Você está na página 1de 21

Ocean Modelling 14 (2006) 81101 www.elsevier.

com/locate/ocemod

Tropical cyclone induced asymmetry of sea level surge and fall and its presentation in a storm surge model with parametric wind elds
Machuan Peng
a b

a,*

, Lian Xie a, Leonard J. Pietrafesa

Department of Marine, Earth and Atmospheric Sciences, North Carolina State University, Box 8202, Raleigh, NC 27695-8208, United States College of Physical & Mathematical Sciences, Box 8201, North Carolina State University, Raleigh, NC, United States Received 22 July 2005; received in revised form 14 March 2006; accepted 15 March 2006 Available online 11 May 2006

Abstract The asymmetry of tropical cyclone induced maximum coastal sea level rise (positive surge) and fall (negative surge) is studied using a three-dimensional storm surge model. It is found that the negative surge induced by oshore winds is more sensitive to wind speed and direction changes than the positive surge by onshore winds. As a result, negative surge is inherently more dicult to forecast than positive surge since there is uncertainty in tropical storm wind forecasts. The asymmetry of negative and positive surge under parametric wind forcing is more apparent in shallow water regions. For tropical cyclones with xed central pressure, the surge asymmetry increases with decreasing storm translation speed. For those with the same translation speed, a weaker tropical cyclone is expected to gain a higher AI (asymmetry index) value though its induced maximum surge and fall are smaller. With xed RMW (radius of maximum wind), the relationship between central pressure and AI is heterogeneous and depends on the value of RMW. Tropical cyclones wind inow angle can also aect surge asymmetry. A set of idealized cases as well as two historic tropical cyclones are used to illustrate the surge asymmetry. 2006 Elsevier Ltd. All rights reserved.
Keywords: Asymmetry; Storm surge; Tropical cyclone; Parametric wind

1. Introduction Tropical cyclone (TC) induced storm surge is a major threat to coastal residents. Measures of the impacts of these storms on coastal communities are in the loss of human lives and property destruction. Fortunately, the

Corresponding author. Tel.: +1 919 515 1436; fax: +1 919 515 7802. E-mail address: mpeng@ncsu.edu (M. Peng).

1463-5003/$ - see front matter 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.ocemod.2006.03.004

82

M. Peng et al. / Ocean Modelling 14 (2006) 81101

improved forecast of TC track and intensity of the National Oceanic and Atmospheric Administration (NOAA) of US has reduced such impacts. Historically, thirty or more people perished in each single event of the top thirty US deadliest TCs. The very deadliest one was unnamed, which struck Galveston Island, Texas, and claimed more than 8000 lives in September, 1900 (NOAA webpage 1, 2). Unlike human death toll which has been sharply reduced over time due to the NOAA warning system, TC induced property damage has been growing. The most costly one was Hurricane Katrina in 2005. The total loss as a result of Katrina is estimated to exceed $100 billion. While high impact events will undoubtedly occur in the future, the advent and further improvement of the coastal warning system may greatly reduce the loss of lives, and potentially reduce the property damage. TC winds have been forecast by NOAAs National Hurricane Center for almost half a century (NOAA webpage 1). The primitive hurricane wind models were developed in the 1960s, and storm surge models came into being a few years later. In the early stage, the surge models were two-dimensional (Welander, 1961; Jelesnianski, 1967; Reid and Bodine, 1968). The National Weather Service (NWS) adopted the SPLASH model (Jelesnianski, 1972; Wanstrath et al., 1976) for storm surge forecasting. However, this model worked well only in regions that are not too close to the shoreline. Big mistakes may occur along the coast, where nonlinear eect is pronounced, and high horizontal resolution is required. In addition, sealand boundary was not movable in the model. So inundation process was not considered. The SLOSH model (Jelesnianski and Chen, 1984; Jelesnianski et al., 1992) followed SPLASH and considered wetting and drying processes. This 2-D model used a polar grid system to allow greater resolution in the area of forecast interest, retained some nonlinear terms in the equations of motion, and considered such sub-grid scale features as channels, barriers and coastal rivers. Pietrafesa et al. (1986, 1988) utilized a 3-D linear numerical model to simulate storm surges in a large coastal lagoon system, Albemarle-Pamlico Sound, NC, and found that the 3-D model greatly outperformed 2-D models. To consider ooding and dying, and further consider nonlinear eects in the surge process, Xie et al., 2004 (referred to as the XPP hereafter) introduced a POM-based (Princeton Ocean Model) storm surge and inundation model that incorporates mass conservation and the exibility to allow the inundation speed to be based on the three-dimensional ow elds. A ooding module and a drying module were embedded in the model to reect the movement of sealand boundary. The XPP 3-D model has been applied to several coastal domains, including the Croatan-Albemarle-Pamlico Estuary System (CAPES) and Charleston Harbor (Peng et al., 2004, in press), and the storm surge and inundation results generally agreed well with observations. The XPP model will be employed as the surge model in this paper. As the focus of this study is on the asymmetry of sea level surge and fall under TC condition, the ooding module is set o throughout the study. As draining process may occur in shallow water under strong wind forcing, the drying module is kept on. Any storm surge model, no matter 2-D or 3-D, depends critically on the accurate input of the TC wind elds. In ideal experiments, parametric wind models, such as the Holland (1980) model, are frequently used for wind forcing to drive storm surge models. One advantage of the parametric wind model is its convenience to generate symmetric wind elds, bearing the principal wind distribution characters of the TC. Another advantage is that it can be employed to study the sensitivity of surge to the change of track, central pressure, RWM and other TC parameters. However, if it is used in real TC cases, any misspecication of wind model may lead to serious surge problems. For example, when the Holland model was applied to storm surge hindcast of Hurricane Isabel in 2003 (Fig. 1), the calculated maximum storm surge, generally, agrees well with the observations at the available water level stations (Fig. 2). The discrepancy at Windmill and Solomon Island is due to lack of maximum surge records at the two stations. Prior and up to the sea level maxima, model results and observations at Chesapeake Bridge and Duck Pier agreed well. However, Fig. 2bc indicate an overestimation of negative surge occurs after the maxima. This simulation error is largely due to misspecication of the wind forcing. The winds used in the simulation were symmetric parametric wind elds that cannot capture the asymmetric features of Hurricane Isabel. As will be demonstrated, there exists an asymmetry of sea level surge and fall in responding to symmetric TC wind elds. Oshore winds are more eective in inducing negative surge than the corresponding onshore winds in inducing positive surge. Such kind of overestimation of surge asymmetry induced by symmetric wind forcing was inferred in previous works (e.g. Ruth et al., 2005).

