the modified version was obtained through NOAA Tides Online Historical Data Retrieval. The tidal station ID is 8721604 (Trident Pier) and is located at 28.415 N and 80.5933 W (Figure 4-1). The local times of low tide were obtained on a daily basis for calculations in the modified rip current predictive index. The moon phases for each year of the study were acquired from the U.S. Naval Observatory website http://aa.usno.navy.mil/data/docs/MoonPhase.html. The occurrence of the new and full moon phases were used during the calculations of the ECFL LURCS. For the days before, after, and during a full or new moon the astronomical tide factor was included in the rip current threat level. WAVEWATCH III The WAVEWATCH III (Tolman 1997, 1999a) model is a third generation wave model developed at the Ocean Modeling Branch of the National Center for Environmental Predictions (NCEP/NOAA). WAVEWATCH III solves the spectral action density balance equation for wavenumber-direction spectra. Assumptions within this method limit the model to application outside the surf zone and to spatial scales larger than 1km. The source terms for the model include wave growth and decay due to the actions of wind, nonlinear resonant interactions, dissipation due to white-capping and bottom friction. The model uses a regularly spaced longitude-latitude grid. The spectral discretization of the wave energy applies an invariant logarithmic intrinsic frequency grid to spatially vary the wavenumber (Tolman and Booij 1998). The directional increment of the wave energy spectra is constant and covers all directions. The output of the resulting

PAGE 36

20 wave spectra is at selected locations. In this study the output location used is identical to the position of the NOAA buoy 41009 (Figure 4-1). Lifeguard Observations In an agreement with the University of Florida Department of Coastal Engineering, Brevard County lifeguards were asked to document daily observations of the nearshore conditions. The protocol consisted of completing a prearranged worksheet containing various air and sea parameters (Figure 4-2). The beaches of Brevard County are usually heavily occupied, resulting in an excess of responsibilities for the lifeguards. Therefore, the observations are logged sporadically through October, November and December of 2004, along with May, June and July of 2005. Knowledge of the conditions from a first-hand observer can still prove to be qualitatively useful during the analysis process. Completed ECFL LURCS Worksheets Completed ECFL LURCS worksheets (Figure 4-3) were obtained from the National Weather Service (NWS) extending from April to October of 2005. These worksheets were filled out by NWS employees and used to distinguish if a warning should be issued. Melbourne Beach and NOAA Data Buoys There is one existing wave buoy on the central east coast of Florida that measures wave direction. This buoy is funded through the beaches and shores division of Florida State University (FSU) and maintained by Dally at Surfbreak Engineering. It was deployed off the coast of Melbourne Beach in a depth of approximately 8m (Figure 4-1). The nearshore wave information obtained from this buoy was used to corroborate the observations made by the Brevard County Lifeguards. The NOAA Data Buoy (# 41009,

PAGE 37

21 Figure 4-1) was utilized as a qualitative reference for oceanographic and meteorological conditions throughout the study. Figure 4-2. Lifeguard observation worksheet completed on August 15, 2005.

PAGE 38

22 Figure 4-3. A completed ECFL LURCS worksheet from September 2, 2005.

PAGE 39

CHAPTER 5 DEVELOPEMENT OF IMPROVED RIP CURRENT INDEX Analysis Lushine (1991), Lascody (1998) and Engle (2003) all used rescue and drowning incidences to develop and test their respective rip current predictive scales. Lushine (1991) obtained medical examiners information, beach patrol rescue logs and newspaper clippings in Dade and Broward Counties from 1979 to 1988. Lascody (1998) acquired similar information in Volusia, Brevard, Indian River, St. Lucie and Martin Counties from 1989 to 1997. Engle (2003) utilized lifeguard rescue logs in Volusia County for only one year, 1996. The aim of this study is to verify the inclusion of wave direction, proposed by Engle (2003), and the adjustment of other parameters by exploring a long-term data set (1997). Lifeguard rescue logs have proved to be useful in developing a rip current prediction scheme. Lushine (1991) and Lascody (1998) showed that rip current rescues are a good qualitative representation of rip currents themselves. The obvious drawback to this method is that the data can be strongly dependent on the population of ocean-goers. If there are no people in the water, then there will be no evidence of a rip current. This doesnt necessarily mean there was no occurrence of a rip current; it might only mean there werent any bathers to be rescued from one. A good example of this phenomenon is the winter season. On the central east coast of Florida, the water becomes relatively cold in the winter months. Therefore considerably less people enter the water and fewer rescues are logged. In examination of Figure 5-1, the majority of rescues occur during the 23

PAGE 40

24 summer months. This does not necessarily indicate a reduction of rip current activity in the winter months, but is far more likely the effect of the dependency of rip current rescues on ocean-goers. Figure 5-1. Time-series plot of daily rescues for each individual year (top) and the total daily rescues (bottom) as logged by the Volusia County lifeguards (19982003). Limiting the examination and analysis of rip current rescues to the summer season mitigates the issue concerning a lack of bathers during the winter. The assumed peak days of attendance are from day 75 to day 250 of a given year. This time frame translates into mid-march until early September. The month of March signifies the start of spring break vacations and the east coast of Florida is considered a prime destination. The rising influx of tourists is directly associated with an increase in rescues (Figure 5-1). The

PAGE 41

25 rescues then become more inconsistent in September as the water temperature starts to decrease and hurricane season is ongoing. As seen in Table 5-1, the majority of the rescues occur during the previously defined summer season. When rescues are totaled over the years examined, excluding 2002 for reasons explained later, almost 87 percent of the rescues occurred in the summer. Table 5-1. Percentage of rip-related rescues occurring in the summer (defined as day 75250), 1998 to 2003. 1998 1999 2000 2001 2002 2003 Total rescues 2058 1799 1232 2399 226 1135 Summer rescues 1887 1633 972 2256 201 723 Percent summer 0.92 0.91 0.79 0.94 0.89 0.64 This study was also quantitatively limited to the years 1998, 1999, 2000, 2001, and 2003. The exclusion of 1997 and 2004 is due to the lack of data coverage in both of these years. The rescue data from 1997 is limited and the hindcast data from 2004 is only available for the first half of the year. Both years were still examined on a daily basis, investigating the high rescue days and qualitative correlations, but overall summer statistics were withdrawn. The omission of 2002 is because of the significant lack of rescues occurring during this year. In Table 5-1 a dramatic decrease in rescues from the other years can be seen. The reason for this is unknown, but one assumption is the considerable decrease of tourism travel in the year following the September 11 th terrorist attacks. Independent of cause, the data demonstrates an unnatural decrease in rescues for 2002, and consequently the year was removed from the study. Another difficulty when using rescues to mark rips is the effect of rough water conditions and/or inclement weather. If the surf zone is violent people are hesitant to enter the water and if the weather conditions are poor (ex. rain, clouds) people are even less likely to make the trip to a beach. This predicament is not as easily remedied as the

PAGE 42

26 previous situation. The only assurance is that the ECFL LURCS and the newly modified index were exposed to the same problem. Therefore, any such disadvantages are assumed to equally challenge the predictive capability of both indexes. The first step in improving the present index was to develop a preliminary new index. The new index was built on the same foundation as the two previous indexes. The LURCS and ECFL LURCS assess given input conditions and then return a rip current threat level. Each input parameter affects the index by increasing or decreasing the resulting threat level. An example of a completed ECFL LURCS worksheet is shown in chapter 4 (Figure 4-3). Approximate ranges for each parameter and their respective threat values are already established. When the conditions (e.g., Wave height) are found to lie in a given range, a value is assigned to that particular factor. After completing the list of factors, their respective values are summed to obtain a threat level. The severity of rip current danger is dependent on the threat level, and a predetermined threshold establishes if it is advisable to issue a warning. The adjustments to the new index were loosely based on an approach established in previous work done by Engle (2003), which served to eliminate the wind parameter, incorporate wave direction, and modify the tide factor. The aforementioned data was then used to test the performance of the new rip current prediction index (RIPDEX) against the ECFL LURCS currently applied by the NWS. Each new (e.g., direction and tide) and old (e.g., swell height and period) parameter was tested and modified in a cyclic process to ensure the index reached maximum performance levels. The previous work by Engle (2003) initiated the idea of applying these new factors (e.g., wave direction) to increase the accuracy of the ECFL LURCS. Now a far longer data set can be used to better

PAGE 43

27 establish the correct ranges for each input parameter and to clarify each of their roles in a more optimal prediction scheme. Ocean Correlations The oceanographic parameters used in index computations were examined to determine their importance in the formation of rip currents. These parameters include wave direction, wave height, wave period and the time of low tide. Each parameter was first examined from a qualitative viewpoint. Figure 5-2 is a time-series plot of daily wave height, period and direction along with the amount of rip current incidents (rips) and the number of victims rescued from each rip (rescues) for 2001. The wave direction is given in meteorological convention, which is measured clockwise from true north (recall that 62 represents shore-normal for Volusia County). The x marks on the plots of height, period and direction correspond to days with more than 15 rescues. These days along with other spikes in the number of rescues were used to identify associations between the wave characteristics and rescues. In Figure 5-2 the high rescue days tend to occur during peaks in the wave period. An example of this can be seen in the beginning of the month of May. The wave period increases from 7 to 11 seconds and the result is multiple days consisting of rescue numbers greater than 15. Another aspect possibly effecting the same time period is the synchronized spike in wave height. Larger waves with longer periods (e.g., distant swell) are directly attributed to an increase in rip current intensity (Shepard et al. 1941). An interesting observation is the successive days of high rescues during the subsidence of this wave height spike. A possible mechanism for such observations may include both physical and human behavioral effects. The increase in wave height and period may lead to the formation of rip channels in the bar, but it is probably too rough for most beach

PAGE 44

28 patrons to confidently enter the water. When the severity of the conditions begins to subside the people waiting become more inclined to enter the surf zone. However, the channels remain intact and the rip currents still pose a considerable threat to bathers. Such events indicate the importance of knowing the ocean conditions and rip current threat for the previous days and lend credence to the use of persistence factors in prediction. Figure 5-2. Time-series plots of wave height (ft), wave period (sec), wave direction (deg) and rip current incidents with associated rescues for 2001. The x marks correspond to days with more than 15 rescues. In 1998 there is a dramatic spike in rescues in the beginning of August and it seemed justified to investigate it further (Figure 5-3). On August 4 th 5 th and 6 th there were 65, 175, and 252 rescues respectively. These three days accounted for 24% of the

PAGE 45

29 rescues recorded in 1998. All of the days landed in the middle of the week, so the often-observed weekend population effect was not a relevant issue. The next logical step was to examine the conditions prior to the incident. In the days leading up to this spike there was a period of relatively consistent swell periods and heights, and the wave direction was slightly north of shore-normal. These consistent rip current generating conditions likely resulted in the formation of localized rip channels, and were verified by the rescues on days prior. Then, as seen in Figure 5-3, the wave direction suddenly changed to the southeast and as the direction migrates back to its original values the abnormal spike in rescues occurred. It is hypothesized that the large change in wave direction over a relatively short time period may magnify the instabilities of a longshore bar and rip current system. This magnification results in more hazardous conditions for beach-goers. Similar plots were generated for the other 7 years included in this study (see appendix A). Each plot was also qualitatively examined to assist in investigating the connection between each parameter and the occurrence of rip-related rescues. The parameters of the rip current predictive index were then further explored to find the range of conditions that constituted the greatest association with rip current development. Each parameter was dissected into distinct ranges, and then associated with the rips and rescues occurring on those days when the conditions are in their particular ranges. A histogram is plotted for every year, grouping the normalized frequency of the oceanographic parameter along with the rip current incidents and the related number of rescues.

PAGE 46

30 Figure 5-3. Time-series plots of wave height (ft), wave period (sec), wave direction (deg) and rip current incidents with associated rescues for 1998. The x marks correspond to days with more than 15 rescues. Wave height The first correlation discussed is between rip currents and the offshore significant wave height (H O ). The record of wave height was divided into one-foot categories ranging from zero to ten feet (0.05m). Figure 5-4 displays three bars plotted in each range. The first (blue) bar represents the percentage of days the significant wave height provided from the hindcast was in the specified range. The second (green) bar represents the percentage of rip-related rescues occurring on the days when the associated wave height was in the respective range. The third (red) and final bar represents the percentage of rip current observations made by the lifeguards when a rescue occurred, which is

PAGE 47

31 independent of the number of rescue victims. A comparison of the latter two bars provides some insight of how severe the rip currents were in a specified condition range. Figure 5-4. Correlation histogram-plot of offshore significant wave height (ft) along with rip current incidents and associated rescues for years 1998 to 2003. The blue (1 st ) bar represents the percentage of days that the wave height was within the respective range. The green (2 nd ) and red (3 rd ) bars represent the percentage of total rescues and the percentage of rip current incidents respectively, that occurred when the wave height was within that range. The numbers on the top of each plot correspond to the ratio of the blue and green bar magnitudes. A trend can be seen from year to year that the majority of the wave heights occur in the 2 to 6 foot (0.61.83m) range. An interesting characteristic of Figure 5-4 is the high number of rescues occurring in the 3 to 4 foot (0.91.22m) range of wave height. If the percentages are combined over all the years, excluding 2002, approx 52% of the rescues occur when the wave heights are between 3 and 4 feet. When examining the histograms, an important aspect is the difference in magnitude within each group of bars. If the rescue

PAGE 48

32 and/or rip current bar is greater than the actual occurrence of the parameter, then this constitutes a higher risk of rip-related rescues on days with those conditions. The difference in the magnitude of the first two bars is represented by the ratio value displayed above each group of bars. This ratio value is calculated as the magnitude of the green (2 nd ) bar divided by the magnitude of the blue (1 st ) bar. A ratio value greater than one corresponds to a higher relative risk of rip-related rescues occurring. In 1998 for example, 39% of the documented wave heights and 60% of the total rescues occurred in the 3 to 4 foot (0.91.22m) range. However, 20% of the wave heights occurred in the 2 to 3 foot (0.61.91m) range, along with only10% of the total rescues. Therefore, on days when the wave height range was 2 to 3 feet there was an average of 3.3 rescues, but on days with a height range of 3 to 4 feet there was an average of 10.4 rescues. The relative importance of each range to the threat level is well represented in the ECFL LURCS and therefore was not changed. The contribution of the swell height parameter to the computation of the modified index is as follows: H O < 1ft swell height factor = 0 1 ft H O < 2 ft swell height factor = 0.5 2 ft H O < 3 ft swell height factor = 1 3 ft H O < 5 ft swell height factor = 2 5 ft H O < 8 ft swell height factor = 3 8 ft H O swell height factor = 4 Wave period The next oceanographic parameter reviewed is the peak wave period associated with the incoming swell (T P ). In Figure 5-5 the wave period is divided into 1-second intervals ranging from 4 to 13 seconds. The groups displayed outside this range include

PAGE 49

33 all occasions below 4 seconds and above 13 seconds respectively. The majority of the recorded wave periods occur in the range of 5 to 9 seconds, approximately 77% of all the years combined. Remember the analysis is limited to the summer months when the average wave period is generally shorter in comparison with the winter. It is apparent there is a high ratio of rescues on days when the wave period is between 6 and 8 seconds. In 1999 and 2003 there is a high risk of rescues between 7 and 8 seconds, represented by ratio values of 1.7 and 1.5 respectively. In 1999, 2000, 2001, and 2002 there is a high risk of rescues between 6 and 7 seconds, represented by ratio values of 1.9, 1.4, 1.3 and 1.5 respectively. This trend varies a bit from year to year, but overall there was an average of 9.1 rescues per day in the 6 to 8 second range, which is slightly above normal. The average for all the summers combined, excluding 2002, was 8.5 rescues per day. The ECFL LURCS only accounts for swells with periods longer than 8 seconds. However evidence has shown reason to include a slightly shorter period swell, especially as a result of the high beach attendance in the summer. The longer period swell remains an integral part in the progression of a hazardous surf environment (Figure 5-5). Basically the swell period factors were shifted down two seconds, still adhering to the structure of assigning a larger threat value for an increased swell period. The resulting contribution of the swell period parameter to the modified index calculation is as follows: T P < 6s swell period factor = 0 6s T P < 7s swell period factor = 0.5 7s T P < 9s swell period factor = 1 9s T P < 11s swell period factor = 2 11s T P swell period factor = 3

PAGE 50

34 Figure 5-5. Correlation histogram-plot of peak wave period (sec) along with rip current incidents and associated rescues for years 1998 to 2003. The blue (1 st ) bar represents the percentage of days that the peak wave period was within the respective range. The green (2 nd ) and red (3 rd ) bars represent the percentage of total rescues and the percentage of rip current incidents respectively, that occurred when the peak wave period was within that range. The numbers on the top of each plot correspond to the ratio of the blue and green bar magnitudes. Wave direction The first additional parameter introduced to the new index is offshore wave direction (D O ). Direction is considered to be a valuable factor when determining the level of rip current formation and subsequent danger to ocean-goers (McKenzie 1958, Sonu 1972, and Engle 2003). The next step is achieving a way to include swell direction into the index and presumably improve the accuracy.

PAGE 51

35 Figure 5-6. Correlation histogram-plot of offshore wave direction (deg) along with rip current incidents and associated rescues for years 1998 to 2003. The blue (1 st ) bar represents the percentage of days that the wave direction was within the respective range. The green (2 nd ) and red (3 rd ) bars represent the percentage of total rescues and the percentage of rip current incidents respectively, that occurred when the wave direction was within that range. The numbers on the top of each plot correspond to the ratio of the blue and green bar magnitudes. The shoreline azimuth where the lifeguards rescue information was obtained (Volusia County) is approximately 62. When viewing Figure 5-6, the incident wave field is considered orthogonal to shore if the angle is relatively close to this value. The figure shows a large number of incidences when the wave angle is greater than 90 degrees, accounting for the high number of southeast summer swells. Although more waves originate from the southeast in the summer, the greatest association with risk or danger to swimmers applies to shore normal conditions. Through general inspection of the size differences within bar groups, the greatest risk was associated with angles in the range of

PAGE 52

36 40 to 80 degrees. This is justified by the high ratio values within this degree range, which translates to approximately 20 degrees north and south of shore-normal. In the summer of 1999, when the wave angle was between 60 and 80 degrees, there was an average of 37 rescues per day. Although 1999 is probably an extreme case in comparison to the other years, it still illustrates how large of an effect wave direction can have. Direction was first incorporated into the new index in the same manner as wave height and period. The closer the wave direction was to shore normal, the larger the value of the direction factor. Then this factor was directly applied, through summation, to the rip current threat level. After the first couple of tests, which are discussed later in this chapter, there was little progression in the probability of detection and a slight increase in false alarms. The decision was then made to approach the inclusion of wave direction in another manner. The new approach consisted of using the wave direction factor as a multiplier of the other swell conditions (height and period). In this method the wave direction worked together with the other swell parameters to indirectly affect the threat level. The indirect association to the threat level was an attempt to decrease the false alarms occurring on days with an exceedingly small swell directed onshore. Also incorporated into the multiplicative parameter was a reduction of the threat level due to an oblique incident wave field. The results of the testing showed improvements in the overall performance of the index, not just the false alarm ratio. The positive response initiated additional effort to refine the method, and the contribution of the swell direction parameter was finalized as follows: -20 D O < 30 swell direction multiplier = 0.75 30 D O < 45 swell direction multiplier = 1.5

PAGE 53

37 45 D O 75 swell direction multiplier = 2 75 < D O 100 swell direction multiplier = 1.5 100 < D O 150 swell direction multiplier = 0.75 Else swell direction multiplier = 1 Low tide In the ECFL LURCS the tidal factor only pertains to an increase in the tidal range due to astronomical effects (Figure 4-3). Rip current researchers agree that low tide directly affects the formation of rip currents on a barred beach (Shepard 1941 and Brander 1999). Additionally, if a rip channel was already established, the decrease in water depth will intensify the rip current flow (Shepard 1941, Sonu 1972 and Brander 1999). Both produce an increased hazard for ocean-goers. Engle (2003) attempted to change the tidal parameter through a relation to the actual tidal level. This proved valuable when analyzing rescues taking place at different times throughout the day. The conclusion was the majority of the rescues occurred during the rising tide. However, the incorporation of tidal level into the index is difficult because the threat assessment is on a daily basis and the tidal level is changing periodically throughout the day. To account for the effect of low tide, the new tide factor will adjust depending on the time of day low tide occurs. In this study the greatest risk to swimmers occurs when low tide is between 10 a.m. and 12 p.m. (Figure 5-7). In 1998 there was an average of almost 21 rescues per day when low tide occurred between 10 a.m. and 12 p.m. This justifies the previous results from Engle (2003) because the rising tide would then occur during the middle of the day, when the beach is most populated.

PAGE 54

38 Figure 5-7. Correlation histogram-plot of the time of low tide (24hr) along with rip current incidents and associated rescues for each year from 1998 to 2003. The blue (1 st ) bar represents the percentage of days that low tide occurred within the respective time range. The green (2 nd ) and red (3 rd ) bars represent the percentage of total rescues and the percentage of rip current incidents respectively, that occurred when low tide occurred within that time range. The numbers on the top of each plot correspond to the ratio of the blue and green bar magnitudes. The increased risk of rescues on the rising tide could be attributed to mental aspects as well as physical. During low tide the waves break more violently on the sandbar and aid in the development of rip channels. The intense breaking detours people from swimming, then when the tide slowly raises the breaking intensity visually decreases and people feel secure enough to enter the water. The incident wave conditions might have only experienced minimal changes, persistently forcing the nearshore circulation system. Yet, more people are entering the water in confidence, which leads to an increased

PAGE 55

39 number of rescues. The new tidal parameter was adjusted to raise the rip current thrlevel if low tide coincides with the daytime beach attendance. The scale is slightly asymmetrical due to the increased rescue dependen eat ce found on the rising tide, and the contributitor = 0.5 the ECFL LURCS to validate the adjustments made, and observe any ents. Testi ulations indexc. on o the index calculation is as follows: Low Tide < 9am tidal factor = 0 9am Low Tide < 1pm tidal factor = 1 1pm Low Tide < 5pm tidal fact 5pm Low Tide tidal factor = 0 With the ranges of each parameter and their respective influence over the rip current index calculations established, the modified index was then tested against the existing form of improvem ng The ECFL LURCS and the newly modified index, with the adjusted parameters discussed in the previous section, were tested on a data set ranging from 1998 to 2003. Each index was computed on a daily basis using the oceanographic and meteorological conditions given by the OWI 3-G hindcast. The daily rip current threat level was attainedfrom each respective index and then compared with the lifeguard records of rip currents and their related rescues. Figure 5-8 is an example of the results of the index calcand the concurrent lifeguard records for the year 1998, after the changes to each parameter were finalized (for remaining years see appendix B). The horizontal lines represent the respective warning thresholds. If the plot of each respective rip current is above this threshold then it is recommended to issue a warning to the publi

PAGE 56

40 The purpose of the testing process was to substantiate the inclusion of wave direction and the modifications made to the other parameters in the rip current predictivindex. To assess the performance of both the ECFL LURCS and the modified indstatistical parameters were used (1) Alarm Ratio, (2) False Alarm Ratio Method 1, (3)False Alarm Ratio Method 2, (4) Correct Alarm Ratio (CAR), (5) Probability of Detection Method 1 e ex, six and (6) Probability of Detection Method 2. Each was computed using the rip current incidents as well as the related rescues documented by the Volusia County lifeguards. Figure 5-8. The daily-calculated threat values of both the ECFL LURCS and the ted rescues (bottom) for 1998. The horizontal lines represent the warning modified index (top) along with the daily rip current incidents with associathreshold for each respective index.

PAGE 57

41 Engle (2003) introduced the Alarm Ratio (AR) as a control when comparing a modified rip current predictive scale with an existing scale. The AR is defined as the percen tage of days each respective index would issue a warning. Days Index ThresholdTotal DaysThe False Alarm Ratio (FAR) is the measure of over warning. The first method(FAR1) is commonly used by the National Weather Service (NWS) and is calculatthe number AR = ed as of days an index issued a warning without the occurrence of rescues normalized by the total number of days the index issued a warning. The result is the percentage of warnings given by the respective rip current index when no rescues occurred. Days Index Threshold w/ no RescuesDays Index ThresholdThe Correct Alarm Ratio (CAR) is identified as the percent of wa FAR1 = rnings issued that coincide with rescues. The CAR is calculated as the number of days an index issued a warning and a rescue occurred normalized by the total number of days the index issued a warni ng. It is also calculated through the subtraction of FAR1 from 1. Days Index Threshold w/ RescuesDays Index Threshold CAR = = FAR1 1 The second method (FAR2) was developed by the author for use in the cyclic testing process as a measure of improvement or deterioration. The FAR2 is defined as the perce nt of the days in which no rescues occurred yet the index issued a warning. Days Index Threshold w/ no RescuesDays w/ no Rescues FAR2 =

PAGE 58

42 The Probability of Detection (POD) is approached through two methods. The firsmethod (POD1) was developed to assess the importance of the events detected b t y the rip current index. The POD1 is computed as the sum of rip related rescues on days the index issued a warning, normalized by the total amount of rescues. This statistical parameter establishes the im portance of detecting the days with high numbers of rescues. Sum of Rescues on Days Index ThresholdPOD1 = Total Rescues The second method (POD2) was directly adapted from the NWS and is defined as the pe rcentage of rip current events detected by the index. Sum of Days Index Threshold w/ RescuesPOD2 = Sum of Days w/ RescuesThe percentage growth of the POD compared with the FAR is reflected through theadditional POD/FAR ratio parameters. These ratios give the relative improvements of anyadjustments made to the index, and were used to find a warning threshold that would produce the maximum levels of performance. The POD2/FAR2 ratio proved hto the large number of rescues that occur during the summer season. Upon averaging all the years in the data set, excluding 2002, it was concluded that approximately 67%days in the summer season (day 75) exhibited documented rescues. The POD2/FAR2 ratio compares the percent of rescue days detected with the percent of nonrescue days also detected. If this ratio value is greater than 1, then the rip current predictive index is identifying a higher percentage of the days with rescues than the days without rescues. Theoretically if a warning was issued everyd elpful due of -ay of the summer, a 100 perce0 nt probability of detection would result. However this would also result in a 10

PAGE 59

43 percent of the false alarms. The POD2/FAR2 ratio helps to establish balance betweenincreases in the probability of detection and the false alarms. During the testing process four trials were completed in which an individual parameter was adapted based on the correlation analysis discussed earlier. After each adjustment, the modified index was computed from 1998 to 2003. The results of the modified index were then compar the ed with the performance of the ECFL LURCS to adjusted parameter. Comparisons between the perfo wind rameters displad pprox. S longshore wind factor only index values. The warning threshold for the modified index was adjusted to 3.5 to ascertain any improvements made by the rmances of each respective index were analyzed using the previously discussed statistical parameters. If the results were positive after each trial, the change remained and carried over to the next trial. Trial 1: Extraction of the wind factor The first of four trials in changing the ECFL LURCS was the removal of thefactor. After the removal, the daily threat values were computed by modified index over all the years. The performance changes are represented by the statistical pa yed in Table 5-2. Each parameter was calculated for all the years and then averagetogether. A positive percent change value is regarded as an improvement of the modified index over the ECFL LURCS. Again, the statistics were calculated for the year 2002 for qualitative inquisition but excluded from the averaging process. After examination of Table 5-2, it is evident that this adjustment is detrimental tothe performance of the index. The FAR1 for rescues increased from 0.25 to 0.35 (a21%), and the FAR1 for rip currents increased from 0.21 to 0.29 (approx. 20%). A slight increase in false alarms is expected as the ECFL LURC negates from the total threat value. Deleting this parameter will undoubtedly raise the

PAGE 60

44 account for mained significantly lower than the ECFL LURCS. -2. results of the modified index after trial 1 (extraction of the wind and the ol ECFLCS avever the years from 1998 to luding A pos changsponds provement CFL LURCS. fied ence nge the increase in threat values, yet the performance levels of the modified index re Table 5 Performance factor) rigina LUR raged o 2003 (exce E 2002) itive % e corre to an im over th Modi ECFL Differ % Cha AR 0.28 0.31 -0.02 -05.4 Rescues 0.35 0.25 -0.06 -21.4 FAR1 Rips 0.29 0.21 -0.05 -20.3 Rescues 0.29 0.23 -0.06 -25.3 FAR2 Rips 0.24 0.18 -0.05 -26.9 Rescues 0.27 0.34 -0.05 -14.4 POD2 Rips 0.28 0.34 -0.04 -12.7 Rescues 0.31 0.43 -0.09 -21.4 POD1 Rips Rescues 0.30 2.11 3.31 -1.20 0.42 -0.08 -21.7 -36.3 CF AR/ AR1Rescues 1.00 1.53 -0.54 -35.0 Rips 3.16 4.59 -1.43 -31.2 POD2/ Rescues 0.87 1.53 -0.65 -42.7 FAR2 Rips 1.39 2.00 -0.62 -30.8 POD2/ FAR1 Rips 1.19 1.96 -0.77 -39.3 ns limited examination of the correlation betwe Although the removal of the wind parameter was intended to improve the performance of the rip current predictive index, the opposite occurred. The longshore wind factor actually increases the accuracy of the ECFL LURCS predictions. These unexpected results might be the effect of human behavior in addition to unknown physical processes. The range of wind velocity that negates the rip current threat begiat 10 knots (11.5 mph) and extends to more than 25 knots (28.7 mph) (Figure 4-3). Theelevated levels of wind intensity might detour people from visiting the beach and therefore correlate to lower rescue numbers. Further analysis of the wind conditions would be needed to verify this assumption. The en the wind conditions and the rip current rescues in this study did not justify the exclusion of the wind parameter. Subsequently, it remained in the prediction scheme.

