www.nature.

com/scientificreports

OPEN Atlantic hurricane activity during

the last millennium
Michael J. Burn1 & Suzanne E. Palmer2

received: 13 February 2015

Hurricanes are a persistent socio-economic hazard for countries situated in and around the
accepted: 10 July 2015 Main Development Region (MDR) of Atlantic tropical cyclones. Climate-model simulations have
Published: 05 August 2015 attributed their interdecadal variability to changes in solar and volcanic activity, Saharan dust flux,
anthropogenic greenhouse gas and aerosol emissions and heat transport within the global ocean
conveyor belt. However, the attribution of hurricane activity to specific forcing factors is hampered
by the short observational record of Atlantic storms. Here, we present the Extended Hurricane
Activity (EHA) index, the first empirical reconstruction of Atlantic tropical cyclone activity for the
last millennium, derived from a high-resolution lake sediment geochemical record from Jamaica.
The EHA correlates significantly with decadal changes in tropical Atlantic sea surface temperatures
(SSTs; r = 0.68; 1854–2008), the Accumulated Cyclone Energy index (ACE; r = 0.90; 1851–2010), and
two annually-resolved coral-based SST reconstructions (1773–2008) from within the MDR. Our results
corroborate evidence for the increasing trend of hurricane activity during the Industrial Era; however,
we show that contemporary activity has not exceeded the range of natural climate variability
exhibited during the last millennium.

The extent to which the climate dynamics of the Main Development Region (MDR) of Atlantic tropical
cyclone activity are controlled by natural or anthropogenic climatic forcing factors remains unclear1,2.
This uncertainty has arisen because of the reliance on historical meteorological records, which are too
short to capture the natural long-term variability of climatic phenomena as well as a lack of understand-
ing of the physical link between tropical Atlantic SSTs and tropical cyclone variability3,4. Consequently,
the fifth assessment report of the Intergovernmental Panel on Climate Change (IPCC) concluded that
there is low confidence in region-specific projections of tropical cyclone activity and that it remains
uncertain whether recent changes in Atlantic tropical cyclone activity lie outside the range of natural
variability2. While much progress has been made in climate-model projections of global tropical cyclone
activity1, such uncertainty at the basin-wide scale makes it difficult for countries, particularly Small
Island Developing States, to plan for the significant socio-economic consequences that are likely to arise
due to 21st Century global warming.
Atlantic tropical cyclones occur within the Atlantic warm pool, a region of warm water (> 28.5 °C)
that encompasses the Caribbean, the Gulf of Mexico, and the western tropical North Atlantic (Fig.  1)
and varies spatially on interannual and interdecadal timescales5. Their activity is controlled by thermo-
dynamic factors that govern heat exchange between the surface ocean and atmosphere, and dynamical
factors that control vertical wind shear, a measure of the difference in average wind speed between the
upper and lower layers of the atmosphere5–8. In general, a large (small) warm pool is associated with
higher (lower) SSTs, tropospheric instability and precipitation, and lower (higher) vertical wind shear5
resulting in enhanced (subdued) tropical cyclone activity.
On longer timescales, Atlantic tropical cyclone activity is influenced primarily by the El Niño Southern
Oscillation (ENSO) and the Atlantic Multidecadal Oscillation (AMO)5,6,9–11, which combine to modu-
late SSTs and vertical wind shear in the MDR. ENSO affects Atlantic climate dynamics on interannual

1
Department of Geography and Geology, The University of the West Indies, Mona Campus, Kingston 7, Jamaica,
W.I. 2Centre for Marine Sciences, The University of the West Indies, Mona Campus, Kingston 7, Jamaica, W.I.
Correspondence and requests for materials should be addressed to M.J.B. (email: [email protected].
jm)

Scientific Reports | 5:12838 | DOI: 10.1038/srep12838 1

www.nature.com/scientificreports/

40

Latitude (°N)
30

20 2
1 4 5 MDR Atlantic hurricanes
3
10

0
100 90 80 70 60 50 40 30 20 10 0
Longitude (°W)

Figure 1.  SST map of the tropical North Atlantic and Caribbean Regions. Map showing location of sites
discussed in the text situated within the Main Developing Region (MDR; 10°–20°N; 15°–85°W) of Atlantic
tropical cyclone activity. 1. Grape Tree Pond, Jamaica27; 2. Yucatan Peninsula, Mexico30; 3. Los Roques,
Venezuela31; 4. SW Puerto Rico32 5. MDR. Map created by Suzanne Palmer using Adobe Illustrator CC
2014.1.0 Release.

