Received: 14 September 2021 Revised: 25 May 2022 Accepted: 23 June 2022 Published on: 15 July 2022

DOI: 10.1002/joc.7774

RESEARCH ARTICLE

Heat stress in the Caribbean: Climatology, drivers,

and trends of human biometeorology indices

Claudia Di Napoli1,2,3 | Theodore Allen4
zaro5

| Pablo A. Méndez-La |
3
Florian Pappenberger

1
School of Agriculture, Policy and
Development, University of Reading, Abstract
Reading, UK Forty years (1980–2019) of reanalysis data were used to investigate climatology
2
Department of Geography and and trends of heat stress in the Caribbean region. Represented via the Universal
Environmental Science, University of
Thermal Climate Index (UTCI), a multivariate thermophysiological-relevant
Reading, Reading, UK
3
European Centre for Medium Range
parameter, the highest heat stress is found to be most frequent and geographi-
Weather Forecasts, Reading, UK cally widespread during the rainy season (August, September, and October).
4
Caribbean Institute for Meteorology and UTCI trends indicate an increase of more than 0.2
Cdecade−1, with southern
Hydrology, St James, Barbados
Florida and the Lesser Antilles witnessing the greatest upward rates
5
Environmental Health Department,
(0.45
Cdecade−1). Correlations with climate variables known to induce heat
Graduate School of Public Health,
University of Puerto Rico, San Juan, stress reveal that the increase in heat stress is driven by increases in air tempera-
Puerto Rico ture and radiation, and decreases in wind speed. Conditions of heat danger, as

Correspondence
depicted by the heat index (HI), have intensified since 1980 (+1.2
C) and are
Claudia Di Napoli, School of Agriculture, found to occur simultaneously to conditions of heat stress suggesting a synergy
Policy and Development, University of between heat illnesses and physiological responses to heat. This work also
Reading, Whiteknights, PO Box
237, Reading RG6 6EU, UK. includes the analysis of the record-breaking 2020 heat season during which the
Email: [email protected] UTCI and HI achieved above average values, indicating that local populations
most likely experienced heat stress and danger higher than the ones they are
Funding information
Royal Society, Grant/Award Number: FOS used to. These findings confirm the gradual intensification of heat stress in the
\R1\191010; Wellcome Trust, Grant/ Caribbean and aim to provide a guidance for heat-related policies in the region.
Award Number: 209734/Z/17/Z
KEYWORDS
bioclimatology, Caribbean, climate adaptation, climate change, heat hazards, human
health, preparedness, resilience

1 | INTRODUCTION With regards to human health, heat stress was first

identified as a major health issue for the region in 2003
In the recent decades heat stress has become an emerg- (Aron et al., 2003). Heat stress may lead to illnesses such
ing threat in the Caribbean region with coral reefs, live- as sunstroke, sunburn, heat exhaustion, and dehydration
stock, labour productivity, and human health being the (Ebi et al., 2006). In a survey study conducted in 2020,
most affected sectors (Kjellstrom et al., 2018; Lallo health professionals from 12 countries, including
et al., 2018; Méndez-L
azaro et al., 2018b; Muniz-Castillo Jamaica, reported on heat-related illnesses already
et al., 2019). adversely affecting local populations, and expect these

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided
the original work is properly cited.
© 2022 The Authors. International Journal of Climatology published by John Wiley & Sons Ltd on behalf of Royal Meteorological Society.

Int J Climatol. 2023;43:405–425. wileyonlinelibrary.com/journal/joc 405

 10970088, 2023, 1, Downloaded from https://rmets.onlinelibrary.wiley.com/doi/10.1002/joc.7774 by Aristotle University Of, Wiley Online Library on [28/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
406 DI NAPOLI ET AL.

illnesses to become more frequent or severe over the next which describes the physiological heat stress the human
10 years (Kotcher et al., 2021). A 2015 focus group of body experiences in the attempt to maintain a thermal
Caribbean health-care providers perceived “hotter than equilibrium with the surrounding outdoor environment
usual” temperatures as contributing to more hospitaliza- (Błażejczyk et al., 2013). Combining an advanced body's
tions for dehydration and sunburn in Grenada, and an physiological model with a state-of-the-art clothing insula-
increase in respiratory-related illnesses in Trinidad and tion model for outdoor climates (UTCI-Fiala model), the
Tobago (Macpherson and Akpinar-Elci, 2015). Excessive UTCI has long been considered superior to other indices
heat exposure has also been linked to reduced children's in representing thermal conditions and their variability
learning ability at school, increased demand for cooling, (Blazejczyk et al., 2012; Fiala et al., 2012; Havenith
and a re-envisioning of the urban environment in eastern et al., 2012). Yet, its application in the Caribbean region
Caribbean states (Wilkinson et al., 2021). remains limited and mostly focused on the assessment of
From a climatological point of view, previous research bioclimatic comfort in touristic settings, such as coastal
on historical climate data confirmed an increase in the resorts (Rutty and Scott, 2014; Rutty and Scott, 2015).
occurrence of extreme high temperatures across the region The UTCI needs four climate variables—2 m air tem-
as well as in single islands, from Hispaniola to Barbados, perature, relative humidity, 10 m wind speed, and
Cuba, Puerto Rico, Trinidad, and Tobago (Singh, 1997; radiation—to be calculated (Brode et al., 2012). This has
Naranjo-Diaz and Centella, 1998; Peterson et al., 2002; been perceived as a limiting factor from local stake-
Aguilar et al., 2005; Pérez and Jury, 2012; Beharry holders, as most of environmental data, especially those
et al., 2015; Jury, 2015; Méndez-L
azaro et al., 2015; Jones from ground meteorological stations, are unavailable
et al., 2016b; Dookie et al., 2018; Mohan et al., 2020). In (Climate Studies Group Mona, 2020). Maintaining a con-
2020 the Caribbean region witnessed a record-breaking sistent and continuous daily recordkeeping of ground
heat season. Air temperatures set new national records in weather observations has historically been considered
Cuba in April, and in Dominica, Grenada, and Puerto challenging in the Caribbean region, leading to gaps in
Rico in September (WMO, 2021). Being one of the three the recorded time series as well as in those geographical
warmest years for the Caribbean region compared with areas where meteorological stations are lacking (Peterson
the average temperature for the 1981–2010 period, 2020 et al., 2002). This limitation can be overcome by using cli-
placed itself in the upward trend in observed air tempera- mate reanalysis data. Obtained by applying forecast
tures since the mid-1990s (Angeles-Malaspina et al., 2018; models and data assimilation systems on past observa-
WMO, 2021). As for climate projections, these foresee a tions, a climate reanalysis provides a complete and con-
further increase in the duration, intensity, and frequency sistent description of the Earth's climate and its evolution
of extreme heat events in the Caribbean region regardless in recent decades in the form of gridded time-stepped
of the future scenario, with positive trends predicted for data (Dee et al., 2016). To the authors' knowledge, only a
warm days and nights (Biasutti et al., 2011; Campbell few studies so far have used climate reanalysis data to
et al., 2011; Hall et al., 2012; Taylor et al., 2012; Karmalkar investigate historical heat stress in the Caribbean region
et al., 2013; McLean et al., 2015; Jones et al., 2016a; 2016b; and none of them makes use of the UTCI (Ramirez-
Stennett-Brown et al., 2017; Angeles-Malaspina Beltran et al., 2017; Angeles-Malaspina et al., 2018).
et al., 2018; Taylor et al., 2018). In view of this, the present paper aims to assess clima-
As awareness grows on the linkages between heat and tology and trends of human heat stress in the Caribbean
human health, a need has arisen to estimate what humans region by means of climate reanalysis data. Using
feel under different climatic conditions. Biometeorological ERA5-HEAT, a derived product from the ERA5 climate
indices have recently been used for the scope. One of these reanalysis dataset (Di Napoli et al., 2021a), pan-
indices is the heat index (HI), which combines air temper- Caribbean maps and temporal trends were obtained for
ature and relative humidity to posit a human-perceived the UTCI over a 40-year historical period and are pre-
apparent temperature from a body's heat transfer balance sented here for the first time. The paper also examines
model (Steadman, 1979). Historical trends and climate the climate variables known to contribute to human heat
projections indicate increasing patterns for the HI as well stress, namely 2 m air temperature, relative humidity,
as intensification of HI-defined extreme events across the 10 m wind speed, and mean radiant temperature. Called
Caribbean region (Ramirez-Beltran et al., 2017; Angeles- heat stress drivers hereafter, their climatology and trends
Malaspina et al., 2018). More sophisticated biometeorolog- are also investigated over the same period. As the HI has
ical indices have also be investigated with the aim to previously been used as an indicator of heat stress for the
evidence the relevance and impact of the climate from a region, climatology and trends of the HI are included and
human thermophysiological perspective. These indices discussed against the UTCI. Finally, the extreme heat epi-
include the Universal Thermal Climate Index (UTCI), sode affecting the Caribbean region in 2020 is examined.
 10970088, 2023, 1, Downloaded from https://rmets.onlinelibrary.wiley.com/doi/10.1002/joc.7774 by Aristotle University Of, Wiley Online Library on [28/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
DI NAPOLI ET AL. 407

35°N
which can be controlled and/or prevented by adopting
specific behavioural and clothing styles (Havenith and
30°N
Fiala, 2015). The UTCI can be calculated solving either
25°N the UTCI-Fiala model or the UTCI operational proce-
dure. The former may be computationally intensive and
20°N time-consuming, so the latter is generally preferred and
Latitude

is here used. The operational procedure computes the off-

15°N
set between the UTCI and 2 m air temperature via a
10°N
6-order polynomial equation in the climate variables of
2 m air temperature, relative humidity, 10 m wind speed,
5°N and mean radiant temperature. The procedure approxi-
mates the UTCI-Fiala model within a root-mean-square
0°
100°W 90°W 80°W 70°W 60°W 50°W error of 1.1
C (Brode et al., 2012).
Longitude
Previous literature made use of the HI to assess heat
F I G U R E 1 Map showing the Caribbean region considered in as a hazard in the Caribbean region (Ramirez-Beltran
this study (black rectangle). Blue and red lines delimit the Greater et al., 2017; Angeles-Malaspina et al., 2018). For compari-
Antilles and the Lesser Antilles study areas, respectively [Colour son reasons, the present study considers the HI too.
figure can be viewed at wileyonlinelibrary.com] Unlike the UTCI, the HI is valid only in the upper range
of thermal exposure, that is, heat. The HI classification
and corresponding physiological responses are reported
2 | DATA AND METHODS in Table 1. The HI is calculated from 2 m air temperature
and relative humidity using the regression equation of
Reanalysis data served as the basis for (a) estimating the Rothfusz (NOAA, 2014),
thermal bioclimate of the Caribbean region in terms of
human heat stress and (b) describing the spatial and tem- HI = −42:379 +2:04901523  T +10:14333127  RH
poral variability of the UTCI, HI, and heat stress drivers −0:22475541  T  RH− 0:00683783  T  T
across the same area. −0:05481717  RH  RH +0:00122874  T  T  RH
+0:00085282  T  RH  RH
2.1 | Study area −0:00000199  T  T  RH  RH,
ð1Þ
The Caribbean region encompasses large (Greater Antil-
les) and small (Lesser Antilles) islands as well as coast- where T is temperature in degrees Fahrenheit (
F) and
lines along Central and South America. Figure 1 shows RH is relative humidity in percent. Its output is HI, also
the geographical area considered in the present study. known as apparent temperature, expressed in
F. If rela-
It includes the Caribbean Sea, parts of Central America tive humidity is less than 13% and air temperature is
and Florida, and the northern part of South America between 80 and 112
F (26.7 and 44.4
C), then the
(9
–28
N and 91
–58
W). To capture climatic variability following is subtracted from Equation (1):
within the Caribbean, two subregions corresponding

 rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
to the Greater Antilles and the Lesser Antilles are 13 −RH ½17 − jðT −95Þj
considered. They are defined as shown in Figure 1. Adjustment =  : ð2Þ
4 17

