1
Detecting inhomogeneities in Caribbean and adjacent Caribbean temperature
2
data using sea surface temperatures
3 4
T. S. Stephenson1, C. M. Goodess2, M. R. Haylock3, A. A. Chen1 and M. A. Taylor1
5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
1. Climate Studies Group Mona, Physics Department, The University of the West Indies, Mona, Kingston 7, Jamaica, West Indies. 2. Climatic Research Unit, School of Environmental Sciences, University of East Anglia, Norwich, Norfolk, NR4 7TJ, United Kingdom. 3. PartnerRe Ltd., P.O. Box 857, Bellerivestrasse 36, CH-8034, Zurich, Switzerland.
2 1
Abstract
2
This study presents a systematic evaluation of the homogeneity of daily
3
surface temperature observations for the Caribbean and neighbouring regions on a
4
monthly timescale. The reference series are developed using adjacent sea surface
5
temperatures (SSTs). This novel approach is undertaken instead of the conventional
6
use of highly correlated nearby stations, given the sparse station network for the
7
Caribbean and adjacent Caribbean. The temperature data are from the regional climate
8
change workshops held for the Caribbean, and Central and Northern South America,
9
in 2001 and 2004 respectively, complemented with data from the National Climatic
10
Data Center, Caribbean Institute for Meteorology and Hydrology and Caribbean
11
meteorological stations. Correlations are used to explore the degree of association
12
between the maximum and minimum temperatures and SSTs, and homogeneity tests
13
are performed on their individual and difference series (e.g. maximum temperature
14
minus SSTs). The results suggest SSTs as a viable option for use in evaluating
15
homogeneity in the data sparse region of the Caribbean. Common statistically
16
significant change points identified across at least three stations are investigated using
17
composite analysis to determine links to large-scale atmospheric circulation patterns.
18
The study identifies two homogeneous periods from the analyses, i.e. 1970-92 and
19
1984-98, with the former used to reanalyze some extreme temperature trends for the
20
Caribbean and adjacent Caribbean. The results are found to be consistent with those
21
obtained from the 2001 Caribbean data workshop.
22 23 24 25
3 1
1.
Introduction
2
Socio-economic sectors within the Caribbean such as agriculture, fisheries and
3
tourism are inextricably linked to climate variability and change and are particularly
4
vulnerable to extreme events such as droughts, floods or temperature extremes
5
(Mahlung 2001). High quality daily data are necessary to assess the changes in
6
extremes that have occurred for the Caribbean and adjacent regions, and to make
7
projections regarding their occurrence in the future using statistical downscaling
8
techniques. To this end a data workshop was hosted in Kingston Jamaica in 2001
9
where daily precipitation and maximum and minimum temperatures across 30 stations
10
within the Caribbean, Belize and Florida (c.f. Figure 1) were brought together. The
11
workshop was invaluable in analyzing changes in climate extremes for the Caribbean
12
and adjacent Caribbean (i.e. Caribbean and neighbouring countries with a Caribbean
13
coastline found in a 55o – 90oW and 5o – 30oN domain) over the 1958-1999 period
14
(Peterson et al. 2002 hereafter PTD).
15
PTD employed a variety of quality control (qc) procedures on the data
16
including checks for physically unreasonable values, unreasonably long consecutive
17
occurrences of the same value, daily maximum temperatures less than minimum
18
temperatures, imperial to metric conversion problems, extreme outliers and very long
19
zero precipitation spells (Peterson et al. 1998a). No specific homogeneity tests were
20
run. Instead there was a qualitative evaluation of the calculated climate extremes
21
indices for each station and those with obvious discontinuities were excluded from
22
subsequent analyses for the given variable. Similar workshops were held for Africa
23
(Easterling et al. 2003, Mokssit 2003, New et al. 2006), Central and South America
24
(Aguilar et al. 2005 hereafter APO, Vincent et al. 2005, Haylock et al. 2006), and
25
Asia (Zhang et al. 2005, Peterson 2005, Klein Tank 2006, Sensoy et al. 2007).
4 1
The exclusion of specific homogeneity tests on the Caribbean data by PTD
2
may have been the result of the data sparsity across the Caribbean region. This paper
3
aims firstly to extend the work of PTD by attempting to systematically evaluate the
4
homogeneity of Caribbean surface temperature data. The surface temperature
5
observation network is essentially that used in PTD but has been expanded to include
6
additional data that were also available to the authors. Data from the National
7
Climatic Data Center, the Caribbean Institute for Meteorology and Hydrology, local
8
meteorological services, as well as data provided for the Central and northern South
9
America workshop (APO) are also utilized (c.f. Figure 1).
10
Even with the expanded dataset, the station density is still sparse and therefore
11
an impediment to the use of conventional homogeneity techniques which exploit
12
neighbouring stations (Peterson et al. 1998b). In light of this, the opportunity is taken
13
in this study to investigate the use monthly sea surface temperatures (SSTs) from the
14
1o-grid HadISST1 dataset (Rayner et al. 2003) to obtain reference series for each
15
station. That is, a reference time series is constructed by averaging SST gridpoint
16
values in 1o proximity to a given station. The difference series is then computed with
17
respect to both the maximum and minimum temperature series, and homogeneity
18
assessments are conducted using the RHTest Software (Wang and Feng 2004). This
19
approach is considered to be particularly useful for islands/coastal areas like the
20
Caribbean with a sparse network of daily temperature observations but where adjacent
21
SST gridded data are available. Additionally the use of the HadISST1 data provides a
22
high resolution independent dataset that correlates robustly with a number of station
23
observations in the Caribbean and adjacent Caribbean as shown in Section 3. As a
24
precursor to the primary analyses, quality checks on the surface temperature data are
5 1
repeated using RClimDex (Zhang and Yang 2004) and an outlier test (described in
2
detail in Section 2).
3
We note that the comparison of surface station temperature observations with
4
SSTs or other gridded temperature dataset is not new in itself. The interannual
5
variations and trends of homogenized maximum, minimum and mean temperature
6
data have been examined in relation to SST and nighttime marine air temperature data
7
for the South Pacific islands (Folland et al. 2003, Folland et al. 1997) and New
8
Zealand (Folland and Salinger 1995). Rusticucci and Kousky (2002) compare
9
temperature observations from selected stations in Argentina with National Centers
10
for Environmental Prediction-National Center for Atmospheric Research (NCEP-
11
NCAR) reanalysis (Kalnay et al. 1996) to determine how well statistics relating to
12
extremes in reanalysis 2-m temperatures correlate with observations. Kushnir (1994)
13
used land-based temperature observations surrounding the North Atlantic region to
14
verify the interdecadal signal that was evident in adjacent SSTs. However, in the
15
context of assessing the homogeneity of the surface temperature data, the proposed
16
method is a novel one or is at least not well documented in the literature. A review of
17
the conventional methodologies is found in Peterson (1998b).
18
The paper secondly attempts to investigate whether statistically significant
19
change points commonly identified across a number of stations, are linked to regional
20
climate change. These change points are identified from the homogeneity tests and are
21
time points where the statistical characteristics of the station series are significantly
22
different before and after. Synoptic conditions over the Atlantic and Pacific before
23
and after the change points are examined using NCEP reanalysis. This latter approach
24
is employed given the absence of available metadata for many of the stations used in
25
the study.
6 1
Finally, a third aim of this study involves investigating whether some of the
2
trends identified over the Caribbean by PTD would hold true for the homogenized
3
data. A number of the indices of PTD are reanalysed and the results for a few of them
4
(maximum and minimum temperature above the 90th percentile and below the 10th
5
percentile) are presented.
6
Section 2 describes data and statistical analyses, section 3 results of the quality
7
control procedures and homogeneity tests, section 4 synoptic relationships, section 5
8
some results of the reanalysed trends for the Caribbean and adjacent Caribbean and
9
section 6 discusses the primary findings. No adjustments of the data are attempted but
10
a homogeneous period is identified which may be exploited in further use of the data
11
for climate change studies for the Caribbean and adjacent Caribbean. This includes
12
additional trend analysis work or statistical downscaling as performed for north-
13
eastern Mexico (Cavazos 1997) and some Caribbean islands (Chen et al. 2006),
14
especially in relation to the National Communications to the United Nations
15
Framework Convention on Climate Change.
16 17
2.
Datasets and Methods
18
2.1
Datasets
19
Daily maximum and minimum temperature data for forty-two stations in the
20
Caribbean, Florida, Central America and northern South America are investigated
21
(c.f. Figure 1 and Table 1). The Caribbean islands include the Bahamas, Barbados,
22
Cayman Islands, Cuba, Dominican Republic, Guadeloupe, Jamaica, Puerto Rico, St.
23
Lucia, St. Vincent and Trinidad. The Central American and northern South American
24
countries comprise Belize, Costa Rica, Guatemala, Honduras, Nicaragua, Panama and
25
Venezuela – all countries with a Caribbean coastline. The data are obtained from the
7 1
Caribbean, and Central and northern South America daily data workshops (PTD,
2
APO), the National Climatic Data Center, Caribbean Institute for Meteorology and
3
Hydrology and local meteorological services1. The PTD and APO data largely span
4
1958-1999 and 1961-2003 respectively, while the other datasets span 1970-2006 (c.f.
5
Table 1). The presence of at least 80% nonmissing data was required before
6
homogeneity tests were applied to the stations. All stations are located in a 55o –
7
90oW and 5o – 30oN area.
8
SSTs are obtained from the HadISST1 dataset2 (Rayner et al. 2003). The
9
dataset is a combination of global SST and sea ice concentration recorded monthly on
10
1o latitude-longitude grids from 1871 to 2004. HadISST1 is constructed using reduced
11
space optimal interpolation and improves upon the representation of local SSTs in
12
comparison to previous global sea ice and SST (GISST) datasets: GISST1 (Parker et
13
al. 1995), GISST2 (Rayner et al. 1996) and GISST3. HadISST1 includes individual
14
ships’ observations from the Met Office Marine Data Bank and monthly median SSTs
15
from 1871-1995 from the Comprehensive Ocean-Atmosphere Data Set (COADS)
16
(Woodruff et al. 1987, 1998). The COADS has been validated over Caribbean waters
17
using correlations with independent temperature measurements off the coast of Puerto
18
Rico (Watanabe et al. 2002).
