References

Energy2green Wind And Solar Power System

Wind Energy DIY Guide

Get Instant Access

Bye, J.A.T., 1996: Coupling Ocean-Atmosphere Models. Earth-Science Reviews, 40, 149-162.

Bye, J.A.T., Keay, K., 2006: A Global Relation for Tropical Cyclone Development. Proceeedings of the 8 th Intl. Conf. On Southern Hemisphere Meteorology and Oceanography, Foz do Iguasu, Brazil, Amer. Meteorol. Soc.

Bye, J.A.T., Keay, K., 2008: A New Hurricane Index for the Caribbean. Interciencia, 33(8), 556-560.

Camargo, S., K. A. Emanuel, et al., 2007: Use of a Genesis Potential Index to Diagnose ENSO Effects on Tropical Cyclone Genesis. J. Climate, 20, 4819-4834 doi: 10.1175/JCLI4282.1.

Chauvin F., J. F. Royer, et al., 2006: Response of Hurricane-type Vorticies to Global Warming as Simulated by ARPEGE-Climat at High Resolution. Climate Dynamics, 27, 377-399.

Gordon A., W. Grace, P. Schwerdtfeger, R. Byron-Scott, 1998: Dynamic Meteorology: A Basic Course. Arnold, London

Gordon H. B., L. D. Rotstayn, et al., 2002: The CSIRO Mk3 Climate System Model [Electronic publication], CSIRO Atmospheric Research (CSIRO Atmospheric research Tech Pap no 60).

Gray, W. M., 1975: Tropical Cyclone Genesis, Dept. of Atmospheric Science Paper 234, Colorado State University, Fort Collins, CO.

Kalnay, E., M. Kanamitsu, et al., 1996: The NCEP/NCAR 40-year Reanalysis Project. Bull. Amer. Meteorol. Soc, 77, 437-471.

Kanamitsu, M., W. Ebisuzaki, et al., 2002: NCEP-DOE AMIP-II Reanalysis (R-2) Bull. Amer. Meteorol. Soc, 83, 1631-1643.

Powell, M. D., P. J. Vickery, et al., 2003: Reduced Drag Coefficients for High Wind Speeds in Tropical Cyclones, Nature, 422, 279-283.

Rayner, N. A., D. E. Parker, et al., 2003: Global Analyses of Sea Surface Temperature, Sea Ice, and Night Marine Air Temperature since the late Nineteenth Century. J. Geophys. Res., 108, doi 10.1029/2002JD002670.

Royer, J. F., F. Chauvin, et al., 1998: A GCM Study of the Impact of Greenhouse Gas Increase on the Frequency of Occurrence of Tropical Cyclones, Climatic Change, 38, 307-343.

Uppala, S., P. Kallberg, et al., 2004: ERA-40: ECMWF 45-year Reanalysis of the Global Atmosphere and Surface Conditions, 1957-2002 ECMWF Newsletter No 101 - Summer-Autumn 2004, 2-21.

Watterson, I.G., J. L. Evans, et al., 1995: Seasonal and Interannual Variability of Tropical Cyclogenesis: Diagnostics from Large-scale Fields. J. Climate, 8, 3052-3066.

Probability of Hurricane Intensification and United States Hurricane Landfall under Conditions of Elevated Atlantic Sea Surface Temperatures

Peter S. Dailey, Greta Ljung, Gerhard Zuba, and Jayanta Guin

Introduction

The genesis, intensification, and eventual demise of tropical cyclones (TCs) involve complex and dynamic interactions between the ocean, the atmosphere, and sometimes the land surface. Two key interactions are the favorable effects of sensible and latent heat transferred from the ocean's surface layer to the atmosphere [e.g., Emanuel, 2005], and the detrimental effects of vertical wind shear in the troposphere [e.g., Elsberry and Jeffries, 1996; Emanuel et al., 2004; Emanuel, 2005] on TC development. The competition amongst these factors can be viewed at a high level to determine seasonal activity levels. Most discussions surrounding tropical cyclogenesis, however, focus on the need for sufficiently warm sea surface temperatures (SSTs) [e.g. Chan et al., 2001]. Some of the latest research has shown that Tropical Cyclone Heat Potential (TCHP), which is a measure of upper layer ocean temperature, rather than that of the ocean's surface, is more highly correlated with TC intensification than SSTs alone, especially episodic and rapid intensification [Shay et al., 2000; Scharroo et al., 2005]. Unfortunately, this type of oceanographic data has only become available over the last ten years. Clearly, the enormous quantity of heat stored within the ocean serves as a reservoir of energy from which cyclones can develop and intensify. Other climate factors, such as the El Nino-Southern Oscillation (ENSO) cycle and its impact on Atlantic wind shear, tend to modulate the underlying capacity of the ocean-atmosphere system to support tropical activity, with ocean heat being the primary driver of activity.