2.2 | Heat metrics If relative humidity is greater than 85% and air tem-
perature is between 80 and 87
F (26.7 and 30.6
C), then
The present study focuses on the UTCI, a thermal stress the following is added to Equation (1):
metrics that has been evaluated and studied over a vari-



ety of climate regions as well as spatial and temporal RH −85 87− T
Adjustment =  : ð3Þ
scales (Coccolo et al., 2016). The UTCI is valid across the 10 5
entire range of thermal exposure, that is, from cold to
hot, and is classified into a 10-category scale (Table 1). If conditions of air temperature and relative humidity
Each category corresponds to a well-defined set of human warrant a HI value below 80
F, a simpler formula is applied
physiological responses to the outdoor environment, to calculate values consistent with Steadman's results:
 10970088, 2023, 1, Downloaded from https://rmets.onlinelibrary.wiley.com/doi/10.1002/joc.7774 by Aristotle University Of, Wiley Online Library on [28/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
408 DI NAPOLI ET AL.

TABLE 1 Category scales of the UTCI and HI, and corresponding physiological responses

Heat metrics Range (
C) Category Physiological responses

Universal Thermal Climate Index (UTCI) Above 46 Extreme heat stress Increase in rectal temperature time gradient
Steep decrease in total net heat loss
Averaged sweat rate >650 ghr−1, steep increase
38–46 Very strong heat stress Low core to skin temperature gradient
Increase in rectal temperature at 30 min
32–38 Strong heat stress Averaged sweat rate >200 ghr−1
Increase in rectal temperature at 120 min
Instantaneous change in skin temperature
26–32 Moderate heat stress Change of slopes in sweat rate, rectal and skin
(mean, face, hand) temperature
Occurrence of sweating at 30 min
Steep increase in skin wettedness
9–26 No thermal stress Averaged sweat rate >100 ghr−1
Plateau in rectal temperature time gradient
0–9 Slight cold stress Local minimum of hand skin temperature
−13 to 0 Moderate cold stress Vasoconstriction
Face skin temperature at 30 min < 15
C (pain)
−13 to −27 Strong cold stress Numbness
Increase in core to skin temperature gradient
−27 to −40 Very strong cold stress Frostbite, numbness, shivering
Steeper decrease in rectal temperature
Below −40 Extreme cold stress Frostbite
Decrease in rectal temperature time gradient
Heat Index (HI) Below 26.7 No caution None
26.7–32.2 Caution Fatigue possible with prolonged exposure and/or
physical activity
32.2–39.4 Extreme caution Heat stroke, heat cramps, or heat exhaustion
possible with prolonged exposure and/or
physical activity
39.4–51.1 Danger Heat cramps or heat exhaustion likely, and heat
stroke possible with prolonged exposure and/or
physical activity
Above 51.1 Extreme danger Heat stroke highly likely

Source: Adapted from Blazejczyk et al. (2012) and NOAA NWS (2021b).

HI =0:5  fT +61:0 + ½ðT −68:0Þ  1:2+ ðRH  0:094Þg:

Medium-Range Weather Forecasts (ECMWF) within the
ð4Þ
Copernicus Climate Change Service (Hersbach
et al., 2020). UTCI reanalysis data were retrieved from
Finally, HI is converted to
C. ERA5-HEAT via the Copernicus Climate Data Store
(CDS, 2020).
As for the heat stress drivers, reanalysis data of mean
2.3 | Reanalysis datasets radiant temperature (MRT) were also retrieved from
ERA5-HEAT as representative of the solar and thermal
The source of UTCI data is ERA5-HEAT, a quality- radiation irradiating, both directly and diffusely, a
controlled reanalysis consisting of global gridded maps of human subject placed in an outdoor environment
the index from 1979 to present (Di Napoli et al., 2021a). (Di Napoli et al., 2020). ERA5 was used as the reanalysis
In ERA5-HEAT, the UTCI is derived using the opera- data source for 2 m air temperature, 2 m dew point tem-
tional procedure with climate variables from the ERA5 perature, and 10 m wind speed (CDS, 2018). Both ERA5
reanalysis, which is produced by the European Centre for and ERA5-HEAT are publicly and freely available for
 10970088, 2023, 1, Downloaded from https://rmets.onlinelibrary.wiley.com/doi/10.1002/joc.7774 by Aristotle University Of, Wiley Online Library on [28/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
DI NAPOLI ET AL. 409

download from the Copernicus Climate Data Store distribution observed in the Caribbean region, that is, an
(CDS, 2018; 2020). early (MJJ) and late (ASO) rainy season interrupted by a
Reanalysis data were used to derive climate variables dry spell in July (mid-summer drought; Stephenson
not directly provided by reanalysis datasets. Relative et al., 2014). Dry seasons, NDJ and FMA, correspond to
humidity, for instance, was computed from 2 m air tem- the time of the year where the lowest rainfall intensity
perature and 2 m dew point temperature via the Clausius– events are observed (Giannini et al., 2000; Chen and
Clapeyron relation (Alduchov and Eskridge, 1996). The HI Taylor, 2002). The thermal bioclimate of FMA, MJJ,
was calculated from 2 m air temperature and relative ASO, and NDJ was obtained by averaging UTCI daily
humidity using Equations (1)–(4). minima, maxima, means across each season in the 1980–
All reanalysis data covered a period of 41 years 2019 period. This produced a set of pan-Caribbean UTCI
(1980–2020) at 1-hr time step, were cropped to the study maps of season-specific heat stress.
area and, at each time step, consisted of 7,956 grid cells at Monthly UTCI averages were computed over the
0.25
spatial resolution (31 km circa). same period to investigate intraseasonal changes in the
UTCI climatology. They were also aggregated for geo-
graphical zones, that is, the entire study area as well as
2.4 | Statistical methods the regions corresponding to the Greater Antilles and the
Lesser Antilles (Figure 1). This allowed for comparison
The following statistical methods were used to investigate between region-wide trends and trends derived subre-
heat stress conditions and associated trends in the gionally. The same methodology was applied to derive
Caribbean region. climatologies for heat stress drivers and the HI.
Finally, the monthly frequency of UTCI stress catego-
ries and HI danger categories was calculated. This offers
2.4.1 | Climatologies insights on the local bioclimate in terms of the number of
occurrences of heat stress/danger conditions to which
The thermal bioclimate of the Caribbean, that is, the local population are exposed across the year.
long-term occurrence of heat stress in the region, was
obtained from the daily minima, maxima, means of
hourly UTCI reanalysis data for the 1980–2019 historical 2.4.2 | Trends
period. These daily statistics were chosen based on their
relevance in a human health perspective. Specifically, Linear trends for the UTCI were computed in order to
daily maxima are considered representative of the day- check for temporal changes in heat stress. The Sen's slope
time heat individuals might be exposed to, with potential estimator was used to evaluate the rate of changes and
direct detrimental effects (e.g., heatstroke); daily minima the Mann–Kendall nonparametric statistical test was
are considered representative of night-time conditions used to evaluate the significance level of the trends
which, for example, during heatwaves may deteriorate (Mann, 1945; Sen, 1968; Kendall, 1975). The combination
sleep quality and prevent individuals from recovering of Mann–Kendall test and Sen's estimator has been suc-
from daytime heat exposure; daily means are considered cessfully used in multiple climate trends studies focussed
representative of the average heat experienced within a on the Caribbean (Aguilar et al., 2005; Beharry
full day (McGregor et al., 2015). These statistics have all et al., 2015; Angeles-Malaspina et al., 2018; Cavazos
been associated to negative health impacts as well as et al., 2019; Mohan et al., 2020). UTCI trends were calcu-
implemented in operational heat health watch warning lated from averaged seasonal values for each year for the
systems (Kotharkar and Ghosh, 2022). To address the 1980–2019 historical period. The year-by-year change in
objectives of the present study, UTCI daily minima, max- heat stress, represented by Sen's slope coefficients, was
ima, means were first computed for each grid cell during then multiplied by 10 so to be represented as a decade-
the hours from 2300 to 2200 UTC (Coordinated Universal by-decade variation. Linear trends were calculated at
Time) and then averaged across months and seasons to each grid cell, thus allowing for investigation in UTCI
obtain corresponding climatologies of heat stress and trends' spatial patterns. The same linear trend analysis
in agreement to previous literature on the topic was applied to heat stress drivers and the HI.
(e.g., Ramirez-Beltran et al., 2017; Angeles-Malaspina Pearson correlations between mean yearly averages of
et al., 2018). the UTCI and mean yearly averages of 2 m air tempera-
Seasons are defined as February–April (FMA), ture, relative humidity, 10 m wind speed, and MRT were
May–July (MJJ), August–October (ASO), and November– computed. This was done to highlight, both spatially and
January (NDJ). This follows the bimodal rainfall temporally, the relationships between the UTCI and each
 10970088, 2023, 1, Downloaded from https://rmets.onlinelibrary.wiley.com/doi/10.1002/joc.7774 by Aristotle University Of, Wiley Online Library on [28/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
410 DI NAPOLI ET AL.