19
The monthly 2.5o gridded NCEP-NCAR reanalysis data (Kalnay et al. 1996) is
20
used to provide air temperature, divergent wind, vorticity and pressure vertical
21
velocity data3. A 30oS-70oN and 180o-355oE domain is extracted which covers the
22
tropical Atlantic and Pacific. NCEP-NCAR reanalysis data has been used by Wang 1
The PTD and APO datasets can be obtained by contacting
[email protected] and
[email protected] respectively. Data are provided only after consultation with the national meteorological services that provided the data. CIMH data requests can be made through their website at www.cimh.edu.bb. NCDC data can be downloaded from ftp://ftp.ncdcngov/pub/data/ghcn/daily. 2 The HadISST data can be downloaded from http://hadobs.metoffice.com/hadisst/. 3 Data are provided by the NOAA/OAR/ESRL PSD, Boulder, Colorado, USA from their website at http://www.cdc.noaa.gov/.
8 1
(2002a,b) to isolate circulation cells over the Atlantic and Pacific and by Stephenson
2
et al. (2007) to relate similar circulation patterns to Caribbean winter rainfall
3
extremes.
4 5
2.2
Method
6
2.2.1 Quality Control
7
PTD and APO conducted a series of quality control (qc) checks on the daily
8
surface temperature and precipitation data as previously outlined in Section 1. For
9
quality assurance the qc tests are repeated here using the RClimDex (version 1.0)
10
Software (Zhang and Yang 2004). The RClimDex qc procedure includes checks for
11
physically unreasonable values such as daily precipitation amounts less than zero or
12
daily maximum temperatures less than minimum temperatures. Additionally, values
13
that are greater than n standard deviations from the climatological mean for the day
14
(i.e. mean of all the January 1s, 2s, etc. respectively) are flagged, where n is a user
15
defined integer. Analyses were done using n values of four and five (Boyer and
16
Levitus 1994, Alexander et al. 2006).
17
Outliers greater than five standard deviations from the climatological mean of
18
the day were also isolated following variance comparisons with temperatures of the
19
surrounding three days, i.e., before and after the day being analysed. The day being
20
tested was excluded so as not to bias the calculation of the standard deviation. The
21
identified outliers were set to missing (-99.9) and manually checked, i.e. visually
22
compared with values of the surrounding days that may have also been flagged. This
23
allows for the identification a probable spell of some extreme weather events.
24
The latter outlier identification technique was more rapidly executed than the
25
RClimDex routine. The initial execution of the RClimDex tests using five standard
9 1
deviations from the mean of the day yielded a single outlier in maximum temperature
2
for the Dominican Republic station and Puerto Rico (Utuado) in all the maximum and
3
minimum timeseries assessed. However, for the same stations, the second technique
4
in a single run, flagged twenty-seven outliers in maximum temperatures and two in
5
minimum temperatures for the Dominican Republic; and sixteen outliers in the
6
maximum temperature for Puerto Rico (Utuado). Albeit an iterative application of the
7
RClimDex tests would yield the same result.
8
A monthly timeseries of maximum and minimum temperatures respectively
9
was created from the quality-controlled daily data by averaging over days in a given
10
month. This is for comparison with the monthly SST data. The SST grid values in 1-
11
degree proximity to a station are averaged to obtain a reference time series.
12
Anomalies of maximum temperature, minimum temperature and SSTs are then
13
calculated by removing the respective variable’s monthly climatology. A series of
14
plots are constructed and examined. These include: 1) anomalies of SST, maximum
15
and minimum temperatures; 2) difference series between SST and maximum and
16
minimum temperatures; 3) anomalies of the difference series and 4) standardized
17
values of the difference series. Anomalies of the difference series are calculated by
18
subtracting the monthly means of the difference series. This is done to remove the
19
seasonal cycle that was evident in the difference series. Each series is also
20
standardized by dividing by its monthly standard deviation.
21
The plots initially allowed easy identification of stations with very short
22
timeseries, i.e. less than approximately fifteen years, such as Puerto Rico (Lares) -
23
which was omitted from subsequent analyses. Secondly, the plots of the anomalous
24
SST, maximum and minimum temperatures provide a visual indication of the degree
25
of covariability among the timeseries. Correlation coefficients are computed between
10 1
monthly SSTs and maximum and minimum temperature to quantify their linear
2
association. The 1970-1998 period is used given the essentially good data coverage
3
across datasets for this period. Statistical significance at the 95% level is assessed
4
using the random phase method (Ebisuzaki 1997) to allow for any serial correlation in
5
the series. SST timeseries that are well correlated with station observations, i.e.,
6
greater than or equal to 0.70 (Malcher and Schönwiese 1987, Sala et al. 2000, Pielke
7
et al. 2007), are employed as reference series. Malcher and Schönwiese (1987)
8
explain that the correlations greater than or equal to 0.70 indicate that at least 50% of
9
the variability in the surface observations is captured by the reference series.
10
Finally the plots of the difference series anomalies help to identify any
11
divergence or differences in the behaviour of maximum and minimum temperatures
12
with respect to SSTs.
13 14
2.2.2 Homogeneity Tests
15
The homogeneity tests are applied to all the individual monthly series, i.e., for
16
SST, maximum and minimum temperature timeseries. Peterson et al. (1998b) suggest
17
that this test in itself is problematic since any common change point identified here
18
could be caused by or masked by real climatic fluctuations. The homogeneity tests are
19
therefore also applied to the difference series anomalies which should better isolate
20
the effects of station inhomogeneities versus regional climate change (Peterson
21
1998b). This is done for those station series that are well correlated with SSTs (≥
22
0.70).
23
The tests are conducted using the RHTest (version 0.95) Software (Wang and
24
Feng 2004). The RHTest is designed to detect multiple step change points that exist in
25
a timeseries based on the comparison of a two-phase regression model with a linear
11 1
trend for the entire series (Wang 2003). Change points that are significant at the 99%
2
level are highlighted for further investigation, focusing on those that are common
3
across a number of stations. The plots are re-examined and composites of atmospheric
4
variables before and after the change point are constructed to evaluate whether the
5
identified change point is associated with any significant shifts in the climate signal
6
over the Pacific and Atlantic in relation to the Caribbean. Several studies have
7
established the link between Caribbean atmospheric circulation and climate
8
variability, and conditions over the Pacific and Atlantic (Chen et al. 1997; Taylor
9
1999, Giannini et al. 2000; Chen and Taylor 2002; Taylor et al. 2002, Spence et al.
10
2004, Ashby et al. 2005, Stephenson et al. 2007). This approach is employed given
11
the absence of metadata for many of the stations.
12 13
2.2.3 Composites
14
We use composites to investigate commonly identified significant change
15
points and their possible links to changes in circulation patterns over the Caribbean
16
and adjacent regions. Difference maps are constructed between 5-year composites
17
before and after the year of the identified change points, i.e. if a breakpoint is
18
identified in 1983, composites of selected variables are calculated over 1978-1982 and
19
1984-1988, and then subtracted. The variables analysed are 1000 mb air temperature,
20
500 mb pressure vertical velocity, and divergent wind and vorticity at sigma levels
21
0.995 (surface) and 0.2101 (upper troposphere). Pressure vertical velocity and
22
divergent wind are used following Wang (2002a,b) who suggests these variables are
23
prerequisite to isolating atmospheric circulation cells. The divergent component of
24
wind
25
v = vψ + vΦ = k × ∇ψ + ∇Φ where ψ is the stream function and Φ is velocity
is
identified
by
the
second
component
in
the
equation
12 1
potential. The first term represents the rotational component which, though the larger
2
term, is not essential for identifying atmospheric cells (Krishnamurti 1971, Wang
3
2002a, b). The pressure vertical velocity (VV) at 500 mb indicates the mean vertical
4
motion at the mid-tropospheric level. Significant differences between the 5-year
5
composites before and after the identified change points are assessed for the 95% and
6
99% significance levels using the Student’s t-test (Panofsky and Brier 1968, Knaff
7
1997).
8 9
3.
Results
10
3.1
Quality Control
11
Figure 2 shows the plots of anomalies of SSTs, maximum temperatures and
12
minimum temperatures for selected stations. The stations shown are representative of
13
the good agreement evident between variations of the surface temperature timeseries
14
and adjacent SSTs for most stations. Correlation coefficients calculated between
15
monthly SSTs and maximum and minimum temperatures for 1970-1998 are shown in
16
Table 2. The temperature-SST associations are particularly strong for the Bahamas,
17
Cayman Islands, Cuba (Casa Blanca), Dominican Republic, Florida, Guadeloupe,
18
Jamaica (Sangster), Puerto Rico, St. Lucia, and St. Vincent, where correlations of
19
0.71-0.93 are obtained for both maximum and minimum temperatures. Interestingly,
20
the relationship between Central American temperatures and adjacent SSTs appears to
21
be stronger with respect to minimum temperatures, as seen for Guatemala and
22
Honduras (Table 2). The strong correlations (≥ 0.70), imply that SSTs are good
23
candidates for reference series for many of the surface temperature observations in
24
this study and these series are retained. Additionally, in cases where a correlation of
25
0.65-0.69 is obtained for one of the temperature series and is at least 0.70 for the other
13 1
series, both series are retained. There are some stations where neither maximum nor
2
minimum temperatures are well correlated with adjacent SSTs. These stations are
3
located in Belize - PSWGIA, Costa Rica, Honduras - Catacamas, Panama and
4
Venezuela.
5
Figure 2 also highlights some large temporal variations in the maximum and
6
minimum temperature anomalies. These are likely due to anomalous climatic
7
influences which in some cases are evident on a regional scale. For example, the large
8
negative minimum temperature anomaly evident for the Bahamas in January 1981
9
(c.f. Figure 2a) is also seen for Honduras (Figure 2b), Belize, Cuba, the Cayman
10
Islands, Guatemala, Honduras, Nicaragua and Florida (stations not shown). This is
11
consistent with anomalously low temperatures that were evident across the eastern
12
and southern United States and the Caribbean, as a result of several polar continental
13
air masses intruding into the region (Walker et al. 1982, Snedaker 1995). The ‘spikes’
14
in the minimum temperature anomaly timeseries are primarily evident during the
15
boreal winter months (November-March).
16
Figure 3 shows plots of the anomalies (i.e. with the monthly mean removed) of
17
the difference series between SST and maximum and minimum temperatures
18
respectively for the same stations shown in Figure 2. The plots again indicate
19
convergence, i.e., similarity in the relationship between SSTs and maximum and
20
minimum surface temperatures at the different stations. This is versus the divergence
21
evident for the 1958-60 period for Ponce, Puerto Rico (Figure 3c).