(2–8yr) timescales and its amplitude has been shown to be modulated on interdecadal (50–90yr) times-
cales during the last millennium12. It influences the tropical Atlantic climate by controlling variations
in sea level pressure across the equatorial Pacific, which modulate the strength of the Pacific Walker
circulation13 and upper-level westerly wind flow into the Atlantic MDR14. The corresponding changes in
vertical wind shear in the tropical Atlantic parallel changes in the strength of the Atlantic NE trade winds
caused by fluctuations in the surface pressure gradient between the North Atlantic and equatorial Pacific.
Thus in the MDR, El Niño (La Niña) events are typically associated with stronger (weaker) vertical wind
shear, which suppresses (enhances) rainfall and tropical cyclone formation11,15. The AMO is a basin-wide
index of Atlantic SST anomalies16, which varies periodically on interdecadal timescales (60–80 years)
and controls the thermodynamic environment that underpins hurricane development6. Its influence on
hurricane activity is manifested in the interdecadal behaviour of Atlantic Accumulated Cyclone Energy
(ACE; 1851–2013)5,6, a composite measure of the frequency, intensity and duration of tropical cyclones17.
While contemporary tropical cyclone dynamics are relatively well understood, few observational
records of their long-term natural variability extend back beyond the instrumental period. Existing
historical documentary accounts rely heavily on colonial archives and are restricted to relatively nar-
row windows, spatially and temporally18–21. Similarly, of the few existing geological records of washover
deposits within the MDR22–25, some have been shown to under-represent hurricane activity26 and their
interpretation can be complicated by difficulties in distinguishing between storm and tsunami-induced
events.
To avoid these shortcomings, we adopt a novel approach to the long-term reconstruction of Atlantic
hurricane activity. We calibrate a high-resolution time series of lake level change from a previously pub-
lished sediment geochemical record from Jamaica27 against the instrumental ACE index for the geo-
graphical area defined by the grid cell 15–25°N; 70–80°W for the period 1851–1969. While it has been
established that the basin-wide ACE index underestimates tropical cyclone activity during the earlier
parts of the record28, the passage of storms through the selected grid cell is well-sampled because the
area encompasses Cuba, Jamaica, Hispaniola, the Bahamas, and the Cayman Islands, for which detailed
meteorological and archival records are available back to the 19th Century (C. Landsea, pers. comm.).
Thus, we argue the selected grid cell is both representative of activity in the region and relatively com-
plete back to 1851. Given the established connection between precipitation and tropical cyclone activity
in the MDR29 and the strong and statistically significant correlation between local rainfall, the sediment
geochemical record and the ACE index on multiple timescales (see below), we demonstrate that the
Jamaican time series is a robust proxy for hurricane activity and extend it back to the beginning of the
last millennium. In doing so, we address the following research questions: (1) How does tropical cyclone
activity change under contrasting mean climatic states of the northern hemisphere? (2) Are the interan-
nual and interdecadal patterns of contemporary tropical cyclone activity persistent and stationary during
the last millennium? (3) What does the extended ACE index reveal about the relative contributions of
natural forcing factors to Atlantic tropical cyclone activity? (4) Do the observed trends in Atlantic trop-
ical cyclone activity during the Industrial Era exceed their longer-term natural variability?
Our reconstruction of hurricane activity is derived from the third Principal Component (PC3) of
mm-scale ITRAX μ -XRF core scans of a ~1200-year long sediment record of lake-level change from
Grape Tree Pond in southern Jamaica27 (Figs 1,2a,3d and 4c). The pond is a freshwater-fed and anoxic
mangrove lagoon, which is sensitive to precipitation variability on multiple timescales. High rates of sed-
iment accumulation (~2.5 mm−yr) combined with low levels of bioturbation result in sedimentation rates
sufficient to capture geochemical changes at an annual resolution. We improve the original 14C-based age

Scientific Reports | 5:12838 | DOI: 10.1038/srep12838 2

www.nature.com/scientificreports/

a. Jamaica, West Indies

4
c.
3
r = 0.64
p < 0.001

Standarsised Units
2
50 km
1 780 f.
Grape Tree Pond (17°53'N, 76°37'W) 1933
0 720
and Easington weather station Easington, Jamaica
(17°55'N, 76°35'W)
b. 1881-2010 -1 660
270 Record Highs:
Jamaica, West Indies Annual Rainfall = 3148 mm
Average Rainfall (mm)

240 1900 1910 1920 1930 1940 1950 1960 600 Basin-wide ACE = 259 10 kt
210 Year (CE)
180 540
Rainfall at Easington
150

Average Rainfall (mm)

(June-Nov;1893-1966)
120 Frequency of tropical cyclones 480
90 (15-25°N; 70-80°W; 1893-1966)
60 420
30
Hurricane Season Dry Season 360
J F M A M J J A S O N D

300 300 300

d. e. 1881-1966
270 270
1914 Easington, Jamaica
Average Rainfall (mm)
Average Rainfall (mm)

240 240 240

Easington, Jamaica (17°55'N, 76°35'W)
210 210
(17°55'N, 76°35'W)
180 180 180
Record Lows:
150 150
Annual Rainfall = 561 mm
120 Basin-wide ACE = 0 120 120
90 90
60 60 60
30 30

J F M A M J J A S O N D J F M A M J J A S O N D J F M A M J J A S O N D

Figure 2.  Precipitation data for Jamaica. (a) Map showing location of Grape Tree Pond in Jamaica and
the location of Easington meteorological station in the Parish of St. Thomas. (b) Average rainfall (mm)
for Jamaica (1881–2010). (c) Relationship between total summer rainfall (June-Nov) at Easington weather
station and the frequency of known storms passing through the geographic area defined by the grid box
15–25°N, 70-80°W for the period 1893–1966. (d) Record low measurements of total rainfall at Easington
and basin-wide hurricane activity during 1914. (e) Average rainfall measured at Easington for the period
1881–1966. (f) Record high measurements in total rainfall at Easington and basin-wide hurricane activity
during 1933. Map created by Suzanne Palmer using Adobe Illustrator CC 2014.1.0 Release. Precipitation
data provided by the Jamaican Meteorological Service.