of its drivers during the 1980–2019 historical period. The is justified by their linkages to human health in
strength of the correlation, represented by the absolute heat-related impacts studies (Di Napoli et al., 2019).
value of the Pearson correlations r, follows the classifica-
tion by (Evans, 1996): 0.00–0.19 is “very weak”; 0.20–0.39
is “weak”; 0.40–0.59 is “moderate”; 0.60–0.79 is “strong” 3 | RESULTS
and 0.80–1.0 is “very strong.”
Geographical areas where trends are not statistically 3.1 | Heat stress climatology of the
significant (p ≥ .05, where p is the probability level) were Caribbean
excluded from the analysis.
Additionally, anomalies by decades were computed The yearly climatology derived from 1980 to 2019
for year- and season-averaged UTCI and HI daily means ERA5-HEAT data show that heat stress is moderate,
with respect to the 1981–2010 climatological baseline. strong, and very strong in UTCI maxima, whereas UTCI
minima are characterized by no thermal stress and UTCI
means are associated with both no thermal stress and
2.4.3 | Multi-index heat information moderate heat stress (Figure 2a).
Season-averaged UTCI maps confirm that UTCI min-
As described in section 2.2, UTCI values are categorized ima are mostly associated to no thermal stress through-
in terms of thermal stress, that is, of the physiological out the year (Figure 2b). UTCI maxima reveal extended
responses put in place by the human body to maintain its moderate heat stress across the region with levels of
core temperature within the range of optimal physiologi- strong and very strong heat stress achieved in islands
cal performance (McGregor and Vanos, 2018). The HI such as Cuba, Hispaniola, Puerto Rico, and Bahamas as
classification reflects the onset of more dangerous heat well as the northern part of South America and the
disorders under prolonged exposure and/or physical Yucat
an peninsula. This translates into UTCI means
activity in the heat (NOAA NWS, 2021b). The co- associated to prevalent conditions of no thermal stress in
occurrence of UTCI thermal stress categories for different NDJ and FMA, of moderate heat stress in ASO, and
HI danger categories was assessed in the present study. between absent and moderate heat stress in MJJ. The
This allowed for investigating the UTCI and the HI as lowest UTCI values occur at latitudes above 25
N circa,
heat metrics providing complementary information on concurrently with the winter and spring seasons in the
the effects of heat on human health. The joint occurrence Northern Hemisphere. The highest UTCI values are
of UTCI stress categories and HI danger categories was widespread in the rainy season, ASO specifically, and
determined at each grid cell of the seasonal averages indicate this as the heat stress season for the Caribbean.
(daily maxima) across the 1980–2019 historical period. The monthly climatology of the UTCI maximum for
Outputs were later aggregated for geographical zones. the average of the entire study area provides further
The analysis focuses on daily maxima to account for insight into the heat stress occurrence throughout the
conditions of highest exposure to heat. year (Figure 2c). The UTCI monthly maximum starts
ascending in March, it plateaus in June, it increases again
reaching the highest value in September, then it descends
2.4.4 | A case study: The 2020 heat season until December and reaches a second plateau from
January to March (or minimum in February), completing
Two heatwaves affected several Caribbean states in April the annual cycle. The pattern is similar to the climatol-
and September 2020. They are here analysed as a case ogies of UTCI minima and mean values. From June to
study of heat extremes in the region. The monthly aver- September, the Greater Antilles exhibit higher UTCI
aged values achieved by 2 m air temperature, the UTCI, values than both the average of the Caribbean region and
and the HI were computed for the months when the the Lesser Antilles (33.9
C against 32.6 and 31.4
C,
heatwave occurred. The exceptionality of the events was respectively, in September). For the Lesser Antilles, the
assessed in terms of the anomalies of those variables to UTCI increase between June and September is greatest
the 1980–2019 historical period. For the UTCI and HI, (1.5–2.1
C against 0.7–1.1 and 1–1.3
C for the entire
anomalies were expressed in differences of values in
C region and the Greater Antilles, respectively), but the
or of stress/danger categories. For the 2 m air tempera- UTCI varies the less between September and February
ture, which is not classified in categories, anomalies were (2.1–5.3
C against 5–6.9
C for the Caribbean region and
expressed with respect to the exceedance of climatologi- the Greater Antilles).
cally defined percentile thresholds, namely the 95th The thermal bioclimate of the Caribbean is also
and 99.5th percentiles. The choice of these percentiles reflected by the monthly frequencies of occurrence of
 10970088, 2023, 1, Downloaded from https://rmets.onlinelibrary.wiley.com/doi/10.1002/joc.7774 by Aristotle University Of, Wiley Online Library on [28/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
DI NAPOLI ET AL. 411

UTCI 0 9 26 32 38 46
(°C)

Minimum Maximum Mean

(a)
Year

(b)
FMA
MJJ
ASO
NDJ

(c) 35

30
UTCI (°C)

25

20 Caribbean Region
Greater Antilles
Lesser Antilles
15
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

F I G U R E 2 Yearly (a), seasonal (b), and monthly (c) UTCI climatology for the 1980–2019 historical period. Background colours refer to
the UTCI stress category scale at the top of the figure [Colour figure can be viewed at wileyonlinelibrary.com]

different heat stress categories (Figure S1, Supporting more than 71% of days in August and September. In the
Information). Conditions of no thermal stress prevail for Greater Antilles, the high frequency (74–99%) of days
UTCI minima throughout the year with moderate heat with UTCI maxima in the strong heat stress category is
stress achieved in less than 10% of the days between May observed over a longer period, from June to October. In
and October (peak in September) in the Greater and the Lesser Antilles, moderate heat stress prevails in the
Lesser Antilles. UTCI maxima are predominantly charac- UTCI maxima throughout the year and strong heat stress
terized by heat stress. In the whole region, moderate heat occurs with the highest frequency (30%) in September. As
stress occurs most frequently (68–100% of days) from a consequence, moderate heat stress in UTCI means rep-
October to June whereas strong heat stress is achieved in resents the dominant condition between July and
 10970088, 2023, 1, Downloaded from https://rmets.onlinelibrary.wiley.com/doi/10.1002/joc.7774 by Aristotle University Of, Wiley Online Library on [28/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
412 DI NAPOLI ET AL.

September in the Caribbean region (77–97% of days), equator. Relative humidity is at its highest in MJJ and
between May and October in the Greater Antilles (63– ASO, in agreement with the bimodal rainfall distribution.
100% of days) and between August and October in the The strongest wind speeds for the study area are observed
Lesser Antilles (70–82% of days). during FMA and are collocated with the Caribbean Low-
Level Jet that peaks during the same season (Whyte et
al., 2008). The lowest wind speeds are instead achieved in
3.2 | Climatology of heat stress drivers ASO. Together, season-averaged maps of heat stress
drivers highlight ASO as the season characterized by gen-
Figure 3 provides a pan-Caribbean overview of the mean eral conditions of higher air temperatures, humidity, and
climatology of 2 m air temperature, relative humidity, radiation, and lower wind speeds.
10 m wind speed, and MRT as derived from ERA5 and Monthly climatologies, aggregated across the Carib-
ERA5-HEAT data for the 1980–2019 historical period. bean region, further support the heat characterization of
Season-averaged maps (Figure 3a) show that air tem- the ASO season (Figure 3b). They show a unimodal dis-
perature and MRT are at their maximum in ASO and tribution for 2 m air temperature and MRT, with the two
minimum in NDJ, reflecting the amount of solar insola- climate variables reaching a maximum in August (27.2
tion the region receives because of its proximity to Earth's and 34.3
C) and a minimum in January (24 and 29.8
C).

Air temperature (°C) Relative humidity (%) Wind speed (m·s−1) MRT (°C)
15 20 25 30 50 60 70 80 90 100 0 2 4 6 8 10 12 20 25 30 35

(a)
FMA
MJJ
ASO
NDJ

°C % m/s °C
(b)
27 81 8 34

6 32
78
Caribbean Region
24 4 30 Greater Antilles
Lesser Antilles
75
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 J F M A M J J A S O N D

F I G U R E 3 Mean climatology of UTCI drivers by seasons (a) and months (b) over the 1980–2019 historical period [Colour figure can be
viewed at wileyonlinelibrary.com]
 10970088, 2023, 1, Downloaded from https://rmets.onlinelibrary.wiley.com/doi/10.1002/joc.7774 by Aristotle University Of, Wiley Online Library on [28/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
DI NAPOLI ET AL. 413

Relative humidity exhibits a bimodal distribution that Antilles. Furthermore, heat stress trends differ according
increases in March, reaches a first peak in June (82%), to the season. Statistically significant increases in the
decreases and plateaus in July–August, increases again to UTCI (minimum, mean, and maximum) are found in
a second peak in September (82%) and then decreases ASO for the whole Caribbean region, with the Lesser
until March (76%), completing the annual cycle. Wind Antilles experiencing rates exceeding 0.45
Cdecade−1.
speeds attain a minimum in September (5.1 ms−1), a pri- The same rate is observed in southern Florida during the
mary maximum in January (6.7 ms−1), and a secondary Caribbean dry season, specifically FMA. It is worth not-
maximum in July (5.9 ms−1). It is worth noting that the ing that the Lesser Antilles show a statistically significant
secondary maximum coincides with the occurrence of increase in all seasons and statistics (minima, maxima,
the Caribbean mid-summer drought and wind speeds are and means).
higher during the dry season (6–6.7 ms−1) than the rainy A further insight into how heat stress has changed in
season. the Caribbean region across 40 years is provided by UTCI
At the subregional level, the climatologies for 2 m air anomalies relative to the 1981–2010 climatology baseline
temperature, relative humidity, 10 m wind speed, and (Figure 4b). In the 2010–2019 decade, yearly data indicate
MRT are similar in pattern between the Greater Antilles that region-averaged UTCI was 0.51
C higher than the
and the Caribbean-averaged ones. The unimodal distribu- baseline and 0.71
C higher than the 1980–1989 decade.
tion of air temperature (resp. MRT) reaches a minimum In the 1990–1999 and 2000–2009 decades, the UTCI was
of 24.1
C (resp. 29.9
C) in January and a maximum of closest to the baseline although the Lesser Antilles show
27.7
C (resp. 34.7
C) in July–August. These values are an increasing trend throughout the entire baseline
above the Caribbean-averaged ones. With regards to rela- period. These findings can also be observed in season
tive humidity and wind speed, the Greater Antilles are data, namely in FMA, MJJ, and NDJ. For ASO, there is
drier (80.9% as maximum value) and less windy (4.3– not increase hiatus during the 1990–2009 period and the
5.7 ms−1) than the region overall. UTCI increases from the 1980–1989 decade to the 2010–
The climatology of each heat stress driver in the 2019 throughout. Such a continuous increase over time is
Lesser Antilles displays region-specific distributions. Air observed for the Lesser Antilles, the Greater Antilles, and
temperature there reaches its maximum in September the entire Caribbean region.
(27.3
C), 1 month later than the Greater Antilles. It is
also characterized by a higher minimum (25.1
C in
February). MRT follows a similar distribution with maxi- 3.4 | Trends in heat stress drivers
mum and minimum values equal to 34.2 and 31.4
C in
September and February, respectively. It follows that the Figure 5a presents linear trends for each of the UTCI heat
change in air temperature and MRT (2.2 and 2.8
C) dur- stress drivers. Since 1980 the Caribbean region has been
ing the year is less marked across the Lesser Antilles than warming by up to 0.18
Cdecade−1, with some areas such
the Greater Antilles (3.6 and 4.8
C). As for relative as the northwestern Cuba, the Yucat
an peninsula and
humidity, a primary maximum is observed during July– the northern part of South America experiencing
August (81.7–81.8%), followed first by a small reduction increases in air temperature maxima above
and then a slight increase in October (81.3%), after which 0.25
Cdecade−1. MRT has also been increasing in the
it descends completing the annual cycle in February– region, with trends up to 0.24
Cdecade−1 in the minima
March. Wind speeds follow the same distribution as the in the Lesser Antilles. The region has become drier with
Caribbean- and the Greater Antilles-averaged ones, with minima of relative humidity decreasing up to 0.53%dec-
higher values throughout the year (6.1–7.9 ms−1). ade−1. As for the wind speed, this has been increasing in
some parts of the Caribbean region, such as Cuba,
Jamaica, and Hispaniola, and decreasing in others, for
3.3 | Trends in heat stress example, the Lesser Antilles and southern Florida. This is
particularly evident in wind speed minima. Trends by
Linear trends of the UTCI are calculated for the Carib- season indicate ASO as the season when air temperature
bean region (Figure 4a), and an increasing trend for the and MRT have increased at the highest rates across most
year-round heat stress is observed. During the 1980–2019 of the Caribbean, and wind speed has decreased more
historical period, UTCI minima have increased more than 0.15 ms−1decade−1 in the Lesser Antilles; similar
than 0.2
Cdecade−1 in 81% of the Caribbean. UTCI max- rates and trends are observed in southern Florida during
ima have increased more than 0.2
Cdecade−1 in 49% of FMA (Figures S2–S5).
the region with increases above 0.3
Cdecade−1 being Figure 5b illustrates the correlation between mean
observed in areas such as southern Florida and the Lesser yearly averages of the UTCI and mean yearly averages
 10970088, 2023, 1, Downloaded from https://rmets.onlinelibrary.wiley.com/doi/10.1002/joc.7774 by Aristotle University Of, Wiley Online Library on [28/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
414 DI NAPOLI ET AL.