22 23 24 25
14 1
3.2
Homogeneity Results
2
3.2.1 SSTs, Maximum Temperature and Minimum Temperatures
3
Table 3 shows results of the homogeneity tests conducted on the individual
4
SST, maximum and minimum temperatures (see Section 2). The assessments reveal
5
no change points for SSTs adjacent to Bahamas, Florida and Cayman. Apart from
6
Cuba, these are the northern-most countries represented in the study. SSTs adjacent to
7
thirteen of the other nineteen countries indicate steps for 1970/71 of -0.24 to -0.80 oC
8
and for 1983/84 of -0.22 to -0.83 oC. (Negative change points indicate a shift towards
9
lower temperatures.) The change points are not significant at the 99% level, except in
10
1970/71 for Barbados, Guadeloupe, Panama, St. Lucia, St. Vincent, and Venezuela
11
(Tumeremo) – essentially the easternmost or southernmost stations in the study; and
12
in 1983 for Belize and Honduras. Other change point years are identified in 1973 and
13
1981. The significant change points noted, i.e., for 1970/71, and 1983, may be
14
consistent with variations in the tropical Atlantic meridional gradient mode,
15
characterized by oppositely signed SST anomalies in the tropical North Atlantic and
16
tropical South Atlantic. The mode was mainly positive for periods pre-1970 and 1976-
17
1983 and negative during 1971-75 and 1984-89 (Wang 2002b). This is explored
18
further in Section 4.
19
Homogeneity tests on the maximum and minimum temperatures (Table 3)
20
indicate no step change points for Freeport (Bahamas), Husbands (Barbados), Grand
21
Cayman (Cayman Islands), Flores (Guatemala), Key West (Florida) and St. Vincent.
22
This is also the case with regard to maximum temperatures for PSWGIA (Belize),
23
David (Panama) and GFL Charles Airport (St. Lucia), and minimum temperatures for
24
Everglades (Florida). Significant change point years commonly identified across three
25
or more surface temperature observations are noted for 1976/77, 1979 and 1983/84.
15 1
The 1976/77 and 1983/84 change point years are also consistent with phases of the
2
tropical Atlantic meridional gradient mode. The 1976/77 change point also coincides
3
with the 1976-77 abrupt climate shift in Pacific circulation centred in the Tropics
4
which significantly influenced subsequent El Niño evolutions (Trenberth 1990,
5
Trenberth and Hoar 1996, Zhang et al. 1997, Guilderson and Schrag 1998, Urban et
6
al. 2000). Additionally the Pacific/North Atlantic (PNA) teleconnection pattern which
7
describes the position, strength and orientation of a trough and ridge pattern over the
8
northern Pacific Ocean and North America shifted to a positive phase in 1976
9
(Schmidt 2003). These may suggest that some of the change points identified in the
10
individual series are related to changes in the climate system (Peterson et al. 1998b).
11
The 1976/77 and 1983/84 change points are oppositely signed with the latter being
12
negative. The 1976/77 change points are obtained for La Ceiba (Honduras), San Juan
13
(Puerto Rico) and Mene Grande and Guiria (Venezuela). The 1983/84 change points
14
are evident for Grantley Adams (Barbados), Catacamas (Honduras), and San Juan
15
(Puerto Rico). The 1979 change points are positive and are obtained for Fabio (Costa
16
Rica), Maiquetia Apt. Bolivar and La Carlota (Venezuela).
17 18
3.2.2 Difference Series
19
The step change points evident in the difference series anomalies are listed in
20
Table 4. Firstly, it is noted that in most cases, there is consistency in the direction of a
21
change point (i.e. positive or negative) identified for a given year (± 1 year), in a
22
station’s maximum and minimum temperature timeseries. This is evident for Cuba,
23
Dominican Republic, Guadeloupe, Florida (Key West) and Puerto Rico (Ponce and
24
San Juan), with the Bahamas (Freeport), Jamaica (Worthy Park) and St. Lucia as the
25
exceptions. Additionally, although there is consistency within most stations, this is not
16 1
necessarily the case across stations in different countries. For example, whereas
2
positive change points are identified for Cuba and Puerto Rico (San Juan) for
3
1981/82, they are negative for Dominican Republic and Florida (Key West). In this
4
case, the absence of any consistent spatial pattern given the difference in signs,
5
suggests that a regional climatic influence may not be an underlying cause. A re-
6
examination of the plots (not shown) confirms this, in that, while there were
7
similarities in the distribution of values before and after the negative change points
8
identified for Dominican Republic and Florida (Key West), they were very dissimilar
9
with respect to the temporal patterns surrounding 1981/82 for both Cuba and Puerto
10
Rico (San Juan).
11
Significant change point years common across three or more stations are
12
identified mainly for 1968/69, 1970/71, 1979 and 1983. The 1968/69 steps are
13
obtained for Cuba (Guantanamo Bay), Florida (Everglades), Guadeloupe (Le Raizet)
14
and Puerto Rico (Ponce and Utuado). These steps are also evident in their individual
15
series (Table 3), and are negative except for Guadeloupe – the southernmost of these
16
stations. The 1970/71 change points are positive and are evident for Cuba
17
(Guantanamo Bay), Dominican Republic, Jamaica (Worthy Park) and Puerto Rico
18
(San Juan) (though not identified in their individual series). The 1978/79 change
19
points are obtained for Barbados (Husbands), Guadeloupe, Florida (Everglades) and
20
Honduras (La Mesa). The change points are positive except for Honduras. The 1983
21
change points are negative and are obtained for Puerto Rico (Ponce and Utuado) and
22
Guatemala. Therefore these change point years (1968/69, 1970/71, 1978/79 and 1983)
23
are investigated further.
24 25
17 1
4.
Analysis of shifts
2
4.1
Data Issues
3
For 1968/69 and 1970/71, the plots for Cuba (Guantanamo Bay), Dominican
4
Republic, Guadeloupe and Puerto Rico (San Juan) (not shown) showed divergence in
5
the maximum and minimum temperature anomalies. This was also the case for
6
Barbados (Husbands) with respect to 1979. The plots for Puerto Rico (Ponce) (c.f.
7
Figure 2c and 3c) reveal a noticeable increase in maximum and minimum temperature
8
anomalies between 1958 and 1968 in comparison to the rest of the timeseries. These
9
inconsistencies in maximum and minimum temperature series are identifiable around
10
the years of the identified step change, but are largely absent from the rest of the
11
temperature timeseries. They are therefore more likely due to station inhomogeneities
12
than regional climate shifts. Consequently, in identifying homogenous periods for the
13
data in Section 5, the change point years identified above for Barbados (Husbands),
14
Cuba (Guantanamo Bay), Dominican Republic, Guadeloupe and Puerto Rico (San
15
Juan) are excluded. Figure 4 shows the location of these stations. There appears to be
16
no distinct regional pattern linking the stations. More detailed investigations would
17
necessitate the use of metadata.
18 19
4.2
Climate shifts
20
The plots for Florida, Guadeloupe, Honduras (La Mesa) and Puerto Rico
21
(Utuado) do not reveal the divergent behaviour noted in the previous section around
22
the 1968 and 1978/79 change points This is the case as well for Guatemala and Puerto
23
Rico (San Juan and Ponce) for 1983. Therefore the 1968, 1978/79 and the 1983
24
change points are further investigated using composite analysis. As noted in the
18 1
previous section, there does not appear to be a distinct spatial relationship across
2
stations.
3
Synoptic analyses are undertaken to determine whether the apparent climate
4
shifts are evident in large-scale circulation patterns. Composites of annual 1000 hPa
5
air temperature, 500 hPa pressure vertical velocity, and divergent wind and vorticity
6
at sigma levels of 0.995 (surface) and 0.2101 (upper troposphere) are constructed for
7
five years before and after the identified step changes. The difference maps are
8
produced with areas significant at the 95% and 99% levels highlighted (see Section
9
2). We note that the significance testing done here is to primarily highlight the
10
difference between periods before and after the identified change points. More
11
rigorous field significance testing, for e.g. see Ventura et al. (2004) could be done and
12
is the subject of future work.
13 14
4.2.1 1968
15
Composites constructed around 1968, i.e. 1969-73 minus 1963-67, are shown
16
in Figure 5. Significant positive changes in surface air temperatures are evident over
17
the Caribbean - south of 18oN, and over Florida and Belize (Figure 5a). This indicates
18
relatively warmer surface air temperatures post-1968 and is not consistent with the
19
change point shift towards lower temperatures identified for 1968 for Florida and
20
Puerto Rico (Table 4).
21
between the North Atlantic and South Atlantic, the former (latter) associated with
22
relatively cooler (warmer) temperatures post-1968. This gradient agrees with the
23
meridional gradient mode presented by Wang (2002b) which is in its negative phase
24
during 1971-75, characterised by negative (positive) SST anomalies over the tropical
25
north (south) Atlantic. Other studies [such as Oort et al. (1987), Kushnir (1994) and
However, Figure 5a reveals a gradient in temperatures
19 1
Grosfeld et al. (2007)] also suggest a relatively cooler north Atlantic during the 1970s,
2
with Levitus (1989 a,b) documenting a cooling of the first 1000m of the North
3
Atlantic ocean post-1970.
4
The relatively cooler surface temperatures of the North Atlantic post-1968, do
5
not appear to penetrate the Caribbean in the composite map shown in Figure 4a as it
6
does for the NCEP SST composite difference maps (1965-70 vs. 1971-75) shown by
7
Wang (2002b) (see author’s Figure 7). This may be as a result of the different periods
8
used for constructing the composite difference maps. Wang’s composite difference
9
maps in SST anomalies do however show: (1) oppositely signed SST anomalies
10
around Florida and northern Central America (i.e. with respect to the North Atlantic),
11
consistent with Figure 5a and (2) strong SST differences over the eastern and southern
12
Caribbean which support the significant change points evident in SSTs for 1970/71
13
over the eastern Caribbean and Panama, i.e. versus the rest of the Caribbean and
14
adjacent regions where the 1970/71 change point was not significant (c.f. Table 3).