M. Peng et al. / Ocean Modelling 14 (2006) 81101

83

S8 S7

39.0N
S6

S5

38.0N

37.0N Hurricane Isabel 36.0N 2003

S4

S3

35.0N S1 34.0N

S2

33.0N

79.0W 78.0W 77.0W 76.0W 75.0W 74.0W 73.0W


Fig. 1. The track of Hurricane Isabel and the locations of eight water level stations. Sl-8 are Beaufort, Cape Hatteras, Duck Pier, Chesapeake Bridge, Windmill Point, Solomons Island, Baltimore, and Chesapeake City, respectively. The box shows the inner domain of the nesting system in the model.

In fact, the asymmetric sea level response to symmetric wind forcing can be explained in a simple 1-D case. As the pressure gradient force required to balance a given wind is proportional to (h + n)dn/dx (where h, n and x are respectively the undisturbed water depth, sea surface elevation, and distant away from the coast), a larger sea level drop is required during fall (when n is negative) than the corresponding surge (when n is positive), provided that the model cross-shore domain is larger than the barotropic radius. Of course, Coriolis force, bottom friction and the other nonlinear features are not considered in the above assumption. In this study, the Holland parametric wind model is employed for all hypothetical TCs. It is also used, based on observed data, for the historic TCs to assess the errors in negative surge due to misspecication of the wind model. This explains why more accurate HRD (Hurricane Research Division) Realtime Hurricane Winds, or H*winds, should be used in real TC cases. This paper contains ve sections. In Section 2, the three-dimensional surge model is briey described. The asymmetry of sea level rise and fall under steady onshore and oshore-wind forcing is examined in Section 3. Section 4 discusses how TC parameters, such as translation speed, RMW and inow angles, aect the water level risefall asymmetry. Two historic TCs are presented in Section 5 to further demonstrate that accurate parameterization of wind model may greatly lessen the deciencies in storm surge results. Finally, discussion and conclusion are presented in Section 6.

84

M. Peng et al. / Ocean Modelling 14 (2006) 81101

Maximum Storm Tide (Meter above N.G.V.D.)

3 2.5 2 1.5 1 0.5 a 0 S1 S2 S3 S4 S5* S6* S7 S8


Observed Simulation

Storm Tide (m)

2 1 0 1

Observed Simulation

Chesapeake Bridge (S4)

Storm Tide (m)

2 1 0 1 17th00Z

Observed Simulation

Duck Pier (S3)

c 12Z 18th00Z 12Z 19th00Z 12Z

Fig. 2. (a) The observed and simulated overall maximum storm tide at eight stations. Time series of storm tide at (b) Chesapeake Bridge and (c) Duck Pier.

2. The storm surge model The storm surge model employed in this study is described in detail in Xie et al. (2004) and Peng et al. (2004). The hydrodynamic component of the modeling system is based on the POM Model (Mellor, 1996), which uses a terrain-following sigma (r) coordinate in the vertical and a staggered Arakawa C grid in the horizontal plane. An embedded second moment turbulence closure scheme is used to compute the vertical mixing coecients. The model uses a free surface, and thus allows explicit prediction of sea level change. In POM, the horizontal nite dierencing is explicit, whereas the vertical dierencing is implicit. The latter eliminates time-step constraints on the vertical resolution and permits the use of ne vertical resolution near the surface and in shallow water regions. A three-time level leapfrog scheme is used for temporal integration. The surface kinematic wind stress is applied at the sea surface through the following boundary condition: ! ~ K M oW s at r 0 1 or H ~ and s are respectively the water depth, vertical eddy viscosity, horizontal velocity and the where H, KM, W ~ w jV w , surface kinematic wind stress, which is computed by using the conventional bulk formula s qC d jV where Vw is wind speed at height of 10 m, q is air density and the drag coecient, Cd, is assumed to vary with wind speed.

M. Peng et al. / Ocean Modelling 14 (2006) 81101

85

8 2:16 > > > > > ~ > > < 0:49 0:065jV w j 3 10 C d 1:14 > > > > ~w j 0:62 1:56=jV > > > : 2:18

~w j P 26 m s1 jV ~w j < 26 m s1 10 m s1 6 jV ~w j < 10 m s1 3 6 jV ~ w j < 3 m s 1 1 6 jV ~w j < 1 m s1 jV 2

This Cd formula follows Large and Pond (1981) when the wind speed is less than 26 m s1, otherwise, it is assumed as a constant as indicated in Powell et al. (2003). At the sea bed, vertical velocity is assumed to be zero, and bottom stress is determined by the horizontal current velocity (U, V) at the lowest vertical layer through   K M oU oV 1=2 ; 3 C z U 2 V 2 U ; V at r 1 H or or where the bottom drag coecient, Cz, is determined by " # k2 ; 0:0025 C z MAX 2 lnf1 rkb1 H =z0 g

where k (= 0.4) is the Von Karman constant, kb, the number of vertical layers and z0 is the roughness length. Zero mass and momentum uxes are the solid boundary conditions. Passive open boundary conditions (OBCs) are employed in the study, which means the dynamics of the ow at the open boundary are determined by the interior circulation. Storm surge results are sensitive to the OBCs used in the model. The details of the sensitivity study can be found in Palma and Matano (1998, 2000). The OBCs employed in our model follows their works. For external mode, Flather radiation condition is used to determine normal ows along the open boundary. This OBC has good behavior in storm surge simulation. It not only allows conservation of mass in the interior domain, but also handles well the passage of dispersive waves. It can be written as U U 0 t C g g0 t H 5