°C/10yr < 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 > 0.45

Caribbean Region
Greater Antilles
Lower Antilles
1

-1

-1

UTCI anomaly (°C)

1

-1

-1

-1

(a) (b)

2000s
1980s
1990s

2010s
F I G U R E 4 Spatiotemporal trends of the UTCI in the Caribbean region. (a) Decadal trends, as represented by Sen's slope coefficients, in
minimum, maximum and mean UTCI for whole years and by seasons over the 1980–2019 period. Grey areas indicate grid cells where trends
are not statistically significant according to the Mann–Kendall test (p ≥ .05). (b) Decadal anomalies of mean UTCI to the 1981–2010
climatological baseline by regions. Error bars represent one standard deviation [Colour figure can be viewed at wileyonlinelibrary.com]

of each of its drivers. The UTCI and 2 m air tempera- the Greater Antilles, particularly the Bahamas and
ture (resp. MRT) were found to be strongly positively Puerto Rico, whereas weak to strong negative correla-
correlated during the 1980–2019 historical period, with tions are observed in the southern part of the Lesser
an average correlation coefficient r of 0.91 (resp. 0.92) Antilles.
in more than 89% (resp. 94%) of the Caribbean region.
On the contrary, the UTCI and wind speed were found
to be negatively correlated. The magnitude of the corre- 3.5 | Heat danger and heat stress
lation is strongly negative for the Lesser Antilles
(r = −.77 ± .01) and moderately negative for the Greater HI climatology derived from ERA5 data for the 1980–
Antilles (r = −.45 ± .01). UTCI and relative humidity 2019 historical period shows yearly HI maxima mostly
trends were found to be both positively and negatively associated to caution and extreme caution, and it iden-
correlated depending on the sub-regional location. tifies ASO as the season when exposure to heat may be
Weak to very strong positive correlations are found for more detrimental (Figure 6a,b). In ASO daily maxima, an
 10970088, 2023, 1, Downloaded from https://rmets.onlinelibrary.wiley.com/doi/10.1002/joc.7774 by Aristotle University Of, Wiley Online Library on [28/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
DI NAPOLI ET AL. 415

(a) Air temperature Relative humidity Wind speed MRT

(°C/10yr) (%/10yr) (m/s/10yr) (°C/10yr)
0 0. 1 0.2 0.3 0.4 0.5 0.6 -3 -2.5 -2 -1.5 -1 -0.5 0 0.5 1 -0.2 -0.1 0 0.1 0.2 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
Minimum
Maximum
Mean

(b) UTCI and Air UTCI and Relative

humidity UTCI and Wind speed UTCI and MRT
temperature
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
Minimum
Maximum
Mean

F I G U R E 5 Spatiotemporal trends of heat stress drivers in the Caribbean region. (a) Decadal trends, as represented by Sen's slope
coefficients, in minimum, maximum and mean yearly averages of 2 m air temperature, relative humidity, 10 m wind speed and MRT over
the 1980–2019 historical period. Grey areas indicate grid cells where trends are not statistically significant according to the Mann–Kendall
test (p ≥ .05). (b) Pearson correlation between mean yearly averages of the UTCI and its drivers over the 1980–2019 historical period.
Geographical areas with not statistically significant correlation (p ≥ .05) are shown in grey [Colour figure can be viewed at
wileyonlinelibrary.com]

extreme caution area spans from the Greater Antilles to Aruba and Curacao. HI values associated with caution
Florida, the Yucat
an peninsula and part of the southern dominate the rest of the region for both daily maxima
American Caribbean coastline, including the islands of and minima. Conditions of no heat-related danger are
 10970088, 2023, 1, Downloaded from https://rmets.onlinelibrary.wiley.com/doi/10.1002/joc.7774 by Aristotle University Of, Wiley Online Library on [28/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
416 DI NAPOLI ET AL.

0 26.7 32.2 39.4

HI (°C)

Minimum Maximum Mean

(a)

Year

(b)
FMA
MJJ
ASO
NDJ

(c) 35
Heat Index (°C)

30

25

Caribbean Region
20
Greater Antilles
Lesser Antilles
15
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

F I G U R E 6 Yearly (a), seasonal (b), and monthly (c) heat index climatology for the 1980–2019 historical period. Background colours
refer to the heat index danger category scale at the top of the figure [Colour figure can be viewed at wileyonlinelibrary.com]

mostly prevalent in NDJ and FMA. As for the monthly experience the maximum peak in the HI later in the year,
climatology, the HI follows a unimodal distribution that in September.
has a minimum in January–February and a maximum in Monthly frequencies of HI danger categories are con-
August (Figure 6c). For most of the year the Greater sistent with above described monthly climatologies
Antilles experience higher maximum and mean HI (Figure S6). Daily minima associated with the caution
values than both the entire averaged Caribbean region category occur most frequently between July and October
and the Lesser Antilles. On the contrary, the Lesser Antil- across the Caribbean (66–100% of days) and the Greater
les are exposed to higher minimum HI values than the Antilles (69–99% of days), and between May and
rest of the region. Furthermore, the Lesser Antilles November (70–98% of days) in the Lesser Antilles. In
 10970088, 2023, 1, Downloaded from https://rmets.onlinelibrary.wiley.com/doi/10.1002/joc.7774 by Aristotle University Of, Wiley Online Library on [28/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
DI NAPOLI ET AL. 417

T A B L E 2 Joint distribution in percentage of UTCI stress categories and HI danger categories across the whole Caribbean region during
the 1980–2019 historical period for ASO maxima

HI/UTCI No danger Caution Extreme caution Danger Extreme danger

No thermal stress 0 0 0 0 0
Moderate heat stress 0.8 [0.5; 0] 45.5 [7.9; 97.2] 19.2 [9.2; 0.2] 0 0
Strong heat stress 0.2 [0.1; 0] 12.4 [7.5; 1.5] 20.6 [74; 1.2] 0 0
Very strong heat stress 0 0 1.3 [0.9; 0] 0 0
Extreme heat stress 0 0 0 0 0

Note: Values in squared brackets refer to corresponding percentages across the Greater Antilles (left) and the Lesser Antilles (right).

daily maxima, extreme caution is the most frequent con- heat stress” and “extreme caution” (12.4%). For the
dition in August and September across the Caribbean Greater Antilles, the co-occurrence of “strong heat stress”
(56–68% of days) and between June and September (70– and “extreme caution” dominates (74%) whereas for the
99% of days) in the Greater Antilles. On the other hand, Lesser Antilles conditions of “moderate heat stress” and
days with HI maxima associated with caution dominate “caution” are more prevalent (97.2%).
the climatology of the Lesser Antilles throughout the
year, with extreme caution observed in no more than
42% of days. Caution is the most frequent category in 3.6 | Case study: The 2020 heat season
daily means in at least 7 months of the year, in agree-
ment with the HI monthly climatology, and reaches Figure 7 illustrates the record-breaking heat season of
extreme levels in up to 16% (resp. 6%) of days in August 2020 in terms of the UTCI, HI and 2 m air temperature,
(resp. September) in the Great Antilles (resp. the Lesser and of the months, April and September, when two heat-
Antilles). waves hit the Caribbean.
Positive trends are observed across the region for the In April 2020 monthly averages of daily maxima high-
HI during the 1980–2019 historical period (Figure S7). light, for Cuba, air temperatures between 33 and 39
C as
According to annual data, increases greater than well as conditions of strong to very strong heat stress in
0.2
Cdecade−1 occurred in all statistics (minima, maxima, the UTCI and of extreme caution in the HI. The excep-
and means) in more than 91% of the reanalysis grids that tionality of the April 2020 heatwave is reflected in the
compose the Caribbean region, with the highest trends anomalies to the 1980–2019 historical period. An area
experienced in the southern part of the Lesser Antilles, the stretching from southern Florida to Yucat
an through the
Yucat
an Peninsula, and the northern part of South Amer- Bahamas, Cuba, Jamaica and Cayman Islands, experi-
ica. Seasonal data reveal trends with statistically signifi- enced higher-than-average air temperatures and heat
cant increases greater than 0.45
Cdecade−1 in 43% of the with the UTCI and HI up to 6
C above their 1980–2019
region during ASO, and in southern Florida during FMA. values. Heat stress and heat danger were up to 1 category
In terms of anomalies to the 1981–2010 climatology base- higher than expected from the period, and air tempera-
line, these were, in the decade 2010–2019 and for region- tures exceeded the 99.5th percentile demonstrating the
averaged HI, between +0.5
C and +0.8
C with the latter exceptionality of the event.
achieved in ASO. In the Lesser Antilles, the HI has been In September 2020 conditions associated to strong
continuously increasing and, in the late rainy seasons of heat stress and extreme caution reached the Lesser Antil-
2010s, was +1.2
C higher than the 1980–1989 decade. On les. Here anomalies in UTCI and HI values were up to
the contrary, in the Greater Antilles and at the region 2
C and, although more contained than the anomalies in
level, an increase hiatus can be observed in FMA, MJJ and April, they brought heat stress and heat danger categories
NDJ during the 1990–1999 and 2000–2009 decades. up to 1 category with respect to the 1980–2019 historical
Table 2 shows the joint occurrence of UTCI stress cat- period. Air temperatures exceeded 30
C and the 95th per-
egories and HI danger categories for daily maxima in centile in the southern part of the Lesser Antilles.
ASO. When the exposure to heat is at its highest, condi-
tions of “moderate heat stress” occur simultaneously to
conditions of “caution” in 45.5% of the Caribbean region. 4 | DISCUSSION
Conditions of “strong heat stress” are associated to condi-
tions of “extreme caution” in 20.6% of the region, imme- This study presents the first assessment of human heat
diately followed by the joint occurrence of “moderate stress in the Caribbean region. Using climate reanalysis
 10970088, 2023, 1, Downloaded from https://rmets.onlinelibrary.wiley.com/doi/10.1002/joc.7774 by Aristotle University Of, Wiley Online Library on [28/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
418 DI NAPOLI ET AL.