PAGE 61

45 Trial 2: Inclusion of a wave direction factor The second trial consisted of integrating the wave direction. The first attempt wasthe inclusion of a factor that directly added to the overall threat value, similar to the swell factors of height and period. This was achieved by increasing the direction factor as thewaves became more orthogonal to the shoreline, with a maximum value of 4 nearly shore-normal wave conditions. The approach was based on previous work doneEngle (2003). After the modification to the scale, the results showed an increase inPOD. However, the POD/FAR ratios were not convincing, denoting that the improvements in detection were counter-balance depicting by the d by an increase in false alarms. The correl a t reat level in concert with the swell condithe values of the POD/FAR and the CAR/FAR ratios. ations in this study and previous studies (see Engle 2003), illustrate a good association between wave direction and rip current events. The difficulty is within the incorporation of the wave direction parameter. After careful deliberation, the idea was developed to include the wave direction asmultiplier of the other swell parameters. The thought process behind the idea was torestrict an orthogonal incident wave field from offsetting the occurrence of swell conditions unfavorable for rip current formation. For example, assume the swell heighand period factors are both less than or equal to one and the direction is completely onshore. In an additive scheme, the direction parameter itself would have the ability to push the threat level into warning status. However, the wave conditions are likely too small for rip current formation and the resulting effect could be a false alarm. If the direction was a multiplier, it will indirectly raise the th tions. The performance results of this modification are exhibited in Table 5-3. The warning threshold was kept at 3 to generate maximum performance levels represented in

PAGE 62

46 There were slight improvements in the FAR1, FAR2 and POD2, but considerable enhancement in the POD1 when compared with the ECFL LURCS. The modified index showed an increase of approximately 28% and 24% in the POD1 analysis for rescues arips respectively. The larger increase in the POD1 symbolizes that the index is detecting agreater amount of the high-rescue days. The progress in both the false alarms and the detections is well portrayed by the POD2/FAR2 ratio values. There was an approxincrease of 10% and 19% in the ratio of general detections to false alarms (POD2/FARfor rescues and rip respectively. The ratio of detection importance to false alarms (POD2/FAR1) also experienced increases of 11% and almost 19% for rescue and rip analysis. This exemplifies the advances in the capability of the modifie nd imate 2) d index to detect e original ECable 5-3. erformance results of the modiex afclue rection), and thinal ECF 1998 xcludin2). A pos chanespondsement over the ECFL LURCS. fied ence nge rip current events when the multiplicative direction factor is applied compared with thFL LURCS, which contains no wave direction parameter. T P fied ind ter trial 2 (in sion of wav di e orig L LURCS averaged over the years from to 2003 (e g 200 itive % ge corr to an improv Modi ECFL Differ % Cha AR 0.32 0.31 0.01 03.7 Rescues 0.24 0.25 0.02 06.1 FAR1 Rips 0.20 0.21 0.01 04.9 Rescues 0.22 0.23 0.00 02.1 FAR2 Rips 0.18 0.18 0.00 00.1 Rescues 0.36 0.34 0.02 05.7 POD2 Rips 0.36 0.34 0.02 04.6 Rescues 0.55 0.43 0.12 28.1 POD1 Rips 0.53 0.42 0.10 24.3 Rescues 3.64 3.31 0.32 09.7 CAR/ FAR1 Rips 5.56 4.59 0.96 21.0 Rescues 1.69 1.53 0.16 10.4 P OD2/ 0.39 19.3 Rescues 1.70 1.53 0.17 11.2 FAR2 Rips 2.39 2.00 POD2/ FAR1 Rips 2.32 1.96 0.37 18.7

PAGE 63

47 Trial 3: Modification of the swell period factor The third trial included the modification of the wave period factor to account for slightly shorter period swells. It was shown in the previous section that there was a highrisk for swimmers when the wave period is between 6 and 8 seconds. However, the ECFL LURCS assigns a zero to this factor for wave conditions with a period less thaseconds. The adjustment to the index begins the wave period factor with a valfor a 6 second wave and n 8 ue of 0.5 increases accordingly. The statistical results after accounting for warning thbsequent changed. Table 5-4. results of the mo index al 3 (moion of the od factod the o ECFLS averar the years to 200luding. A pos changeponds to an ent over the ECFL LURCS. fied ence nge the shorter period waves are shown in Table 5-4. After a series of computations the reshold showed maximum performance at a value of 3.5, and was su ly Performanceeri dified after tri dificat swell p r), an riginal LURC ged ove from 1998vem 3 (exc 2002) itive % corres impro Modi ECFL Differ % Cha AR 0.34 0.31 0.03 09.6 Rescues 0.26 0.25 -0.00 -01.5 FAR1 Rips 0.22 0.21 -0.01 -05.4 Rescues 0.26 0.23 -0.03 -13.0 FAR2 Rips 0.22 0.18 -0.03 -18.0 Rescues 0.37 0.34 0.03 09.0 POD2 Rips 0.37 0.34 0.03 07.9 Rescues 0.58 0.43 0.15 35.0 POD1 Rips Rescu 0.56 3.17 0.42 3.31 0.13 -0.15 31.7 -04.4 es CF AR/ AR1 Rips 4.28 4.59 -0.31 -06.8 Rescues 1.48 1.53 -0.05 -03.3 POD2/ FAR2 Rips 1.86 2.00 -0.15 -07.5 Rescues 1.57 1.53 0.04 02.9 POD2/ FAR1 Rips 1.92 1.96 -0.03 -01.7 the The POD1 and POD2 for both rips and rescues are the only improvements overresults of trial 2. Both the FAR1 and the FAR2 experienced considerable deteriorationThe simultaneous increases in the probability of detection and the false alarm ratio are

PAGE 64

48 best examined through the POD2/FAR ratios. All of the ratios demonstrated inferior performance compared with the ECFL LURCS, and even greater reductions from the results of trial 2. After interpretation of the yearly performance results summar ized in ation of the swell period factor diminishes the quval of the astronomical tide factor and th -s nditions are only slightly favorable for rip formation, yet might trial, it f Table 5-4, it was concluded that the reorganiz ality of the rip current predictive index. Therefore, the changes of the wave period factor will not be incorporated into the modified index for the remaining trial. Trial 4: Redevelopment of the tidal factor The initial change in the fourth trial was the remo e integration of a low tide parameter. However, after a series of preliminary teststhe index exhibited superior results when both tidal factors were accounted for, and therefore the final trial was completed including both. Rip currents are known to intensify during low tide, and if this occurs while the beaches are heavily populated then the result is an increase in rip current risk to oceangoers. To account for the effect of low tide, a small factor was included into the index that is dependent on the time at which the low tide occurs. The max value of this factor i1. Therefore, it does not have as much influence as the other factors. Its purpose is to raise the threat level when the co need a lower water level to achieve potentially dangerous rip current activity. Asshown by the scale presented in the previous section, the threat level is affected only if low tide occurs during the day. The adjustments of the previous trial (wave period factor) were not retained and therefore comparisons should be made with the results of trial 2 in order to ascertain any improvements in the index due to the addition of a low tide parameter. For this was found favorable to keep the warning threshold to a value of 3.5. With the inclusion o

PAGE 65

49 the low tide factor, the overall performance of the modified index increased (Table 5-5). The FAR1 and FAR2 both improved from the results of trial 2 and continue to outperform the ECFL LURCS. The PO D1 and POD2 values decreased slightly (approx. ECFL LUR hich the r current risk i-5. results of the mo index ial 4 (redment of the ctor), anoriginal ELURCSged ovears from 03 (eing 2002). A positive ge corres to an ent over the ECFL LURCS fied ence nge 2% each) from trial 2, however the POD1 values remain considerably higher than the CS. This illustrates the ability of the modified index to detect the days in w ip s greater. Table 5 Performance dified after tr evelop tidal fa d the CFL avera r the ye 1998 to 20em xclud % chan spond improv Modi ECFL Differ % Cha AR 0.30 0.31 -0.00 -01.5 Rescues 0.23 0.25 0.03 10.3 FAR1 Rips 0.19 0.21 0.02 09.0 Rescues 0.21 0.23 0.02 09.4 FAR2 Rips 0.17 0.18 0.01 07.7 Rescues 0.35 0.34 0.01 02.2 POD2 Rips 0.35 0.34 0.00 00.7 Rescues 0.55 0.43 0.12 26.7 POD1 Rips Rescues 0.52 3.72 0.42 3.31 0.09 0.41 22.0 12.4 CF AR/ AR1Rescues 1.78 1.53 0.24 15.7 Rips 6.31 4.59 1.72 37.5 POD2/ Rescues 1.66 1.53 0.13 08.7 FAR2 Rips 2.74 2.00 0.73 36.5 POD2/ FAR1 Rips 2.49 1.96 0.53 27.3 the POD2/FAR1 for rescues, which decreased by only 2.5% from trial 2. However, this The relative improvement and/or worsening of the probability of detection and false alarm values is better viewed through the POD/FAR ratios. The POD2/FAR2 for rescues increased from 1.53 to 1.69 in trial 2, and then further improved to 1.78 in trial 4This indicates that the advances in the FAR2 values surmounted the small decrease in thePOD2 values from trial 2. There was significant progression in all the other ratios except

PAGE 66

50 ratio value still remained almost 9% better than the ECFL LURCS. The outcome of 4 does reinforce the assertion th trial at the occurrence of low tide during the day has an effect ards ocean-goers. Therefore, the adjusted tidal factor was retain. The of a high-risk day is broken into three categories (1) Days with at least 5 resculue CS. tremely on the rip current risk tow ed in the modified index. High-risk examination After the testing process and subsequent verification of the two modifications (direction and tide) made to the index, further analysis was completed to asses its capability in detecting the particularly important high-risk days. Table 5-6 displays the percentage of high rescue/rip days detected by each index for years 1998 to 2003classification es/rips, (2) Days with at least 15 rescues/rips, and (3) days with at least 25 rescues/rips. For days with at least 5 rescues the modified index detection percentage is an average of 49.5%, and the ECFL LURCS forecasted 39.3%. The average detection vais appropriately weighted for each year depending on how many rescues occurred in thatyear. Over the next two rescue categories both indexes increase their detection ratesHowever, the percentages of the modified index remain higher than the ECFL LURThe modified index detected 64.9% of the days when there was at least 15 rescues, compared to 50.4% by the ECFL LURCS. The modified index, but not the ECFL LURCS, detected the single rescue day in this category for 2002. In the final category, days with more than 25 rescues are examined. This category is associated with exdangerous rip current conditions. The modified index detected an average of 70.3% of those days, and the ECFL LURCS detected 52.7%. Overall, the modified index

PAGE 67

51 outperformed the ECFL LURCS by approximately 10% in the first category. In the final ategory the modified index increased to high-risk c almost 20% better performance. able 5-6: The percef th r r e 1998 to 2003. As well as the average of all the years weighted by the er of rips/rescues for each year. 8 9 0 1 2 3 ge tal T ntage o he hig ip current rescue and incident days fo ach year from numb 199 199 200 200 200 200 Avera / To Rescues >= 5 ECFL Rescu es 43.2 46.3 32.2 43.0 35.7 38.6 39.3 Mod. Rescu es 55.4 66.7 49.2 40.7 64.3 52.3 49.5 # of days 74 54 59 86 14 44 331 Rips >= 5 ECFL Rips 46.8 52.8 32.5 42.6 42.3 44.1 35.8 Mod. Rips 59.7 77.8 57.5 45.9 50.0 55.9 47.4 # of days 62 36 40 61 52 34 285 Rescues >= 15 ECFL Rescu es 46.2 59.1 46.2 47.6 00.0 64.3 50.4 Mod. Rescu es 64.1 90.9 76.9 47.6 100 71.4 64.9 # of days 39 22 13 42 01 14 131 Rips >= 15 ECFL Rips 66.7 46.2 42.9 53.8 35.7 50.0 43.6 Mod. Rips 83.3 100 100 50.0 71.4 25 50.0 59.0 # of days 12 13 07 26 14 06 78 Rescues >= ECFL Rescu es 50.0 57.1 50.0 53.6 50.0 52.7 Mod. Rescu es ips >= 25 66.7 100 100 53.6 50.0 70.3 # of days 18 14 08 28 00 06 74 R ECFL Rips 100 16.7 00.0 33.3 50.0 37.5 Mod. Rips 100 100 100 50.0 100 62.5 # of days 07 06 01 12 06 00 32 25 was approximately 5.4 and the average for the days with more than 25 rescues was Another facet of the high rescue examination was the determination of a warning threshold for higher risk rip current conditions. The daily threat level of the modifiedindex within each rescue category was determined. Figure 5-9 and Figure 5-10 display the threat values of each year for the days in which there were more than 15 andrescues respectively. The average threat value for the days with more than 15 rescues

PAGE 68

52 approximately 5.8. It was ascertained that the majority of high-rescue days occurred when the threat level was close to 5.5. As a result, it is recommended to issue a severe rip current warning if the index value is 5.5 or greater. Figure 5-9. Calculated rip current threat values for days with more than 15 rescues (1998). The horizontal line represents the weighted average value. Figure 5-10. Calculated rip current threat values for days with more than 25 rescues (1998). The horizontal line represents the weighted average value.

PAGE 69

53 Summary The parameters and their respective ranges were first analyzed through a subjective histogram approach. The summation of all the years for each parameter is presented in Figure 5-12. The swell height was found to have greatest rip current risk to ocean-goers in the 3 to 4 foot (0.91.22m) range. The swell period exhibited an increase in risk starting above 6 seconds, which differs from the 8-second cutoff used in the ECFL LURCS. The wave direction showed a high risk associated with a shore-normal incident wave field, which was in agreement with previous theory and observations. Figure 4-10 displays the increased risk associated with the direction of the incident wave field ranging from 40 to 80, or within 20 North or South of shore-normal. This high risk is exemplified by exhibiting the greatest overall ratio values as compared with the other parameters. As a result, wave direction has proven to be a crucial component when ascertaining hazardous rip current conditions. However, complications arose when attempting to include wave direction into the index, but were eventually overcome with the use of a multiplicative (rather than an additive) factor. The tidal influence differed from year to year, but in examination of the overall trend the greatest risk to ocean-goers occurred when low tide was in the late morning (10 a.m.). After each parameter change was incorporated into the modified index, it was then tested against the performance of the original ECFL LURCS. As a result of the testing procedure, the wind parameter and wave period factor remained unchanged. Reversely, the inclusion of the wave direction factor and the low tide parameter were justified. The final results of the testing process summarized over the years 1998 to 2003 (excluding 2002) are observed in Table 5-6. The performance of the newly developed rip current predictive index has shown considerable improvements when compared with the ECFL

PAGE 70

54 LURCS. To visually ascertain the difference between the two indexes the daily forecasts computed by the ECFL LURCS and the modified index for 1998 are displayed graphically in Figure 4-13 and Figure 4-14 respectively. The horizontal line represents the warning threshold, and each marker is a daily rip current threat value. The different markers signify the amount of rescues documented for that day. Figure 5-12. Correlation histogram-plot of the offshore wave height (ft), peak wave period (sec), offshore wave direction (deg), and time of low tide (24hr) along with rip current incidents and associated rescues combined over the years from 1998 to 2003 (excluding 2002). The blue (1 st ) bar represents the percentage of days that each parameter occurred within the respective range. The green (2 nd ) and red (3 rd ) bars represent the percentage of total rescues and the percentage of rip current incidents respectively, that occurred when the parameter occurred within that range. The numbers on the top of each plot correspond to the ratio of the blue and green bar magnitudes.

PAGE 71

55 A difference is noticed between the two indexes after examination of the higher risk days. The threat values calculated by the modified index have shifted vertically, accounting for the severity of conditions on those days. Also observed is the large number of the zero rescue days that remained under the warning threshold. There are some increases in threat values for the zero rescue days, but since the statistics are based on rescues and therefore ocean-goers, these anomalies could be the effect of inclement weather conditions. For similar plots of the remaining years (including separate analysis of both rip current incidents and rescues) see appendix-D. Although, not all the initially assumed parameter changes were incorporated into the modified index, the important assumptions of wave direction and low tide having an effect on rip current formation were resolved. The resulting rip current predictive index (RIPDEX) incorporating all observations can be seen in Figure 5-15. Figure 5-13. ECFL LURCS daily rip current threat levels for the summer of 1998. The daily rip current rescue totals are indicated by the marker symbols.

PAGE 72

56 Figure 5-14. The modified index daily rip current threat levels for the summer of 1998. The daily rip current rescue totals are indicated by the marker symbols.

PAGE 73

57 Figure 5-15. RIPDEX (rip current predictive index) worksheet.

PAGE 74

CHAPTER 6 USE OF INDEX AS FORECASTING TOOL Analysis The purpose of this project was not only to improve the existing rip current index (ECFL LURCS), but also to discover more efficient methods of implementing it as a forecasting tool. The use of the WAVEWATCH III model as an input to the index incorporates a forecasting ability, as well as the inclusion of wave direction; the latter of which proved to be beneficial to the performance of the index (see Chapter 5), and both of which had proved to be significant stumbling blocks previously. Application of the WAVEWATCH III model makes it possible to forecast the rip current threat in advance. With this capability the National Weather Service will have additional options when alerting the public of upcoming severe conditions. Since the forecasts are computed in advance, any warnings can be issued the day before a severe rip current event. Therefore the warning can be publicized on the evening news and in the daily newspaper. The knowledge of dangerous conditions in advance allows people to either change their plans accordingly, or at least be more prepared for the situation. Another advantage is the ability to re-establish staffing needs of the local lifeguards. If the index forecasts extreme rip current conditions, the beach safety division has the opportunity to place additional lifeguards on duty to compensate for the expected additional number of rescues. In this study the WAVEWATCH III outputs of wave height, period and direction were used to compute daily rip current threat levels with the modified index for the summer of 2005. The results of the 12pm model simulation from the day prior to the 58

PAGE 75

59 forecast day were used in the calculation of the rip current threat level. Due to the irregularity of the WAVEWATCH III output, the modified index values were calculated by the author. The moon phase and time of low tide was also integrated into the calculations, but the wind field was not included. The analysis was again limited to the summer season to minimize the spurious effects of low beach population issues. In 2005 there were 2,657 rip-related rescues documented by the Volusia County Lifeguards, with approximately 89% of those rescues occurring in the summer season. The summer season for this study is defined as day 79 until day 250, which translates into March 20 th until September 7 th The 4 day lag in the start of the summer season, when compared with the previous study, is due to the lack of WAVEWATCH III data for days prior. The resulting daily threat levels were then compared with the documented rip current rescues in Volusia County, along with the completed ECFL LURCS worksheets (Figure 4-3) and lifeguard observations (Figure 4-2) for the same time period. The ECFL LURCS worksheets and the lifeguard observations are sporadic over the summer of 2005 and consequently, the comparison with these two data sets was restricted to the days with a high number of rescues. The first analysis of the WAVEWATCH III rip current forecasting capabilities was an assessment of the overall performance using the statistical parameters described in chapter 5 (e.g., AR, FAR, POD). The results are displayed in Table 6-1. The Alarm Ratio was approximately 0.48, which is higher than the results of the ECFL LURCS and the modified index from the previous study (see chapter 5). However, the percentage of days with rip-related rescues in 2005 was higher than the average for 1998 to 2003 (89% compared with 68%). The value of the False Alarm Ratio Method 1 remained low at

PAGE 76

60 0.17, yet the value of the False Alarm Ratio Method 2 increased to 0.38. This was expected because of the increase in the number of days with rescues, and corresponding decrease in the number of non-rescue days. The Probability of Detection Method 2 showed considerable improvement over the results from chapter 5, increasing to 0.51. The Probability of Detection Method 1 also showed improvements, increasing to 0.61 and 0.57 for rescues and rips respectively. The POD/FAR ratio values were comparable to those from the previous multi-year analysis. The POD2/FAR2 ratio decreased slightly, yet it was expected because of the high FAR2 value resulting from a decrease in days in which no rescues occurred. The POD2/FAR ratio increased, representing an enhancement in the detection of days with a greater amount of rescues. Table 6-1. Performance results of the WAVEWATCH III forecasting the daily rip current threat levels for the summer of 2005. WAVEWATCH III AR 0.48 Rescues 0.17 FAR1 Rips 0.17 Rescues 0.38 FAR2 Rips 0.38 Rescues 0.51 POD2 Rips 0.51 Rescues 0.61 POD1 Rips 0.57 Rescues 4.88 CAR/ FAR1 Rips 4.88 Rescues 1.35 POD2/ FAR2 Rips 1.35 Rescues 3.03 POD2/ FAR1 Rips 3.03 Overall, the performance results appear to compare well with the statistical values obtained in the multi-year study discussed in chapter 5. A plot containing the daily threat levels and concurrent rescues is presented in Figure 6-1. The horizontal line represents the warning threshold and each marker signifies the index value as well as the amount of

PAGE 77

61 rescues occurring on that specific day. A large majority of the higher rescue days are above the warning threshold, and many of the non-rescue days remain below. In general, this represents a good performance by the modified index and the WAVEWATCH III data. There is a noticeable discrepancy occurring in the month of April, in which there are a few non-rescue days that have high index values. The ECFL LURCS worksheets displayed similar heightened threat levels for these days as well. This could be the result of stormy conditions, in which the rough surf made it unappealing for beach-goers to enter the ocean. Another possibility could be that the inclement weather kept them from going to the beach altogether. Independent of cause, the lack of rescues may likely represent a decrease in ocean-goers and does not necessarily guarantee an absence of rip current activity. Figure 6-1. Daily rip current threat levels computed with the modified index using the WAVEWATCH III oceanographic output information for the summer of 2005. The daily rip current rescue totals are indicated by the marker symbols.