where U and g are respectively the p velocity normal to the open boundary, and sea surface elevation, U0 and g0 are prescribed values, C gH is the phase speed of the incoming wave. Since in the case of passive OBCs the exact values of U0 and g0 are unknown, they are assumed to be zero at the boundary. The OBC of internal mode has not so much inuence on model results as of external mode. The conventional radiation OBC for internal ows in Mellor (1996) is employed in this study. For the two historic TCs (Hurricane Isabel and Charley), the domain of interest is oneway nested within a larger domain to improve the inner horizontal resolution, and to lessen the spurious eects of the outer open boundaries. On the land side, runos from major rivers can be prescribed, but are not used in this study. As mentioned before, the ooding module of the model is purposely shut o while the drying module is kept on. 3. The asymmetry of surge and fall with steady wind forcings It is indicated in Section 1 that symmetric TC wind forcing may produce an asymmetry sea level surge and fall. This section is to investigate the extent of such asymmetry with dierent symmetric wind forcings and various bathymetry conditions. Steady onshore and oshore winds with the same magnitude of speed oer ideal symmetric wind forcings, and will be employed in this section. A square region (100 100 grid points) with grid size of 5 km in both directions, as shown in Fig. 3, is chosen as the study domain. The water depth contours in the gure show the original bathymetry conguration, which linearly increases from zero at the shoreline to the ocean at a slope of 1/10,000. Such a slope is much smaller than on typical continental shelf. It is used to emphasize the shallow water eects, and to make storm surge more striking. Location A in Fig. 3 is where the model results are given. In the experiments, 10, 20, 30

86

M. Peng et al. / Ocean Modelling 14 (2006) 81101

100

90 40

80

Land

70 30

60

50

20

40 10

30

20

Land

10

10

20

30

40

50

60

70

80

90

100

Fig. 3. The bathymetry contours (in meter) of the ideal square domain, and A is the location where model output is given.

and 50 m are arbitrarily chosen as the minimum water depth to readjust the original bathymetry. These four cases are respectively named Case 1, Case 2, Case 3 and Case 4. In Case 1, the water depth is assumed to be 10 m at the location where the original value is less than it. Similar are the other cases. A steady onshore/ oshore wind with speed of 30 m s1 is the driving force. An asymmetry index is introduced as: AI = (jfallmaxj surgemax)/surgemax 100%, where fallmax and surgemax are respectively the maximum sea level fall under the oshore-wind forcing, and the maximum surge under the onshore-wind forcing. As shown in Fig. 4, there is no apparent asymmetry of sea level surge and fall in Case 3 and Case 4 (AI is less than 10%). The spin-up (or spin-down) time for these two deep water cases is around 10 h, after which the steady state is reached due to the balance of the wind stress and the horizontal pressure gradient in the vicinity of the location A. But for the shallow water cases, the oshore wind induces more negative surge than the onshore wind induces positive surge. This is especially true in Case 1, where the maximum surge is less than 3.5 m, but the maximum fall is more than 5 m. AI is as large as 56% in this case. This indicates the asymmetry of the maximum surge and fall is pronounced in shallow water regions. The balance between the wind stress and the horizontal pressure gradient is also reected in the kinks on the curves. In Fig. 4, the sudden sea level drop/rise at kinks indicates the horizontal pressure gradient is greater/ smaller than the wind stress at the time. If there is a kink and how pronounced it is depend on the sea surface distribution near the concerned location, which, at any time, is a function of wind stress force, domain conguration and bathymetry. This explains why the kinks in Fig. 5 (as will be shown later) are not so noticeable, when the bathymetry and wind force are changed. To further investigate the asymmetry of sea level surge and fall with dierent wind forcings, wind speeds of 20, 25, 30 and 35 m s1 are respectively used in 4 model runs. The minimum water depth is set to 10 m in all cases. The asymmetry of surge and fall, as illustrated in Fig. 5, is not evident when the wind speed is 20 or 25 m s1, and AI in both cases is less than 30%. When the wind speed increases to 30 m s1, AI increases

M. Peng et al. / Ocean Modelling 14 (2006) 81101


4

87

Sea Level (m)

a
0 0

Sea Level (m)

Case 1 Case 2 Case 3 Case 4 10 hour 20 hour 30 hour 40 hour 50 hour

Fig. 4. Spin-up/spin-down experiments with dierent bathymetry. Case 14 respectively takes 10, 15, 30 and 50 m as the minimum water depth: (a) spin-up with 30 m s1 onshore wind and (b) spin-down with 30 m s1 oshore wind.

accordingly to 56%. Under 35 m s1 wind forcing, the dierence of the maximum fall and surge increases to 4.1 m, with AI as large as 98%. The extent of the asymmetry, therefore, increases with wind forcing. However, as will be shown in the next section, stronger TC does not necessarily induce larger AI since wind duration of a TC may be not long enough to reach the steady state for sea surface elevation. As indicated in Fig. 5, spin-up time increases moderately with wind forcing. A balance is nearly reached between the wind forcing and the horizontal pressure gradient less than 10 h with 20 or 25 m s1 steady wind. Even for a higher speed wind (e.g. 30 or 35 m s1), the contribution of sea level increase after 10 h is less than 15% of the total surge. The spin-down time is dierent, though the dierence is not evident when the wind speed is small. Sea surface elevation is far from steady state after 10 h of 30 or 35 m s1 wind forcing. The contribution of sea level fall after 10 h is about 1.5 m for 30 m s1 wind, or 29% of its total fall. For 35 m s1 oshore wind, the steady state is not reached after the rst 50 h wind forcing as illustrated in Fig. 5. The above experiments imply the diculties to achieve sound sea level simulation in a TCs oshore-wind quadrant, where wind speed or duration mistake is prone to lead to more sea surface elevation errors. As the wind speed, duration, and fetch of a TC are the functions of RMW, translation speed, and wind inow angle, accurate specication of the parametric wind model is the key to achieve good storm surge results.

88
5

M. Peng et al. / Ocean Modelling 14 (2006) 81101

Sea Level (m)

a
0 0 1 2 3

Sea Level (m)

4 5 6 7 8 9 10 wind speed=35m/s wind speed=30m/s wind speed=25m/s wind speed=20m/s 10 hour 20 hour 30 hour 40 hour 50 hour

Fig. 5. Spin-up/spin-down experiments with dierent wind forcings: (a) spin-up with onshore winds and (b) spin-down with oshore winds.