(a) UTCI HI Air Temperature

9°C 26°C 32°C 38°C 46°C 0°C 26.7°C 32.2°C 39.4°C 51.7°C 12°C 21°C 30°C 39°C

April
September

(b) UTCI anomaly (degrees) HI anomaly (degrees) Air temperature anomaly (degrees)
-2°C -1°C 0°C 1°C 2°C 3°C 4°C 6°C
April
September

(c) UTCI anomaly (classes) HI anomaly (classes) Air temperature anomaly (perc.)
-1 0 1

99.5th
April

95th
September

F I G U R E 7 Bioclimate analysis of the heatwaves that affected the Caribbean region in April and September 2020. (a) Monthly means of
daily maxima for the UTCI, HI and 2 m air temperature. (b) Anomalies to the 1980–2019 historical period in
C. (c) Anomalies to the 1980–
2019 period in classes for the UTCI and HI, and with respect to climatological-defined percentiles for air temperature [Colour figure can be
viewed at wileyonlinelibrary.com]

datasets (ERA5 and ERA5-HEAT), it explores heat stress Considering the Universal Thermal Climate Index
bioclimatology and trends over the period from 1980 to (UTCI) as a representative measure for the heat stress
2019 from a spatiotemporal perspective. endured by the human body when exposed to outdoor
 10970088, 2023, 1, Downloaded from https://rmets.onlinelibrary.wiley.com/doi/10.1002/joc.7774 by Aristotle University Of, Wiley Online Library on [28/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
DI NAPOLI ET AL. 419

conditions, pan-Caribbean bioclimatology maps of the people living in the Caribbean chronically experience
UTCI denote August–September–October (ASO) as the moderate to very strong heat stress. Because of this, they
season when conditions of heat stress up to very strong may require less warming to achieve lethal heat stress
levels dominate the region, particularly the Greater Antil- levels and may incur in heat-related risks despite being
les, as well as the Yucat
an peninsula and the Caribbean used to high level of heat stress. The upward trend of the
coast of South America. This is reflected in the frequency UTCI also calls for the further investigation of sudden
of days when heat stress is achieved. During ASO, the onset heat events, namely heatwaves, during ASO as it
Greater and Lesser Antilles experience UTCI minima has been shown that heatwaves are more likely to occur
associated to moderate heat stress in up to 10% of days, during the rainy season and less likely to take place dur-
and UTCI maxima in the strong heat stress category ing the dry season (Ramirez-Beltran et al., 2017). Further-
between 30 and 99% of days. The analysis of climate vari- more, ASO falls in the hurricane season, and heatwaves
ables known to drive heat stress reveals the meteorologi- have been observed occurring immediately after major
cal reason why the UTCI reaches the highest values hurricanes (Matthews et al., 2019; NOAA NWS, 2021a).
during ASO. From March, 2 m air temperature, relative Compound heatwave-hurricane events may expose local
humidity, and mean radiant temperature (MRT) increase population to an exacerbated heat risk as hurricanes are
concurrently until June, when they plateau and remain responsible for major disruptions in essential services
at their maximum, with peaks in August and September. such as electricity and potable water that are critical for
During the same months, 10 m wind speed decreases, survivability during heat extremes (Méndez-L
azaro
reaching its lowest in September. From October, the joint et al., 2021). Heatwaves as compound hazards should
decrease of 2 m air temperature, relative humidity, and therefore also be investigated in the Caribbean region.
MRT and increase of wind speed drives the descending of This will assess whether heat and its extremes can be a
the UTCI which reaches a minimum in February. This limitation to sustainable development in the Caribbean
finding reflects the thermophysiological nature of the as is cold stress in high-latitude countries (Varentsov
UTCI. High temperatures, exposure to insolation, poor et al., 2020; Vinogradova, 2021). Considering regional
ventilation, and high levels of atmospheric moisture con- economic grouping such CARICOM (Caribbean Commu-
tribute synergistically to raise the heat content of the nity and Common Market; Jones et al., 2016b) could help
human body (McGregor and Vanos, 2018). towards this as well as extend the present analysis to
This study also identifies ASO as the season when the countries in the southern Caribbean (Guyana and
UTCI has been increasing the most since 1980. In the last Suriname).
decade (2010–2019) the Caribbean region has experi- With regards to heat stress drivers, the Caribbean
enced the highest heat stress levels of the past 40 years, region has become warmer and drier. These findings,
with the UTCI 0.51
C higher than the 1981–2010 clima- based on reanalysis data, agree with previous studies
tology baseline. In a global context, such an increase is based on observations from meteorological stations
lower than that observed at higher latitudes, for example, which also observed statistically significant increases in
in Europe (Di Napoli et al., 2018; Antonescu et al., 2021). air temperature during the wet season in Barbados and
Long-term observations from Poland, however, indicate a Tobago (Singh, 1997; Peterson et al., 2002; Aguilar
decadal trend like the Caribbean one for mean UTCI et al., 2005; Pérez and Jury, 2012; Stephenson et al., 2014;
values during the Northern Hemisphere's summer season Beharry et al., 2015; Dookie et al., 2018; Cavazos
(0.55 ± 0.10
C) (Kuchcik et al., 2021). This supports the et al., 2019; Mohan et al., 2020).
importance of the subcontinental/subregional analysis Decadal trends in MRT indicate that the total radia-
here presented. It is worth noting that high rates in the tion (solar and thermal) irradiating a human subject
increase of the UTCI (0.45
Cdecade−1) are observed in placed in an outdoor environment has been increasing.
southern Florida during the Caribbean dry season, specif- An upward trend has been found for solar radiation in
ically FMA. This agrees with the decreasing trends in Puerto Rico (Jury, 2015). As for the thermal component
cold stress observed at higher latitudes in the winter/ of MRT, which refers to radiation emitted and absorbed
early spring season (Błażejczyk and Twardosz, 2010; Var- by land surface, clouds, and other components of the
entsov et al., 2020; Kuchcik et al., 2021; atmosphere–land system, cloud cover data have shown
Vinogradova, 2021). that cloudiness has been increasing in the Caribbean
Increasing exposure to extreme heat has been recently region, especially at night, thus blocking more thermal
identified as a slow onset event with negative implica- radiation from land surface and contributing to a net
tions worldwide but particularly in tropical regions heating effect (Singh, 1997; Yu et al., 2017).
(Oppermann et al., 2021). The bioclimatology analysis Spatial patterns and decadal trends of wind speed are
here presented (UTCI maxima; Figure 2) reveals that in agreement with previous findings (Chadee and
 10970088, 2023, 1, Downloaded from https://rmets.onlinelibrary.wiley.com/doi/10.1002/joc.7774 by Aristotle University Of, Wiley Online Library on [28/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
420 DI NAPOLI ET AL.