Tropical latitudes within the North Atlantic Ocean have historically served as a fertile breeding ground for TCs, especially during the late summer and early autumn months, when SSTs are most elevated. The historical record indicates SSTs in the North Atlantic undergo fluctuations about a long-term average in phases that can last several decades. The physical cause of such multi-decadal fluctuations is a matter of debate [e.g., Mehta, 1998; Xue et al., 2003; Dima and Lohmann, 2007]. It is generally recognized, however, that fluctuations in SSTs are

J.B. Eisner and T.H. Jagger (eds.), Hurricanes and Climate Change, 121

doi: 10.1007/978-0-387-09410-6, © Springer Science + Business Media, LLC 2009

related to long-term fluctuations in tropical cyclone activity [Shapiro and Goldenberg, 1998].

Because the U.S. coastline lies in the path of many Atlantic TCs, and because population density has been trending upwards along the US coastline, it is important to study not only tropical activity as it occurs over the open ocean, but also the risk to life and property for those storms that make their way to the coastline. Historically, only about 12% of TCs reaching tropical storm strength (winds > 35 knots) have struck the U.S. coastline as hurricanes, with an average annual frequency of about 1.5 landfalling hurricanes per year. Despite these relatively low numbers, it is well understood that a single intense landfalling event like Hurricane Andrew (1992) or Hurricane Katrina (2005) can cause catastrophic loss. The human impact of even one intense landfalling hurricane is often far greater than an entire season of TCs that remain at sea.

Thus, this paper will focus on the relationship between sea-surface temperatures (SST) in the North Atlantic basin and the propensity of TCs to make landfall along the US coastline. In order for a tropical cyclone to make landfall as a hurricane, two conditions are required. First, the TC must intensify to hurricane strength (winds > 64 knots), and second, the hurricane must reach the U.S. coastline having maintained that strength. The paper will therefore focus on these two aspects of the TC evolution. Based on the historical record, we will examine the probability of a TC reaching hurricane strength, and the probability of the storm maintaining that strength through to the US coastline. By estimating these probabilities conditioned on regions where storms develop, one can assess how hurricane and landfall probabilities are modulated for individual seasons or individual events. In the end the comparison of warm SST years to climatology shows significant regional shifts conditioned on genesis for storms reaching the U.S. mainland as hurricanes.

Section 2 describes the data used in the study and the general approach to the analysis. The relationship between activity levels in the North Atlantic basin and the proportion of storms making U.S. landfall is analyzed in Section 3. Section 4 describes the balance between storms' probability to intensify and their probability to make landfall, based on the formation region.

Data

The analysis m+akes use of tropical cyclone track and intensity characteristics from the North Atlantic hurricane database (HURDAT) available from the National Hurricane Center [Jarvinen et al., 1984; http://www.aoml.noaa.gov/hrd/hurdat]. The analysis is limited to the period from 1948 to 2006 to eliminate uncertainty in storm counts and tracks in the North Atlantic basin in the early 1900s. Genesis locations for storms of at least tropical storm strength (named storms) are considered for the North Atlantic basin and are derived using the first track location in the data set. A small number of storms that cross into the Atlantic from the Pacific are excluded. In the end, we use a total of 626 genesis locations over 59 North Atlantic seasons capturing an average annual frequency of 10.6 named storms per year.

For purposes of determining the point at which basin storms reach hurricane strength, maximum wind speed from HURDAT is considered for each event. We also make use of the HURDAT landfall point, if any, along the U.S. coastline by interpolating the intersection of HURDAT six hourly track points with the U.S. coastline. For purposes of this paper, we only considered storms that explicitly cross the U.S. mainland. For example, storms that only cross the Florida Keys and the outer banks of North Carolina were not counted as landfalls (such storms are fairly uncommon, occurring one about every 10 years).

For sea surface temperatures, the analysis is based on the Hadley Center HadSST2 [Rayner, 2006] supplemented with National Oceanic and Atmospheric Administration's (NOAA) optimum interpolated (OI) SSTs for the most recent years [Reynolds, 2002]. We also make use of the SST spatial distributions available from the National Center for Atmospheric Research/National Center for Environmental Prediction (NCAR/NCEP) Reanalysis Project, available for the period from 1948 to 2006 [Kalnay et al., 1996]. The average of the SST anomalies during the months of August, September, and October (ASO) are used to quantify variability for the analysis period. The Hadley SST anomaly data series is shown in Fig. 1 since 1900. It shows multi-decadal periods in which anomalies are persistently colder or warmer than average. For purposes of this analysis, years in which North Atlantic SST anomalies are greater than or equal to zero are considered warm years and all other years are considered cold years. Only in one year (1954) was the anomaly equal to zero. Over the 58 remaining seasons from 1948 to 2006, there are 31 years classified as warm and 27 years classified as cold.

ï :

i 'r j

¡

1!