model for the site by anchoring the PC3 record to the annually resolved coral-based SST reconstruction
from the eastern Yucatan Peninsula30 for the period 1775–2009 (Fig. 1; see methods). Of the first three
principal components of the geochemical data, PC1 explains 47.6% of the variance within the dataset
and represents a gradient of sediment deposition from organic sediments to a combination of inorganic
authigenic and detrital elements, the latter interpreted to reflect the aerial deposition of Saharan dust27.
PC2 explains 11.9% of the variance and represents a salinity gradient. PC3 explains 8.9% of the variance
and represents a gradient of redox conditions at the sediment-water interface of the lagoon and exhibits
a positive relationship with lake level change27.
Historical rainfall measurements (1881–1966) from Easington, a town located 4.6 km to the north-
east of Grape Tree Pond (Fig. 2a), demonstrate that more than 75% of average annual precipitation falls
at the study site during the hurricane season (June-November; Fig.  2e) compared with 50% of average
annual precipitation for Jamaica (Fig. 2b). Very little precipitation occurs outside of the hurricane season
because of the influence of the rain shadow of the Blue Mountains. Comparison between rainfall meas-
urements at Easington and the frequency of tropical cyclones passing through the grid cell 15–25°N;
70–80°W for the period 1893-1966, demonstrates a strong positive relationship (r =  0.64; p <  0.001;
Fig. 2c) between the passage of storms and local rainfall. Further, qualitative comparisons between local
rainfall and the basin wide ACE index show that the lowest total annual rainfall for Easington (561 mm)
occurred in 1914, which coincided with the least active hurricane season on record as represented by a
basin-wide ACE value of zero (Fig. 2d). In contrast, the record high for Easington (3148 mm) occurred in

Scientific Reports | 5:12838 | DOI: 10.1038/srep12838 3

www.nature.com/scientificreports/

Age (CE)
1750 1775 1800 1825 1850 1875 1900 1925 1950 1975 2000
0.4

Increasing 0.2
hurricane a. MDR precipitation Warmer

(mm d1)
NOAA-CIRES 20CR 0
activity SSTs
-0.2

-0.4
160

120
b.

104 kt2
Atlantic
ACE index
80

40 28.4

28.0

c.

SST (ºC)
ERSST v.3b
Hurricane MDR 27.6
SST (July-Nov)
27.2
3
Wetter
2 26.8

1
PC3

-1 d. Lake level change

Grape Tree Pond -1.6
e.

Coral growth anomaly (mm)

-2 Jamaica SST NE Yucatan
-1.2
Drier Peninsula
-3
Mexico -0.8
-0.4
0.0
0.4
0.8
-4.3 1.2
Coral δ18O (‰ VPDB)

-4.2

-4.1

-4.0 f. SST Los Roques

N Venezuela
-3.9

-3.8 -4.6

Coral δ18O (‰ VPDB)

-4.4

Decreasing -4.2 Cooler

hurricane SSTs
-4.0
activity g. SST SW Puerto Rico
-3.8

1750 1775 1800 1825 1850 1875 1900 1925 1950 1975 2000
Age (CE)

Figure 3.  Comparison of environmental parameters and proxy-based reconstructions of MDR climatic
phenomena during the modern historical period. (a) MDR precipitation for the interval 1871–2012
(modified from Dunstone29). Reprinted by permission from Macmillan Publishers Ltd: Nature Geoscience
6, 534-539, copyright 2013. (b) Revised Accumulated Cyclone Energy (ACE) index for the MDR from the
National Hurricane Center’s Hurricane Database (HURDATv2). (c) MDR SSTs (July-November) from the
NOAA NCDC extended reconstructed global sea surface temperature data (ERSSTv.3b)49. (d) PC3 sequence
of lake-level change at Grape Tree Pond, Jamaica27. (e–g) Coral-based SST reconstructions from the Yucatan
Peninsula, Mexico30, Los Roques, Venezuela31 and SW Puerto Rico32. All time-series were filtered using an
11-point running mean except (b) which was filtered using a 21-point running mean.

Scientific Reports | 5:12838 | DOI: 10.1038/srep12838 4

www.nature.com/scientificreports/

Age (CE)
1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000

Volcanic forcing
-1.6
-4.8
a.

(W/m2)
-8.0
0.64
-11.2 Oort Wolf Spörer Maunder Dalton
0.32
b.

Solar forcing
(W/m-2)
0.00

-0.32
-0.64
3
100
2

(15-25ºN; 70-80ºW)
c. 90
Grape Tree Pond

1
80

EHA Index
0

(102 kt2)
70
-1
PC3

60
-2 50
-3 40
-4 30
-5
d. 13

Annual TC counts
12
11
10
9
8
e. 7

4
7.8 8

Period (years)
Signal Power

16
4.4
32
1.1 64

-2.3 128
256

1.5
0.75
f. 1.0
0.50

AMO (54 - 80 yr)

Bandpass filter
Bandpass filter

0.25 0.5
ENSO (5-8yr)

0.00 0.0
-0.25
-0.5
-0.50
-0.75 -1.0
-1.00 -1.5

MCA LIA
1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000
Age (CE)

Figure 4.  Time-series analyses of the Extended Hurricane Activity (EHA) index and comparison with
natural external forcing (a) Changes in volcanic forcing52 (black line, upper panel) and (b) Total Solar
Irradiance (TSI)50,51 (yellow and black). (c) PC3 sequence of lake-level change at Grape Tree Pond, Jamaica27
(grey) and the Extended Hurricane Activity (EHA) index (blue) calculated for the geographic region defined
by the grid box 15–25°N, 70–80°W. The black line represents the section of the PC3 series anchored to
the annually-resolved coral-based SST reconstruction from the Yucatan Peninsula30 (d) Annual basin-wide
tropical cyclone counts from a coupled ocean-atmosphere climate model simulation35 (light blue), and
from statistical model reconstructions driven by estimates of landfalling hurricane strikes from overwash
sediment records36 (pink). (e) Morlet wavelet analysis applied to the PC3 sequence from Grape Tree Pond,
Jamaica. Colours represent the signal strength, which is the squared correlation strength with the scaled
morlet wavelet. The 95% confidence level is plotted as a contour and the cone of influence identifies the
region of the wavelet spectrum in which boundary effects are present. (f) Bandpass filters applied to the PC3
sequence centred on interannual (5–8 years; grey) and interdecadal timescales (54–80 years; green). Intervals
associated with the five grand solar minima are shaded in grey and those of the Medieval Climate Anomaly
(MCA) and Little Ice Age (LIA) are represented in open boxes with dashed lines extended to the top of the
figure.