4. The asymmetry of surge and fall with parametric wind forcings 4.1. Parametric wind models Many parametric models have been developed in TC wind studies (Holland, 1980; Harper and Holland, 1999; Vickery et al., 2000). In this study, as in Xie et al. (2004) and Peng et al. (2004), the Holland (1980) parametric wind model is employed to generate symmetrical, 0th order (circular) wind elds: P P c P n P c expA=rB V ABP n P c expA=r =qr A RB max B eqV 0wmax =P n P c
0 w B B 1=2

6 7 8 9

where q is the air density, P, the atmospheric pressure at radius r, Pc and Pn are respectively central and ambient pressures, A and B are scaling parameters, Rmax is RMW, V 0w is the pressure gradient induced wind velocity and V 0wmax is its maximum value at Rmax, and e is the base of natural logarithms. For the hypothetical TCs in Sections 4.24.4, the above parameters are set as: Pn = 1010 mb, q = 1.2 kg m3, and B = 1.9, an arbitrary value from its typical range (1.52.5) in Holland (1980), and Rmax is specied in the experiments. For the two

M. Peng et al. / Ocean Modelling 14 (2006) 81101

89

historic TCs in Section 5, P and q are the same as for the hypothetical TCs, B is determined by (9), where V 0wmax is the maximum symmetric wind speed, which is achieved from the dierence of the observed maximum wind speed and the translation speed, and Rmax is deduced from observations. The way to choose these parameters in historic TCs is due to the fact that V 0wmax is generally easier to obtain from the observations. The wind elds used in the storm surge model are the combination of (7) and the TCs translation speed VH as in (10). For symmetry purpose, VH is set to zero for the hypothetical TCs, but it takes the observed value for the two historic TCs. ~w V ~0 V ~H V w 10

The parametric wind model, assuming to the lowest order, gives a circular wind ow pattern around its center. This does not adequately reect the actual surface wind directions, and wind inow angles should be considered in real cases (Huston et al., 1999). However, as one major purpose of Sections 4.2 and 4.3 is on the asymmetry of sea level surge and fall with symmetric wind forcing, asymmetric factors, such as inow angle and VH, are not considered in these two subsections. 4.2. Translation speed To assess the asymmetry of the TC induced sea level surge and fall, the idealized square domain (Fig. 3), again, is employed with the hypothetical track running along the shoreline from south to north with land on the left. The minimum water depth is set to 20 m in this and the next two subsections to avoid potential drying

25km/h
4 2

20km/h

Storm Surge (m)

0 2 4 6 8 6 970mb 960mb 950mb 940mb 930mb 920mb 970mb 960mb 950mb 940mb 930mb 920mb

15km/h
4 2

10km/h

Storm Surge (m)

0 2 4 6 8 970mb 960mb 950mb 940mb 930mb 920mb 20 hour 40 hour 60 hour 80 hour 970mb 960mb 950mb 940mb 930mb 920mb 20 hour 40 hour 60 hour 80 hour

c
100 hour

d
100 hour

Fig. 6. The eect of tropical cyclones translation speed on the asymmetry of the maximum surge and fall in various central pressure conditions: (a) 25 km h1, (b) 20 km h1, (c) 15 km h1 and (d) l0 km h1.

90

M. Peng et al. / Ocean Modelling 14 (2006) 81101

that makes it dicult to evaluate the maximum sea level fall. Central pressure varies from 970 to 920 mb. RMW remains at 50 km though, statistically, RMW varies from 46 to 51 km according to Hsu and Yan (1998) for TCs with the specied intensity. The hypothetical TC starts 300 km away from the southern border of the domain, and the translation speed is specied to 25, 20, 15 or 10 km h1 respectively. Time series of sea surface elevation at location A are shown in Fig. 6. One has to bear in mind that a slower translation speed implies more wind duration for both onshore and oshore winds, though such winds are not always perpendicular to the coast as in the previous section. For TCs with xed central pressure, the slower the translation speed, the greater the asymmetry of the maximum surge and fall. For instance, AI for a 950 mb TC traveling at 25 km h1 is only 32%, but when its speed decreases to 10 km h1 the value of AI will increase to 71%. This is because the maximum fall, as presented in the previous section, is more slowly achieved, and its magnitude is larger than the maximum surge. Translation speed has almost no inuence on the maximum surge, but has signicant eect on the maximum fall (Fig. 6). This indicates that a steady state for positive surge has been reached at some time for every case. A steady state for negative surge, however, has never been reached. When a 920 mb TC slows down from 15 to 10 km h1, for example, the maximum surge is not changed, while the maximum fall (its absolute value) increases from 5.51 to 6.02 m. A stead state of fall, therefore, is not obtained. As a result, AI is increased. The details on how AI changes with translation speed are shown in Fig. 9a. For those with the same translation speed, a weaker TC is expected to gain a higher AI value though its induced maximum surge and fall are relatively smaller. Of the group with translation speed of 20 km h1, for instance, the maximum surge and fall for a 970 mb TC are only 1.55 m and 2.46 m respectively (see

RMW=30km
4

RMW=50km

Storm Surge (m)

4 970mb 960mb 950mb 940mb 930mb 920mb 970mb 960mb 950mb 940mb 930mb 920mb

12 8

RMW=70km
4

RMW=90km

Storm Surge (m)

4 970mb 960mb 950mb 940mb 930mb 920mb 20 hour 40 hour 60 hour 80 hour 970mb 960mb 950mb 940mb 930mb 920mb 20 hour 40 hour 60 hour 80 hour

c
100 hour

d
100 hour

12

Fig. 7. The eect of tropical cyclones RMW on the asymmetry of the maximum surge and fall in various central pressure conditions: (a) 30 km, (b) 50 km, (c) 70 km and (d) 90 km.