{ I { i

j j

.JJIÜ

L i

nIL 1

1_I

u

[1

iï il mi

1

OTITirPW

lllllllf

L :

rl

j

|„ ||

B

1

IIP

i

i u

■|i»■ «m- ru

n

:

1

-

Fnr

:

1110

1020

1830

1810

1960

1 wo

2000

HtiRtean* Season

1900

1110

1020

1830

1810

1960

1 wo

2000

HtiRtean* Season

Fig. 1 Hadley Centre Sea Surface Temperature Anomalies Data Set (HadSST2-ASO). The figure shows by year the mean Atlantic Ocean anomaly (degrees oC) for the months of August, September and October (ASO), the core of the Atlantic hurricane season, from 1900 to 2006. There are multi-decadal episodes in which anomalies are consistently colder or warmer than average, including the current warm period which began in the mid-1990s

Relationship between Basin and Landfall Activity

To examine the physical relationship between basin and landfall activity, we begin by estimating the probability of a storm making U.S. landfall - climatologically and under anomalous SST conditions. The landfall probability is estimated by the proportion of basin TCs that have historically made landfall. To illustrate, consider a TC forming in the basin. Once it becomes a named (tropical) storm, it has some climatological probability of making a U.S. hurricane landfall, and that probability is estimated by the long-term ratio of hurricane landfalls to the number of named storms that developed in the basin over the same period.

Landfall Probability

Table 1 shows the estimated landfall probability for storms of various intensities. Column A shows that the historical proportion of storms making landfall at tropical storm strength or greater is 31.0% in warm SST years and slightly higher (31.7%) in cold years. Column B shows that landfall proportion at hurricane strength or higher is about 12.2% in warm years and the cold year proportion is slightly higher (12.6%). For stronger hurricanes, shown in columns C (winds > 89 knots; strong hurricanes) and D (winds > 96 knots; major hurricanes), the cold year proportion is again higher than the warm year proportion, and the marginal difference grows with increasing intensity. For storms with a landfall intensity of at least 89 knots, the landfall proportion in warm years is notably smaller than in cold years, and it can be shown that this difference is statistically significant for certain genesis sub-regions within the North Atlantic basin [Dailey et al., 2007]. For the proportions computed in columns (B) to (D), it should be noted that the sample size is small. For example,

Table 1 Landfall proportion based on landfall counts for storms of different intensities. The table shows the proportion of named historical storms that make U.S. landfall over the period from 1948 to 2006. Numerator in each ratio is the number of landfalls at a given wind speed or higher; denominator is the total number of basin storms under the cold or warm SST condition. The landfall proportion is the ratio of the landfall count to the basin storm count for a given landfall wind speed, using (A) tropical storm strength (B) hurricane strength, (C) 89 knots (referred to in the text as strong hurricane strength), and (D) major hurricane strength (Saffir Simpson category 3 to 5). The third column, using a threshold of 89 knots, is consistent with later analyses and takes advantage of a larger number of hurricanes than is available in column (D)

Table 1 Landfall proportion based on landfall counts for storms of different intensities. The table shows the proportion of named historical storms that make U.S. landfall over the period from 1948 to 2006. Numerator in each ratio is the number of landfalls at a given wind speed or higher; denominator is the total number of basin storms under the cold or warm SST condition. The landfall proportion is the ratio of the landfall count to the basin storm count for a given landfall wind speed, using (A) tropical storm strength (B) hurricane strength, (C) 89 knots (referred to in the text as strong hurricane strength), and (D) major hurricane strength (Saffir Simpson category 3 to 5). The third column, using a threshold of 89 knots, is consistent with later analyses and takes advantage of a larger number of hurricanes than is available in column (D)

(A)

(B)

(C)

(D)

Landfalls > 35 kt

Landfalls > 64 kt

Landfalls > 89 kt

Landfalls > 96 kt

Warm SST

31.0% (117/378)

12.2% (46/378)

5.8% (22/378)

4.8% (18/378)

Years

Cold SST

31.7% (78/246)

12.6% (31/246)

6.5% (16/246)

5.3% (13/246)

Years

there have only been 31 landfalling hurricanes since 1948, and only 13 of them have achieved major hurricane status.

These results, which are based on data for the entire U.S., raise several questions. First, would a lower landfall proportion in warm SST seasons versus cold SST seasons have some physical explanation, and second, would physical factors account for regional differences in the landfall probability? If there are locally strong signals in the historical data, this might shed light on the suppression of overall proportions as noted in Table 1.

Physical Factors Influencing Landfall Probability

To address these questions from a physical perspective, one must examine the complete life cycle of TCs as it relates to landfall probability. Given that a TC forms in the North Atlantic basin, its probability of making landfall as a hurricane is dependent on three fundamental factors, namely (a) genesis (where the storm is born), (b) intensification and lyses (the intensification life cycle), and (c) tracking or steering (the storm's ability to approach and potentially cross the coastline). Modulation of any one of these can bring about significant changes in landfall probability and the resulting proportion of storms making landfall. Genesis determines a TC's initial proximity to land, but increasing proximity also limits the time it has to grow and intensify. Thus, genesis and intensification are intimately related. Steering currents, which relate to the atmospheric circulation, determine how far a TC deviates from its expected or "climatological track'', and hence the uncertainty in estimating landfall probability based solely on the climatological record. By closely examining each of these characteristics—genesis, intensification, and tracking—one can better understand which aspects of the TC life cycle are most sensitive to climate and most critical to landfall risk.