15
Significantly stronger mid-tropospheric ascent (or lower magnitude descent)
16
post-1968 (c.f. Figure 5b) is evident over northern South America into the
17
southernmost Caribbean; over Florida and Central America. Over northern South
18
America the mid-level ascent is coupled with low level convergence (c.f. Figure 5c)
19
and upper level divergence (c.f. Figure 5d) – all of which supports the picture of
20
warmer temperatures post-1968. This is not however consistent with the anomalously
21
dry climate observed over the Central American-Caribbean region during the early
22
1970s (Hastenrath 1976, Peterson et al. 2002) which implies increased subsidence,
23
drier mid levels and increased atmospheric cooling post-1970 (Knaff 1997).
24
Other potentially relevant changes in the climate system have also been noted
25
since 1970. These include: (1) increasing positive values of the winter atmospheric
20 1
North Atlantic Oscillation (NAO) which is associated with higher than normal North
2
Atlantic high, hence stronger trade winds and cooler SSTs (Giannini et al. 2001,
3
Hurrell 1995, Higuchi et al. 1999); (2) onset of the negative phase of the oceanic
4
Atlantic multidecadal oscillation – associated with cooler than normal North Atlantic
5
SSTs (Enfield et al., 2001). A plot of annual NAO4 anomalies versus an 11-station
6
annual average temperature index (also anomalies) is shown in Figure 6. Evident is
7
the post - 1970 shift towards positive NAO, and a change towards less synchronized
8
variations of the average temperatures over the Caribbean and adjacent Caribbean
9
with the NAO. Correlation between the indices pre -1970, i.e. 1961 - 1970 (0.43) is
10
greater than post -1970, i.e. 1971 - 1990 (0.38), the latter value being statistically
11
significant.
12
The idea is that though the NCEP reanalysis does not support the negative
13
breakpoint identified in 1968, given the other evidence/studies that exist to point to its
14
climate-relatedness, this feature is not considered as an inhomogeneity. Furthermore it
15
may be useful to note that for the late 1960’s certain NCEP Reanalysis variables
16
including 2m air temperature over southern Europe into the North Atlantic have
17
shown some discontinuity (Pohlmann and Greatbatch 2006). Additionally with the
18
introduction of satellite measurements in the late 1970’s, systematic errors have been
19
evident in the tropical atmosphere observations in some reanalyses products,
20
including. NCEP and the European Centre for Medium-Range Weather Forecast 40-
21
years (ERA-40) (Santer et al. 1999, Trenberth et al 2001, Sturaro 2003, Sterl 2004,
22
Kinter et al. 2004, Trenberth and Smith 2005, Greatbach and Rong 2006). This has
23
had implications for climate trend analyses in that apparent and significant shifts were
24
possibly linked to changes in the observing system and/or the data assimilation
4
NAO values can be downloaded from http://www.cru.uea.ac.uk/cru/data/nao.htm
21 1
procedures (Kinter et al. 2004). For the purpose of this study the shift towards warmer
2
post - 1970 temperatures that is evident in the NCEP Reanalysis, but absent in other
3
datasets, may be related to this issue of discontinuity.
4 5
4.2.2 1983
6
Analysis of the 1983 change point reveals relatively lower temperatures post-
7
1983 but differences are weak and not significant over the Caribbean (Figure 7a). A
8
gradient in surface temperatures is again visible, but with lower (higher) temperatures
9
over the tropical North (South) Atlantic for post-1983. Stronger and significant
10
medium level descent is evident over southeastern Caribbean for post-1983 (Figure
11
7b) and is coupled with low-level divergence (Figure 7c) and upper-level convergence
12
(Figure 7d) just north of South America into the southern Caribbean. This is
13
consistent with the gradient mode pattern presented by Wang (2002b), and with the
14
anomalously dry Caribbean observed in the late 1980s (Peterson et al. 2002). The
15
1980s have also been associated with an intensified North Atlantic high (i.e. high
16
NAO index) (Drinkwater 1996, Curry et al. 1998). Therefore the negative step in
17
Puerto Rico and Guatemala temperatures for 1983 may be associated with this large-
18
scale shift in circulation. It is important to note that this negative shift in temperatures
19
for 1983 is also evident in the individual temperature series for Barbados, Costa Rica,
20
Dominican Republic, Honduras and Panama (Table 3), suggesting a Caribbean –
21
Central American wide influence.
22 23
4.2.3 1979
24
Composite difference maps for 1979 indicate significantly higher air
25
temperatures post-1979 for the central to southern Caribbean but not for Florida
22 1
where a hint of slightly lower temperatures is visible (Figure 8a). Interestingly, the
2
gradient in surface temperatures between the North and South Atlantic is not evident
3
in this case. Significantly stronger low-level convergence and upper-level divergence
4
over the Caribbean post-1979 are indicated (Figures 8c-d) but are not associated with
5
significant changes in medium-level air motion over the main Caribbean (Figure 8b).
6
The implication is that the significant changes evident in the atmospheric circulation
7
do not appear to penetrate the entire troposphere. Therefore, the significant positive
8
step changes evident identified for Barbados and Trinidad for 1978/9 can be
9
associated with changes in the large-scale circulation. As noted in Section 4.2.1, the
10
1978/79 change point coincides with the introduction of satellite measurements.
11
Therefore it is plausible that the 1978/79 change point may be linked to an NCEP
12
discontinuity.
13 14
5.
Reanalysis of some Caribbean and adjacent Caribbean Trends
15
The analyses described in the previous sections allow for the identification of
16
a primary common homogeneous period across the data. This period spans 1970-1992
17
for Grantley Adams (Barbados), Catie (Costa Rica), Casa Blanca (Cuba), Everglades
18
(Florida), Key West (Florida), Le Raizet (Guadeloupe), Flores (Guatemala), Santa
19
Rosa de Copan (Honduras), Ponce (Puerto Rico) and Freeport (Bahamas); with
20
maximum temperatures for Nassau Airport (Bahamas) and minimum temperatures for
21
Worthy Park (Jamaica). A secondary homogeneous period can be deduced for 1984-
22
1998 for PSWGIA (Belize), Guantanamo Bay (Cuba), La Ceiba (Honduras), Tela
23
(Honduras), San Juan (Puerto Rico), Piarco (Trinidad) with maximum temperatures
24
for Grand Cayman (Cayman Islands) Santo Domingo (Dominican Republic) and
25
Worthy Park (Jamaica). Figure 9 shows the location of these stations.
23 1 2
A pertinent question to be posed is: Would the results of PTD change using the
3
homogeneous data for this period? To address this question, indices of temperature
4
extremes were reconstructed for each of stations. The indices examined include
5
percent of days maximum and minimum temperature were greater than or equal to the
6
90th percentile, and the number of days maximum and minimum temperature were
7
less than or equal to the 10th percentile. Following the PTD approach, the numerical
8
average of the index results from each station was calculated. Figure 10 shows the
9
averaged time series and their associated regression lines to indicate the presence of
10
any trend. The statistical significance of the trend is also tested. Although a 23-year
11
dataset can be considered too short for trend analyses, it can justifiably be used in a
12
comparative sense to investigate changes in extremes identified from the PTD work.
13
The results indicate that the percent of days at or above the 90th percentile has
14
increased over the base period (1970-1992) while the percent of days at or below the
15
10th percentile has decreased. All the regression slopes are significant at the 1% level.
16
The trends obtained are consistent with those presented by PTD. Interestingly the
17
annual variations in the indices also resemble the variability obtained by PTD, for e.g.
18
the peaks between 1985 and 1990 for the percent of days when maximum and
19
minimum temperatures are at or above the 90th percentile. For other indices the trends
20
are very similar to PTD and are therefore not discussed.
21 22
6.
Summary and Conclusions
23
This study provides a systematic evaluation of the homogeneity of daily
24
temperature data on a monthly timescale for the Caribbean, Florida, and countries in
25
Central America and northern South America with a Caribbean coastline. The method
24 1
uses adjacent SSTs as reference series, and presents an alternative approach for
2
islands/coastal regions that possess a low density of temperature stations, but where
3
well correlated SST data are available. In this study a strong association was observed
4
between SSTs and station maximum and minimum temperatures, primarily for the
5
Caribbean islands and Florida, and secondly for the minimum temperatures of Central
6
America (see Section 3). This provided a strong basis for the use of SSTs in
7
developing reference series for the homogeneity assessments and purports the
8
possible use of SSTs as a proxy for Caribbean and adjacent Caribbean surface
9
temperatures.
10
A reasonable question to be considered is whether the use of SSTs as a
11
reference series is comparable to the use of neighbouring stations towards the
12
development of reference series. Correlation analysis is used to investigate this
13
question. Station-based reference series were constructed using an average of those
14
stations that were deduced as homogeneous for the 1970-1992 period (see Section 5).
15
The correlation coefficients were calculated between the station average reference
16
series and the maximum and minimum temperature series for each island/country.
17
Where the station of interest was also one of the homogeneous series, it was omitted
18
from the average before calculating the correlation coefficient. The difference
19
between the correlations obtained with respect to SSTs (c.f. Table 1) and the station
20
average reference series were calculated and the significance assessed at the 99%
21
level using a t-test (Chen and Popovich 2002).
22
The results (not shown) indicate that 17 (12) of the 34 stations listed in Table 1
23
exhibited correlations that were not significantly different from correlations obtained
24
with respect to the station average reference series. In these cases the correlations
25
were minimally higher with respect to the station average series. For those cases
25 1
where the difference was statistical significant, again most of the correlations were
2
greater with respect to the station average series when compared to SSTs. These
3
results indicate that the use of SSTs towards developing reference series is at least a
4
viable option in assessing the homogeneity of station series for areas of low station
5
density and which have strong associations with SST variability.
6
The change points identified using the SST-based reference series were also
7
investigated using composite analysis, and two overlapping homogeneous periods
8
were identified across the data. They include: (1) 1970-1992: for Grantley Adams
9
(Barbados), Catie (Costa Rica), Casa Blanca (Cuba), Everglades (Florida), Key West
10
(Florida), Le Raizet (Guadeloupe), Flores (Guatemala), Santa Rosa de Copan
11
(Honduras), Ponce (Puerto Rico) and Freeport (Bahamas); with maximum
12
temperatures for Nassau Airport (Bahamas) and minimum temperatures for Worthy
13
Park (Jamaica); (2) 1984-1998: PSWGIA (Belize), Guantanamo Bay (Cuba), La
14
Ceiba (Honduras), Tela (Honduras), San Juan (Puerto Rico), Piarco (Trinidad) with
15
maximum temperatures for Grand Cayman (Cayman Islands) Santo Domingo
16
(Dominican Republic) and Worthy Park (Jamaica). It is important to note however
17
that the absence of proof of an inhomogeneity is not necessarily proof of absence. The
18
results suggest there is evidence of homogeneity but more data work is needed.