Scientific Reports | 5:12838 | DOI: 10.1038/srep12838 5

www.nature.com/scientificreports/

Environmental parameters Period (yrs) Pearson’s r r2 p-value Edf

ACE index (15–25°N;
1851–1969 0.634 0.4 < 0.1 5
70–80°W)†
Basinwide ACE index
1851–2010 0.591 0.35 < 0.05 10
(10–20°N, 15–85°W)*
Basinwide ACE index
1851–2010 0.706 0.50 < 0.1 5
(10–20°N, 15–85°W)†
Basinwide ACE index
1851–2010 0.816 0.67 < 0.1 4
(10–20°N, 15–85°W)‡
Basinwide ACE index
1851–2010 0.902 0.81 < 0.05 3
(10–20°N, 15–85°W)^
ERSSTAv.3b MDR (Jun–Nov;
1854–2008 0.615 0.38 < 0.1 5
10–20°N, 15–85°W)*
ERSSTAv.3b MDR (Jun–Nov;
1854–2008 0.677 0.46 < 0.15 4
10–20°N, 15–85°W)†
Coral-based SSTs (Los Roques,
1918–2003 − 0.836 0.7 < 0.005 8
Venezuela)*
Coral-based SSTs (Los Roques,
1918–2003 − 0.956 0.91 < 0.001 6
Venezuela)†
Coral growth-rate anomalies
1773–2008 − 0.597 0.36 < 0.001 26
(Yucatan Peninsula)
Coral growth-rate anomalies
1850–2008 − 0.796 0.63 < 0.1 4
(Yucatan Peninsula)*
Coral growth-rate anomalies
1850–2008 − 0.832 0.69 < 0.15 3
(Yucatan Peninsula)†

Table 1.  Correlation between the Jamaican PC3 series and the following environmental metrics of the
Main Development Region (MDR) on inter-annual to decadal timescales (time-series were smoothed
using 5-*, 11-†, 21-‡, and 41^ year running means, where indicated). Edf: Effective degrees of Freedom

1933, which coincided with the most active hurricane season during the 20th Century (ACE ~ 259 104 kt2;
Fig. 2c,f). Consequently, lake level change at Grape Tree Pond occurs in response to rainfall variability
associated with the passage of storms during the hurricane season and we therefore interpret the PC3
series to reflect changes in tropical cyclone-induced precipitation.
The Jamaican PC3 series correlates significantly with different tropical cyclone-related environmen-
tal parameters across the MDR for the modern historical period on decadal and interdecadal times-
cales (Fig. 3; Table 1). Strong and statistically significant relationships exist between PC3 and the locally
derived ACE index (15–25°N; 70–80°W) on decadal timescales (1851–1969; r =  0.63, r2 =  0.4, p <  0.1)
and basin-wide on high- (r =  0.71, r2 =  0.5, p <  0.1) and low frequency (r =  0.9, r2 =  0.81, p <  0.05)
timescales from 1851–2010. It correlates significantly with two independent coral-based SST reconstruc-
tions from the Yucatan Peninsula30 and Venezuela31 (Fig. 3e,f; Table 1), which extend back to 1773 and
1918, respectively. Further qualitative similarities are observed between the low-frequency trends of the
PC3 series and those of a SST reconstruction from Puerto Rico (Fig.  3g), which dates back to 175032.
Moreover, PC3 captures well-known historical climatological events including the interval of intense
hurricane activity during the 1870 s and 1880 s and periods of quiescence during the 1850 s and 60 s and
at the turn of the 20th Century33,34. The apparent centennial-scale increase, and interdecadal variability
of hurricane activity spanning the 20th and early 21st Centuries is supported by similar variability in
Caribbean-wide SST (Fig. 3) and the PC3 time series. However, caution should be taken in the interpre-
tation of trends exhibited by the basin wide ACE index, which becomes increasingly inaccurate prior to
the advent of aircraft reconnaissance in 194428.
Given the statistically significant relationships between hurricane activity (ACE), rainfall at Easington
and the PC3 timeseries on different spatial and temporal scales, we calculate the Extended Hurricane
Activity index (EHA) by regressing PC3 against the ACE index (15–25°N; 70–80°W) for the period
1851–1969 using a linear least squares model:

EHA = 11.596 (± 1.3091) × [PC 3] + 71.054 (± 1.7651) ( 1)

The EHA index (Fig. 4c) spans three distinct climatological intervals of the last millennium; the rela-
tively warm mean climatic state of the Medieval Climate Anomaly (MCA; ~900–1350 CE), the relatively
cool mean state of the Little Ice Age (LIA; ~1450–1850 CE), and the period of centennial warming, which
distinguishes the Industrial Era since 1870. In general, average Caribbean-wide storm activity during
the MCA was lower (EHA ~ 65.1 ×  102 kt2) than that of the LIA (~73.3 ×  102 kt2) and the 20th Century
average (EHA ~ 74.7 ×  102 kt2; ACE ~ 67.3 ×  102 kt2). However, hurricane activity during the LIA is char-
acterised by a greater amplitude of variation where increased frequency of short intervals of enhanced