With regard to genesis, warm SSTs should have an enhancing effect since one of the necessary conditions for TC formation is the presence of sufficiently warm ocean temperatures. One expects warm SSTs to bring about increased intensification as well. This is not very clear from the historical data, however, since on average the probability of a TC intensifying to hurricane strength when the North Atlantic is anomalously warm is less than 2% higher than the long-term average. The explanation may lie in the fact that intensification is more sensitive to SST gradients than to SSTs themselves [Hennon, 2006; Chan et al., 2001], thus a uniform increase in ocean temperatures may not substantially modify intensification patterns. Finally, the degree to which TC steering responds to a warm ocean environment is difficult to quantify. Theoretically, a warmer than average ocean will translate into a warmer than average atmosphere, which in turn should induce a negative pressure perturbation on the semi-permanent high pressure situated over the North Atlantic. In fact, in an analysis using data from 1948 to 2005, we found a weak though statistically significant negative correlation between North Atlantic SST anomalies and sea-level pressure over the North Atlantic Ocean (not shown).

To quantify the impact of warm SSTs on storm tracks, one can apply stochastic or numerical modeling techniques to determine how historical storms would be steered differently under such SST conditions. A combination of long time scale fluctuations in ocean heat content and shorter time scale factors influencing TC steering, this complex subject will be considered in future work. In this study, we will limit the analysis to estimating hurricane probability and landfall probability conditioned on genesis alone.

Of course, if ocean anomalies were randomly distributed across the North Atlantic, or if they tended to concentrate outside the Main Development Region [MDR, see Emanuel, 2005], there might be little hope in tying SST anomalies to anomalies in genesis. But there are several plausible physical explanations for the spatial structure of ocean anomalies being systematic and non-random. First, the intensity of currents within the North Atlantic may limit the ability of anomalies to accumulate and stabilize. Modeling studies have shown that inter-annual variation in the upper ocean's heat content is linked to the advection of anomalous temperatures via the Gulf Stream [Dong and Kelly, 2003]. Within the Gulf Stream, ocean currents tend to be strong, thus, local anomalies may be short-lived. There is evidence, for example, that the strength of the Gulf Stream can reduce the persistence timescale of SST anomalies to just a few days [Gilman and Rothstein, 1994]. Shallow layers of ocean warmth may be transient not only because of swift currents, but also because passing disturbances tend to replace that warmth with cooler waters from below through the upwelling process. More stagnant regions of the North Atlantic, for example within parts of the Gulf of Mexico and along concave portions of the South American coastline, tend to accumulate warmth more readily than more dynamic regions of the open ocean. At the same time, during active periods of the "loop current'' (the clockwise flow that extends northward into the Gulf of Mexico and joins the Yucatan Current and the Florida Current), North Atlantic anomalies are much less relevant to TCs within the Gulf of Mexico. The relationship between the distribution of North Atlantic anomalies and their impact on TC development and intensification is clearly complex, but it is expected that SST anomaly patterns can be coherent and associated with the dynamics of the large scale ocean system. Principal Components Analysis (PCA) of the SST anomaly structure and its association with hurricane tracks has been the subject of limited research [e.g., Xie, et al., 2005] and certainly merits additional investigation. While anomaly patterns may not be predictable on inter-annual timescales, for purposes of this study it will be assumed that tropical SST anomalies are coherent within a season.

Figure 2 shows the spatial distribution of tropical storm genesis location, along with corresponding contours of genesis density, for all tropical cyclones in the North Atlantic. Genesis for all tropical storms and hurricanes is considered since subsequent intensification occurs somewhat independently of genesis location. The genesis density was estimated using a spatial kernel smoothing procedure described by Hall and Jewson [2005]. The optimal radius of influence (ROI), defining the circle of points used to estimate genesis density locally, (usually about 200 km) was roughly doubled in Fig. 2 to draw attention to larger-scale features in the genesis

-100

Fig. 2 Spatial Distribution and Density of Genesis for All Tropical Cyclones (1948-2006). Each dot represents a historical starting point for a TC that eventually reaches at least tropical storm strength. Grey scale contours show the mean genesis density in storms per square kilometer per year. In the top panel, the mean climatological genesis density is characterized by two hot spots, one within the Gulf and western Caribbean, and another along the Main Development Region (MDR). The center panel shows genesis distribution in warm years. Here, the pattern shifts with more focused genesis within the Gulf of Mexico and an eastward expansion of genesis along the MDR. The cold year counterpart is shown in the bottom panel