PAGE 78

62 In order to further validate the use of the WAVEWATCH III data in the rip current prediction scheme an evaluation with existing methods (ECFL LURCS) for the same time period was needed. However, the ECFL LURCS worksheets recorded by the National Weather Service are discontinuous. Therefore, the comparison between the documented ECFL LURCS worksheets and the results from the modified index implementing the WAVEWATCH III data was limited to the high-rescue days. The lifeguard observations are even more intermittent than the worksheets. Subsequently, they were only used as a qualitative confirmation of dangerous rip current conditions. A total of 36 days occurred over the summer of 2005 in which there were more than 15 rescues and also an available ECFL LURCS worksheet. Examination of these days was used to determine if the application of the WAVEWATCH III model is a viable method of forecasting rip current conditions. Table 6-2 presents a summary of the WAVEWATCH III output conditions and subsequent threat levels, along with the threat levels of the ECFL LURCS worksheets for the high rescue days. In addition, the table also displays whether or not each respective index recommends the issuance of a warning. If a yes is followed by a w it indicates the weekend and therefore a lower warning threshold was used. The ECFL LURCS worksheets detected 22 of the 36 high-rescue days (approx. 61%) and the WAVEWATCH III forecasts detected 23 out of the 36 days (approx. 64%). Therefore, the forecasts using the WAVEWATCH III data actually performed slightly better than existing methods represented by the worksheets. In addition to the improved prediction quality, the WAVEWATCH III forecast gives the benefit of advanced notice. The analysis of the high-rescue days also emphasized the importance of incorporating

PAGE 79

63 wave direction into the prediction scheme. The average deviation of the wave direction from shore-normal (Volusia County, 62) on all 36 high-rescue days, weighted by the number of rescues, is approximately 23 degrees. Table 6-2. The WAVEWATCH III forecasted wave conditions and associated rip current threat levels, as well as the threat levels documented by the National Weather Service on the ECFL LURCS worksheets for the high-rescue days in the summer of 2005 (w corresponds to a weekend and subsequent decrease in the warning threshold). WAVEWATCH III Modified Index ECFL LURCS Month Day Rips Rescues Height (ft) Period (sec) Direction (deg) Threat Warning? Threat Warning? 4 21 8 17 1 8 57 3.5 yes 2.5 no 5 8 16 38 3 10 40 6.0 yes 5 yes 5 10 15 27 1 8 38 3.0 no 3.5 yes 5 12 11 25 3 9 48 7.0 yes 3.5 yes 5 13 8 27 2 8 50 4.0 yes 3 yes 5 22 19 35 2 8 102 2.6 no 3 yes 5 27 10 21 3 10 38 5.5 yes 3.5 yes 6 5 7 16 2 11 60 6.0 yes 2 no 6 6 28 90 2 10 60 5.5 yes 3 yes 6 7 49 113 2 9 103 3.0 no 3.5 yes 6 12 9 15 3 7 82 3.5 yes 2 no 6 21 13 21 3 7 30 4.0 yes 3.5 yes 6 22 22 47 3 8 40 5.3 yes 4 yes 6 23 11 17 2 7 44 3.0 no 2.5 no 6 25 36 66 4 8 88 4.8 yes 1 no 6 26 22 35 3 8 102 2.4 no 2 no 7 19 11 22 3 8 103 3.4 no 2.5 no 7 20 18 43 3 10 100 3.8 yes 3 yes 7 21 23 47 4 9 101 3.8 yes 4 yes 7 22 25 46 3 9 101 3.8 yes 3.5 yes 7 23 13 23 3 8 101 3.4 yes (w) 4 yes 7 24 8 15 1 9 103 2.1 no 2.5 yes (w) 7 25 22 40 3 6 31 4.0 yes 1.5 no 7 26 74 136 3 7 47 4.5 yes 2.5 no 8 2 10 16 3 8 101 3.4 no 3.5 yes 8 3 20 41 3 8 101 3.4 no 3 yes 8 4 40 74 2 7 99 2.5 no 3 yes 8 5 26 44 1 8 105 1.8 no 2.5 no 8 6 30 48 1 7 107 1.4 no 2 no 8 7 46 88 1 7 68 2.0 no 2.5 yes (w) 8 9 16 28 1 9 73 3.5 yes 3 yes 8 13 17 44 2 9 97 3.5 yes 3 yes 8 14 21 39 3 8 98 4.3 yes 4 yes 9 3 18 37 1 9 98 3.8 yes 1.5 no 9 4 39 73 5 6 60 7.5 yes 2 no 9 5 77 147 7 6 60 7.5 yes 2 no

PAGE 80

64 Scrutiny of the high-rescue days that were undetected by the modified index revealed an interesting characteristic. In the study, some of high-rescue days (May 22 nd ; June 7 th and 26 th ; July 19 th and 24 th ; August 2 nd 3 rd 5 th and 6 th ) that went undetected by the index exhibited a wave direction slightly above 100 degrees, representing a southeast swell. If the wave direction is between 100 and 150, the result is a swell reduction multiplier of 0.75. Rip currents are not entirely understood and although shore normal waves were shown to have a large effect on the formation of a rip (see chapter 5), a consistent long period swell from any direction could contribute to rip-rescue conditions. It can be seen that when there is a constant swell for multiple days, there are usually rescues somewhat independent of the wave direction (e.g., July 19 th to the 24 th Table 6-2). After further deliberation it was decided to lift the destructive multiplier if the period factor was greater than 0. This was an attempt to isolate the longer period swell and ignore the locally generated wind waves. As a result of the changes, the overall performance of the index improved (Table 6-3). The FAR and FAR2 values did not change, but both the POD1 and POD2 increased from the previous results. This enhancement in the detection capability of the modified index is better elucidated through the POD/FAR ratios. The POD2/FAR2 ratio increased from 1.35 to 1.47 (approx. 8.9%), and the POD2/FAR ratio increased from 3.03 to 3.53 (approx. 16.5%). The slight modification to the WAVEWATCH III index calculations also increased the detection of the previously discussed high-rescue days. Table 6-4 presents the WAVEWATCH III forecasts after the modification of the direction multiplier. The updated index detected 29 out of the 36 days (approx 81%), increasing 6 days from

PAGE 81

65 before. The results of the index after the modifications are also displayed graphically in Figure 6-2. Table 6-3. Performance results of the WAVEWATCH III forecasting the daily rip current threat levels for the summer of 2005, after the modification to the multiplicative direction factor. WAVEWATCH III AR 0.52 Rescues 0.16 FAR Rips 0.16 Rescues 0.38 FAR2 Rips 0.38 Rescues 0.56 POD2 Rips 0.56 Rescues 0.70 POD1 Rips 0.66 Rescues 5.25 CAR/ FAR Rips 5.25 Rescues 1.47 POD2/ FAR2 Rips 1.47 Rescues 3.53 POD2/ FAR Rips 3.53 Another interesting aspect of the results was the accuracy of employing both indexes in the rip current prediction scheme. Of the 36 high rescue days examined, only 3 days were undetected by both indexes (almost 92% accuracy). Implementation of the WAVEWATCH III model has exhibited promising results in comparison with the ECFL LURCS, however complete dependency on this new method is not yet recommended. At present utilizing both methods collectively can further increase the accuracy of the overall process. It seems each method possesses similar as well as dissimilar deficiencies; therefore it would prove beneficial to exercise both methods, with the result that the high-risk days undetected by one index would have the chance to be detected by the other index.

PAGE 82

66 Table 6-4. The WAVEWATCH III forecasted wave conditions and associated rip current threat levels (after modification to the multiplicative direction factor), as well as the threat levels documented by the National Weather Service on the ECFL LURCS worksheets for the high-rescue days in the summer of 2005 (w corresponds to a weekend and subsequent decrease in the warning threshold). WAVEWATCH III Modified Index ECFL LURCS Month Day Rips Rescues Height (ft) Period (sec) Direction (deg) Threat Warning? Threat Warning? 4 21 8 17 1 8 57 3.5 yes 2.5 no 5 8 16 38 3 10 40 6.0 yes 5 yes 5 10 15 27 1 8 38 3.0 no 3.5 yes 5 12 11 25 3 9 48 7.0 yes 3.5 yes 5 13 8 27 2 8 50 4.0 yes 3 yes 5 22 19 35 2 8 102 3.0 yes (w) 3 yes 5 27 10 21 3 10 38 5.5 yes 3.5 yes 6 5 7 16 2 11 60 6.0 yes 2 no 6 6 28 90 2 10 60 5.5 yes 3 yes 6 7 49 113 2 9 103 3.5 yes 3.5 yes 6 12 9 15 3 7 82 3.5 yes 2 no 6 21 13 21 3 7 30 4.0 yes 3.5 yes 6 22 22 47 3 8 40 5.3 yes 4 yes 6 23 11 17 2 7 44 3.0 no 2.5 no 6 25 36 66 4 8 88 4.8 yes 1 no 6 26 22 35 3 8 102 3.0 yes (w) 2 no 7 19 11 22 3 8 103 4.0 yes 2.5 no 7 20 18 43 3 10 100 4.5 yes 3 yes 7 21 23 47 4 9 101 4.5 yes 4 yes 7 22 25 46 3 9 101 4.5 yes 3.5 yes 7 23 13 23 3 8 101 4.0 yes 4 yes 7 24 8 15 1 9 103 2.5 no 2.5 yes (w) 7 25 22 40 3 6 31 4.0 yes 1.5 no 7 26 74 136 3 7 47 4.5 yes 2.5 no 8 2 10 16 3 8 101 4.0 yes 3.5 yes 8 3 20 41 3 8 101 4.0 yes 3 yes 8 4 40 74 2 7 99 2.5 no 3 yes 8 5 26 44 1 8 105 2.0 no 2.5 no 8 6 30 48 1 7 107 1.4 no 2 no 8 7 46 88 1 7 68 2.0 no 2.5 yes (w) 8 9 16 28 1 9 73 3.5 yes 3 yes 8 13 17 44 2 9 97 3.5 yes 3 yes 8 14 21 39 3 8 98 4.3 yes 4 yes 9 3 18 37 1 9 98 3.8 yes 1.5 no 9 4 39 73 5 6 60 7.5 yes 2 no 9 5 77 147 7 6 60 7.5 yes 2 no

PAGE 83

67 Figure 6-2. Daily rip current threat levels computed with the modified index using the WAVEWATCH III data for the summer of 2005 (after modification to the multiplicative direction factor). The daily rip current rescue totals are indicated by the marker symbols. Summary The newly developed index (Figure 5-15) was tested on an independent data set using the WAVEWATCH III forecasts from the previous day. The purposes of this analysis was to verify the improvements made to the ECFL LURCS and assess the forecasting capabilities of a rip current predictive index implementing readily available wave model forecasts. The investigation covered the summer of 2005 for the general performance of the WAVEWATCH III forecasts, but was limited to 36 high-rescue days for comparison with the recorded ECFL LURCS worksheets. The modified index outperformed the ECFL LURCS worksheets for the 36 considered days. Initially the new prediction method correctly detected approximately 64% of the days in which more than 15 rescues occurred. This performance was a slight

PAGE 84

68 improvement over the 61% detection rate of the ECFL LURCS worksheets for the same days. Then after improvements, the new method correctly detected over 80% of the high-rescue days. The modified index calculations for the 2005 summer resulted in a POD2/FAR2 ratio of 1.47 and a POD2/FAR ratio of 3.53. The resulting daily threat levels from the index calculations and the concurrent lifeguard records for the summer of 2005 are displayed in Figure 6-3. Almost every surge in the rips and/or rescues is associated with a rise in the calculated threat level. This parallel movement depicted in the graph explains the increase in the POD2/FAR ratio when compared with the results from the analysis in chapter 5. An increase in the POD2/FAR ratio accounts for an improvement in the detection of days with dangerous rip current activity, illustrated by the surge in rescues. The observed increases in the level of performance as well as the advantages of advanced notice justify the use of the WAVEWATCH III model as an input to the index calculations. However, it was noted that this new method of incorporating wave direction through the WAVEWATCH III model should not yet fully replace the existing ECFL LURCS worksheets. Instead the new method should be an additional tool in the rip current forecasting process to help increase accuracy as well as provide greater lead time for public warnings.

PAGE 85

69 Figure 6-3. The daily-calculated threat values of both the modified index using the WAVEWATCH III oceanographic output information (top) along with the daily rip current incidents with associated rescues (bottom) for the summer of 2005. The horizontal lines represent the warning threshold.

PAGE 86

CHAPTER 7 SUMMARY AND CONCLUSIONS The first part of this study focused on the development of a modified index for the prediction of rip currents, in which wave direction was included as a primary parameter. An examination of all parameters and their correlation with rip-related rescues was completed to establish the relative importance (and hence appropriate weighting values) of each in the modified rip current index. The swell parameter data indicated the greatest relative rip current risk when the swell height ranged between 3 and 4 feet and the swell period was above 6 seconds. The swell height range was well represented in the existing ECFL LURCS and therefore remained unchanged. A slight shift in the swell period factor was thought to improve the performance of the index, yet the testing results proved otherwise. As a result the swell period parameter also remained unchanged. The incident wave direction exhibited the greatest risk to ocean-goers when the angle was between 40 and 80 degrees, representing a range within approximately 20 degrees north and south of shore-normal in the area of Volusia County. These results reinforce the importance of wave direction as an indicator of rip current threat previously noted in a study by Engle (2003). The swell direction was included into the modified index as a multiplicative factor. The tide data indicated the greatest risk for rip currents when low tide occurred during the day, especially between 10 a.m. and 12 p.m. The time of low tide was included into the index in addition to the existing astronomical tide factor. Rip current formation is also influenced by the beach morphology. This becomes a problem for the prediction scheme because the nearshore bathymetry can be inconsistent 70

PAGE 87

71 along the coast of Florida, and transform with varying wave conditions. Therefore some variability will always be present in the correlation and testing results. This is due to the fact that the lifeguard records cover the entire coast of Volusia county, yet there is a near complete absence of time-dependent bathymetry data for the same area. However, the morphological dependence is indirectly incorporated into the index through the persistent swell factor. If there is consistent swell conditions that assist in the formation of a longshore bar, trough and rip channel system, then the index threat value for the days following will be subsequently increased. The overall trends in each parameter are well represented in the modified rip current predictive index, and each change was proven through the testing process. Some of the parameters may appear artificially oriented towards the summer season, which is expected because the analysis was restricted to the summer. The winter season does pose a threat for rip current formation, but there is a considerable decrease in the number of ocean-goers. The analysis was dependent on rip-related rescues serving as a proxy for rip current formation and therefore the examination was limited to the summer season (Day 75). This period corresponds to the highest population of beach-goers and subsequently the majority of the annual rescues. From 1998 to 2005 over 80% of all rip-related rescues occurred during the summer (as defined above). It must be noted that the use of rip current rescues as an indicator or proxy for actual rip current events is not without its drawbacks. This method suffers from the spurious effects of beach population as noted earlier (a lack of swimmers in bad weather may lead to no rescues, but this is not a guarantee that rip currents did not occur). However, in the absence of any direct long-term measurements of rip currents, this

PAGE 88

72 approach provides the best long-term data set to enable the correlation of rip current activity to the relevant meteorological and oceanographic forcing. The restriction of analysis to the summer months (when beach populations are generally large) helps in the reduction of these spurious effects. The new index demonstrated considerable improvements over the original ECFL LURCS when tested on the documented lifeguard rescues from 1998 to 2003. The Probability of Detection Method 1 increased 26.7% and 22.0% for rescues and rips respectively. However, the Probability of Detection Method 2 increased only marginally (2.2% and 0.7%), signifying that the existing ECFL LURCS system detected a similar amount of rescue events. The False Alarm Ratio Method 1 decreased (improved) 10.3% and 9.0% for the rescues and rips. The False Alarm Ratio Method 2 also displayed improvement with decreases of 9.4% for rescues and 7.7% for rips. The combined progression of the False Alarm Ratio and the Probability of Detection is illustrated by each ratio value. The POD2/FAR2 increased 15.7% for rescues and 36.5% for rips. The POD2/FAR increased 8.7% for rescues and 27.3% for rips. In general, the modified index performed more accurately than the ECFL LURCS. There was a decrease in the false alarms and an increase in the detection of the higher risk days, which are represented by an increased number of rescues. The second part of this study was geared toward the implementation of the predictive index as a forecasting tool. This was accomplished by employing a forecast (directional) wave model called WAVEWATCH III, which is a significant step since the east coast of Florida lacks a long-term network of directional wave measurements. The performance of the modified index employing oceanographic conditions given by

PAGE 89

73 WAVEWATCH III compared well with the statistical values obtained from the analysis executed in the first part of the study. The new method was also compared with the ECFL LURCS worksheets on days in which there were more than 15 rescues. After improvements, the modified index detected 81% of the high-rescue days, while the ECFL LURCS worksheets only detected 61% of those same events. Although this new system of rip current prediction shows significant improvements over the existing method, there is still difficulty within the process of creating a rip current forecast. The main problem is the interpretation of the WAVEWATCH III output parameters. Evaluating these parameters can become rather subjective, particularly if there is not an explicitly dominant swell. Care must be taken during this phase and therefore the user should be somewhat aware of the mechanics of rip current formation, as well as other important factors. These factors include, but are not limited to, the swell conditions and the number of rescues occurring in days prior to the forecasted day. To achieve the most consistent results, those who calculate the rip current threat level using the WAVEWATCH III model and the modified index should implement similar protocol. The new method is not intended to replace the existing system completely, however it would certainly assist in the overall process of rip current prediction. It has been shown that shore-normal waves frequently result in a higher risk environment. Therefore, it is beneficial to use the WAVEWATCH III model to incorporate wave direction (which is otherwise unavailable from any real-time measurement platforms). Another significant advantage to this new method is the advanced notice of severe rip current conditions. By using the forecast wave field rather than same-day measurements, the extra time allows for earlier and maybe more effective issuance of warnings to the public through the

PAGE 90

74 media and other venues. It also allows for the beach safety division to staff the local lifeguard stations accordingly, in preparation for the additional rescues anticipated with heightened threat levels. These advancements would hopefully decrease the number of rip current drownings in Florida. The addition of wave direction might also assist in the extension of the application of a rip current predictive index to other coastal locations outside East Central Florida as well. However, further study would likely be needed to re-calibrate the other index parameters in order to reflect local wave and tidal conditions. For now, the next logical step in this process is to use the new predictive index incorporating WAVEWATCH III in an operational mode (perhaps side by side with the ECFL LURC worksheets) to make forecasts and comparative tests on a daily basis.

PAGE 91

APPENDIX A WAVE CONDITIONS AND RIP CURRENT RESCUES Figure A-1. Time-series plots of wave height (ft), wave period (sec), wave direction (deg) and rip current incidents with associated rescues for 1997. The x marks correspond to days with more than 15 rescues. 75

PAGE 92

76 Figure A-2. Time-series plots of wave height (ft), wave period (sec), wave direction (deg) and rip current incidents with associated rescues for 1998. The x marks correspond to days with more than 15 rescues.

PAGE 93

77 Figure A-3. Time-series plots of wave height (ft), wave period (sec), wave direction (deg) and rip current incidents with associated rescues for 1999. The x marks correspond to days with more than 15 rescues.

PAGE 94

78 Figure A-4. Time-series plots of wave height (ft), wave period (sec), wave direction (deg) and rip current incidents with associated rescues for 2000. The x marks correspond to days with more than 15 rescues.

PAGE 95

79 Figure A-5. Time-series plots of wave height (ft), wave period (sec), wave direction (deg) and rip current incidents with associated rescues for 2001. The x marks correspond to days with more than 15 rescues.

PAGE 96

80 Figure A-6. Time-series plots of wave height (ft), wave period (sec), wave direction (deg) and rip current incidents with associated rescues for 2002. The x marks correspond to days with more than 15 rescues.

PAGE 97

81 Figure A-7. Time-series plots of wave height (ft), wave period (sec), wave direction (deg) and rip current incidents with associated rescues for 2003. The x marks correspond to days with more than 15 rescues.

PAGE 98

82 Figure A-8. Time-series plots of wave height (ft), wave period (sec), wave direction (deg) and rip current incidents with associated rescues for 2004. The x marks correspond to days with more than 15 rescues.

PAGE 99

APPENDIX B RIP CURRENT THREAT VALUES AND RESCUES Figure B-1. The daily-calculated threat values of both the ECFL LURCS and the modified index (top) along with the daily rip current incidents with associated rescues (bottom) for 1997. The horizontal lines represent the warning threshold for each respective index. 83

PAGE 100

84 Figure B-2. The daily-calculated threat values of both the ECFL LURCS and the modified index (top) along with the daily rip current incidents with associated rescues (bottom) for 1998. The horizontal lines represent the warning threshold for each respective index.

PAGE 101

85 Figure B-3. The daily-calculated threat values of both the ECFL LURCS and the modified index (top) along with the daily rip current incidents with associated rescues (bottom) for 1999. The horizontal lines represent the warning threshold for each respective index.

PAGE 102

86 Figure B-4. The daily-calculated threat values of both the ECFL LURCS and the modified index (top) along with the daily rip current incidents with associated rescues (bottom) for 2000. The horizontal lines represent the warning threshold for each respective index.

PAGE 103

87 Figure B-5. The daily-calculated threat values of both the ECFL LURCS and the modified index (top) along with the daily rip current incidents with associated rescues (bottom) for 2001. The horizontal lines represent the warning threshold for each respective index.

PAGE 104

88 Figure B-6. The daily-calculated threat values of both the ECFL LURCS and the modified index (top) along with the daily rip current incidents with associated rescues (bottom) for 2002. The horizontal lines represent the warning threshold for each respective index.

PAGE 105

89 Figure B-7. The daily-calculated threat values of both the ECFL LURCS and the modified index (top) along with the daily rip current incidents with associated rescues (bottom) for 2003. The horizontal lines represent the warning threshold for each respective index.

PAGE 106

90 Figure B-8. The daily-calculated threat values of both the ECFL LURCS and the modified index (top) along with the daily rip current incidents with associated rescues (bottom) for 2004. The horizontal lines represent the warning threshold for each respective index.

PAGE 107

APPENDIX C DAILY SUMMER THREAT VALUES, DAILY RESCUE TOTALS AND RESULTING PERFORMANCE STATISTICS

PAGE 108

92 Figure C-1. ECFL LURCS daily rip current threat levels for the summer of 1998. The daily rip current rescue totals are indicated by the marker symbols. Figure C-2. The modified index daily rip current threat levels for the summer of 1998. The daily rip current rescue totals are indicated by the marker symbols.

PAGE 109

93 Figure C-3. ECFL LURCS daily rip current threat levels for the summer of 1999. The daily rip current rescue totals are indicated by the marker symbols. Figure C-4. The modified index daily rip current threat levels for the summer of 1999. The daily rip current rescue totals are indicated by the marker symbols.

PAGE 110

94 Figure C-5. ECFL LURCS daily rip current threat levels for the summer of 2000. The daily rip current rescue totals are indicated by the marker symbols. Figure C-6. The modified index daily rip current threat levels for the summer of 2000. The daily rip current rescue totals are indicated by the marker symbols.

PAGE 111

95 Figure C-7. ECFL LURCS daily rip current threat levels for the summer of 2001. The daily rip current rescue totals are indicated by the marker symbols. Figure C-8. The modified index daily rip current threat levels for the summer of 2001. The daily rip current rescue totals are indicated by the marker symbols.

PAGE 112

96 Figure C-9. ECFL LURCS daily rip current threat levels for the summer of 2002. The daily rip current rescue totals are indicated by the marker symbols. Figure C-10. The modified index daily rip current threat levels for the summer of 2002. The daily rip current rescue totals are indicated by the marker symbols.

PAGE 113

97 Figure C-11. ECFL LURCS daily rip current threat levels for the summer of 2003. The daily rip current rescue totals are indicated by the marker symbols. Figure C-12. The modified index daily rip current threat levels for the summer of 2003. The daily rip current rescue totals are indicated by the marker symbols.

PAGE 114

APPENDIX D DAILY SUMMER THREAT VALUES, DAILY RIP TOTALS AND RESULTING PERFORMANCE STATISTICS

PAGE 115

99 Figure D-1. ECFL LURCS daily rip current threat levels for the summer of 1998. The daily rip current incident totals are indicated by the marker symbols. Figure D-2. The modified index daily rip current threat levels for the summer of 1998. The daily rip current incident totals are indicated by the marker symbols.

PAGE 116

100 Figure D-3. ECFL LURCS daily rip current threat levels for the summer of 1999. The daily rip current incident totals are indicated by the marker symbols. Figure D-4. The modified index daily rip current threat levels for the summer of 1999. The daily rip current incident totals are indicated by the marker symbols.

PAGE 117

101 Figure D-5. ECFL LURCS daily rip current threat levels for the summer of 2000. The daily rip current incident totals are indicated by the marker symbols. Figure D-6. The modified index daily rip current threat levels for the summer of 2000. The daily rip current incident totals are indicated by the marker symbols.

PAGE 118

102 Figure D-7. ECFL LURCS daily rip current threat levels for the summer of 2001. The daily rip current incident totals are indicated by the marker symbols. Figure D-8. The modified index daily rip current threat levels for the summer of 2001. The daily rip current incident totals are indicated by the marker symbols.

PAGE 119

103 Figure D-9. ECFL LURCS daily rip current threat levels for the summer of 2002. The daily rip current incident totals are indicated by the marker symbols. Figure D-10. The modified index daily rip current threat levels for the summer of 2002. The daily rip current incident totals are indicated by the marker symbols.

PAGE 120

104 Figure D-11. ECFL LURCS daily rip current threat levels for the summer of 2003. The daily rip current incident totals are indicated by the marker symbols. Figure D-12. The modified index daily rip current threat levels for the summer of 2003. The daily rip current incident totals are indicated by the marker symbols.

PAGE 121

LIST OF REFERENCES Bowen, A. J., D. L. Inman and V. P. Simmons. 1968. Wave set-down and set-up. Journal of Geophysical Research, 73: 2569-2577. Bowen, A. J. 1969. Rip Currents, 1: Theoretical investigations. Journal of Geophysical Research, 74: 5467-5478. Bowen, A. J. and D. L. Inman. 1969. Rip Currents, 2: Laboratory and field observations. Journal of Geophysical Research, 74: 5479-5490. Brander, R. W. 1999. Field Observations on the morphodynamics evolution of a low-energy rip current system. Marine Geology, 157 (3-4): 199-217. Brander, R. W. and A. D. Short. 2001. Flow kinematics of low-energy rip current systems. Journal of Coastal Research, 17 (2): 468-481. Charles, L., R. Malakar, R. G. Dean. 1994. Sediment data for Floridas east coast. Report, Department of Civil and Coastal Engineering, University of Florida. Cook, D. O. 1970. The occurrence and geological work of rip currents off southern California. Marine Geology, 9: 173-186. Dalrymple, R. A. 1978. Rip currents and their causes. Proceedings 16 th International Conference Coastal Engineering, American Society of Civil Engineers, Hamburg, 1414-1427. Engle, J. A. 2003. Formulation of a rip current forecasting technique through statistical analysis of rip current-related rescues. Master of Science Thesis, University of Florida. Haas, K. A. and I. A. Svendsen, R. W. Brander, and P. Nielsen. 2002. Modeling of a rip current system on Moreton Island, Australia. Proceedings of the 28 th International Conference on Coastal Engineering, Coastal Engineering Research Council, ASCE, Cardiff, Wales, pp. 784-796. Haller, M. C., R. A. Dalrymple, and I. A. Svendsen. 1997. Rip channels and nearshore circulation: Experiments. In Proceedings of the 3 rd International Symposium on Ocean Wave Measurement and Analysis, ASCE, pp. 750-764. Haller, M. C. and R. A. Dalrymple. 2001. Rip current instabilities. Journal Fluid Mechanics, 433: 161-192. 105

PAGE 122

106 Lascody, R. L. 1998. East Central Florida rip current program. National Weather Digest, 22 (2): 25-30. Longuet-Higgins, M. S. and R. W. Stewart. 1964. Radiation stress in water wave, a physical discussion with applications. Deep Sea Research, 11 (4): 529-563. Lushine, J. B. 1991. A study of rip current drownings and related weather factors. National Weather Digest, 16: 13-19. MacMahan, J. H., A. J. H. M. Reniers, E. B. Thornton and T. P. Stanton. 2004. Infragravity rip current pulsations. Journal of Geophysical Research, 109 (C01033), doi: 10.1029/2003JC002068. MacMahan, J. H., E. B. Thornton and A. J. H. M. Thornton. 2006. Rip current review. Journal of Coastal Engineering, 53: 191-208. Mckenzie, R. 1958. Rip current systems. Journal of Geology, 66: 103-113. NOAA. 1997. Station 41009 historical data. National Data Buoy Center, available [on-line], accessed 8/15/05: http://www.ndbc.noaa.gov/station_history.php?station=41009 NOAA. 1997. Station 8721604 (Trident Pier) historical data. Tides Online, available [on-line], accessed 9/17/05: http://co-ops.nos.noaa.gov/data_res.html Shepard, F. P. 1936. Undertow, rip tide or rip current. Science, 84: 181-182. Shepard, F. P., K. O. Emery and E. C. Lafond. 1941. Rip currents: A process of geological importance. Journal of Geology, 49: 338-369. Shepard, F. P. and D. L. Inman. 1950. Nearshore circulation. Proceedings of the 1 st Conference on Coastal Engineering, Council on Wave Research, Berkeley, CA. pp. 50-59. Short, A. D. 1985. Rip current type, spacing and persistence, Narrabeen Beach, Australia. Marine Geology, 65: 47-71. Sonu, C. J. 1972. Field observations of nearshore circulation and meandering currents. Journal Geophysical Research, 77: 3232-3247. Swail, V. R. and A. T. Cox. 2000. On the use of NCEP/NCAR reanalysis surface marine wind fields for a long term north Atlantic wave hindcast. Journal of Atmospheric Technology, 17 (4): 532-545. Tolman, H. L. 1997. User manual and system documentation of WAVEWATCH III version 1.15. NOAA / NWS / NCEP / OMB Technical Note 151, pp. 97.