Clarke, 2014). The increase observed in the 1980–2019 linear dependencies on air temperature). This suggests
historical period across the western Caribbean Sea is the upward trend in heat stress might be driven by
expected to continue in the future (Angeles et al., 2010; increases in air temperature and radiation and decreases
Costoya et al., 2019). To the authors' knowledge, the in wind speed. The latter is particularly evident in the
decrease in wind speed minima observed for the Lesser Lesser Antilles. The correlation between the UTCI and
Antilles has not yet been documented in the literature. relative humidity is both positive and negative, albeit not
To investigate this, future studies can explore wind speed statistically significant in most of the Caribbean region. It
records from weather stations and compare them against is worth noting that the correlation is positive in areas
data from ERA5 reanalysis. This would expand and where relative humidity has no significant decadal trend.
update existing research which reports a good agreement Future research could address this topic by investigating
between weather station observations and data from trends in relative humidity across the Caribbean and
ECMWF previous reanalyses, namely ERA-40 and ERA- their connection to heat stress in the region.
Interim (Jury, 2012; Pérez and Jury, 2012; Chadee and Alongside the UTCI, this study investigates heat-
Clarke, 2014). It is acknowledged that although reanaly- induced danger as expressed by the HI. The 1980–2019
sis datasets have many advantages they may also suffer climatology of the HI, here computed from ERA5 reana-
from biases due to (a) the inherent uncertainties of the lysis data, agrees with the spatiotemporal distribution of
numerical weather prediction model used to generate the HI reported in current literature (Ramirez-Beltran
them, and (b) their nature as a collection of grid cells, et al., 2017; Angeles-Malaspina et al., 2018). To comple-
that is, values averaged over an area (Parker, 2016). ment that, the monthly frequency of heat danger catego-
Biases can be identified by comparing the reanalysis ries is here provided and reveals dominant conditions of
against in situ measurements or observational/satellite caution for the Caribbean, with HI maxima associated to
bias-corrected datasets. Applied to UTCI input variables, extreme caution in up to 42 and 99% of days in the Lesser
the comparison would reveal any bias and assess how and Greater Antilles, respectively. It is also found, for the
this may impact the UTCI in turn. This validation first time, that the UTCI and the HI provide complemen-
approach, which can be explored in future studies, would tary information on the effects of heat on human health.
allow ERA5-HEAT to be corrected where needed, thus Conditions associated to high UTCI values occur simulta-
improving the skill of potential heat stress forecasts neously to conditions characterized by high HI levels.
across the region. This approach has already proved suc- This suggests that dangerous heat disorders occur when
cessful for UTCI-based forecasting systems currently physiological responses to heat are most stressful.
operational in Europe (Di Napoli et al., 2021b). In the 2010–2019 decade, a positive HI change
In this study, reanalysis data are aggregated by geo- (+1.2
C) is found with respect to the 1981–2010 climatol-
graphical zones—the entire Caribbean, the Greater Antil- ogy baseline. Trends greater than 0.45
Cdecade−1 are
les, and the Lesser Antilles—to allow for comparison observed during ASO across most of the Caribbean basin,
between region-wide and subregional derived trends. All with the southern Lesser Antilles experiencing statisti-
three geographical zones include both grid cells on water cally significant increases throughout the year. These
bodies, such as the sea, and grid cells on land. As for the findings are in agreement with previous literature (Lee
latter, they may correspond to coastal or hinterland areas and Brenner, 2015). With regards to the processes and
according to the geographical location and/or size of each patterns responsible for HI trends, HI intensification in
island. Furthermore, some reanalysis grid cells over the Caribbean has been shown to be driven by changes
water may not resolve the land cover from subgrid scale in sea level pressure, sinking dry air enhancement, as
size Caribbean islands like Barbuda or Cayman Island. well as warm advection strengthening and weaker cold
This may influence how heat metrics and heat stress advection (Angeles-Malaspina et al., 2018). These pro-
drivers over islands are represented in the reanalysis. cesses were found to influence humidity and wind speed
Because of this, the representativeness of ERA5-HEAT in patterns, especially during the Caribbean wet season
marine environments, outside populated islands, may be (Angeles et al., 2010; Campbell et al., 2011), and correla-
worth investigating in future studies. tions between air temperature variability and the Atlantic
The spatiotemporal correlation between heat stress multidecadal oscillation (AMO) signal of the North
and heat stress drivers has been assessed via the Pearson Atlantic surface sea temperatures were demonstrated
correlation coefficient. During the 1980–2019 historical (Stephenson et al., 2014). As for the UTCI, evidence on
period the UTCI is found to be negatively correlated with the association between UTCI trends and circulation pat-
wind speed and positively correlated with MRT and 2 m terns is currently missing for the Caribbean. Preliminary
air temperature (Pappenberger et al. (2015) and Zare studies unveiling the role of the North Atlantic Oscilla-
et al. (2018) have previously shown the UTCI shows some tion (NAO) on the variability of the UTCI in Eastern
 10970088, 2023, 1, Downloaded from https://rmets.onlinelibrary.wiley.com/doi/10.1002/joc.7774 by Aristotle University Of, Wiley Online Library on [28/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
DI NAPOLI ET AL. 421

Europe (Głogowski et al., 2020) call for similar investiga- highest, most frequent, and widespread, particularly in the
tions at lower latitudes. maxima. Strong to very strong heat stress is experienced in
During the heatwaves that affected the Caribbean the Greater Antilles, the Yucat
an peninsula, southern
region in 2020, the UTCI and HI raised up to 1 category Florida, and the southern American Caribbean coastline.
higher than their historical averages. This exposed local Besides a spatial variation across the Caribbean, heat
populations to conditions of heat stress and danger they stress has been witnessing a time variation too. From 1980
might not be physiologically used to, with potential con- to 2019, the UTCI has increased more than 0.2
Cdecade−1
sequences to their health. The importance of assessing in almost half of the region, reaching marked above-
the links between extreme heat episodes and mortality climatology values especially in the last 10 years. The ASO
has recently been demonstrated for San Juan, Puerto season is characterized by the strongest upwards trends,
Rico, and contextualized in terms of local vulnerability with rates greater than 0.45
Cdecade−1 in southern Flor-
(Méndez-L
azaro et al., 2015; 2018a; 2018b). Future stud- ida and the Lesser Antilles.
ies can investigate the effects of heat on public health Data from ECMWF ERA5 and ERA5-HEAT reana-
across different Caribbean nations. Complementary to lyses reveal that the Caribbean region has become
that, research efforts could focus on characterizing the warmer and drier, and it has been receiving increasing
years when heat stress anomalies to a climatological radiation. Being temperature, relative humidity, radia-
baseline have been observed or heatwaves have been tion, and wind speed, the drivers of human heat stress, a
reported and on identifying the corresponding driving spatiotemporal correlation between the UTCI and such
mechanisms. Years with the highest incidence of HI- drivers is here investigated and found to be diverse. Dur-
defined extreme events, for instance, have been found to ing the 1980–2019 period heat stress has increased as air
coincide with the cool phase of the El Niño–Southern temperature and radiation have increased, and wind
Oscillation (ENSO; Ramirez-Beltran et al., 2017). speed has decreased. The latter is particularly evident in
Bringing together health and environmental informa- the Lesser Antilles and during ASO. Given the spatial
tion would provide local decision makers with evidence dominance of the marine environment in the subregion,
on how to address and mitigate heat-related health haz- future studies could evaluate reanalysis data against sta-
ards, also via integrated risk monitoring and early warn- tion data in the Caribbean. The correlation between the
ing systems (WHO, 2018). To the authors' knowledge, UTCI and relative humidity is complex exhibiting a mix
Puerto Rico and Saint Lucia are the only Caribbean of positive and negative correlations, which cannot be
nations with an operational early warning system and an easily explained and requires for further investigation.
active response plan for extreme heat events, respectively The present study is also the first to provide a multi-
(Government of Saint Lucia, 2006; NOAA NWS, 2016). index heat assessment of the Caribbean region. Condi-
The increase in heat stress observed across the whole tions of heat danger, as represented by the HI, have
Caribbean calls for the establishment of response plans intensified since 1980 (+1.2
C) and are found to occur
in more nations. Furthermore, heat stress information as simultaneously to conditions of heat stress, as indicated
delivered by the UTCI could help strengthen local meteo- by the UTCI. This supports the idea that the more physio-
climate services (Trotman et al., 2018). A UTCI-based logical responses the human body must put in place to
forecasting system, for example, could provide an added contrast an extremely hot environment, the more danger-
value to the experimental seasonal heat outlook that is ous heat disorders it may develop. The exceptional 2020
currently run by the Caribbean Institute for Meteorology heat season, here investigated as a case study, exposed
and Hydrology (CIMH) and complement its surface air local population to above-average heat stress and danger.
temperature-based forecasts. These findings demonstrate that the Caribbean region
is witnessing a gradual intensification of heat stress on
top of rapid onset extreme heat events. Region-wide maps
5 | C ON C L U S I ON S of heat stress and UTCI trends provide an information
platform that highlights when and where preparedness
Exposure to heat stress based on the UTCI has been and response plans to heat are needed. Further studies
determined and characterized in trends, drivers, and that integrate heat stress data with data on social sectors
extremes across the Caribbean region for the first time. such as public health could help quantify the impacts of
Using 40 years of meteorological data from the heat in the Caribbean.
ECMWF ERA5-HEAT reanalysis, the Caribbean biocli-
mate has been investigated via UTCI maps computed for A C KN O WL ED G EME N T S
the dry (NDJ, FMA) and rainy seasons (MJJ, ASO). ASO This work was supported by a Caribbean Frontiers of Sci-
was identified as the season when heat stress is the ence Follow-on Grant (FOS\R1\191010) from the Royal
 10970088, 2023, 1, Downloaded from https://rmets.onlinelibrary.wiley.com/doi/10.1002/joc.7774 by Aristotle University Of, Wiley Online Library on [28/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
422 DI NAPOLI ET AL.

Society. Claudia Di Napoli was also supported by a Well- Island: Trinidad. Theoretical and Applied Climatology, 122,
come Trust grant (209734/Z/17/Z). Theodore Allen 783–797.
thanks the Climate staff at the Caribbean Institute for Biasutti, M., Sobel, A.H., Camargo, S.J. and Creyts, T.T. (2011) Pro-
jected changes in the physical climate of the Gulf Coast and
Meteorology and Hydrology for their support and effort
Caribbean. Climatic Change, 112, 819–845.
towards building heat stress awareness throughout the
Blazejczyk, K., Epstein, Y., Jendritzky, G., Staiger, H. and Tinz, B.
Caribbean, as well as for the use of station observations (2012) Comparison of UTCI to selected thermal indices. Inter-
to aid in preliminary data analysis. national Journal of Biometeorology, 56, 515–535.
_
BłaZejczyk, K., Jendritzky, G., Bröde, P., Fiala, D., Havenith, G.,
FUNDING INFORMATION Epstein, Y., Psikuta, A. and Kampmann, B. (2013) An introduc-
Royal Society, FOS\R1\191010; Wellcome Trust, 209734/ tion to the Universal Thermal Climate Index (UTCI). Geogra-
Z/17/Z. phia Polonica, 86, 5–10.
_
BłaZejczyk, K. and Twardosz, R. (2010) Long-term changes of bio-
climatic conditions in Cracow (Poland). In: The Polish Climate
CONFLICT OF INTEREST in the European Context: An Historical Overview. Dordrecht:
The authors declare no potential conflict of interest. Springer.
Brode, P., Fiala, D., Blazejczyk, K., Holmer, I., Jendritzky, G.,
ORCID Kampmann, B., Tinz, B. and Havenith, G. (2012) Deriving the
Claudia Di Napoli https://orcid.org/0000-0002-4901- operational procedure for the Universal Thermal Climate Index
3641 (UTCI). International Journal of Biometeorology, 56, 481–494.
Theodore Allen https://orcid.org/0000-0003-4339-0267 Campbell, J.D., Taylor, M.A., Stephenson, T.S., Watson, R.A. and
Whyte, F.S. (2011) Future climate of the Caribbean from a
azaro https://orcid.org/0000-0002-
Pablo A. Méndez-L