1.00

0.80

0.60

0.40

0.20

0.00

-100

Fig. 2 Spatial Distribution and Density of Genesis for All Tropical Cyclones (1948-2006). Each dot represents a historical starting point for a TC that eventually reaches at least tropical storm strength. Grey scale contours show the mean genesis density in storms per square kilometer per year. In the top panel, the mean climatological genesis density is characterized by two hot spots, one within the Gulf and western Caribbean, and another along the Main Development Region (MDR). The center panel shows genesis distribution in warm years. Here, the pattern shifts with more focused genesis within the Gulf of Mexico and an eastward expansion of genesis along the MDR. The cold year counterpart is shown in the bottom panel pattern. The top panel shows the climatological genesis pattern with two known "hot spots'' within the Gulf of Mexico and along the MDR. The relative scarcity of genesis in the vicinity of the Caribbean Islands can be explained by this region's high vertical wind shear and strong teleconnection to ENSO [Aiyyer and Thorn-croft, 2006]. The lower panel shows that, under warm SST conditions, genesis in the Gulf of Mexico shifts south and contracts while genesis in the North Atlantic shifts significantly eastward across the MDR and away from the U.S. coastline. This finding has several implications. First, storms that form in the Gulf are already close to the coastline; thus there is little time for a storm to intensify before landfall. If genesis density here tends to concentrate under warm SSTs, one expects increased probability of weaker hurricanes making landfall in the Gulf. Note that this does not account for less frequent Gulf landfalls with genesis outside the Gulf of Mexico. Though these are less frequent, they can be more intense. Second, storms that form in the MDR, especially off the coast of Africa, have a longer period to intensify, but are also much further from the U.S. coastline. This balance between opportunity to intensify and opportunity to make landfall naturally leads to the analysis carried out in Section 4.

Hurricane Intensification and Landfall Probability

Estimation Method

In this section we describe the method used to evaluate the likelihood of a storm making landfall given its genesis location. The same procedure is then used to evaluate the proportion of storms which later become hurricanes, and the proportion of storms which later make landfall as hurricanes.

One way to estimate the likelihood of storms originating from a specific region making landfall would be to subdivide the North Atlantic basin into sub-regions and compute the ratio of landfalling storms to the total number of storms that form in this region. However, a disadvantage of this method is that the result for a given location may be very sensitive to the size of the sub-region chosen. If the box is too small, it may not capture sufficient historical genesis, and accurate quantification of the proportion is not possible. We therefore prefer a smoothing based approach that allows for estimation of landfall probabilities without the use of grid boxes. The smoothing method used is a variant of a kernel smoothing method used to determine the density contours in Fig. 2.

The first step is to mark genesis locations with an indicator variable ILF that depends on the landfall information for the corresponding storm track. The indicator variable is set to 1 if the storm later makes a U.S. landfall and to 0 if it does not. To estimate landfall probability for a given genesis location, we use a weighted averaging technique that incorporates genesis information from neighboring sites within the radius of influence (ROI). Gaussian weights are applied when averaging within the ROI neighborhood. The length scale used for determining the neighborhood is calculated as follows.

1. ROI or length scale l is chosen, e.g. 300 km, based on Hall and Jewson [2005]

2. Gaussian weights are calculated for all other genesis locations according to the distance di from the current location to a neighboring location according to wi = e~d.2/l2

3. Effective landfall proportion pLF is computed as a weighted average of the indicator variable ILF for all locations within the chosen ROI according to pLF = ^ JLF,wi /

4. Steps 1-3 are repeated for all genesis locations

To estimate the optimum value for the length scale l, a cross validation is performed by looping through all years between 1948 and 2006. For each year, the probability surface is calculated using data from all other years. Landfall probabilities for a given year are the values of the probability surface at the genesis locations of that year. These values are compared to the historical values and the differences in probabilities are accumulated for all locations and over all years. The total accumulation of all differences is minimized by varying the length scale l. After the optimization is complete, the final probability surface is calculated using the optimal l along with the historical data for 1948-2006.1

The optimal value of l depends on the parameter of interest and the value increases with increasing sparseness of the data. In the current analysis, the length scale is about 400 km for genesis locations that produce landfalling TCs and about 800 km for genesis locations that produce strong hurricanes or strong landfalling hurricanes.

Hurricane Intensification Probability

The procedure described above will now be used to analyze the probability of tropical cyclones becoming a hurricane. However, instead of flagging a genesis

1 The outlined procedure can be applied to various tropical cyclones parameters. Parameters analyzed here and their definitions include: Landfalling Tropical Storms: wind speed at landfall >35 knots; Landfalling Hurricanes: wind speed at landfall >64 knots; Landfalling Strong Hurricanes: wind speed at landfall > 89 knots; Storms Becoming Hurricanes: maximum wind speed along track >64 knots; Storms Becoming Strong Hurricanes: maximum wind speed along track > 89 knots. Note that strong hurricanes counts have been chosen for analysis (versus major hurricanes) in order to increase the sample size for the more intense events.

point according to subsequent landfall, each genesis point is now flagged 1 if the storm reaches hurricane strength at any point along the track, and 0 otherwise. After computing the probability surface, the average value at all historical genesis locations for the period 1948-2006 as well as the average value for individual years 2001-2006 are also computed.

The results of this analysis are shown in Table 2. Included in this table are also the historical proportions of storms that reach hurricane status over the period 1948-2006 as well as for the individual years 2001-2006. The table shows that historically 58% of storms have become hurricanes and 32% have become strong hurricanes (wind threshold of 89 knots) in the time frame of 1948-2006. The historical values are close to the predicted value of 59% and 33%, respectively, obtained through smoothing.