Scientific Reports | 5:12838 | DOI: 10.1038/srep12838 6

www.nature.com/scientificreports/

and suppressed activity correspond to periods of low natural radiative forcing. The highest average levels
of activity (EHA ~ 86 ×  102 kt2; 1580–1650 CE) occurred during the late 16th and early 17th Centuries,
a trend corroborated by a coupled ocean-atmosphere climate model simulation of annual basin-wide
tropical cyclone counts35, which indicates heightened activity during the LIA (Fig. 4d). Further support
for enhanced activity during the LIA comes from a record of hurricane deposits in a coastal karst basin
in the Bahamas, which suggests heightened activity occurred between 1350 and 165025. This impor-
tant result suggests that average hurricane activity during the industrial period has not exceeded its
longer-term natural variability during the last millennium. While the general increase in activity revealed
by the EHA index is not consistent with the longer-term decreasing trend of hurricane counts simulated
by the statistical model reconstruction driven by estimates of landfalling hurricane strikes from over-
wash sediment records (Fig  4d), the strong interdecadal variability exhibited by the EHA index is well
represented in both36. This similarity is particularly striking during the modern historical period (1800
– present), where the records are characterised by decadal change superimposed on the longer term
centennial increase in activity.
Comparison between the EHA index and indices of natural radiative forcing suggest changing solar
irradiance has a significant influence on tropical cyclone activity. Spectral analysis of the previously
published PC3 time series revealed periodicities consistent with those of changing solar irradiance (387,
208, 86, 23 years) combined with those of interdecadal variability (57 and 68 years)27 possibly associated
with the AMO or the modulation of ENSO amplitudes on interdecadal timescales. A morlet wavelet
analysis applied to PC3 (Fig.  4e) reveals that the interdecadal variability of hurricane activity (centred
on a ~60-year periodicity) similarly exhibits strong and statistically significant power during intervals of
low natural radiative forcing since the beginning of the LIA. Bandpass filters applied to PC3 to isolate
interannual (5–8 year) and interdecadal (54–80 year) timescales, support this interpretation and suggest
further that the variability is generally greater during radiative forcing minima on both high and low fre-
quency timescales (Fig. 4f), a phenomenon that is particularly conspicuous during the so-called Spörer
and Maunder grand solar minima. Taken together, our analyses suggest that interdecadal hurricane activ-
ity was subdued during the warm mean climatic state of the MCA. It subsequently became more pro-
nounced during the LIA and, after a brief transitional state characterised by enhanced variability between
1850 and 1900, then returned to a more subdued state during the late industrial period (1950–2010), an
episode of distinct climatic warming probably triggered by a combination of anthropogenic activity and
a positive phase of the AMO37. Remarkably, the interdecadal behaviour of hurricane activity does not
appear to persist as strongly during the warmer mean climatic states of the last millennium and therefore
exhibits non-stationarity with respect to its response to changing mean climate states. While the chang-
ing magnitude of interannual variability is not significant at the 95% level it is nevertheless also clearly
subdued during the early and latter parts of the record and more active during the LIA (Fig.  4f). The
similarity in the response of interannual and interdecadal changes suggests that both ENSO and AMO
may share a common response to deviations in external radiative forcing and raises the possibility that
the variability of storms may become more subdued during the anthropogenically-forced warmer mean
climate of the 21st Century.
We propose that external forcing, whether solar, volcanic and/or anthropogenic, exerts a significant
influence on Atlantic hurricane activity by its modulation of both ENSO and AMO. Thus our results
support the hypothesis that changes in radiative forcing may influence ENSO dynamics by means of
the ‘ocean dynamical thermostat hypothesis38–40’. Radiative cooling (warming) of the tropics reduces
(increases) the zonal sea surface temperature gradient in the equatorial Pacific due to greater cooling
(warming) of the western compared with the eastern equatorial Pacific39,41. The latter region is buffered
from further cooling (warming) by strong coastal upwelling. The Walker circulation weakens (strength-
ens) as a consequence and an El Niño-like (La Niña-like) state develops in the tropical Pacific increasing
(decreasing) upper-level westerly flow into the MDR of Atlantic tropical cyclones and in turn suppressing
(promoting) rainfall and hurricane activity. Further observational support for the thermostat hypothesis
comes from the Pacific-based Palmyra coral record42 which suggests not only that the Pacific SST gradi-
ent decreased during El Niño-like periods of the LIA, but also that enhanced ENSO activity dominated
the mid-seventeenth century, consistent with our record of greater variability of hurricane activity in
the tropical Atlantic. The trend to frequent and more intense El Niño-like conditions during the 17th
Century is further supported by lake-sediment evidence of drought conditions in the Yucatan peninsula
of Mexico, the Dominican Republic and the Cariaco Basin, Venezuela43–45, during which subdued hur-
ricane activity was likely.
Our results are also consistent with the external forcing of the AMO, which has been linked to trop-
ical cyclone activity5,46,47 during the modern historical period. Recent climate model simulations45,46 link
long-term AMO dynamics to changes in radiative forcing associated with the combined influence of
solar and volcanic activity, possibly propagated through the Atlantic Meridional Overturning Circulation
(AMOC) to influence SSTs and tropical cyclones in the North Atlantic. While further detection and
attribution studies are required to establish the relative roles of external and internal forcing on tropical
cyclone dynamics, the newly developed EHA index provides significant insights into the historical con-
text within which 21st Century hurricane activity may be understood.

Scientific Reports | 5:12838 | DOI: 10.1038/srep12838 7

www.nature.com/scientificreports/

0.5
0.4 PC3 leads by 2 years r = 0.51
0.3
0.2
0.1

Correlation
0.0
-0.1
-0.2
-0.3
-0.4
-0.5

-100 -75 -50 -25 0 25 50 75 100

Lag

Figure 5.  Cross-correlation between the annually-resolved coral growth rate record from the Yucatan
Peninsula30 and Jamaican PC3 timeseries27 for the period 1773–2008.

