## Abstract

To determine potential changes in the frequency and intensity of future storm events due to climate change in New York City (NYC), a statistical downscaling technique is proposed. First, a historical benchmark was determined using weather station data from the John F. Kennedy (JFK) and La Guardia (LGA) airports for the period 1973–2017. This historical information was used to perform the bias-correction exercise of near-future (2011–2050) global circulation model (GCM) output (ORNL RegCM4; RCP 8.5). Results show that NYC is projected to experience higher wind gusts under a warming climate for the period 2017–2050 in comparison with the historical data period, with the most extreme event projected to produce a maximum wind gust of approximately 110 mph, a significant increase over the past maximum of 80 mph. The historical 700-year return period event was estimated at 115 mph, while the overall 700-year event (historical and projected) is estimated at 124 mph. The most extreme cases of maximum daily wind gusts are projected to occur during the winter and early spring seasons. No increase in the number of projected tropical storms was observed, but the intensity of the storms is projected to be higher than during the historical period. These changes in extreme wind events could have serious implications for NYC in terms of urban planning, potential power outages, transportation disruptions, impacts on building structures, and public safety.

## 1 Introduction

It is widely expected that climate change, in general terms, will produce more intense hurricanes as a warmer ocean provides more energy for tropical storms to intensify and grow [14]. On the other hand, winter storms are not expected to diminish in intensity in a warming climate [5]. In the particular case of New York City (NYC), high-wind gusts are typically produced by a variety of windstorms, which can occur in different forms: straightline storms that blow in one direction (e.g., squall lines), thunderstorms, microbursts, tornadoes, and high-wind events often associated with other storms, such as hurricanes or Nor’easters. They vary in intensity, duration, and geographical extent. They typically have a few hours of lead time and can last for hours, or for up to several days if they result from a large-scale weather system [6]. Here, we focus on the potential changes in frequency and intensity of future storm events due to climate change in the near future, with special interest in tropical events and winter storms. These changes in extreme wind events could have serious implications for NYC in terms of urban planning; potentially needing to update building structural, electrical, and mechanical codes, power outages, transportation disruptions, and public safety. This work is specifically in part motivated by local New York City laws that relate to wind loads impacts on structures.

In the following sections we examine whether the future tropical storm frequency and intensity predicted for the Atlantic Hurricane season will impact in a significant way the city of New York. We also attempt to answer the similar question of impacts of a changing climate on winter storms, with the understanding that these two types of storms are driven by different mechanisms. Section 2 presents the general methodology used in this study to analyze the potential effects of future climate change on the frequency and intensity of extreme wind gust events in NYC. Section 3 presents the results and major findings of the study, with emphasis on separating the data into tropical/non-tropical events. Section 4 summarizes the results and main conclusions, and presents recommendations for future research.

## 2 Methodology and Data

Previous studies on the analysis of wind gusts for building codes in the continental United States of America have relied mostly on historical wind data and on some variation of the Monte Carlo technique, especially for the generation of the necessary statistics in regards to tropical storms (e.g., Refs. [79], among others). Since there is no existing data to perform similar studies on future wind gusts for NYC, here we present a downscaling methodology of global circulation model (GCM) output as an alternative to obtain the necessary data to generate wind distributions for selected climate change scenarios for NYC. Statistical downscaling, or bias-correction, has been used to analyze GCM output locally when high-resolution data is needed [1012]. This study begins with a comprehensive review and analysis of wind gust data with the objective of establishing the historical wind statistics and a benchmark of wind gusts for the city. This was performed using long-term records from ground weather stations located at John F. Kennedy (JFK) and La Guardia (LGA) airports, as archived by the National Oceanic and Atmospheric Administration’s National Climatic Data Center (NOAA-NCDC). The reason for choosing the stations located at the two airports is twofold, not only they provide the most complete and longest historical record (determined after conducting a comprehensive quality control of the data for all stations in the area of interest), but these type of locations are typically used when conducting wind gust studies for structural engineering applications.