19
The 1970-92 period can be considered the more useful period for additional
20
trend analyses and statistical downscaling studies given the length of available
21
observations. This period was used to investigate whether the results of the PTD trend
22
analysis of temperature extremes would vary using the homogeneous data. The results
23
were found to be consistent for the indices characterizing percent of days maximum
24
and minimum temperature were greater than or equal to the 90th percentile, and the
26 1
number of days maximum and minimum temperature were less than or equal to the
2
10th percentile and offers credence to the PTD results.
3
An implicit aim of this study then was to extend the work done by PTD by
4
building upon the daily data archive available for the Caribbean. Even with the
5
possibility of SST use as proxy for surface temperatures for the Caribbean and
6
adjacent Caribbean, there is no substitute for the valuable data recovery that has been
7
done and which needs to continue to advance the trend analysis work for the region.
8
Further efforts must include (i) acquiring additional quality controlled data, and
9
metadata, and (ii) expanding the current database via the application of
10
homogenization techniques. Data adjustment is beyond the scope of the current work
11
since the approaches that could be employed depend on station density, ability to
12
create a reference series, and the availability of metadata (Vincent et al. 2002) - the
13
same critical issues that were encountered in this study. The result would be a data
14
archive for the Caribbean and adjacent Caribbean that could be used with greater
15
confidence for the analyses of trends, climate variability and climate change.
16
Reservations in sharing metadata still exist, that is, in cases where metadata are
17
present and is a challenge that must necessarily be overcome. The homogeneous data
18
isolated in this study is currently being used in statistical downscaling investigations
19
for the Caribbean and adjacent Caribbean.
20 21 22 23 24 25
27 1
Acknowledgement
2
This work was conducted by the lead author during a study visit to the Climatic
3
Research Unit funded by the Caribbean Community Climate Change Centre, Belize
4
and the Department of Physics at the University of the West Indies, Mona, Jamaica.
5
The authors thank Peterson et al. (2002) and Aguilar et al. (2005) for providing the
6
datasets used in the study. The datasets are compiled from meteorological services in
7
nine Caribbean islands, seven Central American countries, Florida and Venezuela.
8
Thanks also to the National Climatic Data Center, Caribbean Institute for
9
Meteorology and Hydrology and the Caribbean meteorological services, particularly
10
St. Vincent and St. Lucia, for data and additional assistance given. Thanks to the
11
reviewers for providing very useful comments.
12 13 14 15 16 17 18 19 20 21 22 23 24 25
28 1
References
2
Ashby, S. A., M. A. Taylor and A. A. Chen (2005), Statistical models for predicting
3
Caribbean rainfall, Theor. Appl. Climatol., 82: 65–80; DOI 10.1007=s00704-004-
4
0118-8.
5 6
Alexander, L.V., et al. (2006), Global observed changes in daily climatic extremes of
7
temperature
8
doi:10.1029/2005JD006290.
and
precipitation,
J.
Geophys.
Res.,
111,
D05109,
9 10
Aguilar, E., et al. (2005), Changes in precipitation and temperature extremes in
11
Central America and northern South America, 1961– 2003, J. Geophys. Res., 110,
12
D23107, doi:10.1029/2005JD006119.
13 14
Aguilar, E., I. Auer, M. Brunet, T. C. Peterson and J. Wieringa (2003), Guidelines on
15
Climate Metadata and Homogenization, WMO/TD No. 1186, Geneva, Switzerland.
16 17
Boyer, T. and S. Levitus (1994), Quality Control and Processing of Historical
18
Oceanographic Temperature, Salinity, and Oxygen Data, NOAA Technical Report
19
NESDIS 81, Washington, D.C., 64pp.
20 21
Cavazos, T. (1997), Downscaling large-scale circulation to local winter rainfall in
22
north-eastern Mexico, Int. J. Climatol., 17, 1069-1082.
23
29 1
Chen, A. A., A. Roy, J. McTavish, M. Taylor, and L. Marx (1997), Using SST
2
Anomalies
3
Technical report No. 49, 24 pp.
to predict flood and drought conditions for the Caribbean, COLA
4 5
Chen, A. A., and M. A. Taylor (2002), Investigating the link between early season
6
Caribbean rainfall and the El Niño+1 year, Int. J. Climatology, 22, 87-106.
7 8
Chen, A. A., C. L. Rhoden, and M. A. Taylor (2006), Climate Change Science and
9
Future Climate Change Scenarios, in Chen, A.A., D.D. Chadee and S. C. Rawlins
10
(eds) Climate Change Impact on Dengue: The Caribbean Experience, Climate Studies
11
Mona, University of the West Indies, pp 66-77, ISBN976-41-0210-7.
12 13
Chen, P. Y., and P. M. Popovich (2002), Correlation: Parametric and nonparametric
14
measures, Newbury Park, CA: Sage Publications, Inc.
15 16
Curry, R. G., M. S. McCartney and T. M. Joyce (1998), Oceanic transport of subpolar
17
climate signals to mid-depth subtropical waters, Nature, 391, 575-577.
18 19
Drinkwater, K. F. (1996), Atmospheric and oceanic variability in the northwest
20
Atlantic during the 1980s and early 1990s, J. Northw. Atl. Fish. Sci., 18, 77-97.
21 22
Easterling, D. R., L. V. Alexander, A. Mokssit, and V. Detemmerman (2003),
23
CCl/CLIVAR workshop to develop priority climate Indices, Bull. Am. Meteorol. Soc.,
24
84, 1403–1407.
25
30 1
Ebisuzaki, W. (1997), A Method to estimate the statistical significance of a
2
correlation when the data are serially correlated, J Climate, 10, 2147-2153.
3 4
Enfield, D. B., A. M. Mestas-Nuñez, and P. J. Trimble (2001), The Atlantic
5
multidecadal oscillation and its relation to rainfall and river flows in the continental
6
U.S., Geophys. Res. Lett., 28, 2077-2080.
7 8
Folland, C., and M. J. Salinger (1995), Surface temperature trends in New Zealand
9
and the surrounding ocean, Int., J., Climatol., 15, 1195-1218.
10 11
Folland, C., M. J. Salinger, and N. Rayner (1997), A comparison of annual South
12
Pacific Island and ocean surface temperatures, Wea. Climate, 17 (1), 23-42.
13 14
Folland, C., M. J. Salinger, N. Jiang, and N. A. Rayner (2003), Trends and variations
15
in South Pacific Island and ocean surface temperatures, J. Climate, 16, 2859-2874.
16 17
Giannini, A., Y. Kushnir, and M. A. Cane (2000), Interannual Variability of
18
Caribbean rainfall, ENSO and the Atlantic Ocean, J. Climate, 13, 297-311.
19 20
Giannini, A., M. A. Cane, and Y. Kushnir (2001), Interdecadal changes in the ENSO
21
teleconnection to the Caribbean region and the North Atlantic Oscillation, J. Climate,
22
14, 2867-2879.
23 24
Greatbach, R. J., and P.-P. Rong (2006), Discrepancies between different northern
25
hemisphere summer atmospheric data products, J. Climate, 19, 1261-1273.
31 1 2
Grosfeld, K., G. Lohmann, N. Rimbu, K. Fraedrich, and F. Lunkeitt (2007),
3
Atmospheric multidecadal variations in the North Atlantic realm: proxy data,
4
observations, and atmospheric circulations model studies, Clim. Past, 3, 39-50.
5 6
Guilderson, T. P., and D. P. Schrag (1998), Abrupt shift in subsurface temperatures in
7
the tropical Pacific associated with changes in El Niño, Science, 281, 240-243.
8 9 10
Hastenrath, S., (1976), Variations in the low-latitude circulation and extreme climatic events in the tropical Americas, J. Atmos. Sci., 33, 202-215.
11 12
Haylock, M. R., et al. (2006), Trends in total and extreme South American rainfall
13
1960–2000 and links with sea surface temperature, J. Climate., 19, 1490-1512.
14 15
Higuchi, K., J. Huang, and A. Shabbar (1999), A wavelet characterization of the
16
North Atlantic Oscillation variation and its relationship to the North Atlantic sea
17
surface temperature, Int. J. Climatol., 19, 1119-1129.
18 19
Hurrell, J. W. (1995), Decadal trends in the North Atlantic Oscillation regional
20
temperatures and precipitation, Science, 269, 676-679.
21 22
Kalnay, E., et al. (1996): The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer.
23
Meteor. Soc., 77, 437–471.
24
32 1
Kinter, J. L., III, M. J. Fennessy, V. Krishnamurthy, and L. Marx (2004), An
2
evaluation of the apparent interdecadal shift in the tropical divergent circulation in the
3
NCEP-NCAR reanalysis, J. Climate, 17, 349-361.
4 5
Knaff, J. A. (1997), Implications of Summertime Sea Level Pressure Anomalies in the
6
Tropical Atlantic Region, J. Climate, 10, 789-804.
7 8
Klein Tank, A. M. G., et al. (2006), Changes in daily temperature and precipitation
9
extremes in central and south Asia, J. Geophys. Res. – Atmos., 111, D16105,
10
doi:10.1029/2005JD006316.
11 12
Krishnamurti, T. N. (1971), Tropical east-west circulations during the northern
13
summer, J. Atmos. Sci., 28, 1342-1347.
14 15
Kushnir, Y., (1994), Interdecadal variations in North Atlantic sea surface temperature
16
and associated atmospheric conditions, J. Climate, 7, 141-157.
17 18
Levitus, S., (1989a), Interpentadal variability of temperature and salinity at
19
intermediate depths of the North Atlantic Ocean, 1970-1974 versus 1955-1959. I.
20
Geophys. Res., 94, 6091-6131.
21 22
Levitus, S., (1989b), Interpentadal variability of salinity in the upper 150 m of the
23
North Atlantic Ocean, 1970-1974 versus 1955-1959, J. Geophys. Res., 94, 9679-9685.
24
33 1
Malcher, J., and Ch.-D. Schönwiese (1987), Homogeneity, spatial correlation and
2
spectral variance analysis of long European and North American air temperature
3
records, Theor. Appl. Climatol., 38, 157-166.