M. Peng et al. / Ocean Modelling 14 (2006) 81101

91

Fig. 6), or AI = 59%. A 920 mb strong TC generates 3.87 m surge and 5.03 m fall. The corresponding AI, however, is only 30%. This is dierent from the case when both onshore and oshore steady winds are long enough to reach the steady states. For most TCs, considering their typical RMW and translation speed, oshore wind is not long enough to reach the steady state for negative surge. For those with xed translation speed, a weaker TC (whose spin-up/spin-down time is shorter) is closer to reach the steady state, so the asymmetry feature is well revealed. 4.3. RMW RWM is another important parameter in the wind model that can inuence TC induced storm surge (Peng et al., 2004). The RMW of an individual TC may be much dierent from its intensity-based statistic value (Hsu and Yan, 1998). Larger RMW not only increases wind duration for both onshore and oshore winds, but also increases wind fetch. The eect of the latter is important and makes peculiar the relationship between RMW and AI as will be discussed later. In our experiments, the TCs translation speed is set to 15 km h1, and RMW are respectively 30, 50, 70 and 90 km. As expected, the extent of both surge and fall increases as RMW is enlarged (Fig. 7). For a 940 mb TC, when RMW is 30 km, the maximum surge and fall are 1.55 and 2.49 m respectively. As RMW increases to 90 km, the maximum surge and fall increases to 4.66 and 7.49 m. This is dierent from when the relationship between translation speed and the maximum surge and fall was addressed in the previous subsection. In

10 degree
4

20 degree

Storm Surge (m)

970mb 960mb 950mb 940mb 930mb 920mb

970mb 960mb 950mb 940mb 930mb 920mb

6 6

30 degree
4

40 degree

Storm Surge (m)

970mb 960mb 950mb 940mb 930mb 920mb 20 hour 40 hour 60 hour 80 hour

c
100 hour

970mb 960mb 950mb 940mb 930mb 920mb 20 hour 40 hour 60 hour 80 hour

d
100 hour

Fig. 8. The eect of tropical cyclones wind inow angle on the asymmetry of the maximum surge and fall in various central pressure conditions: (a) 10, (b) 20, (c) 30 and (d) 40.

92

M. Peng et al. / Ocean Modelling 14 (2006) 81101

the former case, only the maximum fall changes with translation speed. The correlation between RMW and the maximum surge in Fig. 7 is due largely to wind fetch variation as RMW changes. For those with xed RMW, the relationship between central pressure and AI depends on the value of RMW. If RMW is small, AI decreases as the central pressure decreases. This is true when RMW is 30 or 50 km as illustrated in Fig. 9b. However, for TCs with large RMW, there is no clear correlation between central pressure and AI. As RMW increases to 90 km, there is almost no AI change as central pressure varies. This reects that RMW aects AI in a way that is dierent from translation speed. Again, the dierence may be due to the variation of wind fetch. The results in Fig. 9b imply a balance between wind duration (whose increase may increase AI) and TCs intensity and wind fetch (whose increase may decrease AI). 4.4. Tropical cyclones inow angle Another parameter that can change the maximum surge and fall and the consequent AI is the TCs inow angle, which is the angle between the wind direction and the tangent to a circle concentric with the TC center. Unlike translation speed (without considering VH in the wind elds as in the two former subsections) and RMW, which inuence the maximum surge and fall by varying wind duration or fetch, inow angle aects the surge and fall by changing the symmetric nature of the lowest-order wind elds. No matter how much is changed of the wind duration or fetch owing to the change of translation speed or RMW, the variation itself, in terms of the accumulation of wind force on spin-up and spin-down, is symmetric. Holland (1980) model generates symmetric wind elds, so the asymmetry of the maximum surge and fall, as previously addressed, is the shallow waters intrinsic feature of responding to symmetric wind forcing asymmetrically. The consideration of inow angle fundamentally changes the symmetry of wind elds, and provides the surge model with an asymmetrical (with respect to its center) exterior forcing as in real TCs. If inow angle

1.0

10 degree 20 degree 30 degree 40 degree

0.5

Asymmetry Index
0

a
0.5 930 940 950

25km/h 20km/h 15km/h 10km/h


960

b
930 940 950

30km 50km 70km 90km


960

c
930 940 950 960

Central Pressure (mb)


Fig. 9. The relationship of AI with central pressure and (a) translation speed, (b) RMW, and (c) inow angle.

M. Peng et al. / Ocean Modelling 14 (2006) 81101

93

is considered in the model, a TC moving northward with land on its left in the Northern Hemisphere, as will be shown, is expected to generate larger surge and smaller fall. As a result, the overestimation of the asymmetry of surge and fall as discussed in the Introduction may be moderated. Fig. 8 shows how inow angles aect the maximum surge and fall. In our experiments, the RMW and translation speed are respectively xed to 50 km and 15 km h1 for comparison with the former cases. When no inow angles were considered (Fig. 7b), the asymmetry of the maximum surge and fall was evident. For example, the maximum surge, fall and AI for the 970 mb case are respectively 1.51 m, 2.57 m and 70.3%. If 10 is considered as the inow angle homogeneously across the eld (Fig. 8a), the maximum surge, fall and AI turn to 1.78 m, 2.23 m and 25.5%. The asymmetry of the maximum surge and fall is signicantly reduced. As the inow angle increase continues, the asymmetry of surge and fall is further reduced. In fact, as the angle increases to 20 and above, AI becomes negative for all central pressures (Fig. 9c), which indicates the maximum surge surpasses the maximum fall. The reason why inow angle can change the maximum surge and fall and the consequent AI is due to the asymmetry of wind elds and the orientation of the TC track. When inow angles are considered, the TCs moving in the Northern Hemisphere with land on the left may cause larger surge and smaller fall. Hurricane Isabel and Charley that will be addressed in the next section are this kind of TC. Reverse conclusions may hold for a TC with land on its right, or the same orientation in the Southern Hemisphere. In some sense, inow angle is more meaningful on AI adjustment than translation speed and central pressure. This is because inow angle is more elusive to TC forecasters than the other two parameters. More important, it exists in all TCs. In real case, inow angle may vary spatially at a time. Generally, its value near the eye is smaller than around its periphery. Although the upper limit of the inow angle is 25 as suggested in

37.0N Hurricane Charley 36.0N (2004)

35.0N

34.0N

Sunset Beach Springmaid Pier Charleston

33.0N

32.0N Fort Pulaski

31.0N

30.0N

81.0W 80.0W 79.0W 78.0W 77.0W 76.0W 75.0W 74.0W


Fig. 10a. The track of Hurricane Charley and the locations of water level stations. The box shows the inner domain of the nesting system.