Methods
Environmental data and filtering.  Average total annual precipitation data for Jamaica (1881-2010;
Fig.  2) were provided by the Jamaican Meteorological Service and are based on an average of rain
gauges collected from weather stations across the different Parishes of Jamaica. Some spatial bias is to
be expected in these data particularly in the earlier part of the record (~1881–1900) when there were
fewer weather stations. Average annual precipitation data for Easington, St. Thomas (1881–1966) were
made available by NOAA Central library Data Imaging Project (CLDIP; http://docs.lib.noaa.gov/rescue/
data_rescue_jamaica.html). Precipitation and SST data for the Main Development Region of hurricane
activity were obtained from the NOAA Earth System Research Laboratory (ESRL) 2° ×  2° Twentieth-
Century Reanalysis (20CR; 1871–2008)48 and NOAA Extended Reconstructed SST version 3b ERSSTA
v.3b; 1860–2008)49, respectively. Annual data were selected from the geographical area defined by the
grid reference 10°–20°N, 15°–85°W and analyses were performed for the months June to November
(Hurricane season). Data for the Accumulated Cyclone Energy (ACE) index, a composite measure of
the frequency, intensity and duration of tropical cyclones (1851–2013) was obtained from the hurricane
database re-analysis project (HURDATv2; http://www.aoml.noaa.gov/hrd/hurdat/Data_Storm.html). The
Jamaican PC3 time series was obtained from the supplementary material accompanying the original
study27 (http://onlinelibrary.wiley.com/doi/10.1002/jqs.2660/suppinfo). Data for the three coral-based
SST reconstructions30–32 and those of solar and volcanic activity50–52 were recovered from the NOAA
NCDC paleoclimatology database (http://www.ncdc.noaa.gov/data-access). All datasets represented in
Fig. 3 were filtered using an 11-year running mean for the modern historical period (1750–2014) except
for the ACE timeseries, which was filtered using a 21-year running mean.

Chronology.  The age model for the Jamaican PC3 climate series25 was harmonized to annual resolu-
tion to aid direct statistical comparison with other climate time series. Given the strong and significant
correlation between the unfiltered and annually-resolved coral growth rate record from the Yucatan
Peninsula30 and Jamaican PC3 timeseries27 (1773–2008; r =  − 0.6, p <  0.001; Table 1), the PC3 record was
anchored to the coral chronology for the interval 1773–2008 providing excellent chronological control
during the recent historical period. Harmonization of the two chronologies was performed objectively
by calculating Pearson’s cross-correlation coefficients between the two time series and adjusting the PC3
age model by the calculated time lag of 2 years to achieve optimum correlation (r =  − 0.6, p <  0.001)
between the series (Fig.  5). The small age differences between the original 14C-based age model and
the harmonized model support the robustness of the original age model28 and lie well within its errors.

Correlation and time series analyses.  The Extended Hurricane Activity (EHA) index was calcu-
lated by a linear least-squares regression of the 11-year running means of the Jamaican PC3 timeseries27
and the ACE index calculated for the geographical area defined by 15–25°N, 70–80°W for the period
1851–1969 using equation  (1). These spatial and temporal boundaries were selected for calibration in
order to ensure our reconstruction is representative of hurricane activity around Jamaica and to cap-
ture the greatest range of values in both datasets. Correlation between environmental parameters was

Scientific Reports | 5:12838 | DOI: 10.1038/srep12838 8

www.nature.com/scientificreports/

performed on annual datasets, which were smoothed using 5-, 11-, 21- and 41-yr running means. Since
all of the data series exhibit serial autocorrelation and can be described by an autoregressive AR(1) pro-
cess, the number of independent samples (degrees of freedom) are smaller than the sample size (Nobs).
Thus the number of effective degrees of freedom (Neff) was calculated according to ref.  53 using the
equation:

(1 − a1a 2)
N eff = N obs
(1 + a1a 2)
where Nobs is the number of observations and a1, a2 are the lag-1 autocorrelation coefficients for each time
series. Uncertainty in the correlations was quantified by calculating T following ref. 54:
r
T= (N eff − 2)
1 − r2
where r is the correlation and Neff is the effective degrees of freedom. The distribution of T is assumed
to have a t-distribution with N-2 degrees of freedom when the samples are not autocorrelated. This is
used with a one-sided test to estimate the likelihood that the correlation has not occurred by chance.
After harmonization to annual resolution, the PC3 time series was detrended prior to analyses to sup-
press spectral leakage from low frequency (millennial) variations. A continuous wavelet transform was
applied to the PC3 timeseries to detect significant periodicities and test for stationarity. The shape of the
mother wavelet was set to morlet with a wave number ω 0 =  6. Significance levels were calculated using
a chi-squared test55 against a background of white noise. In order to capture inter-annual (ENSO) and
interdecadal (AMO/ENSO) variability within the timeseries, the high and low frequency components of
the PC3 time series were isolated using bandpass filters spanning interannual (5–8 year) and interdecadal
(54–80 year) periodicities, All numerical analyses were performed using the software package PAST56.