Using the historical data as reference, a statistical downscaling for selected GCM ensembles and climate change scenarios for projected future winds was performed. The most comprehensive collection of GCM ensemble runs is contained and archived by the Coupled Model Inter-comparison Project—Community Climate System Model v.4 (CMIP-CCSM4). Because of the coarse horizontal resolution of these global models (∼1 deg, or ∼100 km), the Oak Ridge National Laboratory (ORNL) performed a regional dynamical downscaling of an ensemble of 11 models (including the CCSM4) at 18 km horizontal resolution for the continental US [13], hereafter referred to as the ORNL RegCM4. Since climate models do not produce wind gusts as output, the highest value of the four-times daily instantaneous wind speed simulated for each day was selected and converted to maximum daily wind gust following Cvitan [14] as
$Vg=kgV¯MAX$
(1)
where Vg is the expected maximum daily wind gust; VMAX the maximum 6-hourly wind speed for the day; and kg is the gust factor. kg is then defined as
$kg=1+2.28ln⁡(zzo)$
(2)
here z is the height above ground (10 m); zo is the surface roughness length (0.5 m).
The statistical downscaling consists on a bias-correction technique proposed by Hawkins et al. [11], which corrects the output of large-scale GCMs for location, horizontal resolution, and model uncertainties. The formulation is distribution based for variables following a normal distribution, we have adapted this to use the parameters of the Weibull distribution, namely, shape and scale, as follows:
$VgBC=aOREF+bOREFbVgREF(vg−aVgREF)$
(3)
where VgBC is the bias-corrected projected future daily wind gust; Vg original daily wind gust resulting from Eq. (1); aOREF and aVgREF the distribution scale for the reference period of the observations and projections, respectively; and bOREF and bVgREF the distribution shape for the reference period of the observations and projections, respectively.

In this study, the historical period spans the years 1973–2017, while the GCM output was obtained for the period 2011–2050, which means that the reference period consist of the overlapping years of 2011–2017. Figure 1 shows the complete statistical downscaling flowchart used in the study. Table 1 summarizes the different data sources used for the observations and projected future climate.

Fig. 1
Fig. 1
Close modal
Table 1

Inventory of data sources used in the study presented in this document

Historical observationsStn. IDSourceLatLonTimeframe
JFKNCDC40.63973.7621973–2017
La GuardiaNCDC40.77973.8801973–2017
EnsembleSourceModels usedTime frequencyHorizontal resolutionRCP
Modeling projectionsCMIP/CCSM4NCARVaries3 hr; 6 hr∼1 deg2.6/4.5/6.0/8.5
RegCM4aORNL116 hr18 km8.5
Historical observationsStn. IDSourceLatLonTimeframe
JFKNCDC40.63973.7621973–2017
La GuardiaNCDC40.77973.8801973–2017
EnsembleSourceModels usedTime frequencyHorizontal resolutionRCP
Modeling projectionsCMIP/CCSM4NCARVaries3 hr; 6 hr∼1 deg2.6/4.5/6.0/8.5
RegCM4aORNL116 hr18 km8.5
a

For the specific GCMs included in the dynamical downscaling performed with RegCM4, please see Ref. [13].

## 3 Results

Daily maximum 3-second wind gust data at 10 m above ground level from 1973 to 2017 recorded at the JFK and LGA weather stations were analyzed to determine their probability distribution and seasonal trends (Fig. 2). The occurrence histogram plot for the data shows a classic Weibull distribution, while the monthly average shows that the highest wind gusts occurred during the late hurricane season and winter, with some overlap during early spring. From these data, it is determined that the maximum-recorded wind gusts occurred during super storm Sandy at approximately 80 mph on October 29-30, 2012. The procedure continued by separating the historical Tropical and Non-Tropical events, and return periods for the three datasets (all data; tropical; non-tropical) were calculated following the Gringorten plotting position formula. A calculation of a 700-year event yielded a value of 115 mph for the locations analyzed.

Fig. 2
Fig. 2
Close modal

The statistical downscaling of GCM data was performed using the ORNL RegCM4 6-hourly output for Representative Concentration Pathway 8.5 (RCP 8.5) spanning the period 2011–2050, using the historical 2011–2017 period as reference. The RegCM4 data used here consists of an 11-model ensemble from various sources and horizontal/temporal resolutions, dynamically downscaled to 18 km horizontal resolution and archived four-times daily. The RCPs are family scenarios proposed by the Inter Governmental Panel of Climate Change to represent future climate conditions as possible green house gases (GHG) concentrations, and corresponding global radiation feedback. The RCP 8.5 represents a high-end GHG radiation feedback. Despite the relatively low temporal resolution, and due to the increased horizontal resolution of the modeling system, the RegCM4 bias-corrected output captured the maximum winds observed by the historical data during the reference period (Fig. 3), which validates the downscaling technique used. Results show increased extreme events on the right tail of the distribution when comparing the historical and projected future datasets (Fig. 4). The bias-corrected CCSM4 future projected wind gusts presented a maximum value of 110 ± 28 mph, an event that appeared at both locations in the early spring toward the middle of the projected timeframe. Other extreme events occurred during the late hurricane season and winter, with a considerable dip during the summer months (Fig. 5). It is worth noting that the CCSM4 is a middle road global model in terms of temperature and precipitation [13].