94

M. Peng et al. / Ocean Modelling 14 (2006) 81101

some models (Phadke et al., 2003), it may be far out of the range in real cases. Accurate estimate of the inow angle can greatly improve the storm surge model output in real cases as will be illustrated in the next section. 5. Historic tropical cyclones Any mistakes in wind model specication may lead to errors in surge simulation. The errors, as indicated in the previous section, are more evident in the negative surge phase. As inow angles are not systematically measured and documented as the other parameters, it is singled out in the following hindcast studies to assess its eects on the asymmetry of the surge and fall. Hurricane Charley (2004) and Isabel (2003) are the two historic TCs to be studied, and their track, RMW, and central pressure are obtained from the interpolation of the observed data. 5.1. Hurricane Charley Hurricane Charley passed over the mid-eastern coast of the US on August 14, 2004, as shown in Fig. 10a. This Category One hurricane was not a severe case in this region. However, before landfall at Sunset Beach, North Carolina its track was almost parallel to the coast. So, it was similar to the hypothetical TCs in the previous section. Furthermore, this hurricanes surface wind elds at every 3 h were digitally archived by NOAA/HRD. So, not only its track, central pressure, and RMW, but also inow angles were systematically documented. For instance, at 08/14/1330Z the Hurricanes maximum wind speed, as indicated in Fig. 10b, is about 35 m s1 northeast of its eye. Its RMW is also suggested in the gure. More directly, Fig. 10c provides the observed wind elds by which the wind inow angles may be inferred. Based on the observation, 40 was found to be the best wind inow angle (least square tting for the two components of wind vectors) for the

Fig. 10b. NOAA/HRD H*Wind surface wind analysis of Hurricane Charley on August 14, 2004 1330 UTZ. The contours show wind speed (kt) and the position of the maximum surface wind is indicated by the arrow. The radii of 34, 50, and 64 kt winds for each storm quadrant are shown in the upper left corner of the gure. Winds are maximum 1-min sustained at 10 m and valid for marine and open terrain exposures.

M. Peng et al. / Ocean Modelling 14 (2006) 81101

95

34.0N

30m/s

33.0N

32.0N

82.0W

81.0W

80.0W

79.0W

78.0W

Fig. 10c. Charleys gridded H*Wind analysis wind elds for both marine and open terrain at 08/14/1330Z.

parametric wind elds (Fig. 10d). Consideration of this inow angle greatly improves the reliability of the simulated wind elds. Land friction has been considered for mapping of Fig. 10d, though wind stress over land has no inuence on coastal surge calculation, and this was accomplished by giving a specied friction factor in (10) when wind speed over land was calculated. The same will be performed in Fig. 11c when Hurricane Isabel is addressed in the next subsection. As mentioned before, only wind inow angle is considered as a variable in the experiments while others, such as track, central pressure, translation speed, and RMW, take their interpolated values at each time. Sea level data are available at four stations whose locations are illustrated in Fig. 10a. In the experiments, the inow angles vary from 0 to 50, but for clarity, only the results of 0, 20 and 40 are mapped in Fig. 10e in comparison with the observations. Hurricane Charley induced sea level disturbance was evident at Sunset Beach, Springmaid Pier, and Charleston, but there was almost no perturbation at Fort Pulaski, which is about 300 km south of the landfall location. For the 0 case, in which no wind inow angle is considered, the maximum dierence of the simulated and observed sea level is near 2.0 m at the three northern stations. Severe discrepancy of sea level occurs after the surge maxima. As wind inow angle increases, the dierence is getting smaller. The maximum discrepancy dwindles to about 1.0 m for these stations when the inow angle is 40. As shown in Fig. 10e, when inow angle is considered in the parametric wind model, the maximum negative surge is signicantly reduced, but the maximum positive surge is not signicantly changed. As a result, the asymmetry of the maximum surge and fall is moderated. 5.2. Hurricane Isabel The track of Hurricane Isabel has been shown in Fig. 1. As mentioned previously, the parametric wind elds with no inow angle correction lead to an apparent mistake of negative surge, though the overall

96

M. Peng et al. / Ocean Modelling 14 (2006) 81101

34.0N

30m/s

33.0N

32.0N

82.0W

81.0W

80.0W

79.0W

78.0W

Fig. 10d. Charleys simulated wind elds at 08/14/1330Z with wind inow angle of 40 as a correction of Holland wind elds, and land friction has been considered for the mapping.

maximum positive surge at most stations agreed well with the observations. In this subsection, dierent inow angles will be fed into the model to see how much dierence they can make with regard to sea level surge and fall and their asymmetry. Before landfall, the asymmetry of Hurricane Isabels wind elds was not as noticeable as of Hurricane Charley. This is shown in Fig. 11a and Fig. 11b, which respectively depict the wind structure and eld at 09/18/1330Z. The maximum wind speed is about 44 m s1 in the NE quadrant. In this case, 25 was found to be the best inow angle to match the observation, and the corresponding model wind eld is shown in Fig. 11c. Again, the sea level results at Chesapeake Bridge and Duck Pier of taking 0, 20 and 40 as the wind inow angles are mapped in Fig. 11d to compare with the observations. For the overall maximum sea level surge, the results are the best when the inow angle is 30 (not shown). Taking 20 and 40 as the inow angle respectively underestimates and overestimates the maximum surge. For negative surge, the error is largely reduced as inow angle increases. This is true at least within the selected range of inow angles. Although Fig. 11d illustrates that increasing inow angles moderate the asymmetry of the surge and fall and hence improve the simulation results, it does not indicate that the inow angle of 40 is closer to the truth than 20. In fact, as mentioned before, 25 is the best inow angle to match the observation. It is apparent that the maximum negative surge is still overestimated even with consideration of inow angles. As Hurricane Isabels track, central pressure, and translation speed were accurately measured and employed in the model, the possible mistakes may come from the input of RMW which can only be inferred from observations. Furthermore, Hurricane Charley is not a perfect circle (see Fig. 11a), and its RMW should take its own value in each quadrant at any time. One has to keep in mind that other factors which may also contribute to storm surge, such as precipitation and runos, are not considered in this study. The freshwater contribution may be important in the inner domain of Fig. 1. This is especially true at Chesapeake Bridge which is at the mouth of Chesapeake Bay,