The advantage of having a probability surface is that the probability can be evaluated at any location, not just at the historically observed genesis locations. In addition, one can evaluate the average probability of storms becoming hurricanes given the genesis locations for a particular season. The value is then compared to climatology (long-term mean 1948-2006) and to the actual value realized in that season. This is illustrated by the modeled values for 2001-2006 given in Table 2. These values show that the probability of TCs becoming hurricanes is very close to the climatological value. The differences are larger for strong hurricanes. For example, the modeled value of 21% for 2002 is smaller than the climatological value of 33%, suggesting that the TC genesis in 2002 occurred in places where the chance of TC intensification was below average. The observed proportion of 17% is also smaller than the long term average of 32%. For 2004, on the other hand, the intensification probability of 37% for strong hurricanes exceeds the climatological value indicating that the genesis locations of 2004 were in regions with higher than average probability to intensify. The observed value for 2004 is also higher than the long term average indicating that favorable intensification conditions existed along the storm tracks. In contrast, the 2006 storms had a larger than average potential

Table 2 Probability of storms becoming hurricanes and strong hurricanes for the 1948-2006 analysis period and for select years. The values for individual years are calculated by averaging the estimated climatological values at the genesis locations of that year

Hurricane Strong Hurricane

Table 2 Probability of storms becoming hurricanes and strong hurricanes for the 1948-2006 analysis period and for select years. The values for individual years are calculated by averaging the estimated climatological values at the genesis locations of that year

Hurricane Strong Hurricane

modeled

actual

modeled

actual

Climatology (1948-2006)

0.59

0.58

0.33

0.32

2001

0.60

0.60

0.31

0.33

2002

0.54

0.33

0.21

0.17

2003

0.58

0.44

0.30

0.25

2004

0.61

0.60

0.37

0.47

2005

0.58

0.54

0.30

0.29

2006

0.63

0.50

0.39

0.20

(39%) of becoming strong hurricanes, while only a 20% proportion was observed. This points to unfavorable conditions along the storm tracks in 2006.

In Section 3 we established that the density pattern of genesis locations is different in warm versus cold SST years. To analyze if this has an influence on the probabilities presented in Table 2, we repeated the analysis and selected only genesis location of warm years. Fortunately, all years from 2001 through 2006 shown in Table 2 are all warm years, therefore one can make a direct comparison between conditioned (on warm SSTs) and unconditioned estimates. Table 3 shows the results of this analysis. Observed values for individual years are included for reference. Although there was an apparent regional shift in the genesis location for warm SST years compared to climatology (Fig. 2), the long term average probability of a tropical storm becoming a hurricane or a strong hurricane is essentially the same in Tables 2 and 3). Similarly, differences are small for individual years.

These results, which are based on data for the entire basin, suggest that warm SSTs do not have a significant impact on the basinwide probability of tropical storms becoming hurricanes. However, regional differences may still exist. This is illustrated in Fig. 3, which shows the probability distribution of storms becoming strong hurricanes given the genesis location. Regional differences tend to correspond to shifting genesis within the MDR. During warm SST years more TCs form in the eastern North Atlantic and these storms have a higher likelihood of becoming strong hurricanes contrasted with storms that form in the same region during cold years.

To summarize, the results so far show no basinwide difference in the likelihood of storms to become hurricanes or strong hurricanes during warm SST years. Regional differences presented in Fig. 3, however, imply that there might be larger impacts on landfalls. We will further explore this hypothesis.

Table 3 Probability of storms becoming hurricanes and strong hurricanes in warm SST years for the 1948-2006 analysis period and for select years. The values for individual years are calculated by averaging the estimated climatological values at the genesis locations of that year

Hurricane Strong Hurricane

Table 3 Probability of storms becoming hurricanes and strong hurricanes in warm SST years for the 1948-2006 analysis period and for select years. The values for individual years are calculated by averaging the estimated climatological values at the genesis locations of that year

Hurricane Strong Hurricane

modeled

actual

modeled

actual

Warm Year Climatology

0.60

0.59

0.34

0.34

(1948-2006)

2001

0.59

0.60

0.30

0.33

2002

0.53

0.33

0.19

0.17

2003

0.57

0.44

0.29

0.25

2004

0.62

0.60

0.38

0.47

2005

0.58

0.54

0.29

0.29

2006

0.64

0.50

0.38

0.20

0.36

0.24

(b) Warm years

0.48

0.24

0.00

(c) Cold years

Fig. 3 Probability pattern for genesis locations of storms becoming strong hurricanes for (a) all years 1948-2006, (b) warm and (c) cold years within this period. All genesis locations for the selected years are drawn. Genesis locations that originate strong hurricanes are drawn as squares with a center dot. The darkest shade indicates that 48 to 60 percent of the storms forming in that region became strong hurricanes

0.60

0.48

0.24

Fig. 3 Probability pattern for genesis locations of storms becoming strong hurricanes for (a) all years 1948-2006, (b) warm and (c) cold years within this period. All genesis locations for the selected years are drawn. Genesis locations that originate strong hurricanes are drawn as squares with a center dot. The darkest shade indicates that 48 to 60 percent of the storms forming in that region became strong hurricanes

0.00

0.60

0.36

0.36

0.00

Hurricane Landfall Probability

Table 4 shows the climatological probability of landfall as tropical cyclones (TC), hurricanes, and strong hurricanes, given all genesis locations. As in the previous analysis, estimates are also made for individual years from 2001 to 2006. The results show that genesis in 2002 produced storms with a higher than average likelihood (40% compared to 31%) of producing landfalling TCs and the observed landfall proportion was also above average (58%) in this year. For landfalling strong hurricanes in 2004 and 2005, the modeled probabilities were close to average (6%) but the actual values were well above average (27% and 14%, respectively). This implies "favorable" tracking conditions in these two years.