PAGE 123

107 Tolman, H. L. and N. Booij. 1998. Modeling wind waves using wavenumber-direction spectra and a variable wavenumber grid. The Global Atmospheric and Ocean System, 6: 295-309. Tolman H. L. 1999a. User manual and system documentation of WAVEWATCH III version 1.18. NOAA / NWS / NCEP / OMB Technical Note 166, pp. 110. U.S. Naval Observatory. 1997. Phases of the Moon. Astronomical Applications Department, available [on-line], accessed 4/20/06: http://aa.usno.navy.mil/data/docs/MoonPhase.html WAMDI Group. 1998. The WAM model A third generation ocean wave prediction model. Journal of Physical Oceanography, 18: 1775-1810.

PAGE 124

BIOGRAPHICAL SKETCH Jason Cummins was born on January 1, 1982 in Fort Lauderdale, Florida. Although it was a difficult decision to depart from the coast, in 2000, he moved to Gainesville, Florida to become a gator and an engineer. After four years of football in the swamp and some studying, he graduated with a Bachelors degree in Civil Engineering. He was then presented the opportunity to do research on rip currents under the supervision of Dr. Robert Thieke. The idea of learning about the coastal processes he has been witness to his whole life, as well as the promise of regular visits to the beach definitely caught his attention. He gladly accepted and the research he completed at the University of Florida led to a Master of Science degree in Coastal and Oceanographic Engineering. His future work will certainly involve the application of the knowledge he gained during his six years in Gainesville, as well as his love of the ocean. 108


Permanent Link: http://ufdc.ufl.edu/UFE0016160/00001

Material Information

Title: Improving Rip Current Forecasting Techniques for the East Coast of Florida
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0016160:00001

Permanent Link: http://ufdc.ufl.edu/UFE0016160/00001

Material Information

Title: Improving Rip Current Forecasting Techniques for the East Coast of Florida
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0016160:00001


This item has the following downloads:


Table of Contents
    Title Page
        Page i
        Page ii
    Dedication
        Page iii
    Acknowledgement
        Page iv
        Page v
        Page vi
    List of Tables
        Page vii
        Page viii
    List of Figures
        Page ix
        Page x
        Page xi
        Page xii
        Page xiii
        Page xiv
    Abstract
        Page xv
        Page xvi
    Introduction
        Page 1
        Page 2
        Page 3
    Rip currents
        Page 4
        Page 5
        Page 6
        Page 7
        Page 8
        Page 9
        Page 10
    Forecasting
        Page 11
        Page 12
        Page 13
        Page 14
        Page 15
    Data
        Page 16
        Page 17
        Page 18
        Page 19
        Page 20
        Page 21
        Page 22
    Development of improved rip current index
        Page 23
        Page 24
        Page 25
        Page 26
        Page 27
        Page 28
        Page 29
        Page 30
        Page 31
        Page 32
        Page 33
        Page 34
        Page 35
        Page 36
        Page 37
        Page 38
        Page 39
        Page 40
        Page 41
        Page 42
        Page 43
        Page 44
        Page 45
        Page 46
        Page 47
        Page 48
        Page 49
        Page 50
        Page 51
        Page 52
        Page 53
        Page 54
        Page 55
        Page 56
        Page 57
    Use of index as forecasting tool
        Page 58
        Page 59
        Page 60
        Page 61
        Page 62
        Page 63
        Page 64
        Page 65
        Page 66
        Page 67
        Page 68
        Page 69
    Summary and conclusions
        Page 70
        Page 71
        Page 72
        Page 73
        Page 74
    Appendices
        Page 75
        Page 76
        Page 77
        Page 78
        Page 79
        Page 80
        Page 81
        Page 82
        Page 83
        Page 84
        Page 85
        Page 86
        Page 87
        Page 88
        Page 89
        Page 90
        Page 91
        Page 92
        Page 93
        Page 94
        Page 95
        Page 96
        Page 97
        Page 98
        Page 99
        Page 100
        Page 101
        Page 102
        Page 103
        Page 104
    References
        Page 105
        Page 106
        Page 107
    Biographical sketch
        Page 108
Full Text












IMPROVING RIP CURRENT FORECASTING TECHNIQUES
FOR THE EAST COAST OF FLORIDA

















By

JASON ROLLAND CUMMINS


A THESIS PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
MASTER OF SCIENCE

UNIVERSITY OF FLORIDA


2006

































Copyright 2006

by

Jason Rolland Cummins

































To Arthur and Mildred Cummins.















ACKNOWLEDGMENTS

I would first like to thank all of my family, for their unconditional love and support

throughout my college career.

I thank Dr. Robert Thieke for giving me the opportunity to work on such an

interesting project. His knowledge and guidance were always made available to me. I

also thank Dr. Andrew Kennedy and Dr. Ashish Mehta, for their participation on my

supervisory committee, as well as their insight into the research.

I extend my appreciation to the Florida Sea Grant Program, for their financial

support. I thank Oceanweather Inc., for liberally supplying the hindcast wind and wave

conditions needed for the project. I also thank the Volusia County Beach Safety Division,

for providing the lifeguard rescue information.
















TABLE OF CONTENTS



A C K N O W L E D G M E N T S ................................................................................................. iv

LIST OF TABLES ................................ .. .......... .................................. vii

LIST OF FIGURES ........................................ .............. ix

A B ST R A C T ...................................................................................................... ....... .. xv

CHAPTER

1 IN T R O D U C T IO N ................................................. .............................................. .

2 R IP C U R R E N T S ...................................................................................................4..

F orm ation of a R ip C urrent.......................................... ......................... .............. .4...
C characteristics of a R ip ... ...................................................................... .............. .8...

3 FORECASTING........................ ......... ............... 11

4 D A T A ........................................................................................................ .......... 16

S ite D e scrip tio n .......................................................................................................... 16
Rip Current Rescues ......................... .... .......... ....................... 17
H in d ca st D ata .............................................................................................................. 18
T id e s .......................................................................................................... ....... .. 19
WAVEWATCH III..................... .. ........... ..................................... 19
L ifeguard O b servations ................. ........................................................ ................ 20
Completed ECFL LURCS Worksheets ...................................................20
Melbourne Beach and NOAA Data Buoys............................................................20

5 DEVELOPMENT OF IMPROVED RIP CURRENT INDEX...............................23

A n a ly sis ...................................................................................................................... 2 3
O cean C orrelations .............. .... ............. ................................................ 27
W ave height......................................................................................... .. 30
W ave period ........................................................................................ 32
W ave direction ................ .............. ............................................ 34
L o w tid e ........................................................................................................ 3 7



v









T testing ....................................................... .. .................... ............... 39
Trial 1: Extraction of the w ind factor...................................... ................ 43
Trial 2: Inclusion of a wave direction factor ........................... ................ 45
Trial 3: Modification of the swell period factor......................................47
Trial 4: Redevelopm ent of the tidal factor .............................. ................ 48
H igh-risk exam nation ............................................................. ............... 50
S u m m a ry .................................................................................................................. .. 5 3

6 USE OF INDEX AS FORECASTING TOOL ............................. ..................... 58

A n a ly sis ...................................................................................................................... 5 8
S u m m a ry .................................................................................................................. ... 6 7

7 SUM M ARY AND CONCLUSION S.................................................... ................ 70

APPENDIX

A WAVE CONDITIONS AND RIP CURRENT RESCUES...................................75

B RIP CURRENT THREAT VALUES AND RESCUES .................. ..................... 83

C DAILY SUMMER THREAT VALUES, DAILY RESCUE TOTALS AND
RESULTING PERFORMANCE STATISTICS ...................................................91

D DAILY SUMMER THREAT VALUES, DAILY RIP TOTALS AND
RESULTING PERFORMANCE STATISTICS ...................................................98

LIST O F R EFEREN CE S ... ................................................................... ............... 105

BIOGRAPH ICAL SKETCH .................. .............................................................. 108















LIST OF TABLES


Table page

5-1 Percentage of rip-related rescues occurring in the summer (defined as day 75-
250), 1998 to 2003 ........................................................................................... 25

5-2 Performance results of the modified index after trial 1 (extraction of the wind
factor), and the original ECFL LURCS averaged over the years from 1998 to
2003 (excluding 2002). A positive % change corresponds to an improvement
over the E C FL L U R C S ......................................................................... ................ 44

5-3 Performance results of the modified index after trial 2 (inclusion of wave
direction), and the original ECFL LURCS averaged over the years from 1998 to
2003 (excluding 2002). A positive % change corresponds to an improvement
over the E C FL L U R C S ......................................................................... ................ 46

5-4 Performance results of the modified index after trial 3 (modification of the swell
period factor), and the original ECFL LURCS averaged over the years from
1998 to 2003 (excluding 2002). A positive % change corresponds to an
im provem ent over the ECFL LURCS ................................................ ................ 47

5-5 Performance results of the modified index after trial 4 (redevelopment of the
tidal factor), and the original ECFL LURCS averaged over the years from 1998
to 2003 (excluding 2002). A positive % change corresponds to an improvement
over the E C FL L U R C S .......................................... .......................... ................ 49

5-6 The percentage of the high rip current rescue and incident days for each year
from 1998 to 2003. As well as the average of all the years weighted by the
num ber of rips/rescues for each year................................................... ................ 51

6-1 Performance results of the WAVEWATCH III forecasting the daily rip current
threat levels for the sum m er of 2005................................................... ................ 60

6-2 The WAVEWATCH III forecasted wave conditions and associated rip current
threat levels, as well as the threat levels documented by the National Weather
Service on the ECFL LURCS worksheets for the high-rescue days in the
summer of 2005 ("w" corresponds to a weekend and subsequent decrease in the
warning threshold) ................... .... ........... .........................63









6-3 Performance results of the WAVEWATCH III forecasting the daily rip current
threat levels for the summer of 2005, after the modification to the multiplicative
d irectio n fact r. ........................................................................................................ 6 5

6-4 The WAVEWATCH III forecasted wave conditions and associated rip current
threat levels (after modification to the multiplicative direction factor), as well as
the threat levels documented by the National Weather Service on the ECFL
LURCS worksheets for the high-rescue days in the summer of 2005 ("w"
corresponds to a weekend and subsequent decrease in the warning threshold) .......66















LIST OF FIGURES


Figure page

2-1 Rip current and its component parts (feeders, neck and head) as well as the
commonly associated currents (represented by arrows) (from Shepard et al.
19 4 1) ....................................................................................................... ........... 5

2-2 Profile of the mean water level and the envelope of the wave height for a typical
experiment. Wave period, 1.14 sec; Ho = 6.45cm; Hb = 8.55cm; tan 0 = 0.082
(from B ow en et al. 1968). .............. .............. ...........................................7.. ....

4-1 Map displaying the study site of Volusia and Brevard Counties as well as the
locations of the tidal gauge, the AES40 grid point 3278, the NOAA buoy 41009,
and the M elbourne B each W ave G auge ...................................... ...................... 17

4-2 Lifeguard observation worksheet completed on August 15, 2005........................21

4-3 A completed ECFL LURCS worksheet from September 2, 2005. ........................22

5-1 Time-series plot of daily rescues for each individual year (top) and the total
daily rescues (bottom) as logged by the Volusia County lifeguards (1998-2003). .24

5-2 Time-series plots of wave height (ft), wave period (sec), wave direction (deg)
and rip current incidents with associated rescues for 2001. The "x" marks
correspond to days with m ore than 15 rescues.................................... ................ 28

5-3 Time-series plots of wave height (ft), wave period (sec), wave direction (deg)
and rip current incidents with associated rescues for 1998. The "x" marks
correspond to days with m ore than 15 rescues.................................... ................ 30

5-4 Correlation histogram-plot of offshore significant wave height (ft) along with rip
current incidents and associated rescues for years 1998 to 2003. The blue (1st)
bar represents the percentage of days that the wave height was within the
respective range. The green (2nd) and red (3rd) bars represent the percentage of
total rescues and the percentage of rip current incidents respectively, that
occurred when the wave height was within that range. The numbers on the top
of each plot correspond to the ratio of the blue and green bar magnitudes..............31









5-5 Correlation histogram-plot of peak wave period (sec) along with rip current
incidents and associated rescues for years 1998 to 2003. The blue (1st) bar
represents the percentage of days that the peak wave period was within the
respective range. The green (2nd) and red (3rd) bars represent the percentage of
total rescues and the percentage of rip current incidents respectively, that
occurred when the peak wave period was within that range. The numbers on the
top of each plot correspond to the ratio of the blue and green bar magnitudes. ......34

5-6 Correlation histogram-plot of offshore wave direction (deg) along with rip
current incidents and associated rescues for years 1998 to 2003. The blue (1st)
bar represents the percentage of days that the wave direction was within the
respective range. The green (2nd) and red (3rd) bars represent the percentage of
total rescues and the percentage of rip current incidents respectively, that
occurred when the wave direction was within that range. The numbers on the top
of each plot correspond to the ratio of the blue and green bar magnitudes..............35

5-7 Correlation histogram-plot of the time of low tide (24hr) along with rip current
incidents and associated rescues for each year from 1998 to 2003. The blue (1st)
bar represents the percentage of days that low tide occurred within the respective
time range. The green (2nd) and red (3rd) bars represent the percentage of total
rescues and the percentage of rip current incidents respectively, that occurred
when low tide occurred within that time range. The numbers on the top of each
plot correspond to the ratio of the blue and green bar magnitudes .......................38

5-8 The daily-calculated threat values of both the ECFL LURCS and the modified
index (top) along with the daily rip current incidents with associated rescues
(bottom) for 1998. The horizontal lines represent the warning threshold for each
resp ectiv e in d ex ........................................................................................................ 4 0

5-9 Calculated rip current threat values for days with more than 15 rescues (1998-
2003). The horizontal line represents the weighted average value .......................52

5-10 Calculated rip current threat values for days with more than 25 rescues (1998-
2003). The horizontal line represents the weighted average value .......................52

5-12 Correlation histogram-plot of the offshore wave height (ft), peak wave period
(sec), offshore wave direction (deg), and time of low tide (24hr) along with rip
current incidents and associated rescues combined over the years from 1998 to
2003 (excluding 2002). The blue (1st) bar represents the percentage of days that
each parameter occurred within the respective range. The green (2nd) and red
(3rd) bars represent the percentage of total rescues and the percentage of rip
current incidents respectively, that occurred when the parameter occurred within
that range. The numbers on the top of each plot correspond to the ratio of the
blue and green bar m agnitudes ............................................................ ................ 54

5-13 ECFL LURCS daily rip current threat levels for the summer of 1998. The daily
rip current rescue totals are indicated by the marker symbols ...............................55









5-14 The modified index daily rip current threat levels for the summer of 1998. The
daily rip current rescue totals are indicated by the marker symbols .....................56

5-15 RIPDEX (rip current predictive index) worksheet.............................. ................ 57

6-1 Daily rip current threat levels computed with the modified index using the
WAVEWATCH III oceanographic output information for the summer of 2005.
The daily rip current rescue totals are indicated by the marker symbols..............61

6-2 Daily rip current threat levels computed with the modified index using the
WAVEWATCH III data for the summer of 2005 (after modification to the
multiplicative direction factor). The daily rip current rescue totals are indicated
by the m arker sym bols. ............. ................ .............................................. 67

6-3 The daily-calculated threat values of both the modified index using the
WAVEWATCH III oceanographic output information (top) along with the daily
rip current incidents with associated rescues (bottom) for the summer of 2005.
The horizontal lines represent the warning threshold ......................... ................ 69

A-i Time-series plots of wave height (ft), wave period (sec), wave direction (deg)
and rip current incidents with associated rescues for 1997. The "x" marks
correspond to days with m ore than 15 rescues.................................... ................ 75

A-2 Time-series plots of wave height (ft), wave period (sec), wave direction (deg)
and rip current incidents with associated rescues for 1998. The "x" marks
correspond to days with m ore than 15 rescues.................................... ................ 76

A-3 Time-series plots of wave height (ft), wave period (sec), wave direction (deg)
and rip current incidents with associated rescues for 1999. The "x" marks
correspond to days with m ore than 15 rescues.................................... ................ 77

A-4 Time-series plots of wave height (ft), wave period (sec), wave direction (deg)
and rip current incidents with associated rescues for 2000. The "x" marks
correspond to days with m ore than 15 rescues.................................... ................ 78

A-5 Time-series plots of wave height (ft), wave period (sec), wave direction (deg)
and rip current incidents with associated rescues for 2001. The "x" marks
correspond to days with m ore than 15 rescues.................................... ................ 79

A-6 Time-series plots of wave height (ft), wave period (sec), wave direction (deg)
and rip current incidents with associated rescues for 2002. The "x" marks
correspond to days with m ore than 15 rescues.................................... ................ 80

A-7 Time-series plots of wave height (ft), wave period (sec), wave direction (deg)
and rip current incidents with associated rescues for 2003. The "x" marks
correspond to days with m ore than 15 rescues.................................... ................ 81









A-8 Time-series plots of wave height (ft), wave period (sec), wave direction (deg)
and rip current incidents with associated rescues for 2004. The "x" marks
correspond to days with m ore than 15 rescues.................................... ................ 82

B-I The daily-calculated threat values of both the ECFL LURCS and the modified
index (top) along with the daily rip current incidents with associated rescues
(bottom) for 1997. The horizontal lines represent the warning threshold for each
resp ectiv e in d ex ........................................................................................................ 8 3

B-2 The daily-calculated threat values of both the ECFL LURCS and the modified
index (top) along with the daily rip current incidents with associated rescues
(bottom) for 1998. The horizontal lines represent the warning threshold for each
resp ectiv e in d ex ........................................................................................................ 84

B-3 The daily-calculated threat values of both the ECFL LURCS and the modified
index (top) along with the daily rip current incidents with associated rescues
(bottom) for 1999. The horizontal lines represent the warning threshold for each
resp ectiv e in d ex ........................................................................................................ 8 5

B-4 The daily-calculated threat values of both the ECFL LURCS and the modified
index (top) along with the daily rip current incidents with associated rescues
(bottom) for 2000. The horizontal lines represent the warning threshold for each
resp ectiv e in d ex ........................................................................................................ 8 6

B-5 The daily-calculated threat values of both the ECFL LURCS and the modified
index (top) along with the daily rip current incidents with associated rescues
(bottom) for 2001. The horizontal lines represent the warning threshold for each
resp ectiv e in d ex ........................................................................................................ 8 7

B-6 The daily-calculated threat values of both the ECFL LURCS and the modified
index (top) along with the daily rip current incidents with associated rescues
(bottom) for 2002. The horizontal lines represent the warning threshold for each
resp ectiv e in d ex ........................................................................................................ 8 8

B-7 The daily-calculated threat values of both the ECFL LURCS and the modified
index (top) along with the daily rip current incidents with associated rescues
(bottom) for 2003. The horizontal lines represent the warning threshold for each
resp ectiv e in d ex ........................................................................................................ 8 9

B-8 The daily-calculated threat values of both the ECFL LURCS and the modified
index (top) along with the daily rip current incidents with associated rescues
(bottom) for 2004. The horizontal lines represent the warning threshold for each
resp ectiv e in d ex ........................................................................................................ 9 0

C-i ECFL LURCS daily rip current threat levels for the summer of 1998. The daily
rip current rescue totals are indicated by the marker symbols ...............................92









C-2 The modified index daily rip current threat levels for the summer of 1998. The
daily rip current rescue totals are indicated by the marker symbols ....................92

C-3 ECFL LURCS daily rip current threat levels for the summer of 1999. The daily
rip current rescue totals are indicated by the marker symbols ...............................93

C-4 The modified index daily rip current threat levels for the summer of 1999. The
daily rip current rescue totals are indicated by the marker symbols ....................93

C-5 ECFL LURCS daily rip current threat levels for the summer of 2000. The daily
rip current rescue totals are indicated by the marker symbols ...............................94

C-6 The modified index daily rip current threat levels for the summer of 2000. The
daily rip current rescue totals are indicated by the marker symbols ....................94

C-7 ECFL LURCS daily rip current threat levels for the summer of 2001. The daily
rip current rescue totals are indicated by the marker symbols ...............................95

C-8 The modified index daily rip current threat levels for the summer of 2001. The
daily rip current rescue totals are indicated by the marker symbols ....................95

C-9 ECFL LURCS daily rip current threat levels for the summer of 2002. The daily
rip current rescue totals are indicated by the marker symbols ...............................96

C-10 The modified index daily rip current threat levels for the summer of 2002. The
daily rip current rescue totals are indicated by the marker symbols ....................96

C-11 ECFL LURCS daily rip current threat levels for the summer of 2003. The daily
rip current rescue totals are indicated by the marker symbols ...............................97

C-12 The modified index daily rip current threat levels for the summer of 2003. The
daily rip current rescue totals are indicated by the marker symbols ....................97

D-1 ECFL LURCS daily rip current threat levels for the summer of 1998. The daily
rip current incident totals are indicated by the marker symbols...............................99

D-2 The modified index daily rip current threat levels for the summer of 1998. The
daily rip current incident totals are indicated by the marker symbols...................99

D-3 ECFL LURCS daily rip current threat levels for the summer of 1999. The daily
rip current incident totals are indicated by the marker symbols.......................... 100

D-4 The modified index daily rip current threat levels for the summer of 1999. The
daily rip current incident totals are indicated by the marker symbols................. 100

D-5 ECFL LURCS daily rip current threat levels for the summer of 2000. The daily
rip current incident totals are indicated by the marker symbols...........................101









D-6 The modified index daily rip current threat levels for the summer of 2000. The
daily rip current incident totals are indicated by the marker symbols..................101

D-7 ECFL LURCS daily rip current threat levels for the summer of 2001. The daily
rip current incident totals are indicated by the marker symbols.......................... 102

D-8 The modified index daily rip current threat levels for the summer of 2001. The
daily rip current incident totals are indicated by the marker symbols................. 102

D-9 ECFL LURCS daily rip current threat levels for the summer of 2002. The daily
rip current incident totals are indicated by the marker symbols.......................... 103

D-10 The modified index daily rip current threat levels for the summer of 2002. The
daily rip current incident totals are indicated by the marker symbols................. 103

D-11 ECFL LURCS daily rip current threat levels for the summer of 2003. The daily
rip current incident totals are indicated by the marker symbols.......................... 104

D-12 The modified index daily rip current threat levels for the summer of 2003. The
daily rip current incident totals are indicated by the marker symbols................. 104















Abstract of Thesis Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Master of Science

IMPROVING RIP CURRENT FORECASTING TECHNIQUES
FOR THE EAST COAST OF FLORIDA

By

Jason Rolland Cummins

December 2006

Chair: Robert J. Thieke
Major: Coastal and Oceanographic Engineering

This study documents the development of an improved rip current predictive index

through the detailed examination of a long-term record of oceanographic and

meteorological conditions with concurrent rip current events. Rip current rescue statistics

are used as a proxy for actual in situ rip current measurements; the correlations of various

meteorological and oceanographic parameters with the occurrence of rescues are used to

establish the relative importance (and hence weighting) of these factors in the predictive

scheme. The correlation analysis was conducted on a long-term data set consisting of rip-

related rescues and hindcast wind and wave data in Volusia County, Florida, extending

from 1997 to 2004. In addition to the established dependence on wave height and period

(already incorporated in National Weather Service [NWS] predictions), the relative risk

of daily rip current activity was found to increase during periods of shore-normal waves

and when the occurrence of low tide coincided with times of peak beach attendance. The

existing rip current forecasting methods practiced by the NWS were accordingly









modified to include both a wave direction parameter and a low tide parameter. After

these adjustments, the modified predictive index exhibited significant improvements.

Currently, the NOAA (National Oceanic and Atmospheric Association) wave buoys off

the east coast of Florida (which are used on a same-day basis for NWS rip current

warnings at present) are not equipped to measure the wave direction. To overcome this

obstacle, the wave direction was incorporated into the rip current predictive scheme by

implementing a readily available forecast (directional) wave model called

WAVEWATCH III (currently used by NWS in other connections). An investigation of

the performance of the modified index using oceanographic output information from the

wave model was conducted in "blindfold" fashion using the Volusia County rescue

information for the summer of 2005. The results indicate that the modified index

outperforms the index currently employed by the NWS. These improvements in

performance, as well as the advantages of advanced notice, justify the incorporation of a

forecast wave model (WAVEWATCH III) into the operational rip current prediction

process.














CHAPTER 1
INTRODUCTION

A rip current is a strong, channeled flow of water extending seaward from the

shoreline. According to the United States Lifesaving Association rip currents are the

cause of approximately 80% of their rescues nationwide. Since 1995, over 17,500

documented rescues in the U.S. were directly related to rip current activity. Rip current

activity affects the safety of beach-goers visiting the coastal waters of this nation and

others. As a result of these dangerous conditions, it is reported that in the U.S. alone over

100 deaths annually are attributed to rip currents.

An investigation was conducted at the University of Florida to examine the existing

rip current prediction methods implemented by the National Weather Service. Analysis of

rip-related rescues correlated with oceanographic conditions assisted in improving the

accuracy of the presently employed rip current index. The index is a scale used to

calculate the level of risk for ocean-goers due to local rip current formation. Knowledge

of these conditions aids the local governmental authorities in the issuance of warnings to

the public. Public awareness, as well as the talent and dedication of the beach lifeguards,

plays a vital role in the prevention of rip-related drownings.

The development of a rip current prediction scheme began in south Florida with the

Lushine Rip Current Scale (LURCS), which utilized wind speed and direction along with

swell height and the time of low tide to assess the daily level of risk associated with rip

currents (Lushine 1991). The LURCS prediction scheme was then later adapted for use

on the central east coast of Florida (ECFL LURCS) with the inclusion of the swell period









and a modification to the tidal parameter (Lascody 1998). The ECFL LURCS was later

modified by Engle (2003). Engle (2003) eliminated the wind parameters and also

incorporated the incident wave angle, directional spreading and the tidal level.