M. Peng et al. / Ocean Modelling 14 (2006) 81101

97

Sunset Beach, NC
2 1 0 Observed 0 degree 20 degree 40 degree

Sea level under the influence of Hurricane Charley (m)

Springmaid Pier, SC
2 1 0 1 Observed 0 degree 20 degree 40 degree

Charleston, SC
2 1 0 1 Observed 0 degree 20 degree 40 degree

Fort Pulaski, GA
2 1 0 1 Observed 0 degree 20 degree 40 degree 14/0000Z 14/0600Z 14/1200Z 14/1800Z

Fig. 10e. The results of sea surface elevation time series taking 0, 20 and 40 as the wind inow angle are compared with the observation at four stations for Hurricane Charley.

the largest estuary system in the US. Even at Duck, 100 km south of the estuary mouth, sea level is still, to some extent, under the inuence of the inland runo. This is because freshwater in an estuary in the Northern Hemisphere generally moves to its right after jetting out from its mouth due to the earths rotation (Zhang et al., 1987; Pietrafesa and Janowitz, 1988). This may be another reason why, as shown in Fig. 11d, the maximum sea level fall is overestimated even with consideration of inow angles. The freshwater contribution, of course, may be far less than RMW. 6. Conclusion and discussion TC induced asymmetry of the maximum surge and fall is studied in this paper. It is found that oshore winds are more eective in inducing sea level fall than the corresponding onshore winds in inducing positive surge. As a result, the response of sea level in the TCs oshore-wind quadrant is more sensitive to wind forcing than in the onshore wind quadrant. Any misspecication of wind parameters, such as translation speed, RMW, and inow angle may lead to inaccurate sea level simulation. The study nds that the asymmetry of the maximum surge and fall is evident in shallow water regions. It takes more time for an oshore wind to reach the steady state. Or, the spin-down time is longer than spin-up time. In deep water, the huge original undisturbed water depth subdues the asymmetry of sea level surge and fall. For steady onshore and oshore winds with long duration, the asymmetry increases with wind forcing. Any change of parameters in the TC parametric wind model may lead to changes in sea level surge and fall and the consequent asymmetry. Generally, the parameters fall into two major categories as to how to aect

98

M. Peng et al. / Ocean Modelling 14 (2006) 81101

Fig. 11a. NOAA/HRD H*Wind surface wind analysis of Hurricane Isabel on September 18, 2003 1330 UTZ. The contours show wind speed (kt) and the position of the maximum surface wind is indicated by the arrow. The radii of 34, 50 and 64 kt winds for each storm quadrant are shown in the upper left corner of the gure. Winds are maximum 1-min sustained at 10 m and valid for marine and open terrain exposures.

37.0N

30m/s

36.0N

35.0N

34.0N

33.0N

32.0N

78.0W

77.0W

76.0W

75.0W

74.0W

73.0W

72.0W

Fig. 11b. Isabels gridded H*Wind analysis wind elds for both marine and open terrain at 09/18/1330Z.

M. Peng et al. / Ocean Modelling 14 (2006) 81101

99

37.0N

30m/s

36.0N

35.0N

34.0N

33.0N

32.0N

78.0W

77.0W

76.0W

75.0W

74.0W

73.0W

72.0W

Fig. 11c. Isabels simulated wind elds at 09/18/1330Z with wind inow angle of 25 as a correction of Holland wind elds, and land friction has been considered for the mapping.
Observed 0 degree 20 degree 40 degree

Chesapeake Bridge (S4)

Maximum Storm Tide (m)

Maximum Storm Tide (m)

Observed 0 degree 20 degree 40 degree

Duck Pier (S3)

1 17th00Z 12Z 18th00Z 12Z 19th00Z 12Z

Fig. 11d. The results of sea surface elevation time series taking 0, 20 and 40 as the wind inow angle are compared with the observation at two stations for Hurricane Isabel.

100

M. Peng et al. / Ocean Modelling 14 (2006) 81101

surge and fall. Varying wind duration and fetch is one and varying the symmetry of wind eld is the other. Translation speed and RMW are the parameters of the rst category while the inow angle is of the second. In the rst category, translation speed only changes wind duration for a location, but RMW changes both wind duration and fetch. For TCs with xed intensity, the slower the translation speed, the greater the asymmetry of the surge and fall. This is because in most TC cases the steady state for onshore wind is reached. However, the steady state for oshore wind is not. It is closer to be reached as translation speed decreases. The asymmetry of the maximum surge and fall, therefore, becomes more evident as the translation speed decreases. For TCs with the same translation speed, a weaker one is more likely to gain a higher AI value though its induced maximum surge and fall are relatively smaller. This is dierent from the long steady wind forcing case. It should be noticed that the wind used to drive the storm surge model for the hypothetical TCs is not the combination of the circularly parametric wind and translation speed VH. The latter is ignored for the purpose of obtaining a symmetric wind eld and emphasizing the variation of wind duration due to translation speed change. In the two historic TCs, however, the translation speed is combined with the circular wind as the driving force. If translation speed is xed, a larger RMW oers longer wind duration and larger wind fetch. As a result, the extent of both surge and fall increases with RMW. The relationship between central pressure and AI, however, depends on the value of RMW. If RMW is small, AI decreases with increasing TC intensity. However, for large RMW, there is no distinct correlation between central pressure and AI. These conclusions may be only correct within the specied parameter ranges in the experiments. The results indicate that RMW aects AI in a way that is dierent from translation speed, and it is the changing fetch that makes the dierence. In fact, the relationship among RMW, central pressure, and AI, implies a balance of wind duration, fetch and TCs intensity. Another parameter that can change the asymmetry index is the TCs inow angle, which is the only parameter in the study that changes the circular feature of the parametric wind elds. This is dierent from the other two parameters in the way to induce asymmetry of surge and fall. No matter how much wind duration or fetch is changed owing to the change of a TCs translation speed or RMW, the variation itself is symmetric. The asymmetry of surge and fall reects the shallow waters intrinsic feature of asymmetrical response to symmetric wind forcing. The consideration of TCs inow angle, however, fundamentally changes the symmetry of parametric wind eld and oers an asymmetrical wind forcing. Adding inow angles to symmetric wind as a correction may increase the maximum surge and, simultaneously, decrease its maximum fall. The asymmetry index, as a result, is reduced. This may be only the case for TCs moving northward with land on the left in the Northern Hemisphere. As inow angle is more elusive to TC forecasters than the other two parameters in real cases, its misspecication may lead to large errors in storm surge forecast. The importance of considering a reasonable TC inow angle is illustrated in the hindcast of Hurricane Charley and Hurricane Isabel. In both cases, adding inow angles greatly improves sea level simulation results. The consideration of inow angle for these two TCs results in AI decrease at all coastal stations due to the moderation of the maximum sea level fall. This greatly improves storm surge results. Acknowledgements This study is supported by the Carolina Coastal Ocean Observation and Prediction System (Caro-COOPS) project under NOAA Grant No. NA16RP2543. Caro-COOPS is a partnership between the University of South Carolina and North Carolina State University. We thank Madilyn Fletcher and Earle Buckley for constructive comments and project coordination, Shaowu Bao for helpful discussions on the tropical cyclone wind model. References
Harper, B.A., Holland, G.J., 1999. An updated parametric model of the tropical cyclone. In: Proceedings of the 23rd Conference Hurricanes and Tropical Meteorology, 1015 January 1999. American Meteorological Society, Dallas, TX.