The information in Table 4 is based on data for the entire period 1948-2006. A similar analysis is presented in Table 5, using data from warm SST years only. The differences between the results in the two tables are generally small. The most notable differences in Table 4 can be found for storms that later become strong landfalling hurricanes. For example, for 2003, the potential for strong hurricane landfalls is 5.2% based on data for the entire period. This is 86% of the corresponding climatological value of 6.1%. The percentage drops to 4.4%, which is 79% of the climatological value (5.6%), when the analysis is based on warm SST years in Table 5.

In a separate analysis (Dailey et al., 2007), landfall counts in warm SST years were analyzed revealing marginal significance when considering the entire U.S. coastline. Significant differences in landfall rates were found regionally, however, especially for the Southeast coast of the U.S. from the tip of Florida to Cape Hatteras, where the landfall rates increased during warm SST years. This raises a logical question whether warm ocean conditions can translate to significant shifts in landfall probability. The estimated probability of a storm—having formed

Table 4 Probability of storms making landfall as tropical cyclones, hurricanes, or strong hurricanes for the 1948-2006 analysis period and for select years. The values for individual years are calculated by averaging the estimated climatological values at the genesis locations of that year All Tropical Hurricane Strong Hurricane

Table 4 Probability of storms making landfall as tropical cyclones, hurricanes, or strong hurricanes for the 1948-2006 analysis period and for select years. The values for individual years are calculated by averaging the estimated climatological values at the genesis locations of that year All Tropical Hurricane Strong Hurricane

modeled

actual

modeled

actual

modeled

actual

Climatology

0.31

0.31

0.12

0.12

0.061

0.059

(1948-2006)

2001

0.28

0.20

0.11

0.00

0.049

0.000

2002

0.40

0.58

0.10

0.08

0.037

0.000

2003

0.32

0.31

0.12

0.13

0.052

0.000

2004

0.23

0.53

0.11

0.27

0.056

0.267

2005

0.30

0.25

0.12

0.14

0.060

0.143

2006

0.21

0.20

0.11

0.00

0.065

0.000

Table 5 Probability of storms making landfall as tropical cyclones, hurricanes, or strong hurricanes for warm SST years within the 1948-2006 analysis period and for select years. The values for individual years are calculated by averaging the estimated climatological values at the genesis locations of that year

All Tropical Cyclones (TC)

Hurricane (>64 knots)

Strong Hurricane (>89 knots)

Table 5 Probability of storms making landfall as tropical cyclones, hurricanes, or strong hurricanes for warm SST years within the 1948-2006 analysis period and for select years. The values for individual years are calculated by averaging the estimated climatological values at the genesis locations of that year

All Tropical Cyclones (TC)

Hurricane (>64 knots)

Strong Hurricane (>89 knots)

modeled

actual

modeled

actual

modeled

actual

Warm Year Climatology

0.30

0.30

0.12

0.12

0.056

0.055

(1948-2006)

2001

0.30

0.20

0.10

0.00

0.044

0.000

2002

0.38

0.58

0.09

0.08

0.033

0.000

2003

0.34

0.31

0.11

0.13

0.044

0.000

2004

0.23

0.53

0.11

0.27

0.059

0.267

2005

0.29

0.25

0.12

0.14

0.052

0.143

2006

0.20

0.20

0.10

0.00

0.065

0.000

somewhere in the North Atlantic basin—making a hurricane landfall along the Southeast coast of the U.S. is 4.5%. The estimate increases to 5.5% when the analysis is based on warm SST years. The sample sizes are small, however, and the difference between cool and warm years is not statistically significant at the 10% level. This was established using bootstrapping as opposed to a traditional test on proportions because of the small sample sizes. Though some shifts in landfall risk cannot be firmly demonstrated as significant based on the limited historical record, there are plausible physical explanations for why such shifts may occur under the influence of a non-stationary climate. This is a subject of continuing research. Even a small increase in the underlying probability may be of practical significance to those interested in subtle frequency modulations brought about by climate (e.g., catastrophe risk managers).

The regional pattern of strong hurricane landfall probability is shown in Fig. 4. The figure is based on data for the entire U.S. coastline. Again regional shifts, away from the U.S. coast and towards the eastern part of the North Atlantic, are noticeable. In addition, Gulf coast probabilities decrease when only warm years are considered. It can be seen that strong hurricane landfalls originating from the eastern Atlantic occur almost exclusively during warm years. Not surprisingly, the cold year pattern is a mirror image of the warm with noticeable probability shifts in the Gulf of Mexico and along the MDR.