The intention of this study was to further justify the incorporation of a swell

direction parameter and establish a way to integrate it into the prediction scheme, with

the ultimate goal of using the index as an operational forecasting tool. An analysis using

Volusia County rescue information extending from 1997 to 2004 was completed to

establish the importance and respective range of each oceanographic parameter used in

the rip current prediction scheme. The modifications were then individually tested against

the performance of the existing ECFL LURCS method for the same time period. The

incorporation of wave direction improved the accuracy of the rip current predictions.

However, the NOAA (National Oceanic and Atmospheric Association) data buoys on the

east coast of Florida are presently not capable of measuring wave direction; this

represented a significant stumbling block in the implementation of the modified index as

an operational tool. The application of the WAVEWATCH III model (Tolman 1997,

1999a) was introduced into the prediction scheme in order to resolve this quandary. The

modified prediction index developed here used the oceanographic output information

from the wave model to calculate the daily rip current threat levels for the summer of

2005. The results were then compared with the Volusia County rescue information and

the completed ECFL LURCS worksheets to examine if the model could be used to

accurately predict rip current conditions.

Previous research concerning the driving forces and theoretical underpinnings of

rip current formation, as well as general characteristics of a rip are reviewed in Chapter 2.









An overview of the LURCS, ECFL LURCS and the Modified ECFL LURCS is presented

in Chapter 3, along with motivation for improvements to the existing rip current

prediction scheme. Chapter 4 discusses the data sources used during the analysis process

of this study. The correlation between rescues and specific oceanographic parameters,

and the subsequent modifications to the existing ECFL LURCS is presented in Chapter 5.

A comparison between the resulting modified index implementing the WAVEWATCH

III forecast model and the documented ECFL LURCS worksheets is presented Chapter 6.

Finally, the summarized results of the study and overall conclusions are then presented in

Chapter 7.














CHAPTER 2
RIP CURRENTS

Formation of a Rip Current

In 1936 the term "Rip Current" was suggested by F. P. Shepard to describe the

phenomena observed by lifeguards and others on the coast of California. Originally, the

phenomenon of seaward flow was referred to as a "Rip Tide". Yet the occurrence has

little connection with the tidal flow itself, hence the recommendation by Shepard to

rename the process. Rip currents are defined as narrow, seaward-directed currents that

extend from the inner surf zone out through the line of breaking waves (Haller and

Dalrymple 2001).

Five years later F. P. Shepard, along with K. 0. Emery and E. C. Lafond (1941),

reported various qualitative observations regarding rip currents. They divided the rip

current, also referred to as a rip, into three specific parts: 1) the feeder currents, 2) the

neck and 3) the head (Figure 2.1). The flow was perceived to be strongest in the

nearshore through the surf zone, or "neck", and then the observed speeds reduced as the

current traveled further offshore into the rip "head". Another important observation made

by Shepard et al. was the association of rips with certain meteorological conditions. They

noticed that an increase in rip current intensity was directly associated with an increase in

wave height.

In a later study performed by Shepard and Inman (1950), it was found that rip

currents are an integral component of a larger nearshore circulation system. They

hypothesized the driving force of these circulation cells was the convergence and











divergence of the incoming wave field due to the effect of the refraction, producing a


longshore variation in breaking wave conditions. The resulting wave set-up field creates


conditions where water is driven away from regions of the larger waves towards areas of


smaller waves in the form of a longshore current. These currents eventually converge and


turn seaward in the form of a rip current. The physics of the forcing mechanism was not


completely understood until Longuet-Higgins and Stewart (1964) introduced the concept


of radiation stress, and linked it analytically to the wave set-up. Radiation stress is


defined as the excess momentum flux conveyed by a progressive wave. It is a function of


the wave energy and therefore proportional to the square of the wave's height.


V -

.* .. .. .....
4

S ,, ) HEAD -* '
-. f -

Si

/ I i


BREAKER REAKER
ZONE N -'[ Ki ZONE
> f fI !




,K. E-
::-1^< ?- -

SHORE UNIE


Figure 2-1. Rip current and its component parts (feeders, neck and head) as well as the
commonly associated currents (represented by arrows) (from Shepard et al.
1941).









Originally the development of nearshore currents was attributed to the shoreward

mass-transport of water by waves and the subsequent localized changes in sea level. In a

qualitative field experiment, McKenzie (1958) observed that the water brought in toward

the shore by breakers and translator waves tends to cause longshore currents close to the

beach. At variable intervals the longshore currents turn seaward and form outgoing rip

currents. Longuet-Higgins and Stewart (1964) adopted a different approach by using the

concept of radiation stress to analyze the conservation of momentum flux and observed

changes in sea level. According to the theory, any change in the cross-shore radiation

stress is balanced by a hydrostatic pressure gradient, or change in water level. In their

study they theoretically predicted a decrease in the water level, know as set-down, when

the waves approach the breaking point. Adhering to the continuity of momentum flux, the

decrease in the water level is due to the increase in energy from the shoaling of a wave.

They also predicted an increase in water level, known as set-up, shoreward of the

breaking zone. The increase in water level is attributed to the energy dissipation, and

subsequent decrease in radiation stress, of the wave during breaking. Figure 2.2 displays

measurements from a lab experiment performed by Bowen et al. (1968) and their

calculated results applying the theory developed by Longuet-Higgins and Stewart (1964).

Bowen (1969) investigated the observed nearshore circulation system using the

concept of radiation stress developed by Longuet-Higgins and Stewart (1964). As

mentioned earlier, cross-shore variations of the radiation stress is the cause of set-up and

set-down. Since the radiation stress is proportional to the wave height, a longshore

variation in the incident wave height will result in a longshore variation of set-up and set-

down. The variation in set-up induces a pressure field, driving a flow of water in the surf











zone away from regions of high waves toward the regions of low waves (Bowen 1969).


When two of these flows converge on the same location exhibiting low wave energy,


they turn offshore and exit the surf zone as a confined rip current (MacMahan 2006).





25

BEACH 2.0
o [
1.5

MEAN WATER LEVEL, 1.0
o
0 05
THEORY
S "' "------ SW.L: -- o

EXPERIMENT o 0

BREAK
POINT
PLUNGE
POINT
ENVELOPE OF WAVE FH-EIGHT BA 6
.1 BEACH
WAVE CRESt ,




S.WL 0

I "/ -2

WAVE TROUGH
1 I- -4-
400 300 00 100 0
DISTANCE FROM S ILL WATER LINE ON BEACH, x (cms)


Figure 2-2. Profile of the mean water level and the envelope of the wave height for a
typical experiment. Wave period, 1.14 sec; Ho = 6.45cm; Hb = 8.55cm; tan 3 =
0.082 (from Bowen et al. 1968).

In review, the generation of a rip current begins with longshore variations of the


incoming wave field. These variations in wave height can be derived from incident and


edge wave interactions (Bowen and Inman 1969), the convergence or divergence of wave


rays over offshore bottom topography, and/or the induced variability in wave height









caused by coastal structures (Dalrymple 1978). Independent of cause, the forcing

disparity in the incident wave field drives nearshore circulation cells. These cells exhibit

wide regions of shoreward flow separated by narrow regions of offshore flow. If the

narrow regions are strong enough they will appear as rip currents (Haller and Dalrymple

2001).

Characteristics of a Rip

Rip currents are not confined to a specific type of beach and have been observed on

the east and west coasts of Florida (Sonu 1972, Engle 2003), the coast of California

(Shepard et al. 1941 & 1950, Bowen and Inman 1969, Cook 1970, MacMahan 2004) and

on the coasts of Australia (McKenzie 1958, Short 1985, Brander 1999 & 2001, Haas

2002). Each location contains a different offshore topography and incoming wave field,

yet the same phenomenon was seen in all. Rip currents have also been noted to occur

around man-made structures, such as jetties, groins and piers. In this study the

concentration is on the straight, typically barred beaches seen on the east coast of Florida.

Rips have been observed in such locations and have been established in a similar

controlled lab environment. Experiments conducted by Haller et al. (1997) at the

University of Delaware demonstrated the occurrence of cell circulation on an alternating

bar and channel configured shoreline, similar to the bathymetry on Florida's east coast.

On a barred beach the rip current is typically associated with a rip channel, or a gap

in the bar where the out-flowing current is located. Since the rip itself can be unstable in

its location, the associated channel can wander as well. However, if there is structural

influence, the channel will tend to remain stationary. Generally, rip currents are not

constant features, they can flow intermittently, the head swings back and forth, and their

channels may migrate (Cook 1970). The migration of the channel can be the result of









changing wave conditions. When the incident wave field increases in intensity, the rips

tend to transform from many small rips to a few large rips (McKenzie 1958). Another

mechanism for migration is the variability in strength of the feeder currents, which are

rarely the same length and intensity. Uneven feeder currents can orient the rip obliquely

to the shore, and cause the head to become unstable, moving from side to side. Rip

currents have been known to form both orthogonally and diagonally across the surf zone

(McKenzie 1958).

Due to the many variable factors influencing the nature of a rip current, the flow of

the rip can also be highly unsteady, constantly varying in flow intensity. The changing

speed or pulsation of the rip current is an important factor when evaluating the safety of

beachgoers. A swimmer can be situated in a channel during a lull period and remain in

control, yet when the rip pulsates and increases its flow the swimmer suddenly becomes

in danger. Such unsteadiness has been observed on several occasions. Shepard and Inman

(1950) noted that rip currents tend to register all variations in wave strength with a short

time lag. Typical average rip current velocities are 0(1.0)ft/s (0.3m/s), but on shorter

time scales velocities can reach a max of 6.6ft/s (2.0m/s). Sonu (1972) observed

pulsations at high tide corresponding to variance of the incoming swell, and at low tide

corresponding with surf beat frequencies. Others have observed the latter phenomenon

and generally associate the pulsations with wave groups at the infragravity level (0.004-

0.04Hz) (Shepard et al. 1941, Shepard and Inman 1950, Brander and Short 2001,

MacMahan et al. 2004). Field experiments performed by MacMahan et al. (2004)

confirmed the pulsations were driven by infragravity cross-shore standing waves, also

known as surf beat.









Although short-term pulsations can adversely affect ocean-goers, the overall

strength and severity of a rip current is the primary determinant of ocean safety. It is

almost uniformly agreed that an increase in wave height is directly correlated with an

increase in rip current flow. An experiment conducted by Shepard et al. (1941) showed

an increase of rip intensity with every period of larger waves. When conditions of larger

swell occur, the surf zone increases in width and a system of larger and more active rips

can establish itself (McKenzie 1958). Another important factor contributing to the

intensity of a rip current is the tidal stage. During low tide there are two responses: (1) the

breaking on the bar intensifies and (2) the flow concentrates in the rip channel. Both of

these reactions contribute to the increased intensity of the rip current. Shepard (1941) first

observed the concentration of seaward flow in the rip channel during low tide, preserving

form even in less than ideal conditions. Additional field experiments conducted by Sonu

(1972) helped validate the association of rip current strength with tidal stage, as well as a

correlation with swell direction. During an experiment in Australia, Brander (1999)

measured the greatest flow velocities at low tide. It has been noted and observed by many

researchers, that there is a definite correlation between rip current intensity and wave

height, as well as tidal stage. Therefore, both the wave height and tidal stage become

crucial when attempting to interpret the severity of the rip current conditions, with an

eventual goal of providing appropriate warnings to the public.














CHAPTER 3
FORECASTING

Researchers have long been examining rip currents to identify their forcing

mechanisms and the conditions associated with their occurrence in nature. This

information was incorporated into more recent works in an attempt to correlate rip

currents with specific oceanographic parameters. With such knowledge one could

accurately assess the rip current related hazards in the surf zone and then inform beach-

goers. There are more rip-related deaths in Florida each year than hurricanes, tropical

storms, tornados, severe thunderstorms and lighting combined (Lascody 1998). With a

more accurate prediction system and adequate warning methods hopefully the number of

rip current victims can be decreased.

In one of the first attempts at rip current predictions, Lushine (1991) examined the

reported rip-related drownings and rescues in southeast Florida and the concurrent

oceanographic and meteorological conditions. Since the availability of long-term records

on rip current incidence is scarce, rip-related rescues proved to be a useful indicator of rip

current events. Through his work, an experimental scale (Lushine Rip Current Scale,

LURCS) was developed to calculate the risk level of the surf zone due to rip currents.

The scale ranges from zero to five, zero corresponding to no weather-related rip current

danger and five meaning high danger for all swimmers. In the development of the

LURCS, Lushine (1991) found a strong correlation between rip current rescues and wind

conditions. In southeast Florida wind is the primary source for wave generation, because

the islands of the Bahamas intercept most of the distant swell. Wind then became the









primary foundation of the scale, although it also includes swell height and small factors

for tidal stage and persistence. An increase in wind and/or swell height will raise the

scale's value respectively. The tidal stage factor is only added to the scale's value if the

time is between two hours before and four hours after low tide. The persistence factor

accounts for continuing rip current conditions.

Confirmation of the ability to accurately predict rip currents with the LURCS was

achieved through testing on an independent data sample. Three parameters were used to

interpret the results 1) Probability of Detection (POD), 2) False Alarm Ration (FAR), and

3) Critical Success Index (CSI). The results were convincing, and the LURCS was

recognized as a beneficial approach to forecasting the occurrence of rip currents.

The scale developed by Lushine (1991) was intended for use in southeast Florida,

an area dominated by locally generated wind waves. This is evident in the LURCS,

because of the emphasis placed on wind conditions. The southeast division of the

National Weather Service (NWS) implemented the forecasting technique, and warnings

were issued when rip current activity was calculated high enough to pose a threat. Other

sectors of the NWS noticed the advantage of the LURCS and a modified version was

prepared for use along the central east coast of Florida.

The ECFL (East Central Florida) LURCS, developed by Lascody (1998), was

derived from the original LURCS with the addition of a swell period factor and small

changes to other parameters. Lascody (1998) realized the limitations of the LURCS

application because of the dependence on wind waves, where east central Florida is more

disposed to long period swell conditions. Less emphasis was placed on the wind

conditions, and the inclusion of wave period into the scale assured dependence on swell









conditions. The modified scale predicts well and is used today, yet a high false alarm

ratio in testing showed the ECFL LURCS still needed some improvement (Lascody

1998).

A later investigation by Engle (2003) examined the ECFL LURCS and looked for

such methods of improvement. Engle (2003) proposed two parameters, wave direction

and tidal stage, to be particularly influential when attempting to assess rip current

activity. The modified scale was based on the ECFL LURCS. The first change was

removing the wind speed and direction parameters completely. Secondly, wave direction

and tidal level were incorporated into the index. Tests were completed to compare the

performance of the ECFL LURCS with the modified version, using rescue data and

associated weather conditions. The modified scale showed improvements over the ECFL

LURCS when using the POD and FAR for comparison. Alarm Ratio (AR) was another

statistical parameter introduced by Engle (2003) to develop balance between the two

scales. AR is the percentage of days the scale is predicting rip currents (Engle 2003).

Engle's (2003) modified index showed promising improvements over the original ECFL

LURCS. Another analysis was executed by Schrader (2004) to reinforce the accuracy of

the modifications to the ECFL LURCS. This analysis applied a smaller independent data

set, but still demonstrated the validity of the improvements claimed in earlier work by

Engle (2003).

These more recent studies established the need to include wave direction in a

predictive index, however there was no ready way to implement the new directional

parameters. The existing rip current index applies information obtained from the NOAA

weather buoys located off the east coast of Florida. These buoys measure the wave height









and period, but provide no directional data. The aim of the present study is to overcome

this shortcoming in an effort to make the predictive index operational. Through

correspondence with the scientists operating the ECFL LURCS at the National Weather

Service, the author determined what information was readily accessible to them during

the forecasting process. One such asset is the output of the WAVEWATCH III model

managed by the National Oceanographic and Atmospheric Association (NOAA).

WAVEWATCH III is a global wave model used to predict height, period and direction at

each of its grid points. It is executed every six hours (ex. 12am, 6am, 12pm, and 6pm)

and the output is given on one-hour intervals for the following seven days. The resulting

information could prove valuable by improving the index not only with the inclusion of

wave direction but also through the extension of the forecast. Currently the rip current

index calculation is prepared in the morning, and the rip current threat is ascertained for

the same day. The conditions (and warnings if necessary) are distributed to the public

through different avenues of the media, such as NOAA weather radio. Knowledge of the

rip current threat a day in advance would help in the education and awareness of the

public as well as establishing additional staffing needs of the local beach lifeguards.

This study built upon the previous work of Engle (2004), by the inclusion of a

wave direction parameter and a modified tidal stage parameter. The modified index

created here, and each of its parameters, was refined through testing on a long-term data

set. This data consisted of rip-related rescues and hindcast oceanographic conditions.

After finalization of the modified index, an investigation was conducted implementing

the WAVEWATCH III forecast wave data (which includes wave direction) into the






15


prediction methods. The WAVEWATCH III examination was completed to assess the

forecasting capabilities of a modified prediction scheme.














CHAPTER 4
DATA

Site Description

The study site is separated into two sections, both of which are located on the

central east coast of Florida (Figure 4-1). The northern section is comprised of the

Volusia County beaches ranging from the North Ormond region south to New Smyrna.

The southern section consists of the Brevard County beaches extending from Cape

Canaveral south to Melbourne beach. The two sections are divided by the property of

NASA's Kennedy Space Center.

The coastline of Volusia and Brevard County consists of sandy beaches with a

mean sediment diameter of about 0.23mm and 0.33mm respectively (Charles et al. 1994).

The nearshore bathymetry of both locations typically includes a single shore-parallel bar

and trough configuration. The approximate azimuth of Volusia County is 62 East of

North. The shoreline angle of Brevard County shifts due to a coastline perturbation at the

location of the Kennedy Space Center. The azimuth therefore migrates gradually from

1000 in Cape Canaveral to approximately 650 in Melbourne Beach (Figure 4-1). The

continental shelf in the Volusia County region extends out approximately 80km from

shore and the contours are relatively shore-parallel (Engle 2003). The continental shelf

narrows further south on the Florida coastline, and therefore the width decreases to

approximately 60km in Brevard County. However, the bottom contours remain relatively

shore-parallel. The tides in Volusia and Brevard County are semidiurnal and have a

maximum range of approximately 6-12 ft (2m).








































Figure 4-1. Map displaying the study site of Volusia and Brevard Counties as well as the
locations of the tidal gauge, the AES40 grid point 3278, the NOAA buoy
41009, and the Melbourne Beach Wave Gauge.

Rip Current Rescues

Archived rescue logs were obtained from the Volusia County Beach Safety

Division extending from 1997 through 2005. Each lifeguard reports their daily rescue

activity including the date, location, type of rescue, and number of victims involved. This

study utilizes both the number of daily rip related incidences and the number of victims

rescued in each rip current incident. The reports are divided into six zones: (1) North

Ormond to Flagler County line (2) Ormond Beach (3) Daytona Beach (4) Daytona Beach









Shores (5) Dunlawton Avenue south to Ponce Inlet jetty and (6) New Smyrna. The rip

current predictive index has been formulated for application on the entire central east

coast of Florida. In order to limit the possible spurious correlations due to localized

effects, the present study does not account for the zones separately but compiles the

rescue information from each zone into one comprehensive record.

Hindcast Data

Historic oceanographic (wave height, period, and direction) and meteorological

(wind speed and direction) conditions were provided by the AES40 North Atlantic Wind

and Wave Climatology hindcast model called OWI 3-G. The OWI 3-G model is a direct

spectral type based from the WAM model (WAMDI Group 1988), and was originally

developed by Oceanweather Inc. during a project for the Meteorological Service of

Canada. The model grid spans from the equator in the south to the 75.6250 latitude in the

north, with the North American coastline representing the western boundary and the 200

longitude as the eastern boundary. The grid is spaced at 0.833 increments in the

longitude and 0.6250 increments in the latitude, therefore consisting of 9023 wet grid

points. The information in this study was extracted from grid point #3278, which is

located at 28.75 N latitude and 800 W longitude (Figure 4-1).

The OWI 3-G model has been tested to validate the precision of the model's

deepwater wind and wave output. A quantile-quantile evaluation performed by Swail et

al. (2000) compared the model hindcast with recorded satellite and buoy information. A

quantile-quantile assessment is used to determine if two data sets are comprised of a

common distribution. The study exhibited a good correlation in the 1st to the 99th

percentile between the model output and the documented real-time data.









Tides

The tidal data used in analyzing the ECFL LURCS and the modified version was

obtained through NOAA Tides Online Historical Data Retrieval. The tidal station ID is

8721604 (Trident Pier) and is located at 28.415 N and 80.5933 W (Figure 4-1). The

local times of low tide were obtained on a daily basis for calculations in the modified rip

current predictive index.

The moon phases for each year of the study were acquired from the U.S. Naval

Observatory website "http://aa.usno.navy.mil/data/docs/MoonPhase.html". The

occurrence of the new and full moon phases were used during the calculations of the

ECFL LURCS. For the days before, after, and during a full or new moon the

astronomical tide factor was included in the rip current threat level.

WAVEWATCH III

The WAVEWATCH III (Tolman 1997, 1999a) model is a third generation wave

model developed at the Ocean Modeling Branch of the National Center for

Environmental Predictions (NCEP/NOAA). WAVEWATCH III solves the spectral

action density balance equation for wavenumber-direction spectra. Assumptions within

this method limit the model to application outside the surf zone and to spatial scales

larger than 1km. The source terms for the model include wave growth and decay due to

the actions of wind, nonlinear resonant interactions, dissipation due to white-capping and

bottom friction. The model uses a regularly spaced longitude-latitude grid. The spectral

discretization of the wave energy applies an invariant logarithmic intrinsic frequency grid

to spatially vary the wavenumber (Tolman and Booij 1998). The directional increment of

the wave energy spectra is constant and covers all directions. The output of the resulting









wave spectra is at selected locations. In this study the output location used is identical to

the position of the NOAA buoy 41009 (Figure 4-1).

Lifeguard Observations

In an agreement with the University of Florida Department of Coastal Engineering,

Brevard County lifeguards were asked to document daily observations of the nearshore

conditions. The protocol consisted of completing a prearranged worksheet containing

various air and sea parameters (Figure 4-2). The beaches of Brevard County are usually

heavily occupied, resulting in an excess of responsibilities for the lifeguards. Therefore,

the observations are logged sporadically through October, November and December of

2004, along with May, June and July of 2005. Knowledge of the conditions from a first-

hand observer can still prove to be qualitatively useful during the analysis process.

Completed ECFL LURCS Worksheets

Completed ECFL LURCS worksheets (Figure 4-3) were obtained from the

National Weather Service (NWS) extending from April to October of 2005. These

worksheets were filled out by NWS employees and used to distinguish if a warning

should be issued.

Melbourne Beach and NOAA Data Buoys

There is one existing wave buoy on the central east coast of Florida that measures

wave direction. This buoy is funded through the beaches and shores division of Florida

State University (FSU) and maintained by Dally at Surfbreak Engineering. It was

deployed off the coast of Melbourne Beach in a depth of approximately 8m (Figure 4-1).

The nearshore wave information obtained from this buoy was used to corroborate the

observations made by the Brevard County Lifeguards. The NOAA Data Buoy (# 41009,









Figure 4-1) was utilized as a qualitative reference for oceanographic and meteorological

conditions throughout the study.


University of Florida Wave Conditions and Rip Current
Checklist


Time: \QOO)
Obsrnver: U Ck_-( -
Location- -' ..... nreen
Breaking Vvve Height (ft): [0j) [ [2] [4] [5] [6] [7] [8) [9] [10] (>10}

Wave Period (s): [41 [5] [6] [8] [9] [10] [11] [12] [13] i>131

Wave Direction; [Strong N-S] [Weak N-S] T9 ly Onshorej [Weak S-N] [strong S-.N

Rip Channel Development. [None] [Weaki] rate] [Strong]

Tida Stage' [Extreme High) [High)] Mid-high [Mid-Low] [Lowl [ExtremeLowl

Tidal Directon; [Fallingj [Stationary] [R)( 1

Md-TideBar Depth (ft): [NoBar] [0.5] 1] (15] [21 [2. 32] 13.5] [4] [4.5] [5] [5.5S [6j (>31

Longshore Current (ft/)' 4 N-Sj [3 N-S1 [2 N-8] [1 N-SJ (9 1 S-NJ [2 S-NJ [3 S-N] f4 S-N]

Rip Current Danger: [None] [Low] 9,deratej (Significant] [High] [ExtrrieJm

Surf Danger-, [Very Mild] IMikldQ Coderate] {Rough] [Very Rough] [Extreme]

Weather: [Sun. [Pertianly Cloudy] [Overcast] [Raining]

Air Temperature ('F); [<50] [60-60] [60-65] 65-70] [70-751 D-80] [80-85] [85-90] [90-95] [>95]

Water Temperature ('F) [<50] 150-601 [60-65] [65.701 [70-75] -e80] 18 [85] -0
Comments:


Figure 4-2. Lifeguard observation worksheet completed on August 15, 2005.











NAME Ly (check box if statement issued) 0 DATE q A

CALCULATING DAILY RIP CURRENT THREAT- ECFL LURCS
1. WIND FACTOR ONSHORE LONGSHORE FACTOR
(40-100) (110-160., 340-30w)
10-14 kt 2.0 -0.5
15-19 3.0 -1.0
20-24 4.0 -2.0
25 + 5.0 -3.0
ONSHORE FACTOR LONGSHORE FACTOR
2. SWELL FACTOR (Do nat Include WIND WAVE height)
a) SWELL HEIGHT SWELL HEIGHT FACTOR


2 C1.0)
3-4 2.0
5-7 3.0
8-10 4.0
b) SWELL PERIOD SWELL PERIOD FACTOR
8 sec 0.5
9-10 .
11 1.5
12-13 2.5
>13 3.5
SWELL HEIGHT FACTOR + SWELL PERIOD FACTOR SWELL FACTOR / 0 2 0
3. MISCELLANEOUS FACTORS
If astronomical tides are higher than normal (i.e., near full/new moon), add 0.5
If previous day swell factor z 1.5, add 0.5

MISC. FACTOR .-
4. TODAY'S RIP CURRENT THREAT is summation of LONGSHORE. SWELL and MISC. factors
Do not Include ONSHORE FACTOR RIP CURRENT THREAT /." .T '

5. IF RIP CURRENT THREAT is < 2.5, there is aOWRISK of rip currents, and generic
statement for rip currents near piers/jetties may be me loned for values 1.5 2.5.
If RIP CURRENT THREAT is 3.0 4.5 (2.5 4.5 on weekends):
issue statement for MODERATE RISK of rip currents.
If RIP CURRENT THREAT is > 5.0:
issue statement for HIGH RISK of rip currents (coordinate with DAB Beach Patrol).
IF ONSHORE FACTOR is 3.0 or SWELL FACTOR is > 4.0, highlight rough surf.
IF ONSHORE FACTOR > 4.0 or SWELL FACTOR is > 6.0, consider High Surf Advisory.
IF LONGSHORE FACTOR is < -1, discuss longshore current threat in HWO (depicted as
LOW RISK in gHWO). (Z:kRandyuMarine\RipNew\Oct2003sheet.wpd)

Figure 4-3. A completed ECFL LURCS worksheet from September 2, 2005.