References
1. Knutson, T. R. et al. Tropical cyclones and climate change. Nat. Geosci. 3, 157–163 (2010).
2. Christensen, J. H. et al. in Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment
Report of the Intergovernmental Panel on Climate Change (eds. Stocker, T.F., Qin, D., Plattner, G.-K., Tignor, M., Allen, S.K.,
Boschung, J., Nauels, A., Xia, Y., V., B. & P. M., M.) (Cambridge University Press, Cambridge, United Kingdom and New York,
NY, USA, 2013).
3. Emanuel, K. & Sobel, A. Response of tropical sea surface temperature, precipitation, and tropical cyclone-related variables to
changes in global and local forcing. J. Adv. Model. Earth Syst. 5, 447–458 (2013).
4. Kossin, J. P., & Vimont, D. J. A more general framework for understanding Atlantic hurricane variability and trends. Bull. Amer.
Meteor. Soc., 88, 1767–1781 (2007).
5. Wang, C., Lee, S.-K. & Enfield, D. B. Atlantic Warm Pool acting as a link between Atlantic Multidecadal Oscillation and Atlantic
tropical cyclone activity. Geochem. Geophys. Geosys. 9, Q05V03 (2008).
6. Goldenberg, S. B., Landsea, C. W., Mestas-Nunez, A. M. & Gray, W. M. The recent increase in Atlantic hurricane activity: causes
and implications. Science 293, 474–9 (2001).
7. Aiyyer, A. & Thorncroft, C. Climatology of vertical wind shear over the tropical Atlantic. J. Clim. 19, 2969–2983 (2006).
8. Latif, M., Keenlyside, N. & Bader, J. Tropical sea surface temperature, vertical wind shear, and hurricane development. Geophys.
Res. Lett. 34, L01710 (2007).
9. Wang, C., Dong, S., Evan, A. T., Foltz, G. R. & Lee, S.-K. Multidecadal covariability of North Atlantic sea surface temperature,
African dust, Sahel rainfall, and Atlantic hurricanes. J. Clim. 25, 5404–5415 (2012).
10. Klotzbach, P. J. The influence of El Niño–Southern Oscillation and the Atlantic Multidecadal Oscillation on Caribbean tropical
cyclone activity. J. Clim. 24, 721–731 (2011).
11. Klotzbach, P. J. El Niño–Southern Oscillation’s impact on Atlantic basin hurricanes and U.S. landfalls. J. Clim. 24, 1252–1263
(2011).
12. Li, J. et al. Interdecadal modulation of El Niño amplitude during the past millennium. Nat. Clim. Chang. 1, 114–118 (2011).
13. L’Heureux, M. L., Lee, S. & Lyon, B. Recent multidecadal strengthening of the Walker circulation across the tropical Pacific. Nat.
Clim. Chang. 3, 571–576 (2013).
14. Aiyyer, A. & Thorncroft, C. Interannual-to-multidecadal variability of vertical shear and tropical cyclone activity. J. Clim. 24,
2949–2962 (2011).
15. Gray, W. M. Atlantic seasonal hurricane frequency. Part I: El Niño and 30 mb Quasi-biennial oscillation influences. Mon. Weather
Rev. 112, 1649–1668 (1984).
16. Enfield, D. The Atlantic multidecadal oscillation and its relation to rainfall and river flows in the continental US. Geophys. Res.
Lett. 28, 2077–2080 (2001).
17. Bell, G. D. et al. Climate assessment for 1999. Bull. Am. Meteorol. Soc. 81, 1–50 (1999).
18. Berland, A. J., Metcalfe, S. E. & Endfield, G. H. Documentary-derived chronologies of rainfall variability in Antigua, Lesser
Antilles, 1770–1890. Clim. Past 9, 1331–1343 (2013).
19. García-Herrera, R. New records of Atlantic hurricanes from Spanish documentary sources. J. Geophys. Res. 110, D03109 (2005).
20. Chenoweth, M. A reassessment of historical Atlantic basin tropical cyclone activity, 1700–1855. Clim. Change 76, 169–240
(2006).
21. Chenoweth, M. & Divine, D. A document-based 318-year record of tropical cyclones in the Lesser Antilles, 1690-2007. Geochem.
Geophys. Geosyst., 9, Q08013 (2008).
22. Donnelly, J. P. & Woodruff, J. D. Intense hurricane activity over the past 5,000 years controlled by El Niño and the West African
monsoon. Nature 447, 465–468 (2007).
23. Lane, P., Donnelly, J., Woodruff, J. & Hawkes, A. A decadally-resolved paleohurricane record archived in the late Holocene
sediments of a Florida sinkhole. Mar. Geol. 287, 14–30 (2011).
24. McCloskey, T. A. & Liu, K. A sedimentary-based history of hurricane strikes on the southern Caribbean coast of Nicaragua.
Quat. Res. 78, 454–464 (2012).