regional climate model. International Journal of Climatology,
3059-3295
31, 1866–1878.
Florian Pappenberger https://orcid.org/0000-0003- Cavazos, T., Luna-Niño, R., Cerezo-Mota, R., Fuentes-Franco, R.,
1766-2898 Méndez, M., Pineda Martínez, L.F. and Valenzuela, E. (2019)
Climatic trends and regional climate models intercomparison
R EF E RE N C E S over the CORDEX-CAM (Central America, Caribbean, and
Aguilar, E., Peterson, T.C., Obando, P.R., Frutos, R., Retana, J.A., Mexico) domain. International Journal of Climatology, 40,
Solera, M., Soley, J., García, I.G., Araujo, R.M., Santos, A.R., 1396–1420.
Valle, V.E., Brunet, M., Aguilar, L., Álvarez, L., Bautista, M., CDS. (2018) ERA5 hourly data on single levels from 1979 to present.
Castañ
on, C., Herrera, L., Ruano, E., Sinay, J.J., S
anchez, E., CDS. https://doi.org/10.24381/cds.adbb2d47.
Oviedo, G.I.H., Obed, F., Salgado, J.E., V
azquez, J.L., Baca, M., CDS. (2020) Thermal comfort indices derived from ERA5 reanalysis.
Gutiérrez, M., Centella, C., Espinosa, J., Martínez, D., Olmedo, B., CDS. https://doi.org/10.24381/cds.553b7518.
Espinoza, C.E.O., Núñez, R., Haylock, M., Benavides, H. and Chadee, X.T. and Clarke, R.M. (2014) Large-scale wind energy
Mayorga, R. (2005) Changes in precipitation and temperature potential of the Caribbean region using near-surface reanalysis
extremes in Central America and northern South America, 1961– data. Renewable and Sustainable Energy Reviews, 30, 45–58.
2003. Journal of Geophysical Research, 110, D23107. Chen, A.A. and Taylor, M.A. (2002) Investigating the link between
Alduchov, O.A. and Eskridge, R.E. (1996) Improved Magnus' form early season Caribbean rainfall and the El Niño+1 year. Inter-
approximation of saturation vapor pressure. Journal of Applied national Journal of Climatology, 22, 87–106.
Meteorology, 35, 601–609. Climate Studies Group Mona. (2020) The state of the Caribbean climate.
Angeles, M., Gonz
alez, J., Erickson Iii, D. and Hern
andez, J. (2010) Produced for the Caribbean Development Bank. Available at:
The impacts of climate changes on the renewable energy https://www.caribank.org/sites/default/files/publication-resources/
resources in the Caribbean region. Journal of Solar Energy The%20State%20of%20the%20Caribbean%20Climate%20Report.pdf.
Engineering, 132, 031009. Coccolo, S., Kämpf, J., Scartezzini, J.-L. and Pearlmutter, D. (2016)
Angeles-Malaspina, M., Gonz
alez-Cruz, J.E. and Ramírez- Outdoor human comfort and thermal stress: a comprehensive
Beltran, N. (2018) Projections of heat waves events in the intra- review on models and standards. Urban Climate, 18, 33–57.
Americas region using multimodel ensemble. Advances in Costoya, X., Decastro, M., Santos, F., Sousa, M. and G
omez-
Meteorology, 2018, 1–16. Gesteira, M. (2019) Projections of wind energy resources in the
Antonescu, B., Marmureanu, L., Vasilescu, J., Marin, C., Andrei, S., Caribbean for the 21st century. Energy, 178, 356–367.
Boldeanu, M., Ene, D. and Ţilea, A. (2021) A 41-year bioclima- Dee, D., Fasullo, J., Shea, D., Walsh, J. and National Center for
tology of thermal stress in Europe. International Journal of Cli- Atmospheric Research Staff. (2016) The climate data guide: atmo-
matology., 41, 3934–3952. spheric reanalysis: overview & comparison tables. Available at:
Aron, J.L., Corval
an, C.F. and Philippeaux, H. (2003) Climate vari- https://climatedataguide.ucar.edu/climate-data/atmospheric-
ability and change and their health effects in the Caribbean: reanalysis-overview-comparison-tables [Accessed June 2021].
information for adaptation planning in the health sector. Di Napoli, C., Barnard, C., Prudhomme, C., Cloke, H.L. and
Geneva: World Health Organization. Pappenberger, F. (2021a) ERA5-HEAT: a global gridded histori-
Beharry, S.L., Clarke, R.M. and Kumarsingh, K. (2015) Variations cal dataset of human thermal comfort indices from climate rea-
in extreme temperature and precipitation for a Caribbean nalysis. Geoscience Data Journal, 8, 2–10.
 10970088, 2023, 1, Downloaded from https://rmets.onlinelibrary.wiley.com/doi/10.1002/joc.7774 by Aristotle University Of, Wiley Online Library on [28/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
DI NAPOLI ET AL. 423

Di Napoli, C., Hogan, R.J. and Pappenberger, F. (2020) Mean radi- Keeley, S., Laloyaux, P., Lopez, P., Lupu, C., Radnoti, G.,
ant temperature from global-scale numerical weather predic- Rosnay, P., Rozum, I., Vamborg, F., Villaume, S. and
tion models. International Journal of Biometeorology, 64, 1233– Thépaut, J.N. (2020) The ERA5 global reanalysis. Quarterly
1245. Journal of the Royal Meteorological Society, 146, 1999–2049.
Di Napoli, C., Messeri, A., Nov
ak, M., Rio, J., Wieczorek, J., Jones, P.D., Harpham, C., Burton, A. and Goodess, C.M. (2016a)
Morabito, M., Silva, P., Crisci, A. and Pappenberger, F. (2021b) Downscaling regional climate model outputs for the Caribbean
The Universal Thermal Climate Index as an operational fore- using a weather generator. International Journal of Climatology,
casting tool of human biometeorological conditions in Europe. 36, 4141–4163.
In: Applications of the Universal Thermal Climate Index (UTCI) Jones, P.D., Harpham, C., Harris, I., Goodess, C.M., Burton, A.,
in Biometeorology. Cham, Switzerland: Springer. Centella-Artola, A., Taylor, M.A., Bezanilla-Morlot, A.,
Di Napoli, C., Pappenberger, F. and Cloke, H.L. (2018) Assessing Campbell, J.D., Stephenson, T.S., Joslyn, O., Nicholls, K. and
heat-related health risk in Europe via the Universal Thermal Baur, T. (2016b) Long-term trends in precipitation and temper-
Climate Index (UTCI). International Journal of Biometeorology, ature across the Caribbean. International Journal of Climatol-
62, 1155–1165. ogy, 36, 3314–3333.
Di Napoli, C., Pappenberger, F. and Cloke, H.L. (2019) Verification Jury, M.R. (2012) Representation of the Caribbean mean diurnal
of heat stress thresholds for a health-based heat-wave cycle in observation, reanalysis, and CMIP3 model datasets.
definition. Journal of Applied Meteorology and Climatology, 58, Theoretical and Applied Climatology, 107, 313–324.
1177–1194. Jury, M.R. (2015) Climatic trends in Puerto Rico: observed and pro-
Dookie, N., Chadee, X.T. and Clarke, R.M. (2018) Trends in jected since 1980. Climate Research, 66, 113–123.
extreme temperature and precipitation indices for the Carib- Karmalkar, A.V., Taylor, M.A., Campbell, J., Stephenson, T.,
bean small islands: Trinidad and Tobago. Theoretical and New, M., Centella, A., Benzanilla, A. and Charlery, J. (2013) A
Applied Climatology, 136, 31–44. review of observed and projected changes in climate for the
Ebi, K.L., Lewis, N.D. and Corvalan, C. (2006) Climate variability islands in the Caribbean. Atmosfera, 26, 283–309.
and change and their potential health effects in small Island Kendall, M.G. (1975) Rank Correlation Methods, 4th edition.
states: information for adaptation planning in the health sector. London: Charles Griffin.
Environmental Health Perspectives, 114, 1957–1963. Kjellstrom, T., Freyberg, C., Lemke, B., Otto, M. and Briggs, D.
Evans, J.D. (1996) Straightforward Statistics for the Behavioral Sci- (2018) Estimating population heat exposure and impacts on
ences. Pacific Grove, CA: Thomson Brooks/Cole. working people in conjunction with climate change. Interna-
Fiala, D., Havenith, G., Brode, P., Kampmann, B. and tional Journal of Biometeorology, 62, 291–306.
Jendritzky, G. (2012) UTCI-Fiala multi-node model of human Kotcher, J., Maibach, E., Miller, J., Campbell, E., Alqodmani, L.,
heat transfer and temperature regulation. International Journal Maiero, M. and Wyns, A. (2021) Views of health professionals
of Biometeorology, 56, 429–441. on climate change and health: a multinational survey study.
Giannini, A., Kushnir, Y. and Cane, M.A. (2000) Interannual vari- Lancet Planet Health, 5, e316–e323.
ability of Caribbean rainfall, ENSO, and the Atlantic Ocean. Kotharkar, R. and Ghosh, A. (2022) Progress in extreme heat man-
Journal of Climate, 13, 297–311. agement and warning systems: a systematic review of heat-
Głogowski, A., Bry
s, K. and Bry
s, T. (2020) Influence of NAO on health action plans (1995–2020). Sustainable Cities and Society,
forming the UTCI index in Kłodzko. Theoretical and Applied 76, 103487.
Climatology, 142, 1555–1567. _
Kuchcik, M., BłaZejczyk, K. and Hala
s, A. (2021) Long-term
Government of Saint Lucia. (2006) Response plan for extreme heat changes in hazardous heat and cold stress in humans: multi-
event. Government of Saint Lucia. city study in Poland. International Journal of Biometeorology,
Hall, T.C., Sealy, A.M., Stephenson, T.S., Kusunoki, S., Taylor, M. 65, 1567–1578.
A., Chen, A.A. and Kitoh, A. (2012) Future climate of the Lallo, C.H.O., Cohen, J., Rankine, D., Taylor, M., Cambell, J. and
Caribbean from a super-high-resolution atmospheric general Stephenson, T. (2018) Characterizing heat stress on livestock using
circulation model. Theoretical and Applied Climatology, 113, the temperature humidity index (THI)—prospects for a warmer
271–287. Caribbean. Regional Environmental Change, 18, 2329–2340.
Havenith, G. and Fiala, D. (2015) Thermal indices and thermophy- Lee, D. and Brenner, T. (2015) Perceived temperature in the course
siological modeling for heat stress. Comprehensive Physiology, 6, of climate change: an analysis of global heat index from 1979 to
255–302. 2013. Earth System Science Data, 7, 193–202.
Havenith, G., Fiala, D., Blazejczyk, K., Richards, M., Brode, P., Macpherson, C.C. and Akpinar-Elci, M. (2015) Caribbean heat
Holmer, I., Rintamaki, H., Benshabat, Y. and Jendritzky, G. threatens health, well-being and the future of humanity. Public
(2012) The UTCI-clothing model. International Journal of Bio- Health Ethics, 8, 196–208.
meteorology, 56, 461–470. Mann, H.B. (1945) Nonparametric tests against trend. Econome-
Hersbach, H., Bell, B., Berrisford, P., Hirahara, S., Hor
anyi, A., trica: Journal of the Econometric Society, 13, 245–259.
Muñoz-Sabater, J., Nicolas, J., Peubey, C., Radu, R., Matthews, T., Wilby, R.L. and Murphy, C. (2019) An emerging trop-
Schepers, D., Simmons, A., Soci, C., Abdalla, S., Abellan, X., ical cyclone–deadly heat compound hazard. Nature Climate
Balsamo, G., Bechtold, P., Biavati, G., Bidlot, J., Bonavita, M., Change, 9, 602–606.
Chiara, G., Dahlgren, P., Dee, D., Diamantakis, M., McGregor, G.R., Bessmoulin, P., Ebi, K. and Menne, B. (2015) Heat-
Dragani, R., Flemming, J., Forbes, R., Fuentes, M., Geer, A., waves and Health: Guidance on Warning-System Development.
Haimberger, L., Healy, S., Hogan, R.J., H
olm, E., Janiskov
a, M., Geneva, Switzerland: WMOP.
 10970088, 2023, 1, Downloaded from https://rmets.onlinelibrary.wiley.com/doi/10.1002/joc.7774 by Aristotle University Of, Wiley Online Library on [28/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
424 DI NAPOLI ET AL.