4 5
Mokssit, A. (2003), Development of priority climate indices for Africa: A
6
CCI/CLIVAR workshop of the World Meteorological Organization, in Mediterranean
7
climate: variability and trends, edited by H. J. Bolle, pp. 116–123, Springer, New
8
York.
9 10
Mahlung, C. (2001), Jamaica. Jamaica’s first national communication to the United
11
Nations
12
http://unfccc.int/resource/docs/natc/jamnc1.pdf
Framework
Convention
on
Climate
Change
(UNFCCC).
13 14
New, M., et al. (2006), Evidence of trends in daily climate extremes over Southern
15
and West Africa, J. Geophys. Res., 111, D14102, DOI: 10.1029/2005JD006289.
16 17
Oort, A. H., Y. H. Pan, R. W. Reynolds and C. F. Ropelewski (1987), Historical
18
trends in the surface temperature over the oceans based on COADS, Climate. Dyn., 2,
19
29-38.
20 21
Panofsky, H. A. and G. Brier (1968), Some Applications of Statistics to Meteorology.
22
Pennsylvania: The Pennsylvania State University, 224 pp.
23 24
34 1
Parker, D. E., C. K. Folland, A. Bevan, M. N. Ward, M. Jackson, and K. Maskell
2
(1995), Marine surface data for analysis of climatic fluctuations on interannual to
3
century timescales, in Natural Climate Variability on Decade-to-Century Time Scales,
4
edited by D. G. Martinson et al., pp. 241–250 + color figures on pp. 222– 228, Natl.
5
Acad. Press, Washington, D. C.
6 7
Pielke Sr., R., et al. (2007), Documentation of bias associated with surface
8
temperature measurement sites for climate change assessment. Bull. Amer. Meteor.
9
Soc., accepted.
10 11
Peterson, T., R. Vose, R. Vose, R. Schmoyer, and V. Razuvaëv (1998a), Global
12
historical climatological network (GHCN) quality control of monthly temperature
13
data, Int. J. Climatol, 18, 1169-1179.
14 15
Peterson, T., et al. (1998b), Homogeneity adjustments of in situ atmospheric climate
16
data: a review, Int. J. Climatol., 18, 1493-1517.
17 18
Peterson, T., et al. (2002), Recent changes in climate extremes in the Caribbean
19
region, J. Geophysic. Res. – Atmos., 107, doi:10.1029/2002JD002251.
20 21
Peterson, T. (2005), The workshop on enhancing south and central Asian climate
22
monitoring and indices, Pune, India, February 14 – 19, 2005, CLIVAR Exch., 10, 6
23
35 1
Rayner, N. A., E. B. Horton, D. E. Parker, C. K. Folland, and R. B. Hackett (1996),
2
Version 2.2 of the global sea ice and sea surface temperature data set, 1903–1994,
3
Clim. Res. Tech. Note CRTN74, Hadley Cent., Met Office, Bracknell, UK.
4 5
Rayner, N.A., D. E. Parker, E. B. Holton, C. K. Folland, L. V. Alexander, D. P.
6
Rowell, E. C. Kent and A. Kaplan (2003), Global analyses of se a surface
7
temperature, sea ice, and night marine air temperature since the late nineteenth
8
century, J. Geophys. Res., 108, D14, 4407, doi:10.1029/2002JD002670.
9 10
Rusticucci, M. M., and V. E. Kousky (2002), A comparative study of maximum and
11
minimum temperatures over Argentina: NCEP-NCAR Reanalysis versus station data,
12
J. Climate, 15, 2089-2101.
13 14
Sala, J. Q., A. G. Olcina, A. P. Cuevas, J. O. Cantos, A., R., Amoros and E. M. Chiva
15
(2000), Climatic warming in the Spanish Mediterranean: natural trend or urban effect,
16
Climate Change, 46, 473-483.
17 18
Santer, B. D., J. J. Hnilo, T. M. L. Wigley, J. S. Boyle, C. Doutriaux, M. Fiorino, D.
19
E. Parker, and K. E. Taylor (1999), Uncertainties in observationally based estimates
20
of temperature change in the free atmosphere, J. Geophys. Res., 104, 6305-6334.
21 22 23
Schmidt, N. (2003), A dry El Niño winter, END Insight, 4-6.
36 1
Sensoy, S., T. C. Peterson, L. V. Alexander, and X. Zhang (2007), Enhancing Middle
2
East Climate Change Monitoring and Indices Workshop summary, Bull. Am.
3
Meteorol. Soc., in press.
4 5
Snedaker, S. C. (1995), Mangroves and climate change in the Florida and Caribbean
6
region: scenarios and hypotheses. Hydrobiologia, 295, 43-49.
7 8
Spence, J. M., M. A. Taylor and A. A. Chen (2004), The effect of concurrent sea
9
surface temperature anomalies in the tropical Pacific and Atlantic on Caribbean
10
Rainfall, Int. J. Climatol., 24, 1531-1541, doi:10.1002/joc.1068.
11 12
Stephenson, T. S., A. A. Chen, and M. A. Taylor (2007), Toward the development of
13
prediction models for the primary Caribbean dry season, accepted, Theor. App.
14
Climatol., doi:10.1007/s00704-007-0308-2.
15 16
Sturaro, G. (2003), A closer look at the climatological discontinuities present in the
17
NCEP/NCAR Reanalysis temperature due to the introduction of satellite data, Clim.
18
Dyn., 21, 309-316.
19 20
Taylor, M. A. (1999) October in May: the effect of tropical Atlantic SSTs on early
21
season Caribbean rainfall. Ph. D. thesis, University of Maryland, College Park, 213
22
pp.
23
37 1
Taylor, M. A., D. B. Enfield, and A. A. Chen (2002), The influence of the tropical
2
Atlantic vs. the tropical Pacific on Caribbean Rainfall, J. Geophys. Res., 107, (C9)
3
3127, doi:10.1029/2001/JC001097.
4 5
Trenberth, K. E. (1990), Recent observed interdecadal changes in the Northern
6
Hemisphere. Bull. Amer. Meteor. Soc., 71, 988-993.
7 8
Trenberth, K. E., and T. J. Hoar (1996), The 1990-1995 El Niño Southern Oscillation
9
event: Longest on record. Geophys. Res. Lett., 23, 57-60.
10 11
Trenberth, K. E., J. W. Stepaniak, J. W. Hurrell, and M. Fiorino (2001), Quality of
12
reanalyses in the tropics, J. Climate, 14, 1499-1510.
13 14
Trenberth, K. E., and L. Smith (2005), The mass of the atmosphere: A constraint on
15
global analyses, J. Climate, 18, 864-875.
16 17
Urban, F. E., J. E. Cole, and J. T. Overpeck (2000), Modification of tropical Pacific
18
variability by its mean state inferred from a 155-year coral record, Nature, 407, 989-
19
993.
20 21
Ventura, V., C. J. Paciorek, and J. S. Risbey (2004), Controlling the proportion of
22
falsely rejected hypotheses when conducting multiple tests with climatological data, J.
23
Climate, 17, 4343-4356.
24
38 1
Vincent, L. A., X. Zhang, B. R. Bonsal, and W. D. Hogg (2002), Homogenization of
2
daily temperature over Canada, J. Climate, 15, 1322-1334.
3 4
Vincent, L. A., et al. (2005), Observed trends in indices of daily temperature extremes
5
in South America 1960– 2000, J. Climate, 18, 5011 – 5023.
6 7
Walker, N. D., H. H. Roberts, L.J. Rouse Jr., and O.K. Huh (1982), Thermal history
8
of reef-associated environments during a record cold-air outbreak event, Coral Reefs,
9
1, 83-87
10 11
Wang, C. (2002a), Atmospheric circulation cells associated with the El Niño Southern
12
Oscillation, J. Climate, 15, 399-419.
13 14
Wang, C. (2002b), Atlantic climate variability and its associated atmospheric
15
circulation cells, J. Climate, 15, 1516-1536.
16 17
Wang, X. (2003), Comments on “Detection of undocumented changepoints: A
18
revision of the Two-Phase regression model”, J. Climate, 16, 3383-3385.
19 20
Wang, X., and Y. Feng (2004): RHTest User Manual. Available online at
21
http://cccma.seos.uvic.ca/ETCCDMI/RHTest/RHTestUserManual.doc
22 23
Wantanabe, T., A. Winter, T. Oba, R. Anzai, and H. Ishioroshi, (2002), Evaluation of
24
the fidelity of isotope records as an environmental proxy in the coral Monastraea,
25
Coral Reefs, 21, 169-178, DOI 10.1007/s00338-002-0218-9.
39 1
Woodruff, S. D., R. J. Slutz, R. L. Jenne, and P. M. Steurer (1987), A comprehensive
2
ocean-atmosphere data set, Bull. Am. Meteorol. Soc., 68, 1239 – 1250.
3 4
Woodruff, S. D., H. F. Diaz, J. D. Elms, and S. J. Worley (1998), COADS Release 2
5
data and metadata enhancements for improvements of marine surface flux fields,
6
Phys. Chem. Earth, 23, 517– 526.
7 8
Zhang, X., and F. Yang (2004). RclimDex User Manual. Available online at
9
http://cccma.seos.uvic.ca/ETCCDMI/RClimDex/RClimDexUserManual.doc
10 11
Zhang, X., et al. (2005), Trends in Middle East climate extreme indices from 1950 to
12
2003, J. Geophys. Res., 110, D22104, doi:10.1029/2005JD006181.
13 14
Zhang, Y., J. M. Wallace, and D. S. Battisti (1997), ENSO-like interdecadal
15
variability: 1900-93, J. Climate, 10, 1004-1020.
16 17 18 19 20 21 22 23 24 25
40 1
Figure Captions
2 3
Figure 1. Location of the stations used in the study. Stations obtained from Peterson
4
et al. (2002) and Aguilar et al. (2005) are depicted by circles (14 stations) and
5
triangles (21 stations) respectively. Stations obtained from the National Climatic Data
6
Center, Caribbean Institute for Meteorology and Hydrology and Caribbean
7
meteorological services are represented by crosses (6 stations). Some stations are
8
close enough for the overlap of their symbols.
9 10
Figure 2. Plots of SST, maximum and minimum temperature anomalies (oC) for: (a)
11
Bahamas (Freeport), (b) Honduras (Tela) and (c) Puerto Rico (Ponce). SSTs,
12
maximum temperature and minimum temperature are represented by black, red and
13
blue lines respectively.