M. Peng et al. / Ocean Modelling 14 (2006) 81101

101

Holland, G.J., 1980. An analytic model of the wind and pressure proles in hurricanes. Mon. Wea. Rev. 108, 12121218. Hsu, S.A., Yan, Z., 1998. A note on the radius of maximum wind for hurricanes. J. Coast. Res. 14 (2), 667668. Huston, S.H., Shaer, W.A., Powell, M.D., Chen, J., 1999. Comparisons of HRD and SLOSH surface wind elds in hurricanes: implications for storm surge modeling. Wea. Forecast. 14, 671686. Jelesnianski, C.P., 1967. Numerical computation of storms surges with bottom stress. Mon. Wea. Rev. 95, 740756. Jelesnianski, C.P., 1972. SPLASH (special program to list amplitudes of surges from hurricanes): Landfall storms. NOAA Technical Memorandum NWS TDL-46, National Oceanic and Atmospheric Administration, US Department of Commerce, 33 pp. Jelesnianski, C.P., Chen, J., 1984. SLOSH (Sea, Lake, and Overland Surges from Hurricanes). NOAA Technical Memorandum. Jelesnianski, C.P., Chen, J., Shaer, W., 1992. SLOSH: Sea, Lake, and Overland Surges from Hurricanes. NOAA Technical Report NWS 48. Large, W.G., Pond, S., 1981. Open ocean momentum uxes in moderate to strong winds. J. Phys. Oceanogr. 11, 324336. Mellor, G.L., 1996. Users Guide for a Three Dimensional, Primitive Equation, Numerical Ocean Model. Princeton University, Princeton, NJ. NOAA webpage2 @ <http://www.noaa.gov/galveston1900>. NOAA webpagel @ <http://www.nhc.noaa.gov>. Palma, E.D., Matano, R.P., 1998. On the implementation of passive open boundary conditions for a general circulation model: the barotropic mode. J. Geophys. Res. 103, 13191341. Palma, E.D., Matano, R.P., 2000. On the implementation of passive open boundary conditions for a general circulation model: the threedimensional case. J. Geophys. Res. 105, 86058627. Peng, M., Xie, L., Pietrafesa, L., 2004. A numerical study of storm surge and inundation the Croatan-Albemarle-Pamlico estuary system. Estuarine Coast. Shelf Sci. 59, 121137. Peng, M., Xie, L., Pietrafesa, L., in press. A Numerical study on hurricane induced storm surge and inundation in Charleston Harbor, South Carolina. J. Geophys. Res. Phadke, A., Martino, C., Cheung, K.F., Houston, S.H., 2003. Modeling of tropical cyclone winds and waves for emergency management. Ocean Eng. 30, 553578. Pietrafesa, L.J., Janowitz, G.S., 1988. Physical oceanographic processes aecting larval transport around and through North Carolina Inlets. Am. Fish. Soc. Sympos. 3, 3450. Pietrafesa, L.J., Janowitz, G.S., Chao, T.-Y., Weisberg, R.H., Askari, F., Noble, E., 1986. The physical oceanography of the Pamlico Sound. UNC Sea Grant Publication UNC-WP-86-5, Raleigh, NC. Powell, M.D., Vivkery, P.J., Reinhold, T.A., 2003. Reduced drag coecient for high wind speeds in tropical cyclones. Nature 422, 278283. Reid, E.O., Bodine, B.R., 1968. Numerical model for storms surges in Galveston Bay. Journal of the Waterways and Harbors Division, ASCE, 94, WWL, Proc. Paper 5805, 3357. Ruth, P., Posey., P., Dawson, G., 2005. Hurricane Isabel: a numerical model study of storm surge along the east coast of the United States. In: Sixth Conference on Coastal Atmospheric and Oceanic Prediction and Processes, P2.9. Vickery, P.J., Skerlj, P.F., Steckley, A.C., Twisdale, L.A., 2000. Hurricane wind eld model for use in hurricane simulations. J. Struct. Eng. (2000), 12031221. Wanstrath, J.J., Whitaker, R.E., Reid, R.O., Vastano, A.C., 1976. Storm surge simulation and application. Corps of Engineers Technical Report No. 76-3, US Army Coastal Engineering Research Center, US Department of Defense, 166 pp. Welander, P., 1961. Numerical prediction of storm surges. Adv. Geophys. 8, 315379. Xie, L., Pietrafesa, L.J., Peng, M., 2004. An integrated storm surge and inundation modeling system for lakes, estuaries and coastal ocean. J. Coast. Res. 20, 12091223. Zhang, Q.H., Janowitz, G.S., Pietrafesa, L.J., 1987. The interaction of estuarine and shelf waters: a model and application. J. Phys. Oceanogr. 17, 455469.

Você também pode gostar