Summary

This study has examined a very specific aspect of tropical cyclone risk, namely the risk that storms forming in the North Atlantic basin will become hurricanes and subsequently make landfall as hurricanes along the U.S. coastline. Though much

(b) Warm years

0.12

0.09

0.06

0.03

0.00

(c) Cold years

(c) Cold years

0.12

0.09

0.06

0.03

0.00

Fig. 4 Probability pattern for genesis locations of storms leading to a strong hurricane landfall for (a) all years 1948-2006, (b) warm and (c) cold years within this period. All genesis locations for the selected years are drawn. Genesis locations that originate strong hurricane landfalls are drawn as squares with a center dot. The darkest shade indicates that 12 to 15 percent of the storms forming in that region made landfall as a strong hurricane

X Coord

Fig. 4 Probability pattern for genesis locations of storms leading to a strong hurricane landfall for (a) all years 1948-2006, (b) warm and (c) cold years within this period. All genesis locations for the selected years are drawn. Genesis locations that originate strong hurricane landfalls are drawn as squares with a center dot. The darkest shade indicates that 12 to 15 percent of the storms forming in that region made landfall as a strong hurricane

0.15

0.12

0.09

0.06

0.03

0.00

0.15

0.12

0.09

0.06

0.03

0.00

0.15

0.12

0.09

0.06

0.03

0.00

X Coord attention has been paid to the influence of climate on the frequency of basin storms, less emphasis has been placed on the relationship between basin activity and landfall activity. Three key reasons are (a) the low level of long-term U.S. hurricane landfall frequency and the corresponding difficulty in detecting low level trends, (b) the general lack of fully reliable historical data, especially in the early 1900s, and (c) the complexity of the interaction and feedback amongst various climate signals. Despite these challenges, creative use of analytical tools can foster progress in this developing field. In this paper, we have examined from multiple perspectives the modulation of the regional risk of landfall in the North Atlantic basin. The analysis centers on the impact of warm SSTs on the probability of hurricane intensification (long-term average 58%) and the probability of US hurricane landfall (long-term average 12%). Though tracking mechanisms are very likely modulated by warm SSTs, such relationships are difficult to quantify since steering currents vary on a shorter timescale than the typical hurricane season. The data does, however, indicate that steering may be influenced by climate conditions, and this is a subject of continuing research.

The overall probability of tropical storms becoming hurricanes within the Atlantic basin appears to be stable under both warm and cold SST conditions. In fact, even within the current active period since 1995, data shows that the estimated probability of hurricane intensification lies relatively close to the long-term average despite large variations in seasonal genesis locations. The warm year spatial pattern is not very different than the climatological pattern.

When conditioned on genesis, the findings with regard to regional landfall proportion are noteworthy. Within the Gulf of Mexico, warm SSTs appear to reduce the probability of hurricane landfalls below the U.S. average. A physical explanation for this result is that increased genesis occurring with the Gulf should result in more storms that have a limited ability to intensify to hurricane strength. Thus, one expects a significant increase in the probability of tropical storm landfalls along the Gulf coast, but less so for hurricanes and especially major hurricane landfalls. Of course, not all Gulf landfalls originate within the Gulf of Mexico, but storms that spend their entire life cycle within the Gulf appear to dominate the landfall statistics for the region.

Genesis expansion and increased frequency within the MDR is indicated by the genesis data, but also by the regional warming of the ocean in this part of the North Atlantic during warm SST years. This is important for two reasons. First, this area has historically been a fertile source of major hurricanes, largely due to the longer period of intensification available from this source region. Second, when storms from the MDR become hurricanes and make landfall, they tend to do so along the Southeast U.S. coast. This is not to say that these storms never make their way into the Gulf, and in fact most storms from this region do not make landfall at all. Climatologically, when they do make landfall, they are more likely to land somewhere between Key West, Florida and Cape Hatteras, North Carolina than along any other part of the U.S. coastline. We conclude that there is an increased probability of storms making hurricane landfall along the Southeast coast from this genesis region. Because some of these storms can make their way into the Gulf, however, there is likely a marginal impact on Gulf landfall risk as well.

Because the results are built on climatology, they apply more to a large number of seasons than to single one. Since many scientists expect the current warm phase of North Atlantic SST to continue for many years to come, the results of this study provide a basis for estimating landfall risk conditioned on a warm SST climate. This discussion can also serve to motivate further study of other climate signals, such as ENSO, and their influence on landfall risk. Though this study has focused on risk to the U.S. mainland, its techniques and analyses can be used to study similar aspects of risk in other areas prone to tropical cyclones.

Acknowledgements The authors would like to thank colleagues who contributed to this study. We are particularly grateful to Ioana Dima and Jason Butke for their significant contributions to this work.

Was this article helpful?

0 0
Renewable Energy Eco Friendly

Renewable Energy Eco Friendly

Renewable energy is energy that is generated from sunlight, rain, tides, geothermal heat and wind. These sources are naturally and constantly replenished, which is why they are deemed as renewable.

Get My Free Ebook


Post a comment