CHAPTER 5
DEVELOPMENT OF IMPROVED RIP CURRENT INDEX

Analysis

Lushine (1991), Lascody (1998) and Engle (2003) all used rescue and drowning

incidences to develop and test their respective rip current predictive scales. Lushine

(1991) obtained medical examiner's information, beach patrol rescue logs and newspaper

clippings in Dade and Broward Counties from 1979 to 1988. Lascody (1998) acquired

similar information in Volusia, Brevard, Indian River, St. Lucie and Martin Counties

from 1989 to 1997. Engle (2003) utilized lifeguard rescue logs in Volusia County for

only one year, 1996. The aim of this study is to verify the inclusion of wave direction,

proposed by Engle (2003), and the adjustment of other parameters by exploring a long-

term data set (1997-2004).

Lifeguard rescue logs have proved to be useful in developing a rip current

prediction scheme. Lushine (1991) and Lascody (1998) showed that rip current rescues

are a good qualitative representation of rip currents themselves. The obvious drawback to

this method is that the data can be strongly dependent on the population of ocean-goers.

If there are no people in the water, then there will be no evidence of a rip current. This

doesn't necessarily mean there was no occurrence of a rip current; it might only mean

there weren't any bathers to be rescued from one. A good example of this phenomenon is

the winter season. On the central east coast of Florida, the water becomes relatively cold

in the winter months. Therefore considerably less people enter the water and fewer

rescues are logged. In examination of Figure 5-1, the majority of rescues occur during the











summer months. This does not necessarily indicate a reduction of rip current activity in


the winter months, but is far more likely the effect of the dependency of rip current


rescues on ocean-goers.


250 rI---------I--- I I I
1998
-- 1999
200 2000 -
2001
--2002
S150 2003 -

a 100 -




Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1 Oct 1 Nov 1 Dec 1



250 Li i i I i i i i i i
-- Total

200 -

( 150 -


Figure 5-1. Time-series plot of daily rescues for each individual year (top) and the total
daily rescues (bottom) as logged by the Volusia County lifeguards (1998-
2003).

Limiting the examination and analysis of rip current rescues to the summer season


mitigates the issue concerning a lack of bathers during the winter. The assumed peak


days of attendance are from day 75 to day 250 of a given year. This time frame translates


into mid-march until early September. The month of March signifies the start of spring


break vacations and the east coast of Florida is considered a prime destination. The rising


influx of tourists is directly associated with an increase in rescues (Figure 5-1). The









rescues then become more inconsistent in September as the water temperature starts to

decrease and hurricane season is ongoing. As seen in Table 5-1, the majority of the

rescues occur during the previously defined summer season. When rescues are totaled

over the years examined, excluding 2002 for reasons explained later, almost 87 percent of

the rescues occurred in the summer.

Table 5-1. Percentage of rip-related rescues occurring in the summer (defined as day 75-
250), 1998 to 2003.
1998 1999 2000 2001 2002 2003
Total rescues 2058 1799 1232 2399 226 1135
Summer rescues 1887 1633 972 2256 201 723
Percent summer 0.92 0.91 0.79 0.94 0.89 0.64

This study was also quantitatively limited to the years 1998, 1999, 2000, 2001, and

2003. The exclusion of 1997 and 2004 is due to the lack of data coverage in both of these

years. The rescue data from 1997 is limited and the hindcast data from 2004 is only

available for the first half of the year. Both years were still examined on a daily basis,

investigating the high rescue days and qualitative correlations, but overall summer

statistics were withdrawn. The omission of 2002 is because of the significant lack of

rescues occurring during this year. In Table 5-1 a dramatic decrease in rescues from the

other years can be seen. The reason for this is unknown, but one assumption is the

considerable decrease of tourism travel in the year following the September 11th terrorist

attacks. Independent of cause, the data demonstrates an unnatural decrease in rescues for

2002, and consequently the year was removed from the study.

Another difficulty when using rescues to mark rips is the effect of rough water

conditions and/or inclement weather. If the surf zone is violent people are hesitant to

enter the water and if the weather conditions are poor (ex. rain, clouds) people are even

less likely to make the trip to a beach. This predicament is not as easily remedied as the









previous situation. The only assurance is that the ECFL LURCS and the newly modified

index were exposed to the same problem. Therefore, any such disadvantages are assumed

to equally challenge the predictive capability of both indexes.

The first step in improving the present index was to develop a preliminary new

index. The new index was built on the same foundation as the two previous indexes. The

LURCS and ECFL LURCS assess given input conditions and then return a rip current

threat level. Each input parameter affects the index by increasing or decreasing the

resulting threat level. An example of a completed ECFL LURCS worksheet is shown in

chapter 4 (Figure 4-3). Approximate ranges for each parameter and their respective threat

values are already established. When the conditions (e.g., Wave height) are found to lie in

a given range, a value is assigned to that particular factor. After completing the list of

factors, their respective values are summed to obtain a threat level. The severity of rip

current danger is dependent on the threat level, and a predetermined threshold establishes

if it is advisable to issue a warning.

The adjustments to the new index were loosely based on an approach established in

previous work done by Engle (2003), which served to eliminate the wind parameter,

incorporate wave direction, and modify the tide factor. The aforementioned data was then

used to test the performance of the new rip current prediction index (RIPDEX) against

the ECFL LURCS currently applied by the NWS. Each new (e.g., direction and tide) and

old (e.g., swell height and period) parameter was tested and modified in a cyclic process

to ensure the index reached maximum performance levels. The previous work by Engle

(2003) initiated the idea of applying these new factors (e.g., wave direction) to increase

the accuracy of the ECFL LURCS. Now a far longer data set can be used to better









establish the correct ranges for each input parameter and to clarify each of their roles in a

more optimal prediction scheme.

Ocean Correlations

The oceanographic parameters used in index computations were examined to

determine their importance in the formation of rip currents. These parameters include

wave direction, wave height, wave period and the time of low tide. Each parameter was

first examined from a qualitative viewpoint. Figure 5-2 is a time-series plot of daily wave

height, period and direction along with the amount of rip current incidents (rips) and the

number of victims rescued from each rip (rescues) for 2001. The wave direction is given

in meteorological convention, which is measured clockwise from true north (recall that

62 represents shore-normal for Volusia County). The "x" marks on the plots of height,

period and direction correspond to days with more than 15 rescues. These days along

with other spikes in the number of rescues were used to identify associations between the

wave characteristics and rescues.

In Figure 5-2 the high rescue days tend to occur during peaks in the wave period.

An example of this can be seen in the beginning of the month of May. The wave period

increases from 7 to 11 seconds and the result is multiple days consisting of rescue

numbers greater than 15. Another aspect possibly effecting the same time period is the

synchronized spike in wave height. Larger waves with longer periods (e.g., distant swell)

are directly attributed to an increase in rip current intensity (Shepard et al. 1941). An

interesting observation is the successive days of high rescues during the subsidence of

this wave height spike. A possible mechanism for such observations may include both

physical and human behavioral effects. The increase in wave height and period may lead

to the formation of rip channels in the bar, but it is probably too rough for most beach











patrons to confidently enter the water. When the severity of the conditions begins to

subside the people waiting become more inclined to enter the surf zone. However, the

channels remain intact and the rip currents still pose a considerable threat to bathers.

Such events indicate the importance of knowing the ocean conditions and rip current

threat for the previous days and lend credence to the use of "persistence factors" in

prediction.


2001
I I I I I I I I I
10 -- Height (ft)


SI I I I I I I I I I
Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1 Oct 1 Nov 1 Dec 1


10 li t Period (sec)



Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1 Oct 1 Nov 1 Dec 1


300- |Direction (deg)
100-
100 f

Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1 Oct 1 Nov 1 Dec 1


100 Rips
Rescues



Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1 Oct 1 Nov 1 Dec 1


Figure 5-2. Time-series plots of wave height (ft), wave period (sec), wave direction (deg)
and rip current incidents with associated rescues for 2001. The "x" marks
correspond to days with more than 15 rescues.

In 1998 there is a dramatic spike in rescues in the beginning of August and it

seemed justified to investigate it further (Figure 5-3). On August 4th, 5th, and 6th there

were 65, 175, and 252 rescues respectively. These three days accounted for 24% of the









rescues recorded in 1998. All of the days landed in the middle of the week, so the often-

observed weekend population effect was not a relevant issue. The next logical step was to

examine the conditions prior to the incident. In the days leading up to this spike there was

a period of relatively consistent swell periods and heights, and the wave direction was

slightly north of shore-normal. These consistent rip current generating conditions likely

resulted in the formation of localized rip channels, and were verified by the rescues on

days prior. Then, as seen in Figure 5-3, the wave direction suddenly changed to the

southeast and as the direction migrates back to its original values the abnormal spike in

rescues occurred. It is hypothesized that the large change in wave direction over a

relatively short time period may magnify the instabilities of a longshore bar and rip

current system. This magnification results in more hazardous conditions for beach-goers.

Similar plots were generated for the other 7 years included in this study (see

appendix A). Each plot was also qualitatively examined to assist in investigating the

connection between each parameter and the occurrence of rip-related rescues.

The parameters of the rip current predictive index were then further explored to

find the range of conditions that constituted the greatest association with rip current

development. Each parameter was dissected into distinct ranges, and then associated with

the rips and rescues occurring on those days when the conditions are in their particular

ranges. A histogram is plotted for every year, grouping the normalized frequency of the

oceanographic parameter along with the rip current incidents and the related number of

rescues.











1998

10 Height (ft)











Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1 Oct 1 Nov 1 Dec 1

n300 Mr 1 A Direction (deg) D
200 -
100
0 -
Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1 Oct 1 Nov 1 Dec 1


Rescues
100 I i I I

Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1 Oct 1 Nov 1 Dec 1


Figure 5-3. Time-series plots of wave height (ft), wave period (sec), wave direction (deg)
and rip current incidents with associated rescues for 1998. The "x" marks





The first correlation discussed is between rip currents and the offshore significant

wave height (Ho). The record of wave height was divided into one-foot categories

ranging from zero to ten feet (0-3.05m). Figure 5-4 displays three bars plotted in each

range. The first (blue) bar represents the percentage of days the significant wave height


provided from the hindcast was in the specified range. The second (green) bar represents

the percentage of rip-related rescues occurring on the days when the associated wave


height was in the respective range. The third (red) and final bar represents the percentage

of rip current observations made by the lifeguards when a rescue occurred, which is










independent of the number of rescue victims. A comparison of the latter two bars

provides some insight of how severe the rip currents were in a specified condition range.



r ,: ,, 1998


n ?7 I nAR i 7 77 1 n Rq A I I I n n A I flR
b l 1999



01 051 I r1 12U 033 I 29 036 0072 054


Or2 j
I II I 2000


076 1 1 16 081 0 1 30 2 0 1 0052
05 -



,I 059 1 16 029 I 0 0



I 024 058 I 3 I 1 062 23 I '0 43 27 0I

,, I E Ji EU'in-,-n_ i"
2003

Wave height (ft)


Figure 5-4. Correlation histogram-plot of offshore significant wave height (ft) along with
rip current incidents and associated rescues for years 1998 to 2003. The blue
(1st) bar represents the percentage of days that the wave height was within the
respective range. The green (2n ) and red (3rd) bars represent the percentage of
total rescues and the percentage of rip current incidents respectively, that
occurred when the wave height was within that range. The numbers on the top
of each plot correspond to the ratio of the blue and green bar magnitudes.

A trend can be seen from year to year that the majority of the wave heights occur in

the 2 to 6 foot (0.61-1.83m) range. An interesting characteristic of Figure 5-4 is the high

number of rescues occurring in the 3 to 4 foot (0.91-1.22m) range of wave height. If the

percentages are combined over all the years, excluding 2002, approx 52% of the rescues

occur when the wave heights are between 3 and 4 feet. When examining the histograms,

an important aspect is the difference in magnitude within each group of bars. If the rescue









and/or rip current bar is greater than the actual occurrence of the parameter, then this

constitutes a higher risk of rip-related rescues on days with those conditions. The

difference in the magnitude of the first two bars is represented by the ratio value

displayed above each group of bars. This ratio value is calculated as the magnitude of the

green (2nd) bar divided by the magnitude of the blue (1st) bar. A ratio value greater than

one corresponds to a higher relative risk of rip-related rescues occurring. In 1998 for

example, 39% of the documented wave heights and 60% of the total rescues occurred in

the 3 to 4 foot (0.91-1.22m) range. However, 20% of the wave heights occurred in the 2

to 3 foot (0.61-0.91m) range, along with onlyl0% of the total rescues. Therefore, on days

when the wave height range was 2 to 3 feet there was an average of 3.3 rescues, but on

days with a height range of 3 to 4 feet there was an average of 10.4 rescues.

The relative importance of each range to the threat level is well represented in the

ECFL LURCS and therefore was not changed. The contribution of the swell height

parameter to the computation of the modified index is as follows:

Ho < Ift -- swell height factor = 0

1 ft < Ho < 2 ft -- swell height factor = 0.5

2 ft < Ho < 3 ft -- swell height factor = 1

3 ft < Ho < 5 ft -- swell height factor = 2

5 ft < Ho < 8 ft -- swell height factor = 3

8 ft < Ho -- swell height factor = 4

Wave period

The next oceanographic parameter reviewed is the peak wave period associated

with the incoming swell (Tp). In Figure 5-5 the wave period is divided into 1-second

intervals ranging from 4 to 13 seconds. The groups displayed outside this range include









all occasions below 4 seconds and above 13 seconds respectively. The majority of the

recorded wave periods occur in the range of 5 to 9 seconds, approximately 77% of all the

years combined. Remember the analysis is limited to the summer months when the

average wave period is generally shorter in comparison with the winter. It is apparent

there is a high ratio of rescues on days when the wave period is between 6 and 8 seconds.

In 1999 and 2003 there is a high risk of rescues between 7 and 8 seconds, represented by

ratio values of 1.7 and 1.5 respectively. In 1999, 2000, 2001, and 2002 there is a high risk

of rescues between 6 and 7 seconds, represented by ratio values of 1.9, 1.4, 1.3 and 1.5

respectively. This trend varies a bit from year to year, but overall there was an average of

9.1 rescues per day in the 6 to 8 second range, which is slightly above normal. The

average for all the summers combined, excluding 2002, was 8.5 rescues per day.

The ECFL LURCS only accounts for swells with periods longer than 8 seconds.

However evidence has shown reason to include a slightly shorter period swell, especially

as a result of the high beach attendance in the summer. The longer period swell remains

an integral part in the progression of a hazardous surf environment (Figure 5-5). Basically

the swell period factors were shifted down two seconds, still adhering to the structure of

assigning a larger threat value for an increased swell period. The resulting contribution of

the swell period parameter to the modified index calculation is as follows:

Tp < 6s -* swell period factor = 0

6s < Tp < 7s -- swell period factor = 0.5

7s < Tp < 9s -* swell period factor = 1

9s < Tp < 11 s -* swell period factor = 2

11 s < Tp -* swell period factor = 3













j


11 1 0.46 033 0.64 05 1 8 1. 1 6 0 8 i 3 1 2


0 5 M Record 016 038 19 17 083 1.3 12 03 0.11

II *.I I- _- ,
S -I I-u II ,E I,., I I I" I.


' 076 i 079 1 4 048 09 2.7 072 1 7 '


nfl i~Uirn-rni


2.2 i 0.54


7


2.


0.43 0.53 0.98 I 1.3 I 0.83 0.83 0.53 I 1 I 5.6 I


' 3.1
2001


I-I-J-~I
I'' II I


' 0 44 0.96 1 5 0.95 0.98 0.77 0 1


" I', I I I I "


0 89 029 I 039 I 098 I 15


078 093 I 6 1


I 58 I 0.49 I 0


E-m -


w m i
I,, II


I I .,


Wave period (sec)


Figure 5-5. Correlation histogram-plot of peak wave period (sec) along with rip current
incidents and associated rescues for years 1998 to 2003. The blue (1st) bar
represents the percentage of days that the peak wave period was within the
respective range. The green (2nd) and red (3rd) bars represent the percentage of
total rescues and the percentage of rip current incidents respectively, that
occurred when the peak wave period was within that range. The numbers on
the top of each plot correspond to the ratio of the blue and green bar
magnitudes.

Wave direction

The first additional parameter introduced to the new index is offshore wave

direction (Do). Direction is considered to be a valuable factor when determining the level

of rip current formation and subsequent danger to ocean-goers (McKenzie 1958, Sonu

1972, and Engle 2003). The next step is achieving a way to include swell direction into


the index and presumably improve the accuracy.


&1 05-
CT
ILL


1998



1999



2000


2002



2003


I I










S 044 0.074 0.52 1 11 2.5 1 6 015 28 021 15 0.59 057 0.23

.1 n II11 I 199


,ll i i I,, Ii
04 012 026 049 096 061 093 32 6 069 0 84 034 019 0.13


0 i. .i- E In ..- m ,-.l I. -.iEl.i.u i0. -.



I In) in ,f1 o 11 _I I n,, Ii_,, I ,_,
O 4 Record 0 14 0311 36 078' 1 7 2 3 1 086 95 0 6 0.43'



04 51 053 I 1 I 027 I 7 34 14 068 0 8 I 06 0 9 I 0.18 I


0 I







I. I I.. m. I- .-.l --- M.- inI I lni n


1999



2000



2001



2002



2003


Wave direction (deg)

Figure 5-6. Correlation histogram-plot of offshore wave direction (deg) along with rip
current incidents and associated rescues for years 1998 to 2003. The blue (1st)
bar represents the percentage of days that the wave direction was within the
respective range. The green (2nd) and red (3rd) bars represent the percentage of
total rescues and the percentage of rip current incidents respectively, that
occurred when the wave direction was within that range. The numbers on the
top of each plot correspond to the ratio of the blue and green bar magnitudes.

The shoreline azimuth where the lifeguards rescue information was obtained

(Volusia County) is approximately 62. When viewing Figure 5-6, the incident wave field

is considered orthogonal to shore if the angle is relatively close to this value. The figure

shows a large number of incidences when the wave angle is greater than 90 degrees,

accounting for the high number of southeast summer swells. Although more waves

originate from the southeast in the summer, the greatest association with risk or danger to

swimmers applies to shore normal conditions. Through general inspection of the size

differences within bar groups, the greatest risk was associated with angles in the range of









40 to 80 degrees. This is justified by the high ratio values within this degree range, which

translates to approximately 20 degrees north and south of shore-normal. In the summer of

1999, when the wave angle was between 60 and 80 degrees, there was an average of 37

rescues per day. Although 1999 is probably an extreme case in comparison to the other

years, it still illustrates how large of an effect wave direction can have.

Direction was first incorporated into the new index in the same manner as wave

height and period. The closer the wave direction was to shore normal, the larger the value

of the direction factor. Then this factor was directly applied, through summation, to the

rip current threat level. After the first couple of tests, which are discussed later in this

chapter, there was little progression in the probability of detection and a slight increase in

false alarms. The decision was then made to approach the inclusion of wave direction in

another manner. The new approach consisted of using the wave direction factor as a

multiplier of the other swell conditions (height and period). In this method the wave

direction worked together with the other swell parameters to indirectly affect the threat

level. The indirect association to the threat level was an attempt to decrease the false

alarms occurring on days with an exceedingly small swell directed onshore. Also

incorporated into the multiplicative parameter was a reduction of the threat level due to

an oblique incident wave field. The results of the testing showed improvements in the

overall performance of the index, not just the false alarm ratio. The positive response

initiated additional effort to refine the method, and the contribution of the swell direction

parameter was finalized as follows:

-20 < Do < 300 swell direction multiplier = 0.75

300 < Do < 450 swell direction multiplier = 1.5









450 < Do < 750 -* swell direction multiplier = 2

750 < Do < 1000 swell direction multiplier = 1.5

100 < Do < 1500 swell direction multiplier = 0.75

Else swell direction multiplier = 1

Low tide

In the ECFL LURCS the tidal factor only pertains to an increase in the tidal range

due to astronomical effects (Figure 4-3). Rip current researchers agree that low tide

directly affects the formation of rip currents on a barred beach (Shepard 1941 and

Brander 1999). Additionally, if a rip channel was already established, the decrease in

water depth will intensify the rip current flow (Shepard 1941, Sonu 1972 and Brander

1999). Both produce an increased hazard for ocean-goers. Engle (2003) attempted to

change the tidal parameter through a relation to the actual tidal level. This proved

valuable when analyzing rescues taking place at different times throughout the day. The

conclusion was the majority of the rescues occurred during the rising tide. However, the

incorporation of tidal level into the index is difficult because the threat assessment is on a

daily basis and the tidal level is changing periodically throughout the day. To account for

the effect of low tide, the new tide factor will adjust depending on the time of day low

tide occurs.

In this study the greatest risk to swimmers occurs when low tide is between 10 a.m.

and 12 p.m. (Figure 5-7). In 1998 there was an average of almost 21 rescues per day

when low tide occurred between 10 a.m. and 12 p.m. This justifies the previous results

from Engle (2003) because the rising tide would then occur during the middle of the day,

when the beach is most populated.











05ues 065 19 0.73 081 086
1998


05 12 17 07 063 086 H

SI _I I1999

05- 0 53 09 11 11 2 2000
| 2000


0.5 0 I I 11 0,99 0.96 12 I
SI I I 2001


05 076 I 034 12 12 15 +
M I I -l 2002


065 072 I 16 089 086 1
I I 2003


Time of lowtide (24hr)


Figure 5-7. Correlation histogram-plot of the time of low tide (24hr) along with rip
current incidents and associated rescues for each year from 1998 to 2003. The
blue (1st) bar represents the percentage of days that low tide occurred within
the respective time range. The green (2nd) and red (3rd) bars represent the
percentage of total rescues and the percentage of rip current incidents
respectively, that occurred when low tide occurred within that time range. The
numbers on the top of each plot correspond to the ratio of the blue and green
bar magnitudes.

The increased risk of rescues on the rising tide could be attributed to mental aspects

as well as physical. During low tide the waves break more violently on the sandbar and

aid in the development of rip channels. The intense breaking detours people from

swimming, then when the tide slowly raises the breaking intensity visually decreases and

people feel secure enough to enter the water. The incident wave conditions might have

only experienced minimal changes, persistently forcing the nearshore circulation system.

Yet, more people are entering the water in confidence, which leads to an increased









number of rescues. The new tidal parameter was adjusted to raise the rip current threat

level if low tide coincides with the daytime beach attendance. The scale is slightly

asymmetrical due to the increased rescue dependence found on the rising tide, and the

contribution to the index calculation is as follows:

Low Tide < 9am -* tidal factor = 0

9am < Low Tide < 1pm -- tidal factor = 1

1pm < Low Tide < 5pm -* tidal factor = 0.5

5pm < Low Tide -- tidal factor = 0

With the ranges of each parameter and their respective influence over the rip

current index calculations established, the modified index was then tested against the

existing form of the ECFL LURCS to validate the adjustments made, and observe any

improvements.

Testing

The ECFL LURCS and the newly modified index, with the adjusted parameters

discussed in the previous section, were tested on a data set ranging from 1998 to 2003.

Each index was computed on a daily basis using the oceanographic and meteorological

conditions given by the OWI 3-G hindcast. The daily rip current threat level was attained

from each respective index and then compared with the lifeguard records of rip currents

and their related rescues. Figure 5-8 is an example of the results of the index calculations

and the concurrent lifeguard records for the year 1998, after the changes to each

parameter were finalized (for remaining years see appendix B). The horizontal lines

represent the respective warning thresholds. If the plot of each respective rip current

index is above this threshold then it is recommended to issue a warning to the public.







40


The purpose of the testing process was to substantiate the inclusion of wave

direction and the modifications made to the other parameters in the rip current predictive

index. To assess the performance of both the ECFL LURCS and the modified index, six

statistical parameters were used (1) Alarm Ratio, (2) False Alarm Ratio Method 1, (3)

False Alarm Ratio Method 2, (4) Correct Alarm Ratio (CAR), (5) Probability of

Detection Method 1, and (6) Probability of Detection Method 2. Each was computed

using the rip current incidents as well as the related rescues documented by the Volusia

County lifeguards.

-- Modified
12 -------ECFLLURCS
10 -








Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1 Oct 1 Nov 1 Dec 1



250 r ----
-- Rips
Rescues
200 -

0 15 -

1 100 -

50 -
o0 -, J-, ,J, -. i A 'A L _. - -- . -1
Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1 Oct 1 Nov 1 Dec 1


Figure 5-8. The daily-calculated threat values of both the ECFL LURCS and the
modified index (top) along with the daily rip current incidents with associated
rescues (bottom) for 1998. The horizontal lines represent the warning
threshold for each respective index.









Engle (2003) introduced the Alarm Ratio (AR) as a control when comparing a

modified rip current predictive scale with an existing scale. The AR is defined as the

percentage of days each respective index would issue a warning.

AR = Days Index > Threshold
Total Days

The False Alarm Ratio (FAR) is the measure of over warning. The first method

(FAR1) is commonly used by the National Weather Service (NWS) and is calculated as

the number of days an index issued a warning without the occurrence of rescues

normalized by the total number of days the index issued a warning. The result is the

percentage of warnings given by the respective rip current index when no rescues

occurred.

FARI = Days Index > Threshold w/ no Rescues
Days Index > Threshold

The Correct Alarm Ratio (CAR) is identified as the percent of warnings issued that

coincide with rescues. The CAR is calculated as the number of days an index issued a

warning and a rescue occurred normalized by the total number of days the index issued a

warning. It is also calculated through the subtraction of FAR1 from 1.

CAR = ^ Days Index > Threshold w/ Rescues FAR I
CAR = = FAR1 1
Days Index > Threshold

The second method (FAR2) was developed by the author for use in the cyclic

testing process as a measure of improvement or deterioration. The FAR2 is defined as the

percent of the days in which no rescues occurred yet the index issued a warning.

FAR2 = Days Index > Threshold w/ no Rescues
Days w/ no Rescues









The Probability of Detection (POD) is approached through two methods. The first

method (PODI) was developed to assess the importance of the events detected by the rip

current index. The POD1 is computed as the sum of rip related rescues on days the index

issued a warning, normalized by the total amount of rescues. This statistical parameter

establishes the importance of detecting the days with high numbers of rescues.

POD = Sum of Rescues on Days Index > Threshold
Total Rescues

The second method (POD2) was directly adapted from the NWS and is defined as

the percentage of rip current events detected by the index.