14 15
Figure 3. Plots of monthly difference anomalies (oC) for: (a) Bahamas, (b) Honduras
16
(Tela) and (c) Puerto Rico (Ponce). Differences with respect to maximum and
17
minimum temperatures are represented by red and blue lines respectively.
18 19
Figure 4. Station distribution in relation to homogeneity results. Stations where no
20
significant change points were identified are depicted by circles (4 stations). Stations
21
where change points identified may be related to data issues are represented by
22
crosses (4 stations) and where change points may be related to climate shifts are
23
represented by squares (7 stations).
24
41 1
Figure 5. Composite difference for 1968 of (a) 1000 hPa air temperature (oC), (b) 500
2
hPa vertical velocity (Pa/s), (c) divergent wind (m/s) and vorticity (m2/s) at sigma
3
level 0.995 and (d) divergent wind and vorticity at sigma level 0.2101. Broken
4
contours are negative. Differences in contoured values significant at the 1% (5%)
5
level are indicated by dark (light) shading.
6 7
Figure 6. Plot of annual North Atlantic Oscillation (NAO) anomalies (dashed line and
8
circles) versus annual average temperature anomalies over the Caribbean and adjacent
9
Caribbean (solid line and crosses).
10 11
Figure 7. Same as Figure 5 but for 1983.
12 13
Figure 8. Same as Figure 5 but for 1979.
14 15
Figure 9. Location of station for which common homogeneous periods are identified.
16
Stations with maximum and/or minimum temperatures for 1970-92 are represented by
17
squares (12 stations). Stations with maximum and/or minimum temperatures for 1984-
18
98 are represented by crosses (9 stations).
19 20
Figure 10.
21
(dashed line) temperatures are (a) at or above the 90th percentile) and (b) at or below
22
the 10th percentile. Percentiles determined by homogeneous data from 1970 though
23
1992.
24 25
The percent of days when the maximum (solid line) and minimum
42 1 Tables 2 Table 1. Temperature stations used in the study. Data are sourced from Peterson et al. 3 (2002) (PTD), Aguilar et al (2005) (APO), the National Climatic Data Center (NCDC), 4 the Caribbean Institute for Meteorology and Hydrology (CIMH) and the local 5 meteorological services. 6 Country
Station Name
Latitude
Longitude
Elevation
Data span
Source
(m) Bahamas
FREEPORT
26.55
-78.70
11
1968-1999
PTD
Bahamas
NASSAU
25.05
-77.47
7
1973-2004
NCDC
13.07
-59.48
56
1973-2006
NCDC
AIRPORT Barbados
GRANTLEY ADAMS
Ba8rbados
HUSBANDS
13.17
-59.59
-999
1969-1999
PTD
Belize
PSWGIA
17.53
-88.30
5
1961-2003
APO
Belize
CENTRAL FARM
17.31
-88.12
61
1966-1999
PTD
Cayman
GRAND
19.17
-81.21
3
1976-1999
PTD
Islands
CAYMAN
Costa Rica
CATIE
9.90
-83.75
0
1961-2003
APO
Costa Rica
FABIO BAUDRIT
10.00
-84.25
0
1961-2003
APO
Cuba
CASA BLANCA
23.17
-82.35
50
1961-2003
APO
Cuba
GUANTANAMO
19.90
-75.15
16
1958-1998
PTD
18.48
-69.92
14
1958-1999
PTD
25.83
-81.38
2
1958-1999
PTD
BAY
7
Dominican
SANTO
Republic
DOMINGO
U.S.-Florida
EVERGLADES
43 1
Table 1. (continued)
2 Country
Station Name
Latitude
Longitude
Elevation
Data span
Source
(m) U.S.-Florida
KEY WEST WSO 24.55
-81.75
1
1958-1999
PTD
AIRPORT Guadeloupe
LE RAIZET
16.27
-61.60
11
1951-2000
NCDC
Guatemala
FLORES
16.51
-89.87
123
1961-2003
APO
Jamaica
SANGSTER
18.50
-77.92
8
1973-2006
NCDC
Jamaica
WORTHY PARK
18.15
-77.17
550
1961-1999
Met Service
Honduras
CATACAMAS
14.90
-85.93
442
1961-2003
APO
Honduras
LA CEIBA
15.73
-86.87
26
1961-2003
APO
Honduras
LA MESA
15.45
-87.93
31
1961-2003
APO
Honduras
SANTA ROSA DE 14.78
-88.78
1079
1961-2003
APO
COPAN Honduras
TEGUCIGALPA
13.50
-87.22
1007
1961-2003
APO
Honduras
TELA
15.72
-87.48
3
1961-2003
APO
Nicaragua
RIVAS
11.42
-85.83
70
1961-2003
APO
Panama
ANTON
8.35
-80.27
33
1961-2003
APO
Panama
DAVID
8.40
-82.42
27
1961-2003
APO
U.S.-Puerto
LARES
18.27
-66.85
445
1958-1998
PTD
PONCE
18.02
-66.52
21
1958-1999
PTD
SAN JUAN WSFO
18.43
-66.00
3
1958-1999
PTD
Rico U.S.-Puerto Rico U.S.-Puerto Rico
44 1
Table 1. (continued)
2 Country
Station Name
Latitude
Longitude
Elevation
Data span
Source
1958-1998
PTD
1982-2006
Met
(m) U.S.-Puerto
UTUADO
18.25
-66.68
CHARLES 14.02
-61.00
159
Rico St. Lucia
GFL
AIRPORT St. Vincent
ST. VINCENT
Service 13.13
-61.20
13
1987-2006
PTD, CIMH, Met Service
Trinidad
PIARCO IAP
10.37
-61.21
15
1959-1999
PTD
Venezuela
GUIRIA
10.58
-62.32
14
1961-2003
APO
Venezuela
LA CARLOTA
10.50
-66.88
835
1961-2003
APO
Venezuela
MAIQUETIA Apt. 10.60
-66.98
48
1961-2003
APO
BOLIVAR
3 4 5 6 7 8
Venezuela
MARACAY
10.25
-67.65
437
1961-2003
APO
Venezuela
MENE GRANDE
9.82
-70.93
28
1961-2002
APO
Venezuela
MERIDA
8.60
-71.18
1498
1961-2003
APO
Venezuela
TUMEREMO
7.30
61.45
181
1961-2003
APO
45 1
Table 2. Correlations between sea surface temperatures and maximum and minimum
2
temperatures respectively for 1970-1998. Values in bold are significant at the 95%
3
level.
4 Country
Station
Maximum
Minimum
Temperature
Temperature
Bahamas
Freeport
0.93
0.91
Bahamas
Nassau Airport
0.93
0.91
Barbados
Grantley Adams
0.74
0.66
Barbados
Husbands
0.58
0.72
Belize
PSWGIA
0.64
0.65
Belize
Central Farm
0.45
0.80
Cayman Islands
Grand Cayman
0.89
0.79
Costa Rica
Catie
0.63
0.59
Costa Rica
Fabio Baudrit
0.09
0.57
Cuba
Casa Blanca
0.89
0.94
Cuba
Guantanamo
0.66
0.83
Bay
5
Dominican Republic
Santo Domingo
0.82
0.81
Florida
Everglades
0.91
0.92
Florida
Key West
0.92
0.91
Guadeloupe
Le Raizet
0.83
0.78
Guatemala
Flores
0.47
0.86
Honduras
Catacamas
0.29
0.66
46 1
Table 2. (continued)
2 Country
Station
Maximum
Minimum
Temperature
Temperature
Honduras
La Ceiba
0.58
0.76
Honduras
La Mesa
0.44
0.77
Honduras
Santa
0.49
0.88
Rosa
De
Copan Honduras
Tegucigalpa
0.50
0.82
Honduras
Tela
0.66
0.83
Jamaica
Sangster
0.77
0.85
Jamaica
Worthy Park
0.67
0.74
Nicaragua
Rivas
0.55
0.58
Panama
Anton
0
0.38
Panama
David
0.57
0.68
Puerto Rico
Ponce
0.78
0.81
Puerto Rico
San Juan
0.76
0.84
Puerto Rico
Utuado
0.72
0.86
St. Lucia
GFL
0.74
0.76
Charles
Airport
3
St. Vincent
St. Vincent
0.71
0.85
Trinidad
Piarco IAP
0.41
0.66
Venezuela
Guiria
0.29
0.44
Venezuela
La Carlota
0.08
0.43
47 1
Table 2. (continued)
2 Country
Venezuela
Station
Maiquetia
Apt.
Maximum
Minimum
Temperature
Temperature
0.73
0.69
Bolivar
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Venezuela
Maracay
-0.39
0.34
Venezuela
Mene Grande
-0.10
0.05
Venezuela
Merida
0.29
0.53
Venezuela
Tumeremo
0.51
0.49
48 1
Table 3. Change points obtained from homogeneity assessments of adjacent monthly
2
sea surface temperatures and maximum and minimum temperatures at each station.
3
Values in bold are significant at the 99% level.