Fig. 3
Fig. 3
Close modal
Fig. 4
Fig. 4
Close modal
Fig. 5
Fig. 5
Close modal

In order to separate the future data into tropical and non-tropical, the timing of the occurrence of the events was used, as there is no other information regarding the nature of these events. Events occurring during the hurricane season with a value of or above 38 mph were allocated in the Tropical dataset, events above the same threshold occurring during winter and spring were considered Extra-tropical or Non-tropical. Return periods for the projected future wind gusts were calculated using the same Gringorten plotting position formula as the historical data. The return period comparison between the historical and projected data for the tropical and non-tropical events clearly shows that higher wind gust magnitudes should be expected in the future (Fig. 6). It should be noted that these events are also expected to occur less frequent, as evidenced by the slightly shorter return periods. Return periods for the complete period of study, spanning the years 1973–2050, were obtained and a calculation of a 700-year event yielded a value of 124 mph (Fig. 7). Uncertainties for this study (±28 mph) were estimated by applying the bias-correction procedure to all 11 models downscaled at ORNL and calculating the mean, standard deviation, and spread for each maximum daily wind gust data point. Due to the different types of data used and techniques included in the downscaling exercise, other uncertainties are difficult to estimate for extreme winds.

Fig. 6
Fig. 6
Close modal
Fig. 7
Fig. 7
Close modal

## 4 Conclusions and Recommendations

After performing a historical analysis of wind gusts for NYC, and future projections using bias-corrected (statistically downscaled) GCM output for the same variable, the following conclusions are drawn:

• New York City is projected to experience higher wind gusts under a warming climate for the period 2017–2050 in comparison with the historical data period of 1973–2017. The future maximum wind gusts are expected to reach 110 mph, a significant increase from the recent maximum wind of 80 mph.

• The historical benchmark (1973–2017) estimated 700-year return period is 115 mph, while the future is estimated as 124 mph. This result agrees quite well with the recent analysis by Kossin et al. [15] which projects an 8% per decade of years of wind speed increase in the north Atlantic. This means that since 1979 when their data analysis started till now, 40 years after, we should expect 32% increase which is very close to what has been predicted in this work.

• The most extreme cases of maximum daily wind gusts are projected to occur during the winter seasons.

• There is no significant increase in the number of projected tropical storms for the area of study with maximum wind gusts >38 mph; however, the intensity of the storms is expected to be higher than during the historical period.

Previous studies of wind gust data resulting from tropical storms for building codes in the United States have relied on some variation of Monte Carlo simulations to account for the relatively low number of events needed to perform the statistical tests. In the study presented here such methodology was not followed because there is no information regarding the nature of the events for the projected future climate. Therefore, it is recommended that in order to determine the characteristics of each storm, further analysis is needed, and a high-resolution dynamical downscaling simulation of several specific events be performed. This would not only inform a potential Monte Carlo simulation but will also provide the horizontal distribution of winds across the city for these events. This will be helpful in determining how the city affects each event and explaining why JFK is projected to experience much higher winds and higher frequency of events, at both tails of the distribution, than La Guardia. Past modeling efforts of specific historical events have provided information about high-wind hot spots in the city that might not be at or near the airports, which are the locations analyzed in the study presented here.

## Acknowledgment

This study was sponsored and financially supported by the New York City Department of Buildings (NYC DoB).

## Conflict of Interest

There are no conflicts of interest.

## Data Availability Statement

The authors attest that all data for this study are included in the paper. Data provided by a third party listed in Acknowledgment.