Scientific Reports | 5:12838 | DOI: 10.1038/srep12838 9

www.nature.com/scientificreports/

25. Van Hengstum, P. J. et al. Heightened hurricane activity on the Little Bahama Bank from 1350 to 1650 AD. Cont. Shelf Res. 86,
103–115 (2014).
26. Woodruff, J. D., Donnelly, J. P., Emanuel, K. & Lane, P. Assessing sedimentary records of paleohurricane activity using modeled
hurricane climatology. Geochem. Geophys. Geosyst., 9, 1–12 (2008).
27. Burn, M. J. & Palmer, S. E. Solar forcing of Caribbean drought events during the last millennium. J. Quat. Sci. 29, 827–836
(2014).
28. Landsea, C. W. & Franklin, J. L. Atlantic hurricane database uncertainty and presentation of a new database format. Mon.
Weather Rev. 141, 3576–3592 (2013).
29. Dunstone, N. J., Smith, D. M., Booth, B. B., Hermanson, L. & Eade, R. Anthropogenic aerosol forcing of Atlantic tropical storms.
Nat. Geosci. 6, 534–539 (2013).
30. Vásquez-Bedoya, L. F., Cohen, A. L., Oppo, D. W. & Blanchon, P. Corals record persistent multidecadal SST variability in the
Atlantic Warm Pool since 1775 AD. Paleoceanography 27, PA3231 (2012).
31. Hetzinger, S. et al. Caribbean coral tracks Atlantic Multidecadal Oscillation and past hurricane activity. Geology 36, 11 (2008).
32. Kilbourne, K. H. et al. Paleoclimate proxy perspective on Caribbean climate since the year 1751: Evidence of cooler temperatures
and multidecadal variability. Paleoceanography 23, PA3220 (2008).
33. Vecchi, G. A. & Knutson, T. R. Estimating annual numbers of Atlantic hurricanes missing from the HURDAT database
(1878–1965) using ship track density. J. Clim. 24, 1736–1746 (2011).
34. Nyberg, J. et al. Low Atlantic hurricane activity in the 1970s and 1980s compared to the past 270 years. Nature 447, 698–701
(2007).
35. Kozar, M. E., Mann, M. E., Emanuel, K. A. & Evans, J. L. Long-term variations of North Atlantic tropical cyclone activity
downscaled from a coupled model simulation of the last millennium. J. Geophys. Res. Atmos. 118, 13,383–13,392 (2013).
36. Mann, M. E., Woodruff, J. D., Donnelly, J. P. & Zhang, Z. Atlantic hurricanes and climate over the past 1,500 years. Nature 460,
880–3 (2009).
37. Enfield, D. B. & Cid-Serrano, L. Secular and multidecadal warmings in the North Atlantic and their relationships with major
hurricane activity. Int. J. Climatol. 30, 174–184 (2010).
38. Clement, A., Seager, R. & Cane, M. Orbital controls on the El Nino/Southern Oscillation and the tropical climate. Paleoceanography
14, 441–456 (1999).
39. Cane, M. A. The evolution of El Niño, past and future. Earth Planet. Sci. Lett. 230, 227–240 (2005).
40. Zebiak, S. & Cane, M. A model EI Nino-Southern Oscillation. Mon. Weather Rev. 115, 2262–2278 (1987).
41. Mann, M. E., Cane, M. A., Zebiak, S. E. & Clement, A. Volcanic and solar forcing of the tropical Pacific over the past 1000 years.
J. Clim. 18, 447–456 (2005).
42. Cobb, K. M., Charles, C. D., Cheng, H. & Edwards, R. L. El Niño/Southern Oscillation and tropical Pacific climate during the
last millennium. Nature 424, 271–276 (2003).
43. Hodell, D. A. et al. Climate change on the Yucatan Peninsula during the Little Ice Age. Quat. Res. 63, 109–121 (2005).
44. Lane, C. S., Horn, S. P., Orvis, K. H. & Thomason, J. M. Oxygen isotope evidence of Little Ice Age aridity on the Caribbean slope
of the Cordillera Central, Dominican Republic. Quat. Res. 75, 461–470 (2011).
45. Haug, G. H., Hughen, K. A, Sigman, D. M., Peterson, L. C. & Röhl, U. Southward migration of the intertropical convergence
zone through the Holocene. Science 293, 1304–1308 (2001).
46. Knudsen, M. F., Jacobsen, B. H., Seidenkrantz, M.-S. & Olsen, J. Evidence for external forcing of the Atlantic Multidecadal
Oscillation since termination of the Little Ice Age. Nat. Commun. 5, 3323 (2014).
47. Otterå, O. H., Bentsen, M., Drange, H. & Suo, L. External forcing as a metronome for Atlantic multidecadal variability. Nat.
Geosci. 3, 688–694 (2010).
48. Compo, G.P. et al. The Twentieth Century Reanalysis Project. Q.J.R. Meteorological. Soc. 137, 1–28 (2011).
49. Smith, T. M., Reynolds, R. W., Peterson, T. C. & Lawrimore, J. Improvements to NOAA’s historical merged land–ocean surface
temperature analysis (1880–2006). J. Clim. 21, 2283–2296 (2008).
50. Bard, E., Raisbeck, G., Yiou, F. & Jouzel, J. Solar irradiance during the last 1200 years based on cosmogenic nuclides. Tellus B 52,
985–992 (2000).
51. Steinhilber, F., Beer, J. & Fröhlich, C. Total solar irradiance during the Holocene. Geophys. Res. Lett. 36, L19704 (2009).
52. Crowley, T. J. Causes of climate change over the past 1000 years. Science 289, 270–277 (2000).
53. Bretherton, C. S., Widmann, M., Dymnikov, V. P., Wallace, J.M. & Blade, I. The effective number of spatial degrees of freedom
of a time-varying field, J. Clim., 12, 1990–2009 (1999).
54. McCarthy, G. D., Haigh, I. D., Hirschi, J. J., Grist, J. P. & Smeed, D. A. Ocean impact on decadal Atlantic climate variability
revealed by sea-level observations. Nature 521, 508––510..
55. Torrence, C. & Compo, G. P. A practical guide to wavelet analysis. Bull. Am. Meteorol. Soc. 79, 61–78 (1998).
56. Hammer, Ø., Harper, D. A. T. & Ryan, P. D. PAST: Paleontological statistics software package for education and data analyses.
Palaeontologica Electronica 4, 1–9 (2001).

Acknowledgements
M.J.B. acknowledges support from the UWI New Initiative Programme and UWI Study and Travel
Grant. We thank Jeff Donnelly for his comments and suggestions, and Chris Landsea and James Kossin
for their detailed reviews, which helped improve the manuscript substantially.

Author Contributions
M.J.B. and S.E.P. designed the study and compiled the data. M.J.B. performed the analyses and wrote
the manuscript with significant contributions from S.E.P. Both authors contributed to the interpretation
of results.

Additional Information
Supplementary information accompanies this paper at http://www.nature.com/srep
Competing financial interests: The authors declare no competing financial interests.
How to cite this article: Burn, M. J. and Palmer, S. E. Atlantic hurricane activity during the last
millennium. Sci. Rep. 5, 12838; doi: 10.1038/srep12838 (2015).

Scientific Reports | 5:12838 | DOI: 10.1038/srep12838 10

www.nature.com/scientificreports/

This work is licensed under a Creative Commons Attribution 4.0 International License. The
images or other third party material in this article are included in the article’s Creative Com-
mons license, unless indicated otherwise in the credit line; if the material is not included under the
Creative Commons license, users will need to obtain permission from the license holder to reproduce
the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

Scientific Reports | 5:12838 | DOI: 10.1038/srep12838 11