POD2 = Sum of Days Index > Threshold w/ Rescues
Sum of Days w/ Rescues

The percentage growth of the POD compared with the FAR is reflected through the

additional POD/FAR ratio parameters. These ratios give the relative improvements of any

adjustments made to the index, and were used to find a warning threshold that would

produce the maximum levels of performance. The POD2/FAR2 ratio proved helpful due

to the large number of rescues that occur during the summer season. Upon averaging all

the years in the data set, excluding 2002, it was concluded that approximately 67% of

days in the summer season (day 75-250) exhibited documented rescues. The

POD2/FAR2 ratio compares the percent of rescue days detected with the percent of non-

rescue days also detected. If this ratio value is greater than 1, then the rip current

predictive index is identifying a higher percentage of the days with rescues than the days

without rescues. Theoretically if a warning was issued everyday of the summer, a 100

percent probability of detection would result. However this would also result in a 100









percent of the false alarms. The POD2/FAR2 ratio helps to establish balance between the

increases in the probability of detection and the false alarms.

During the testing process four trials were completed in which an individual

parameter was adapted based on the correlation analysis discussed earlier. After each

adjustment, the modified index was computed from 1998 to 2003. The results of the

modified index were then compared with the performance of the ECFL LURCS to

ascertain any improvements made by the adjusted parameter. Comparisons between the

performances of each respective index were analyzed using the previously discussed

statistical parameters. If the results were positive after each trial, the change remained

and carried over to the next trial.

Trial 1: Extraction of the wind factor

The first of four trials in changing the ECFL LURCS was the removal of the wind

factor. After the removal, the daily threat values were computed by modified index over

all the years. The performance changes are represented by the statistical parameters

displayed in Table 5-2. Each parameter was calculated for all the years and then averaged

together. A positive "percent change" value is regarded as an improvement of the

modified index over the ECFL LURCS. Again, the statistics were calculated for the year

2002 for qualitative inquisition but excluded from the averaging process.

After examination of Table 5-2, it is evident that this adjustment is detrimental to

the performance of the index. The FAR1 for rescues increased from 0.25 to 0.35 (approx.

21%), and the FAR1 for rip currents increased from 0.21 to 0.29 (approx. 20%). A slight

increase in false alarms is expected as the ECFL LURCS longshore wind factor only

negates from the total threat value. Deleting this parameter will undoubtedly raise the

index values. The warning threshold for the modified index was adjusted to 3.5 to









account for the increase in threat values, yet the performance levels of the modified index

remained significantly lower than the ECFL LURCS.

Table 5-2. Performance results of the modified index after trial 1 (extraction of the wind
factor), and the original ECFL LURCS averaged over the years from 1998 to
2003 (excluding 2002). A positive % change corresponds to an improvement
over the ECFL LURCS.
Modified ECFL Difference % Change
AR 0.28 0.31 -0.02 -05.4
Rescues 0.35 0.25 -0.06 -21.4
Rips 0.29 0.21 -0.05 -20.3
Rescues 0.29 0.23 -0.06 -25.3
Rips 0.24 0.18 -0.05 -26.9
Rescues 0.27 0.34 -0.05 -14.4
Rips 0.28 0.34 -0.04 -12.7
Rescues 0.31 0.43 -0.09 -21.4
Rips 0.30 0.42 -0.08 -21.7
CAR/ Rescues 2.11 3.31 -1.20 -36.3
FAR1 Rips 3.16 4.59 -1.43 -31.2
POD2/ Rescues 1.00 1.53 -0.54 -35.0
FAR2 Rips 1.39 2.00 -0.62 -30.8
POD2/ Rescues 0.87 1.53 -0.65 -42.7
FAR1 Rips 1.19 1.96 -0.77 -39.3

Although the removal of the wind parameter was intended to improve the

performance of the rip current predictive index, the opposite occurred. The longshore

wind factor actually increases the accuracy of the ECFL LURCS predictions. These

unexpected results might be the effect of human behavior in addition to unknown

physical processes. The range of wind velocity that negates the rip current threat begins

at 10 knots (11.5 mph) and extends to more than 25 knots (28.7 mph) (Figure 4-3). The

elevated levels of wind intensity might detour people from visiting the beach and

therefore correlate to lower rescue numbers. Further analysis of the wind conditions

would be needed to verify this assumption. The limited examination of the correlation

between the wind conditions and the rip current rescues in this study did not justify the

exclusion of the wind parameter. Subsequently, it remained in the prediction scheme.









Trial 2: Inclusion of a wave direction factor

The second trial consisted of integrating the wave direction. The first attempt was

the inclusion of a factor that directly added to the overall threat value, similar to the swell

factors of height and period. This was achieved by increasing the direction factor as the

waves became more orthogonal to the shoreline, with a maximum value of 4 depicting

nearly shore-normal wave conditions. The approach was based on previous work done by

Engle (2003). After the modification to the scale, the results showed an increase in the

POD. However, the POD/FAR ratios were not convincing, denoting that the

improvements in detection were counter-balanced by an increase in false alarms. The

correlations in this study and previous studies (see Engle 2003), illustrate a good

association between wave direction and rip current events. The difficulty is within the

incorporation of the wave direction parameter.

After careful deliberation, the idea was developed to include the wave direction as a

multiplier of the other swell parameters. The thought process behind the idea was to

restrict an orthogonal incident wave field from offsetting the occurrence of swell

conditions unfavorable for rip current formation. For example, assume the swell height

and period factors are both less than or equal to one and the direction is completely

onshore. In an additive scheme, the direction parameter itself would have the ability to

push the threat level into warning status. However, the wave conditions are likely too

small for rip current formation and the resulting effect could be a false alarm. If the

direction was a multiplier, it will indirectly raise the threat level in concert with the swell

conditions. The performance results of this modification are exhibited in Table 5-3. The

warning threshold was kept at 3 to generate maximum performance levels represented in

the values of the POD/FAR and the CAR/FAR ratios.









There were slight improvements in the FAR1, FAR2 and POD2, but considerable

enhancement in the POD1 when compared with the ECFL LURCS. The modified index

showed an increase of approximately 28% and 24% in the POD1 analysis for rescues and

rips respectively. The larger increase in the POD1 symbolizes that the index is detecting a

greater amount of the high-rescue days. The progress in both the false alarms and the

detections is well portrayed by the POD2/FAR2 ratio values. There was an approximate

increase of 10% and 19% in the ratio of general detections to false alarms (POD2/FAR2)

for rescues and rip respectively. The ratio of detection importance to false alarms

(POD2/FAR1) also experienced increases of 11% and almost 19% for rescue and rip

analysis. This exemplifies the advances in the capability of the modified index to detect

rip current events when the multiplicative direction factor is applied compared with the

original ECFL LURCS, which contains no wave direction parameter.

Table 5-3. Performance results of the modified index after trial 2 (inclusion of wave
direction), and the original ECFL LURCS averaged over the years from 1998
to 2003 (excluding 2002). A positive % change corresponds to an
improvement over the ECFL LURCS.
Modified ECFL Difference % Change
AR 0.32 0.31 0.01 03.7
Rescues 0.24 0.25 0.02 06.1
Rips 0.20 0.21 0.01 04.9
Rescues 0.22 0.23 0.00 02.1
Rips 0.18 0.18 0.00 00.1
Rescues 0.36 0.34 0.02 05.7
Rips 0.36 0.34 0.02 04.6
Rescues 0.55 0.43 0.12 28.1
Rips 0.53 0.42 0.10 24.3
CAR/ Rescues 3.64 3.31 0.32 09.7
FAR1 Rips 5.56 4.59 0.96 21.0
POD2/ Rescues 1.69 1.53 0.16 10.4
FAR2 Rips 2.39 2.00 0.39 19.3
POD2/ Rescues 1.70 1.53 0.17 11.2
FAR1 Rips 2.32 1.96 0.37 18.7









Trial 3: Modification of the swell period factor

The third trial included the modification of the wave period factor to account for

slightly shorter period swells. It was shown in the previous section that there was a high

risk for swimmers when the wave period is between 6 and 8 seconds. However, the

ECFL LURCS assigns a zero to this factor for wave conditions with a period less than 8

seconds. The adjustment to the index begins the wave period factor with a value of 0.5

for a 6 second wave and increases accordingly. The statistical results after accounting for

the shorter period waves are shown in Table 5-4. After a series of computations the

warning threshold showed maximum performance at a value of 3.5, and was

subsequently changed.

Table 5-4. Performance results of the modified index after trial 3 (modification of the
swell period factor), and the original ECFL LURCS averaged over the years
from 1998 to 2003 (excluding 2002). A positive % change corresponds to an
improvement over the ECFL LURCS.
Modified ECFL Difference % Change
AR 0.34 0.31 0.03 09.6
Rescues 0.26 0.25 -0.00 -01.5
Rips 0.22 0.21 -0.01 -05.4
Rescues 0.26 0.23 -0.03 -13.0
Rips 0.22 0.18 -0.03 -18.0
P 2 Rescues 0.37 0.34 0.03 09.0
Rips 0.37 0.34 0.03 07.9
PD Rescues 0.58 0.43 0.15 35.0
Rips 0.56 0.42 0.13 31.7
CAR/ Rescues 3.17 3.31 -0.15 -04.4
FAR1 Rips 4.28 4.59 -0.31 -06.8
POD2/ Rescues 1.48 1.53 -0.05 -03.3
FAR2 Rips 1.86 2.00 -0.15 -07.5
POD2/ Rescues 1.57 1.53 0.04 02.9
FAR1 Rips 1.92 1.96 -0.03 -01.7

The POD1 and POD2 for both rips and rescues are the only improvements over the

results of trial 2. Both the FAR1 and the FAR2 experienced considerable deterioration.

The simultaneous increases in the probability of detection and the false alarm ratio are









best examined through the POD2/FAR ratios. All of the ratios demonstrated inferior

performance compared with the ECFL LURCS, and even greater reductions from the

results of trial 2. After interpretation of the yearly performance results summarized in

Table 5-4, it was concluded that the reorganization of the swell period factor diminishes

the quality of the rip current predictive index. Therefore, the changes of the wave period

factor will not be incorporated into the modified index for the remaining trial.

Trial 4: Redevelopment of the tidal factor

The initial change in the fourth trial was the removal of the astronomical tide factor

and the integration of a low tide parameter. However, after a series of preliminary tests

the index exhibited superior results when both tidal factors were accounted for, and

therefore the final trial was completed including both.

Rip currents are known to intensify during low tide, and if this occurs while the

beaches are heavily populated then the result is an increase in rip current risk to ocean-

goers. To account for the effect of low tide, a small factor was included into the index

that is dependent on the time at which the low tide occurs. The max value of this factor is

1. Therefore, it does not have as much influence as the other factors. Its purpose is to

raise the threat level when the conditions are only slightly favorable for rip formation, yet

might need a lower water level to achieve potentially dangerous rip current activity. As

shown by the scale presented in the previous section, the threat level is affected only if

low tide occurs during the day.

The adjustments of the previous trial (wave period factor) were not retained and

therefore comparisons should be made with the results of trial 2 in order to ascertain any

improvements in the index due to the addition of a low tide parameter. For this trial, it

was found favorable to keep the warning threshold to a value of 3.5. With the inclusion of









the low tide factor, the overall performance of the modified index increased (Table 5-5).

The FAR1 and FAR2 both improved from the results of trial 2 and continue to

outperform the ECFL LURCS. The POD1 and POD2 values decreased slightly (approx.

2-4% each) from trial 2, however the POD1 values remain considerably higher than the

ECFL LURCS. This illustrates the ability of the modified index to detect the days in

which the rip current risk is greater.

Table 5-5. Performance results of the modified index after trial 4 (redevelopment of the
tidal factor), and the original ECFL LURCS averaged over the years from
1998 to 2003 (excluding 2002). A positive % change corresponds to an
improvement over the ECFL LURCS
Modified ECFL Difference % Change
AR 0.30 0.31 -0.00 -01.5
Rescues 0.23 0.25 0.03 10.3
Rips 0.19 0.21 0.02 09.0
Rescues 0.21 0.23 0.02 09.4
Rips 0.17 0.18 0.01 07.7
P 2 Rescues 0.35 0.34 0.01 02.2
Rips 0.35 0.34 0.00 00.7
Rescues 0.55 0.43 0.12 26.7
Rips 0.52 0.42 0.09 22.0
CAR/ Rescues 3.72 3.31 0.41 12.4
FAR1 Rips 6.31 4.59 1.72 37.5
POD2/ Rescues 1.78 1.53 0.24 15.7
FAR2 Rips 2.74 2.00 0.73 36.5
POD2/ Rescues 1.66 1.53 0.13 08.7
FAR1 Rips 2.49 1.96 0.53 27.3

The relative improvement and/or worsening of the probability of detection and

false alarm values is better viewed through the POD/FAR ratios. The POD2/FAR2 for

rescues increased from 1.53 to 1.69 in trial 2, and then further improved to 1.78 in trial 4.

This indicates that the advances in the FAR2 values surmounted the small decrease in the

POD2 values from trial 2. There was significant progression in all the other ratios except

the POD2/FAR1 for rescues, which decreased by only 2.5% from trial 2. However, this









ratio value still remained almost 9% better than the ECFL LURCS. The outcome of trial

4 does reinforce the assertion that the occurrence of low tide during the day has an effect

on the rip current risk towards ocean-goers. Therefore, the adjusted tidal factor was

retained in the modified index.

High-risk examination

After the testing process and subsequent verification of the two modifications

(direction and tide) made to the index, further analysis was completed to asses its

capability in detecting the particularly important high-risk days. Table 5-6 displays the

percentage of high rescue/rip days detected by each index for years 1998 to 2003. The

classification of a high-risk day is broken into three categories (1) Days with at least 5

rescues/rips, (2) Days with at least 15 rescues/rips, and (3) days with at least 25

rescues/rips.

For days with at least 5 rescues the modified index detection percentage is an

average of 49.5%, and the ECFL LURCS forecasted 39.3%. The average detection value

is appropriately weighted for each year depending on how many rescues occurred in that

year. Over the next two rescue categories both indexes increase their detection rates.

However, the percentages of the modified index remain higher than the ECFL LURCS.

The modified index detected 64.9% of the days when there was at least 15 rescues,

compared to 50.4% by the ECFL LURCS. The modified index, but not the ECFL

LURCS, detected the single rescue day in this category for 2002. In the final category,

days with more than 25 rescues are examined. This category is associated with extremely

dangerous rip current conditions. The modified index detected an average of 70.3% of

those days, and the ECFL LURCS detected 52.7%. Overall, the modified index









outperformed the ECFL LURCS by approximately 10% in the first category. In the final

high-risk category the modified index increased to almost 20% better performance.

Table 5-6: The percentage of the high rip current rescue and incident days for each year
from 1998 to 2003. As well as the average of all the years weighted by the
number of rips/rescues for each year.


1998 1999 2000 2001 2002 2003


Rescues >= 5
ECFL Rescues
Mod. Rescues
# of days
Rips >= 5
ECFL Rips
Mod. Rips
# of days
Rescues >= 15
ECFL Rescues
Mod. Rescues
# of days
Rips >= 15
ECFL Rips
Mod. Rips
# of days
Rescues >= 25
ECFL Rescues
Mod. Rescues
# of days
Rips >= 25
ECFL Rips
Mod. Rips


# of days


43.2
55.4
74

46.8
59.7
62

46.2
64.1
39

66.7
83.3
12

50.0
66.7
18


46.3
66.7
54

52.8
77.8
36

59.1
90.9
22

46.2
100
13

57.1
100
14


32.2
49.2
59

32.5
57.5
40

46.2
76.9
13

42.9
100
07

50.0
100
08


43.0
40.7
86

42.6
45.9
61

47.6
47.6
42

53.8
50.0
26

53.6
53.6
28


35.7
64.3
14

42.3
50.0
52

00.0
100
01

35.7
71.4
14


00


38.6
52.3
44

44.1
55.9
34

64.3
71.4
14

50.0
50.0
06

50.0
50.0
06


Average
/ Total

39.3
49.5
331

35.8
47.4
285

50.4
64.9
131

43.6
59.0
78

52.7
70.3
74


100 16.7 00.0 33.3 50.0 37.5
100 100 100 50.0 100 62.5
07 06 01 12 06 00 32


Another facet of the high rescue examination was the determination of a warning

threshold for higher risk rip current conditions. The daily threat level of the modified

index within each rescue category was determined. Figure 5-9 and Figure 5-10 display

the threat values of each year for the days in which there were more than 15 and 25

rescues respectively. The average threat value for the days with more than 15 rescues was

approximately 5.4 and the average for the days with more than 25 rescues was








52



approximately 5.8. It was ascertained that the majority of high-rescue days occurred


when the threat level was close to 5.5. As a result, it is recommended to issue a severe rip


current warning if the index value is 5.5 or greater.


Figure 5-9. Calculated rip current threat values for days with more than 15 rescues
(1998-2003). The horizontal line represents the weighted average value.


Figure 5-10. Calculated rip current threat values for days with more than 25 rescues
(1998-2003). The horizontal line represents the weighted average value.


x 1998
> 25 Rescues + 1999
2 2000
2001
S0 2003
Avg
0


8
+
0 +
+ +
+ + + +

S

+

2

Avg = 5.78

0 2 4 6 10 12 14









Summary

The parameters and their respective ranges were first analyzed through a subjective

histogram approach. The summation of all the years for each parameter is presented in

Figure 5-12. The swell height was found to have greatest rip current risk to ocean-goers

in the 3 to 4 foot (0.91-1.22m) range. The swell period exhibited an increase in risk

starting above 6 seconds, which differs from the 8-second cutoff used in the ECFL

LURCS. The wave direction showed a high risk associated with a shore-normal incident

wave field, which was in agreement with previous theory and observations. Figure 4-10

displays the increased risk associated with the direction of the incident wave field ranging

from 400 to 800, or within 200 North or South of shore-normal. This high risk is

exemplified by exhibiting the greatest overall ratio values as compared with the other

parameters. As a result, wave direction has proven to be a crucial component when

ascertaining hazardous rip current conditions. However, complications arose when

attempting to include wave direction into the index, but were eventually overcome with

the use of a multiplicative (rather than an additive) factor. The tidal influence differed

from year to year, but in examination of the overall trend the greatest risk to ocean-goers

occurred when low tide was in the late morning (10-12 a.m.).

After each parameter change was incorporated into the modified index, it was then

tested against the performance of the original ECFL LURCS. As a result of the testing

procedure, the wind parameter and wave period factor remained unchanged. Reversely,

the inclusion of the wave direction factor and the low tide parameter were justified. The

final results of the testing process summarized over the years 1998 to 2003 (excluding

2002) are observed in Table 5-6. The performance of the newly developed rip current

predictive index has shown considerable improvements when compared with the ECFL







54


LURCS. To visually ascertain the difference between the two indexes the daily forecasts

computed by the ECFL LURCS and the modified index for 1998 are displayed

graphically in Figure 4-13 and Figure 4-14 respectively. The horizontal line represents

the warning threshold, and each marker is a daily rip current threat value. The different

markers signify the amount of rescues documented for that day.


0.6 Record I 0.84 059 I 14 I 072 1 I 1
-.6 PRe uc-,eR


0 -
z: o;


> 0.3
_ 0. -
'--0.1 1


1 1.7 0.5 1 071 014


0 1 2 3 4 5 6 7 8
Wave height (ft)


0.77 0.41 06 1.2 1 1.2 1.3 1.7 2.5


9 10

2.1 1.1


4 5 6 7 8 9 10 11 12 13
Wave period (sec)
0.4
1 4 0.36 0.94 0.86 2 2 2.7 2.8 073 0.99 6.3 0 51 0.29
0.3 -

0

0 10 20 30 40 50 60 70 80 90 100 110 120 130
Wave direction (deg)
0.81 1.5 0.86 0.85 1


17KM Iflfl.n


Time of low tide (24hr)


Figure 5-12. Correlation histogram-plot of the offshore wave height (ft), peak wave
period (sec), offshore wave direction (deg), and time of low tide (24hr) along
with rip current incidents and associated rescues combined over the years
from 1998 to 2003 (excluding 2002). The blue (1st) bar represents the
percentage of days that each parameter occurred within the respective range.
The green (2nd) and red (3rd) bars represent the percentage of total rescues and
the percentage of rip current incidents respectively, that occurred when the
parameter occurred within that range. The numbers on the top of each plot
correspond to the ratio of the blue and green bar magnitudes.


S .1-











A difference is noticed between the two indexes after examination of the higher


risk days. The threat values calculated by the modified index have shifted vertically,


accounting for the severity of conditions on those days. Also observed is the large


number of the zero rescue days that remained under the warning threshold. There are


some increases in threat values for the zero rescue days, but since the statistics are based


on rescues and therefore ocean-goers, these anomalies could be the effect of inclement


weather conditions. For similar plots of the remaining years (including separate analysis


of both rip current incidents and rescues) see appendix-D.


Although, not all the initially assumed parameter changes were incorporated into


the modified index, the important assumptions of wave direction and low tide having an


effect on rip current formation were resolved. The resulting rip current predictive index


(RIPDEX) incorporating all observations can be seen in Figure 5-15.



0 rescues
+ 1-4
x 5-14
+ 15-24
7 0 >25



P+0
5 -



S+ + + x


~ + 4 + + x0 + + +
2 -- + ++ x + *-+ *+ +x- x -+ x
+ x + -+o -0- -H-+ --H-x+ 0 + +
1 x + +
AR=0.347 +I
FAR1=0213 POD1=0 603 +
FAR2=0 .6 POD2=0 381
Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1


Figure 5-13. ECFL LURCS daily rip current threat levels for the summer of 1998. The
daily rip current rescue totals are indicated by the marker symbols.








56






10 ----
0 rescues
9 + 1-4
x 5-14
15-24 +
8 + 0 >25


7 -


6 + SxO>
x + *



4 XA + + -

++ + +- Xx 4[x
+ xc
-P + ++ + + +
2 + x + + ++ x +
+ + + +- ++- + + + +

S+ x +0 + x + 0 +
AR=0.313 + x
FARI -=0.164 POD1 -0.661
FAR2=0.18 POD2=0.365
Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1


Figure 5-14. The modified index daily rip current threat levels for the summer of 1998.
The daily rip current rescue totals are indicated by the marker symbols.












RIPDEX


SWELL HEIGHT_
Height, Ho (ft) Factor
Ho<1 0
1 <= Ho < 2 0.5
2 <= Ho<3 1
3<=Ho<5 2
5 <= Ho<8 3
8 <=Ho 4
Swell Heght Factor _


) SWELL PERIOD
Period, T (sec) Factor
T<8 0
8 <= T < 9 0.5
9 <= T<11 1
11 <=T< 12 1.5
12<= T<13 2.5
13<=T 3.5
Swell Period Factor


3) Swell Height Factor + Swell Period Factor = Swell Factor


SWELL DIRECTION
Direction, Do (deg) Multiplier
340 (-20) <= Do < 30 0.75
30 <= Do < 45 1.5
45 <= Do <= 75 2
75< Do<=100 1.5
100 < Do <= 150 0.75
other 1
Swell Direction Multiplier


WIND FACTOR
Longshore only (110-160, 340-30)
Wind Speed, Ws (kt) Factor
10 <= Ws < 15 -0.5
15 <=Ws < 20 -1
20 <= Ws < 25 -2
25 <= Ws -3
Wind Factor


) TIDE
Time of Low Tide, LT (LST) Factor
9<=LT<1 3 1
13<=LT<17 0.5
LT<9 orLT>17 0
Tide Factor





F) MISCELLANEOUS FACTOR
Condition Factor
If astronomical tides are
higher than normal
(i.e. near full/new moon) 0.5
If previous day swell
factor (3) >= 1.5 0.5
Misc. Factor


Rip Current Threat = (Swell Factor x Swell Direction Multiplier) + Wind Factor
+ Tide Factor + Misc. Factor

RIP CURRENT THREAT =


If RIP CURRENT THREAT is 3.5 5.5 (3.0 5.0 on weekends and holidays):
----issue statement for MODERATE RISK of rip currents.

If RIP CURRENT THREAT is >= 5.5:
-------issue statement for HIGH RISK of rip currents.



Figure 5-15. RIPDEX (rip current predictive index) worksheet.














CHAPTER 6
USE OF INDEX AS FORECASTING TOOL

Analysis

The purpose of this project was not only to improve the existing rip current index

(ECFL LURCS), but also to discover more efficient methods of implementing it as a

forecasting tool. The use of the WAVEWATCH III model as an input to the index

incorporates a forecasting ability, as well as the inclusion of wave direction; the latter of

which proved to be beneficial to the performance of the index (see Chapter 5), and both

of which had proved to be significant stumbling blocks previously. Application of the

WAVEWATCH III model makes it possible to forecast the rip current threat in advance.

With this capability the National Weather Service will have additional options when

alerting the public of upcoming severe conditions. Since the forecasts are computed in

advance, any warnings can be issued the day before a severe rip current event. Therefore

the warning can be publicized on the evening news and in the daily newspaper. The

knowledge of dangerous conditions in advance allows people to either change their plans

accordingly, or at least be more prepared for the situation. Another advantage is the

ability to re-establish staffing needs of the local lifeguards. If the index forecasts extreme

rip current conditions, the beach safety division has the opportunity to place additional

lifeguards on duty to compensate for the expected additional number of rescues.

In this study the WAVEWATCH III outputs of wave height, period and direction

were used to compute daily rip current threat levels with the modified index for the

summer of 2005. The results of the 12pm model simulation from the day prior to the









forecast day were used in the calculation of the rip current threat level. Due to the

irregularity of the WAVEWATCH III output, the modified index values were calculated

by the author. The moon phase and time of low tide was also integrated into the

calculations, but the wind field was not included. The analysis was again limited to the

summer season to minimize the spurious effects of low beach population issues. In 2005

there were 2,657 rip-related rescues documented by the Volusia County Lifeguards, with

approximately 89% of those rescues occurring in the summer season. The summer season

for this study is defined as day 79 until day 250, which translates into March 20th until

September 7th. The 4 day lag in the start of the summer season, when compared with the

previous study, is due to the lack of WAVEWATCH III data for days prior. The resulting

daily threat levels were then compared with the documented rip current rescues in

Volusia County, along with the completed ECFL LURCS worksheets (Figure 4-3) and

lifeguard observations (Figure 4-2) for the same time period. The ECFL LURCS

worksheets and the lifeguard observations are sporadic over the summer of 2005 and

consequently, the comparison with these two data sets was restricted to the days with a

high number of rescues.

The first analysis of the WAVEWATCH III rip current forecasting capabilities was

an assessment of the overall performance using the statistical parameters described in

chapter 5 (e.g., AR, FAR, POD...). The results are displayed in Table 6-1. The Alarm

Ratio was approximately 0.48, which is higher than the results of the ECFL LURCS and

the modified index from the previous study (see chapter 5). However, the percentage of

days with rip-related rescues in 2005 was higher than the average for 1998 to 2003 (89%

compared with 68%). The value of the False Alarm Ratio Method 1 remained low at









0.17, yet the value of the False Alarm Ratio Method 2 increased to 0.38. This was

expected because of the increase in the number of days with rescues, and corresponding

decrease in the number of non-rescue days. The Probability of Detection Method 2

showed considerable improvement over the results from chapter 5, increasing to 0.51.