## References

1.
Vecchi
,
G. A.
, and
Soden
,
B. J.
,
2007
, “
Increased Tropical Atlantic Wind Shear in Model Projections of Global Warming
,”
Geo. Res. Lett.
,
34
(
L08702
), pp.
1
5
. 10.1029/2006GL028905
2.
Bender
,
M. A.
,
Knutson
,
T. R.
,
Tuleya
,
R. E.
,
Sirutis
,
J. J.
,
Vecchi
,
G. A.
,
Garner
,
S. T.
, and
Held
,
I. M.
,
2010
, “
Modeled Impact of Anthropogenic Warming on the Frequency of Intense Atlantic Hurricanes
,”
Science
,
327
(
5964
), pp.
454
458
. 10.1126/science.1180568
3.
Knutson
,
T. R.
,
Sirutis
,
J. J.
,
Vecchi
,
G. A.
,
Garner
,
S.
,
Zhao
,
M.
,
Kim
,
H.
,
Bender
,
M.
,
Tuleya
,
R. E.
,
Held
,
I. M.
, and
Villarini
,
G.
,
2013
, “
Dynamic Downscaling Projections of Late 21st Century Atlantic Hurricane Activity: CMIP3 and CMIP5 Model-Based Scenarios
,”
J. Climate
,
26
(
17
), pp.
6591
6617
. 10.1175/JCLI-D-12-00539.1
4.
Knutson
,
T. R.
,
Sirutis
,
J. J.
, and
Zhao
,
M.
,
2015
, “
Global Projections of Intense Cyclone Activity for Late Twenty-First Century From Dynamical Downscaling of CMIP5/RCP4.5 Scenarios
,”
J. Climate
,
28
(
18
), pp.
7203
7224
. 10.1175/JCLI-D-15-0129.1
5.
Zarzycki
,
C. M.
,
2018
, “
Projecting Changes in Societally Impactful Northeastern U.S. Snowstorms
,”
Geo. Res. Lett.
,
45
(
21
), pp.
12067
12075
. 10.1029/2018GL079820
6.
02 Coastal Storms
,
2019
, “Chapter 3: Selected Hazard Profiles,”
NYC’s Risk Landscape: A Guide to Hazard Mitigation
,
S.
Bednarczyk
, ed.,
NYC
,
New York
.
7.
Vickery
,
P. J.
,
,
D.
,
Powell
,
M. D.
, and
Chen
,
Y.
,
2009
, “
A Hurricane Boundary Layer and Wind Field Model for Use in Engineering Applications
,”
J. Appl. Meteorol.
,
48
(
2
), pp.
381
405
. 10.1175/2008JAMC1841.1
8.
Vickery
,
P. J.
,
,
D.
,
Twisdale
,
L. A.
, Jr
, and
Lavelle
,
F. M.
,
2009
, “
United States Hurricane Wind Speed Risk and Uncertainty
,”
J. Struct. Eng.
,
135
(
3
), pp.
301
320
. 10.1061/(ASCE)0733-9445(2009)135:3(301)
9.
Vickery
,
P. J.
,
,
D.
,
Galsworthy
,
J.
,
Peterka
,
J. A.
,
Irwin
,
P. A.
, and
Griffis
,
L. A.
,
2010
, “
Ultimate Wind Load Design Gust Wind Speeds in the United States for Use in ASCE-7
,”
J. Struct. Eng.
,
136
(
5
), pp.
613
625
. 10.1061/_ASCE_ST.1943-541X.0000145
10.
Piani
,
C.
,
Weedon
,
G. P.
,
Best
,
M.
,
Gomes
,
S. M.
,
Viterbo
,
P.
,
Hagemann
,
S.
, and
Haerter
,
J. O.
,
2010
, “
Statistical Bias Correction of Global Simulated Daily Precipitation and Temperature for the Application of Hydrological Model
,”
J. Hydrology
,
395
(
3–4
), pp.
199
215
. 10.1016/j.jhydrol.2010.10.024
11.
Hawkins
,
E.
,
Osborne
,
T. M.
,
Ho
,
C. K.
, and
Challinor
,
A. J.
,
2013
, “
Calibration and Bias Correction of Climate Projections for Crop Modeling: An Idealized Case Study Over Europe
,”
Agric. For. Meteorol.
,
170
, pp.
19
31
. 10.1016/j.agrformet.2012.04.007
12.
Chen
,
J.
,
Brissette
,
F. P.
,
Chaumont
,
D.
, and
Braun
,
M.
,
2013
, “
Finding Appropriate Bias Correction Methods in Downscaling Precipitation for Hydrologic Impact Studies Over North America
,”
Water Resources Res.
,
49
(
7
), pp.
4187
4205
. 10.1002/wrcr.20331
13.
Ashfaq
,
M.
,
Rastogi
,
D.
,
Mei
,
R.
,
Kao
,
S.-C.
,
,
S.
,
Naz
,
B. S.
, and
Touma
,
D.
,
2016
, “
High-resolution Ensemble Projections of Near-Term Regional Climate Over the Continental United States
,”
J. Geophys. Res. Atmos.
,
121
(
17
), pp.
9943
9963
. 10.1002/2016JD025285
14.
Cvitan
,
L.
,
2003
, “
Determining Wind Gusts Using Mean Hourly Wind Speed
,”
Geofizika
,
20
, pp.
63
74
.
15.
Kossin
,
J. P.
,
Knapp
,
K. R.
,
Olanderc
,
T. L.
, and
Velden
,
C. S.
,
2020
, “
Global Increase in Major Tropical Cyclone Exceedance Probability Over the Past Four Decades
,”
PNAS
,
117
(
22
), pp.
11975
11980
. 10.1073/pnas.1920849117