4 Country
Station
Sea Surface
Maximum
Minimum
Temperatures
Temperature
Temperature
Year
Year
Size
Size
(oC)
Size
(oC)
(oC)
Bahamas
Freeport
-
-
-
-
-
-
Bahamas
Nassau Airport
-
-
1995
-0.96
1987
1.08
Barbados
Grantley Adams
1970
-0.77
1984
-0.63
1983
-0.92
1983
-0.44
1996
-0.54
1970
-0.77
-
-
-
-
1983
-0.44
1971
-0.21
-
-
1974
-0.69
1983
-0.83
1971
-0.21
1985
-1.90
-
-
1983
-0.83
-
-
-
-
-
-
1973
-0.43
1982
-0.22
1983
-0.52
1984
-0.20
1993
1.48
1973
-0.56
1981
0.54
1979
1.62
1984
-0.34
1992
1.18
-
-
-
-
Barbados
Belize
Belize
Husbands
PSWGIA
Central Farm
Cayman Islands
Grand Cayman
Costa Rica
Catie
Costa Rica
Cuba
5
Year
Fabio Baudrit
Casa Blanca
-
-
49 1
Table 3. (continued)
2 Country
Station
Sea Surface
Maximum
Minimum
Temperatures
Temperature
Temperature
Year
Year
Year
Size
Size
(oC) Cuba
Guantanamo
1970
(oC)
-0.40
Bay Dominican
Santo Domingo
(oC)
1968
-1.07
1980
2.22
1981
-0.73
1985
1.06
1972
0.22
1983
-0.53
1970
-0.69
1983
-0.31
States Everglades
-
-
1968
-1.29
-
-
States Key West
-
-
-
-
-
-
1970
-0.74
1967
0.72
1979
0.78
1983
-0.51
1977
0.74
1990
0.69
Republic United
Size
(Florida) United (Florida) Guadeloupe
Guatemala
Flores
1983
-0.52
-
-
-
-
Honduras
Catacamas
1971
-0.42
1976
0.94
1984
-1.23
1983
-0.46
1971
-0.26
1990
1.3
1977
1.51
1983
-0.75
1971
-0.24
1975
-0.62
1973
-0.22
1983
-0.69
1983
-0.85
Honduras
Honduras
3
Le Raizet
La Ceiba
La Mesa
50 1
Table 3. (continued)
2 Country
Station
Sea Surface
Maximum
Minimum
Temperatures
Temperature
Temperature
Year
Year
Size
Size
(oC) Honduras
Santa Rosa De
Year
(oC)
Size (oC)
1973
0.27
1973
-1.04
1984
-0.61
1973
0.46
1983
-0.88
1976
1.56
1990
0.58
1971
-0.24
1975
-0.62
1973
-0.22
1983
-0.69
1983
-0.85
1970
-0.49
1983
-0.42
1970
-0.49
1983
-0.42
1973
0.51
1990
0.84
1971
-0.4
1984
-0.31
1971
-0.33
1984
-0.22
1970 1983
Copan Honduras
Honduras
Jamaica
Jamaica
Nicaragua
Panama
Panama
Puerto Rico
3
Tegucigalpa
Tela
Sangster
Worthy Park
Rivas
Anton
David
Ponce
1992
0.67
1989
0.86
1986
0.68
1976
1.06
1987
1.21
1987
0.68
1987
0.40
1983
-0.75
1993
-0.56
-
-
1983
-0.54
-0.80
1969
-1.58
1969
-1.95
-0.36
1983
-0.83
1983
-0.88
51 1
Table 3. (continued)
2 Country
Station
Sea Surface
Maximum
Minimum
Temperatures
Temperature
Temperature
Year
Size
Year
Size
(oC) Puerto Rico
Puerto Rico
St. Lucia
St. Vincent
Trinidad
Venezuela
Venezuela
Venezuela
Venezuela
3
San Juan
1970
-0.80
1983
-0.36
1970
-0.80
1983
-0.36
1970
-0.78
Airport
1983
-0.50
St. Vincent
1970
-0.77
1983
-0.47
1970
Utuado
GFL
Charles
Year
Size
(oC)
1977
1972
(oC)
1.04
0.85
1973
-0.42
1983
-2.05
1970
-1.20
1983
-0.50
-
-
1995
0.79
-
-
-
-
-0.66
1976
0.71
1979
0.56
1981
-0.35
1987
0.89
1971
-0.61
1976
1.37
1971
-0.89
1983
-0.40
1987
1.09
1981
-1.35
1971
-0.62
1976
1.06
1979
1.29
1983
-0.59
1986
1.36
Maiquetia Apt.
1971
-0.62
1976
1.06
1979
1.29
Bolivar
1983
-0.59
1986
1.36
Maracay
1971
-0.65
1985
0.68
1980
0.92
1983
-0.52
1990
-0.52
Piarco IAP
Guiria
La Carlota
52 1
Table 3. (continued)
2 Country
Station
Sea Surface
Maximum
Minimum
Temperatures
Temperature
Temperature
Year
Size
Year
(oC) Venezuela
Venezuela
Venezuela
3 4 5 6 7 8 9 10 11 12 13 14 15
Mene Grande
Merida
Tumeremo
Size
Year
(oC)
Size (oC)
1971
-0.51
1973
-0.33
1976
0.34
1983
-0.36
1988
-0.91
1989
1.46
1971
-0.51
1980
0.94
1976
0.50
1983
-0.36
1989
0.82
1971
-0.51
1971
-1.54
1972
-0.4
1990
-1.66
1985
0.81
53 1
Table 4. Same as Table 3 but for monthly difference anomalies for maximum
2
temperature and SST, and minimum temperature and SST respectively. Possible only
3
for stations with correlation of greater than 0.70 with respect to SSTs. Italicized
4
values indicate stations with correlations of 0.65-0.69 for one of the temperature
5
series and greater than or equal to 0.70 for the other series.
6 Country
Station
Maximum
Minimum
Temperature
Temperature
Difference
Difference
Anomalies
Anomalies
Size (oC)
Year
Year
Size (oC)
Bahamas
Freeport
1982
-0.41
1982
0.59
Bahamas
Nassau Airport
1994
-0.95
1987
0.92
Barbados
Grantley Adams
1984
-0.1
1983
-0.55
1994
-0.46
1994
-0.14
Barbados
Husbands
1979
0.53
Belize
Central Farm
1987
0.88
Cayman
Grand Cayman
Cuba
Casa Blanca
Cuba
Dominican
Guantanamo Bay
Santo Domingo
Republic Florida
Everglades
-
-
1986
0.55
1971
0.37
1971
0.29
1981
0.43
1993
-0.36
1968
-0.82
1970
0.53
1981
2.21
1982
1.23
1971
0.49
1971
0.51
1981
-0.64
1989
-0.50
1968
-1.17
1979
0.62
54 1
Table 4. (continued)
2 Country
Station
Maximum
Minimum
Temperature
Temperature
Difference
Difference
Anomalies
Anomalies
Year Florida
Guadeloupe
Le Raizet
Year
Size
1968
-0.18
1968
-0.08
1984
0.37
1982
-0.57
1969
0.69
1967
-0.23
1979
0.52
1979
0.61
1989
0.09
Guatemala
Flores
1983
0.65
Honduras
La Ceiba
1977
0.96
1989
0.03
1979
-0.51
1991
-1.24
1973
-0.35
Copan
1984
-0.46
Tegucigalpa
1975
1.25
1985
-0.86
1977
-0.21
1988
0.84
Honduras
Honduras
Honduras
Honduras
Jamaica 3
Key West
Size
La Mesa
Santa
Tela
Sangster
Rosa
De
1975
-1.07
1990
-0.36
1991
0.39
55 1
Table 4. (continued)
2 Country
Station
Maximum
Minimum
Temperature
Temperature
Difference
Difference
Anomalies
Anomalies
Year Jamaica
Puerto Rico
Puerto Rico
Puerto Rico
St. Lucia
Worthy Park
Ponce
San Juan
Utuado
GFL
Charles
Size
Year
Size
1981
-0.59
-1.00
1969
-1.24
1986
-0.59
1983
-0.51
1971
1.56
1970
1.32
1981
0.63
1983
-1.27
1973
0.73
1968
-0.70
1986
-0.27
1992
-0.22
1993
0.47
-
-
-
-
1978
1.21
1971
0.15
1982
0.78
1969
Airport
3 4 5 6 7
St. Vincent
St. Vincent
Venezuela
Maiquetia Apt.
1975
0.79
Bolivar
1985
1.25
56 1
Figures
2
Figure 1. Location of the stations used in the study. Stations obtained from Peterson et
3
al. (2002) and Aguilar et al. (2005) are depicted by circles (14 stations) and triangles
4
(21 stations) respectively. Stations obtained from the National Climatic Data Center,
5
Caribbean Institute for Meteorology and Hydrology and Caribbean meteorological
6
services are represented by crosses (6 stations). Some stations are close enough for the
7
overlap of their symbols.
8 9
10 11 12 13 14 15 16 17 18
57 1
Figure 2. Plots of SST, maximum and minimum temperature anomalies (oC) for: (a)
2
Bahamas (Freeport), (b) Honduras (Tela) and (c) Puerto Rico (Ponce). SSTs,
3
maximum temperature and minimum temperature are represented by black, red and
4
blue lines respectively. Units are oC.
5
(a)
6 7
(b)
8 9
(c)
10
58 1
Figure 3. Plots of monthly difference anomalies (oC) for: (a) Bahamas, (b) Honduras
2
(Tela) and (c) Puerto Rico (Ponce). Differences with respect to maximum and
3
minimum temperatures are represented by red and blue lines respectively. Units are
4
o
5
(a)
C.
6 7
(b)
8 9
10
(c)
59 1
Figure 4. Station distribution in relation to homogeneity results. Stations where no
2
significant change points were identified are depicted by circles (4 stations). Stations
3
where change points identified may be related to data issues are represented by
4
crosses (4 stations) and where change points may be related to climate shifts are
5
represented by squares (7 stations).
6 7
8 9 10 11 12 13 14 15 16 17 18
60 1
Figure 5. Composite difference for 1968 of (a) 1000 hPa air temperature (oC), (b) 500
2
hPa vertical velocity (Pa/s), (c) divergent wind (m/s) and vorticity (m2/s) at sigma
3
level 0.995 and (d) divergent wind and vorticity at sigma level 0.2101. Broken
4
contours are negative. Differences in contoured values significant at the 1% (5%)
5
level are indicated by dark (light) shading.
6 7
(a)
(b)
(c)
(d)
8 9 10
11 12
61 1
Figure 6. Plot of annual North Atlantic Oscillation (NAO) anomalies (dashed line and
2
circles) versus annual average temperature anomalies over the Caribbean and adjacent
3
Caribbean (solid line and crosses).
4
5 6 7 8 9 10 11 12 13 14 15 16 17
62 1
Figure 7. Same as Figure 5 but for 1983.
2 3
(a)
(b)
(c)
(d)
4 5 6
7 8 9 10 11 12 13
63 1
Figure 8. Same as Figure 5 but for 1979.
2 3
(a)
(b)
(c)
(d)
4 5 6
7 8 9 10 11
64 1
Figure 9. Location of station for which common homogeneous periods are identified.
2
Stations with maximum and/or minimum temperatures for 1970-92 are represented by
3
squares (12 stations). Stations with maximum and/or minimum temperatures for 1984-
4
98 are represented by crosses (9 stations).
5 6
7 8 9 10 11 12 13 14 15 16 17 18
65 1
Figure 10. The percent of days when the maximum (solid line) and minimum (dashed
2
line) temperatures are (a) at or above the 90th percentile) and (b) at or below the 10th
3
percentile. Percentiles determined by homogeneous data from 1970 though 1992.
4 5
(a)
6 7 8
9
(b)