McGregor, G.R. and Vanos, J.K. (2018) Heat: a primer for public Parker, W.S. (2016) Reanalyses and observations: What's the dif-
health researchers. Public Health, 161, 138–146. ference? Bulletin of the American Meteorological Society, 97,
McLean, N.M., Stephenson, T.S., Taylor, M.A. and Campbell, J.D. 1565–1572.
(2015) Characterization of future Caribbean rainfall and Pérez, C.R. and Jury, M.R. (2012) Spatial and temporal analysis of
temperature extremes across rainfall zones. Advances in climate change in Hispañola. Theoretical and Applied Climatol-
Meteorology, 2015, 1–18. ogy, 113, 213–224.
Méndez-L
azaro, P.A., Bernhardt, Y.M., Calo, W.A., Pacheco Peterson, T.C., Taylor, M.A., Demeritte, R., Duncombe, D.L.,
Díaz, A.M., García-Camacho, S.I., Rivera-Lugo, M., Acosta- Burton, S., Thompson, F., Porter, A., Mercedes, M., Villegas, E.,
Pérez, E., Pérez, N. and Ortiz-Martínez, A.P. (2021) Environ- Semexant Fils, R., Klein Tank, A., Martis, A., Warner, R.,
mental stressors suffered by women with gynecological cancers Joyette, A., Mills, W., Alexander, L. and Gleason, B. (2002)
in the aftermath of hurricanes Irma and María in Puerto Rico. Recent changes in climate extremes in the Caribbean region.
International Journal of Environmental Research and Public Journal of Geophysical Research: Atmospheres, 107, 1–9.
Health, 18, 11183. Ramirez-Beltran, N.D., Gonzalez, J.E., Castro, J.M.,
Méndez-L
azaro, P.A., Martínez-S
anchez, O., Méndez-Tejeda, R., Angeles, M., Harmsen, E.W. and Salazar, C.M. (2017) Anal-
Rodríguez, E., Morales, E. and Schmitt-Cortijo, N. (2015) ysis of the heat index in the Mesoamerica and Caribbean
Extreme heat events in San Juan Puerto Rico: trends and vari- region. Journal of Applied Meteorology and Climatology, 56,
ability of unusual hot weather and its possible effects on ecol- 2905–2925.
ogy and society. Journal of Climatology & Weather Forecasting, Rutty, M. and Scott, D. (2014) Thermal range of coastal tourism
3, 2. resort microclimates. Tourism Geographies, 16, 346–363.
Méndez-L
azaro, P.A., Muller-Karger, F.E., Otis, D., Mccarthy, M.J. Rutty, M. and Scott, D. (2015) Bioclimatic comfort and the thermal
and Rodríguez, E. (2018a) A heat vulnerability index to perceptions and preferences of beach tourists. International
improve urban public health management in San Juan, Puerto Journal of Biometeorology, 59, 37–45.
Rico. International Journal of Biometeorology, 62, 709–722. Sen, P.K. (1968) Estimates of the regression coefficient based on
Méndez-L
azaro, P.A., Pérez-Cardona, C.M., Rodríguez, E., Kendall's tau. Journal of the American Statistical Association,
Martínez, O., Taboas, M., Bocanegra, A. and Méndez-Tejeda, R. 63, 1379–1389.
(2018b) Climate change, heat, and mortality in the tropical Singh, B. (1997) Climate changes in the greater and southern Carib-
urban area of San Juan, Puerto Rico. International Journal of bean. International Journal of Climatology, 17, 1093–1114.
Biometeorology, 62, 699–707. Steadman, R.G. (1979) The assessment of sultriness. Part I: a
Mohan, S., Clarke, R.M. and Chadee, X.T. (2020) Variations in temperature-humidity index based on human physiology and
extreme temperature and precipitation for a Caribbean Island: clothing science. Journal of Applied Meteorology, 18,
Barbados (1969–2017). Theoretical and Applied Climatology, 861–873.
140, 1277–1290. Stennett-Brown, R.K., Jones, J.J.P., Stephenson, T.S. and Taylor, M.
Muniz-Castillo, A.I., Rivera-Sosa, A., Chollett, I., Eakin, C.M., A. (2017) Future Caribbean temperature and rainfall extremes
Andrade-Gomez, L., Mcfield, M. and Arias-Gonzalez, J.E. from statistical downscaling. International Journal of Climatol-
(2019) Three decades of heat stress exposure in Caribbean coral ogy, 37, 4828–4845.
reefs: a new regional delineation to enhance conservation. Sci- Stephenson, T.S., Vincent, L.A., Allen, T., Van Meerbeeck, C.J.,
entific Reports, 9, 11013. Mclean, N., Peterson, T.C., Taylor, M.A., Aaron-Morrison, A.P.,
Naranjo-Diaz, L.R. and Centella, A. (1998) Recent trends in the cli- Auguste, T., Bernard, D., Boekhoudt, J.R.I., Blenman, R.C.,
mate of Cuba. Weather, 53, 78–85. Braithwaite, G.C., Brown, G., Butler, M., Cumberbatch, C.J.M.,
NOAA. (2014) The heat index equation. Available at: https://www. Etienne-Leblanc, S., Lake, D.E., Martin, D.E., Mcdonald, J.L.,
wpc.ncep.noaa.gov/html/heatindex_equation.shtml. [Accessed Ozoria Zaruela, M., Porter, A.O., Santana Ramirez, M.,
June 2021]. Tamar, G.A., Roberts, B.A., Sallons Mitro, S., Shaw, A.,
NOAA NWS. (2016) Heat risk in Puerto Rico. Available at: https:// Spence, J.M., Winter, A. and Trotman, A.R. (2014) Changes in
www.weather.gov/sju/heat_risk [Accessed August 2021]. extreme temperature and precipitation in the Caribbean region,
NOAA NWS. (2021a) Tropical cyclone climatology. Available at: 1961–2010. International Journal of Climatology, 34, 2957–2971.
https://www.nhc.noaa.gov/climo/ [Accessed August 2021]. Taylor, M.A., Clarke, L.A., Centella, A., Bezanilla, A.,
NOAA NWS. (2021b) What is the heat index?. Available at: Stephenson, T.S., Jones, J.J., Campbell, J.D., Vichot, A. and
https://www.weather.gov/ama/heatindex [Accessed July Charlery, J. (2018) Future Caribbean climates in a world of ris-
2021]. ing temperatures: the 1.5 vs 2.0 dilemma. Journal of Climate,
Oppermann, E., Kjellstrom, T., Lemke, B., Otto, M. and Lee, J.K.W. 31, 2907–2926.
(2021) Establishing intensifying chronic exposure to extreme Taylor, M.A., Stephenson, T.S., Chen, A.A. and Stephenson, K.A.
heat as a slow onset event with implications for health, well- (2012) Climate change and the Caribbean: review and response.
being, productivity, society and economy. Current Opinion in Caribbean Studies, 40, 169–200.
Environmental Sustainability, 50, 225–235. Trotman, A., Mahon, R., Shumake-Guillemot, J., Lowe, R. and
Pappenberger, F., Jendritzky, G., Staiger, H., Dutra, E., Di Stewart-Ibarra, A.M. (2018) Strengthening climate Services for
Giuseppe, F., Richardson, D. and Cloke, H. (2015) Global fore- the Health Sector in the Caribbean. Bulletin of the World Meteo-
casting of thermal health hazards: the skill of probabilistic pre- rological Organization, 67, 2.
dictions of the Universal Thermal Climate Index (UTCI). Varentsov, M., Shartova, N., Grischenko, M. and Konstantinov, P.
International Journal of Biometeorology, 59, 311–323. (2020) Spatial patterns of human thermal comfort conditions in
 10970088, 2023, 1, Downloaded from https://rmets.onlinelibrary.wiley.com/doi/10.1002/joc.7774 by Aristotle University Of, Wiley Online Library on [28/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
DI NAPOLI ET AL. 425

Russia: present climate and trends. Weather, Climate, and Soci- Zare, S., Hasheminejad, N., Shirvan, H.E., Hemmatjo, R.,
ety, 12, 629–642. Sarebanzadeh, K. and Ahmadi, S. (2018) Comparing Universal
Vinogradova, V. (2021) Using the Universal Thermal Climate Index Thermal Climate Index (UTCI) with selected thermal
(UTCI) for the assessment of bioclimatic conditions in Russia. indices/environmental parameters during 12 months of the
International Journal of Biometeorology, 65, 1473–1483. year. Weather and Climate Extremes, 19, 49–57.
WHO. (2018) Climate change and health in small island developing
states: a WHO special initiative. Geneva: WHO.
Whyte, F.S., Taylor, M.A., Stephenson, T.S. and Campbell, J.D. SU PP O R TI N G I N F O RMA TI O N
(2008) Features of the Caribbean low level jet. International Additional supporting information can be found online
Journal of Climatology, 28, 119–128. in the Supporting Information section at the end of this
Wilkinson, E., Arvis, B., Mendler De Suarez, J., Weingartner, L., article.
Jaime, C., Grainger, N., Simonet, C., Bazo, J. and
Kruczkiewicz, A. (2021) Preparing for extreme weather in the
Caribbean: What role for forecast-based early action? London: How to cite this article: Di Napoli, C., Allen, T.,
Oversees Development Institute (ODI). Méndez-L
azaro, P. A., & Pappenberger, F. (2023).
WMO. (2021) State of the climate in Latin America and the Carib- Heat stress in the Caribbean: Climatology, drivers,
bean. Geneva: WMO. WMO No. 1272.
and trends of human biometeorology indices.
Yu, M., Gao, Q., Gao, C. and Wang, C. (2017) Extent of night warm-
ing and spatially heterogeneous cloudiness differentiate tempo-
International Journal of Climatology, 43(1),
ral trend of greenness in mountainous tropics in the new 405–425. https://doi.org/10.1002/joc.7774
century. Scientific Reports, 7, 1–10.