The Probability of Detection Method 1 also showed improvements, increasing to 0.61

and 0.57 for rescues and rips respectively. The POD/FAR ratio values were comparable

to those from the previous multi-year analysis. The POD2/FAR2 ratio decreased slightly,

yet it was expected because of the high FAR2 value resulting from a decrease in days in

which no rescues occurred. The POD2/FAR ratio increased, representing an enhancement

in the detection of days with a greater amount of rescues.

Table 6-1. Performance results of the WAVEWATCH III forecasting the daily rip
current threat levels for the summer of 2005.
WAVEWATCH III
AR 0.48
FARI Rescues 0.17
Rips 0.17
FAR2 Rescues 0.38
Rips 0.38
Rescues 0.51
Rips 0.51
PODi Rescues 0.61
Rips 0.57
CAR/ Rescues 4.88
FAR1 Rips 4.88
POD2/ Rescues 1.35
FAR2 Rips 1.35
POD2/ Rescues 3.03
FAR1 Rips 3.03

Overall, the performance results appear to compare well with the statistical values

obtained in the multi-year study discussed in chapter 5. A plot containing the daily threat

levels and concurrent rescues is presented in Figure 6-1. The horizontal line represents

the warning threshold and each marker signifies the index value as well as the amount of







61


rescues occurring on that specific day. A large majority of the higher rescue days are


above the warning threshold, and many of the non-rescue days remain below. In general,


this represents a good performance by the modified index and the WAVEWATCH III


data. There is a noticeable discrepancy occurring in the month of April, in which there


are a few non-rescue days that have high index values. The ECFL LURCS worksheets


displayed similar heightened threat levels for these days as well. This could be the result


of stormy conditions, in which the rough surf made it unappealing for beach-goers to


enter the ocean. Another possibility could be that the inclement weather kept them from


going to the beach altogether. Independent of cause, the lack of rescues may likely


represent a decrease in ocean-goers and does not necessarily guarantee an absence of rip


current activity.


12
0 rescues
+ 1-4
x 5-14
15-24
10 0 >25
+







o *
*
> 6 +- *

++ x ++ +
+ x +
+ x + x+ + 0 +-+++
4 x 0 + + + +
x x ++ W x0- 0.







F *I .- .c 4 POD20.511 I
+ 0 x 0 + ++ x
0 ++ -% x x ++0+
2 x + x + x++ + +
+ _i x + +

F iFI =,', `,4 POD10.605
F F .-., .'c; POD2=0.511
Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1


Figure 6-1. Daily rip current threat levels computed with the modified index using the
WAVEWATCH III oceanographic output information for the summer of
2005. The daily rip current rescue totals are indicated by the marker symbols.









In order to further validate the use of the WAVEWATCH III data in the rip current

prediction scheme an evaluation with existing methods (ECFL LURCS) for the same

time period was needed. However, the ECFL LURCS worksheets recorded by the

National Weather Service are discontinuous. Therefore, the comparison between the

documented ECFL LURCS worksheets and the results from the modified index

implementing the WAVEWATCH III data was limited to the high-rescue days. The

lifeguard observations are even more intermittent than the worksheets. Subsequently,

they were only used as a qualitative confirmation of dangerous rip current conditions.

A total of 36 days occurred over the summer of 2005 in which there were more

than 15 rescues and also an available ECFL LURCS worksheet. Examination of these

days was used to determine if the application of the WAVEWATCH III model is a viable

method of forecasting rip current conditions. Table 6-2 presents a summary of the

WAVEWATCH III output conditions and subsequent threat levels, along with the threat

levels of the ECFL LURCS worksheets for the high rescue days. In addition, the table

also displays whether or not each respective index recommends the issuance of a

warning. If a "yes" is followed by a "w" it indicates the weekend and therefore a lower

warning threshold was used.

The ECFL LURCS worksheets detected 22 of the 36 high-rescue days (approx.

61%) and the WAVEWATCH III forecasts detected 23 out of the 36 days (approx. 64%).

Therefore, the forecasts using the WAVEWATCH III data actually performed slightly

better than existing methods represented by the worksheets. In addition to the improved

prediction quality, the WAVEWATCH III forecast gives the benefit of advanced notice.

The analysis of the high-rescue days also emphasized the importance of incorporating








63



wave direction into the prediction scheme. The average deviation of the wave direction


from shore-normal (Volusia County, 62) on all 36 high-rescue days, weighted by the


number of rescues, is approximately 23 degrees.


Table 6-2. The WAVEWATCH III forecasted wave conditions and associated rip current
threat levels, as well as the threat levels documented by the National Weather
Service on the ECFL LURCS worksheets for the high-rescue days in the
summer of 2005 ("w" corresponds to a weekend and subsequent decrease in
the warning threshold).
WAVEWATCH III Modified Index ECFL LURCS
Height Period Direction
Month Day Rips Rescues (ft) (sec) (deg) Threat Warning? Threat Warning?
4 21 8 17 1 8 57 3.5 yes 2.5 no


yes 5 yes
no 3.5 yes
yes 3.5 yes
yes 3 yes
no 3 yes
yes 3.5 yes
yes 2 no
yes 3 yes
no 3.5 yes
yes 2 no
yes 3.5 yes
yes 4 yes
no 2.5 no
yes 1 no
no 2 no
no 2.5 no
yes 3 yes
yes 4 yes
yes 3.5 yes
yes (w) 4 yes
no 2.5 yes (w)
yes 1.5 no
yes 2.5 no
no 3.5 yes
no 3 yes
no 3 yes
no 2.5 no
no 2 no
no 2.5 yes (w)
yes 3 yes
yes 3 yes
yes 4 yes
yes 1.5 no
yes 2 no


9 5 77 147 7 6 60 7.5 yes 2 no


5
5
5
5
5
5
6
6
6
6
6
6
6
6
6
7
7
7
7
7
7
7
7
8
8
8
8
8
8
8
8
8
9
9


38
27
25
27
35
21
16
90
113
15
21
47
17
66
35
22
43
47
46
23
15
40
136
16
41
74
44
48
88
28
44
39
37
73


40
38
48
50
102
38
60
60
103
82
30
40
44
88
102
103
100
101
101
101
103
31
47
101
101
99
105
107
68
73
97
98
98
60


6.0
3.0
7.0
4.0
2.6
5.5
6.0
5.5
3.0
3.5
4.0
5.3
3.0
4.8
2.4
3.4
3.8
3.8
3.8
3.4
2.1
4.0
4.5
3.4
3.4
2.5
1.8
1.4
2.0
3.5
3.5
4.3
3.8
7.5









Scrutiny of the high-rescue days that were undetected by the modified index

revealed an interesting characteristic. In the study, some of high-rescue days (May 22nd

June 7th and 26th; July 19th and 24th; August 2nd, 3rd, 5th and 6th) that went undetected by

the index exhibited a wave direction slightly above 100 degrees, representing a southeast

swell. If the wave direction is between 100 and 150, the result is a swell reduction

multiplier of 0.75. Rip currents are not entirely understood and although shore normal

waves were shown to have a large effect on the formation of a rip (see chapter 5), a

consistent long period swell from any direction could contribute to rip-rescue conditions.

It can be seen that when there is a constant swell for multiple days, there are usually

rescues somewhat independent of the wave direction (e.g., July 19th to the 24th, Table 6-

2). After further deliberation it was decided to lift the destructive multiplier if the period

factor was greater than 0. This was an attempt to isolate the longer period swell and

ignore the locally generated wind waves. As a result of the changes, the overall

performance of the index improved (Table 6-3). The FAR and FAR2 values did not

change, but both the POD1 and POD2 increased from the previous results. This

enhancement in the detection capability of the modified index is better elucidated through

the POD/FAR ratios. The POD2/FAR2 ratio increased from 1.35 to 1.47 (approx. 8.9%),

and the POD2/FAR ratio increased from 3.03 to 3.53 (approx. 16.5%).

The slight modification to the WAVEWATCH III index calculations also increased

the detection of the previously discussed high-rescue days. Table 6-4 presents the

WAVEWATCH III forecasts after the modification of the direction multiplier. The

updated index detected 29 out of the 36 days (approx 81%), increasing 6 days from









before. The results of the index after the modifications are also displayed graphically in

Figure 6-2.

Table 6-3. Performance results of the WAVEWATCH III forecasting the daily rip
current threat levels for the summer of 2005, after the modification to the
multiplicative direction factor.
WAVEWATCH III
AR 0.52
FAR Rescues 0.16
Rips 0.16
FAR2 Rescues 0.38
Rips 0.38

POD2 Rescues 0.56
Rips 0.56
PODi Rescues 0.70
Rips 0.66
CAR/ Rescues 5.25
FAR Rips 5.25
POD2/ Rescues 1.47
FAR2 Rips 1.47
POD2/ Rescues 3.53
FAR Rips 3.53

Another interesting aspect of the results was the accuracy of employing both

indexes in the rip current prediction scheme. Of the 36 high rescue days examined, only 3

days were undetected by both indexes (almost 92% accuracy). Implementation of the

WAVEWATCH III model has exhibited promising results in comparison with the ECFL

LURCS, however complete dependency on this new method is not yet recommended. At

present utilizing both methods collectively can further increase the accuracy of the

overall process. It seems each method possesses similar as well as dissimilar deficiencies;

therefore it would prove beneficial to exercise both methods, with the result that the high-

risk days undetected by one index would have the chance to be detected by the other

index.









66




Table 6-4. The WAVEWATCH III forecasted wave conditions and associated rip current

threat levels (after modification to the multiplicative direction factor), as well

as the threat levels documented by the National Weather Service on the ECFL

LURCS worksheets for the high-rescue days in the summer of 2005 ("w"

corresponds to a weekend and subsequent decrease in the warning threshold).

WAVEWATCH III Modified Index ECFL LURCS
Height Period Direction


(ft)


Month Day Rips Rescues
4 21 8 17

5 8 16 38
5 10 15 27

5 12 11 25

5 13 8 27
5 22 19 35

5 27 10 21

6 5 7 16
6 6 28 90

6 7 49 113

6 12 9 15
6 21 13 21

6 22 22 47
6 23 11 17

6 25 36 66

6 26 22 35
7 19 11 22

7 20 18 43

7 21 23 47
7 22 25 46

7 23 13 23
7 24 8 15

7 25 22 40

7 26 74 136
8 2 10 16

8 3 20 41

8 4 40 74
8 5 26 44

8 6 30 48
8 7 46 88

8 9 16 28

8 13 17 44
8 14 21 39

9 3 18 37

9 4 39 73
9 5 77 147


(sec) (deg)
8 57

10 40
8 38

9 48

8 50
8 102

10 38

11 60
10 60

9 103

7 82
7 30

8 40
7 44

8 88

8 102
8 103

10 100

9 101
9 101

8 101
9 103

6 31

7 47
8 101

8 101

7 99
8 105

7 107
7 68

9 73

9 97
8 98

9 98

6 60
6 60


Threat
3.5

6.0
3.0

7.0

4.0
3.0

5.5

6.0
5.5

3.5

3.5
4.0

5.3
3.0

4.8

3.0
4.0

4.5

4.5
4.5

4.0
2.5

4.0

4.5
4.0

4.0

2.5
2.0

1.4
2.0

3.5

3.5
4.3

3.8

7.5
7.5


Warning?
yes

yes
no

yes

yes
yes (w)

yes

yes
yes

yes

yes
yes

yes
no

yes

yes (w)
yes

yes

yes
yes

yes
no

yes

yes
yes

yes

no
no

no
no

yes

yes
yes

yes

yes
yes


Threat
2.5

5
3.5

3.5

3
3

3.5

2
3

3.5

2
3.5

4
2.5

1
2
2.5

3

4
3.5

4
2.5

1.5
2.5
3.5

3

3
2.5

2
2.5

3

3
4

1.5
2
2


Warning?
no

yes
yes

yes

yes
yes

yes

no
yes

yes

no
yes

yes
no

no

no
no

yes

yes
yes

yes
yes (w)

no

no
yes

yes

yes
no

no
yes (w)

yes

yes
yes

no

no
no








67



12
0 rescues
+ 1-4
x 5-14
10 15-24
0 >25
++





8 x
i+ x

++ + x
+ x ++ + +
6. ++- -
4- x x 0+ + + +40 *


2



n


Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1


Figure 6-2. Daily rip current threat levels computed with the modified index using the
WAVEWATCH III data for the summer of 2005 (after modification to the
multiplicative direction factor). The daily rip current rescue totals are
indicated by the marker symbols.


Summary

The newly developed index (Figure 5-15) was tested on an independent data set


using the WAVEWATCH III forecasts from the previous day. The purposes of this


analysis was to verify the improvements made to the ECFL LURCS and assess the


forecasting capabilities of a rip current predictive index implementing readily available


wave model forecasts. The investigation covered the summer of 2005 for the general


performance of the WAVEWATCH III forecasts, but was limited to 36 high-rescue days


for comparison with the recorded ECFL LURCS worksheets.


The modified index outperformed the ECFL LURCS worksheets for the 36


considered days. Initially the new prediction method correctly detected approximately


64% of the days in which more than 15 rescues occurred. This performance was a slight


+ 0 x0 + X + +x +x+ x x
x + +x + +++ +-+ + + +
x + x +x x+* + + 0 + 4
+ x+
X +
F =U I -
-F- l= uI POD1=0.698
FF. .J POD2=0.556









improvement over the 61% detection rate of the ECFL LURCS worksheets for the same

days. Then after improvements, the new method correctly detected over 80% of the high-

rescue days. The modified index calculations for the 2005 summer resulted in a

POD2/FAR2 ratio of 1.47 and a POD2/FAR ratio of 3.53. The resulting daily threat

levels from the index calculations and the concurrent lifeguard records for the summer of

2005 are displayed in Figure 6-3. Almost every surge in the rips and/or rescues is

associated with a rise in the calculated threat level. This parallel movement depicted in

the graph explains the increase in the POD2/FAR ratio when compared with the results

from the analysis in chapter 5. An increase in the POD2/FAR ratio accounts for an

improvement in the detection of days with dangerous rip current activity, illustrated by

the surge in rescues. The observed increases in the level of performance as well as the

advantages of advanced notice justify the use of the WAVEWATCH III model as an

input to the index calculations. However, it was noted that this new method of

incorporating wave direction through the WAVEWATCH III model should not yet fully

replace the existing ECFL LURCS worksheets. Instead the new method should be an

additional tool in the rip current forecasting process to help increase accuracy as well as

provide greater lead time for public warnings.







69


2005
14 -
12 -
10 -



4 v N


140
120
100
S80
On 60
fy


Apr 1 May 1


Figure 6-3. The daily-calculated threat values of both the modified index using the
WAVEWATCH III oceanographic output information (top) along with the
daily rip current incidents with associated rescues (bottom) for the summer of
2005. The horizontal lines represent the warning threshold.


J, A,,A














CHAPTER 7
SUMMARY AND CONCLUSIONS

The first part of this study focused on the development of a modified index for the

prediction of rip currents, in which wave direction was included as a primary parameter.

An examination of all parameters and their correlation with rip-related rescues was

completed to establish the relative importance (and hence appropriate weighting values)

of each in the modified rip current index. The swell parameter data indicated the greatest

relative rip current risk when the swell height ranged between 3 and 4 feet and the swell

period was above 6 seconds. The swell height range was well represented in the existing

ECFL LURCS and therefore remained unchanged. A slight shift in the swell period factor

was thought to improve the performance of the index, yet the testing results proved

otherwise. As a result the swell period parameter also remained unchanged. The incident

wave direction exhibited the greatest risk to ocean-goers when the angle was between 40

and 80 degrees, representing a range within approximately 20 degrees north and south of

shore-normal in the area of Volusia County. These results reinforce the importance of

wave direction as an indicator of rip current threat previously noted in a study by Engle

(2003). The swell direction was included into the modified index as a multiplicative

factor. The tide data indicated the greatest risk for rip currents when low tide occurred

during the day, especially between 10 a.m. and 12 p.m. The time of low tide was included

into the index in addition to the existing astronomical tide factor.

Rip current formation is also influenced by the beach morphology. This becomes a

problem for the prediction scheme because the nearshore bathymetry can be inconsistent









along the coast of Florida, and transform with varying wave conditions. Therefore some

variability will always be present in the correlation and testing results. This is due to the

fact that the lifeguard records cover the entire coast of Volusia county, yet there is a near

complete absence of time-dependent bathymetry data for the same area. However, the

morphological dependence is indirectly incorporated into the index through the persistent

swell factor. If there is consistent swell conditions that assist in the formation of a

longshore bar, trough and rip channel system, then the index threat value for the days

following will be subsequently increased.

The overall trends in each parameter are well represented in the modified rip

current predictive index, and each change was proven through the testing process. Some

of the parameters may appear artificially oriented towards the summer season, which is

expected because the analysis was restricted to the summer. The winter season does pose

a threat for rip current formation, but there is a considerable decrease in the number of

ocean-goers. The analysis was dependent on rip-related rescues serving as a proxy for rip

current formation and therefore the examination was limited to the summer season (Day

75-250). This period corresponds to the highest population of beach-goers and

subsequently the majority of the annual rescues. From 1998 to 2005 over 80% of all rip-

related rescues occurred during the summer (as defined above).

It must be noted that the use of rip current rescues as an indicator or proxy for

actual rip current events is not without its drawbacks. This method suffers from the

spurious effects of beach population as noted earlier (a lack of swimmers in bad weather

may lead to no rescues, but this is not a guarantee that rip currents did not occur).

However, in the absence of any direct long-term measurements of rip currents, this









approach provides the best long-term data set to enable the correlation of rip current

activity to the relevant meteorological and oceanographic forcing. The restriction of

analysis to the summer months (when beach populations are generally large) helps in the

reduction of these spurious effects.

The new index demonstrated considerable improvements over the original ECFL

LURCS when tested on the documented lifeguard rescues from 1998 to 2003. The

Probability of Detection Method 1 increased 26.7% and 22.0% for rescues and rips

respectively. However, the Probability of Detection Method 2 increased only marginally

(2.2% and 0.7%), signifying that the existing ECFL LURCS system detected a similar

amount of rescue events. The False Alarm Ratio Method 1 decreased (improved) 10.3%

and 9.0% for the rescues and rips. The False Alarm Ratio Method 2 also displayed

improvement with decreases of 9.4% for rescues and 7.7% for rips. The combined

progression of the False Alarm Ratio and the Probability of Detection is illustrated by

each ratio value. The POD2/FAR2 increased 15.7% for rescues and 36.5% for rips. The

POD2/FAR increased 8.7% for rescues and 27.3% for rips. In general, the modified index

performed more accurately than the ECFL LURCS. There was a decrease in the false

alarms and an increase in the detection of the higher risk days, which are represented by

an increased number of rescues.

The second part of this study was geared toward the implementation of the

predictive index as a forecasting tool. This was accomplished by employing a forecast

(directional) wave model called WAVEWATCH III, which is a significant step since the

east coast of Florida lacks a long-term network of directional wave measurements. The

performance of the modified index employing oceanographic conditions given by









WAVEWATCH III compared well with the statistical values obtained from the analysis

executed in the first part of the study. The new method was also compared with the ECFL

LURCS worksheets on days in which there were more than 15 rescues. After

improvements, the modified index detected 81% of the high-rescue days, while the ECFL

LURCS worksheets only detected 61% of those same events.

Although this new system of rip current prediction shows significant improvements

over the existing method, there is still difficulty within the process of creating a rip

current forecast. The main problem is the interpretation of the WAVEWATCH III output

parameters. Evaluating these parameters can become rather subjective, particularly if

there is not an explicitly dominant swell. Care must be taken during this phase and

therefore the user should be somewhat aware of the mechanics of rip current formation,

as well as other important factors. These factors include, but are not limited to, the swell

conditions and the number of rescues occurring in days prior to the forecasted day. To

achieve the most consistent results, those who calculate the rip current threat level using

the WAVEWATCH III model and the modified index should implement similar protocol.

The new method is not intended to replace the existing system completely, however

it would certainly assist in the overall process of rip current prediction. It has been shown

that shore-normal waves frequently result in a higher risk environment. Therefore, it is

beneficial to use the WAVEWATCH III model to incorporate wave direction (which is

otherwise unavailable from any real-time measurement platforms). Another significant

advantage to this new method is the advanced notice of severe rip current conditions. By

using the forecast wave field rather than same-day measurements, the extra time allows

for earlier and maybe more effective issuance of warnings to the public through the









media and other venues. It also allows for the beach safety division to staff the local

lifeguard stations accordingly, in preparation for the additional rescues anticipated with

heightened threat levels. These advancements would hopefully decrease the number of

rip current drownings in Florida.

The addition of wave direction might also assist in the extension of the application

of a rip current predictive index to other coastal locations outside East Central Florida as

well. However, further study would likely be needed to re-calibrate the other index

parameters in order to reflect local wave and tidal conditions. For now, the next logical

step in this process is to use the new predictive index incorporating WAVEWATCH III

in an operational mode (perhaps side by side with the ECFL LURC worksheets) to make

forecasts and comparative tests on a daily basis.


















APPENDIX A
WAVE CONDITIONS AND RIP CURRENT RESCUES


1 Feb 1 Mar 1 Apr 1


Rips
i N Rescues



Jun 1 Jul 1 Aug 1 Sep 1 Oct 1 Nov 1 Dec 1


Figure A-1. Time-series plots of wave height (ft), wave period (sec), wave direction
(deg) and rip current incidents with associated rescues for 1997. The "x"
marks correspond to days with more than 15 rescues.








76



1998

10 Height (ft)



Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1 Jull 1 Aug 1 Sep 1 Ocl 1 Nol 1 Dec 1







Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1 Oct 1 Nov 1 Doc 1
Period (sec)




Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1 Oct 1 Nov 1 Dec 1










200 Rips
Rescues
100 -


Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1 Jult Aug 1 Sep 1 Ocl 1 Nov 1 Dec 1



Figure A-2. Time-series plots of wave height (ft), wave period (sec), wave direction

(deg) and rip current incidents with associated rescues for 1998. The "x"
marks correspond to days with more than 15 rescues.















10 ---Height (ft)


S1' 1 I I .I' I MIa I Ju "v
Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1 Oct 1 Nov 1 Dec 1


IeU -- Rips
Rescues

50
0 ..... .... I .... .. .. .+ ..... 1. ... ----- .... ------
Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1 Oct 1 Nov 1 Dec 1



Figure A-3. Time-series plots of wave height (ft), wave period (sec), wave direction
(deg) and rip current incidents with associated rescues for 1999. The "x"
marks correspond to days with more than 15 rescues.





















10 Period (sec)



Jan1 Febi1 Marl1 Apr 1 May 1 JunI Jul 1 Aug 1 Sepi1 Octi1 Novi1 Dec


I Rips
40 Rescues
20- i it

Jan 1 Feb 1 Marl Apr 1 May 1 Jun 1 Jull Aug 1 Sep1 Oct 1 Nov 1 Dec 1



Figure A-4. Time-series plots of wave height (ft), wave period (sec), wave direction
(deg) and rip current incidents with associated rescues for 2000. The "x"
marks correspond to days with more than 15 rescues.








79



2001
1 5I I I I I I I I I
10 -- Height (ft)



Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1 Oct 1 Nov 1 Dec 1




I I I I I I I I I


Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1 Oct 1 Nov 1 Dec 1



Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1 Oct 1 Nov 1 Dec 1
200 Direction (deg)



Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1 Oct 1 Nov 1 Dec 1


100 Rips
Rescues

O -'e- I AL -

Jan 1 Feb 1 Marl Apr 1 May 1 Jun 1 Jull Aug 1 Sep 1 Oct 1 Nov 1 Dec 1



Figure A-5. Time-series plots of wave height (ft), wave period (sec), wave direction
(deg) and rip current incidents with associated rescues for 2001. The "x"
marks correspond to days with more than 15 rescues.








80



2002






Jan 1 Feb 1 Mar 1 ApIr 1 May 1 Jun 1 Jull 1 Aug 1 Sep 1 Ocl 1 N 1 Dec 1
10 Height ft1 No) 1 D)




Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1 Oct 1 Nov 1 Dec 1






1 5 I I I II I- -
Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1 Oct 1 Nov 1 Dec 1




















Figure A-6. Time-series plots of wave height (ft), wave period (sec), wave direction
30marks correspond Direction (15 rescuedeg)
100

Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1 Oct 1 Nov 1 Dec 1



10 rRe cues



Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1 Jul t Aug 1 Sep I Oct 1 Nov 1 Dec I



Figure A-6. Time-series plots of wave height (ft), wave period (sec), wave direction

(deg) and rip current incidents with associated rescues for 2002. The "x"
marks correspond to days with more than 15 rescues.














15 L I"II
J In -- 1 rHeight (ft) l l



J I Ma I Ju 1 Ju
Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1 Oct 1 Nov 1 Dec 1


1 Feb 1 Mar 1 Apr 1 May 1


Rips
Rescues



Jun 1 Jull Aug 1 Sep 1 Oct 1 Nov 1 Dec 1


Figure A-7. Time-series plots of wave height (ft), wave period (sec), wave direction
(deg) and rip current incidents with associated rescues for 2003. The "x"
marks correspond to days with more than 15 rescues.








82



2004


5


Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1







Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1


300 l r 1-- Direction (deg)
200
100





40 Rescues

2 0 -_

Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1



Figure A-8. Time-series plots of wave height (ft), wave period (sec), wave direction
(deg) and rip current incidents with associated rescues for 2004. The "x"
marks correspond to days with more than 15 rescues.


















APPENDIX B
RIP CURRENT THREAT VALUES AND RESCUES


45 -- Rips
40 Rescues
35 -
0 -30
25 -
S20-
1 15
10 -

0 L ------------ I ------- L--- L I J. -- l ----- 6_I
Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1 Oct 1 Nov 1 Dec 1


Figure B-1. The daily-calculated threat values of both the ECFL LURCS and the
modified index (top) along with the daily rip current incidents with associated
rescues (bottom) for 1997. The horizontal lines represent the warning
threshold for each respective index.













Modified
12 -| -------ECFLLURCS

10 -
a)












25 0 I I I I I I I I I I I
Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1 Oct 1 Nov 1 Dec 1




250 r
--Rips
Rescues
200 -

a)


100-
Ct

50 -


Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1 Oct 1 Nov 1 Dec 1



Figure B-2. The daily-calculated threat values of both the ECFL LURCS and the
modified index (top) along with the daily rip current incidents with associated
rescues (bottom) for 1998. The horizontal lines represent the warning
threshold for each respective index.