Water Temperature Variation and the Meteorological and
Hydrographic Environment of Bocas del Toro, Panama
KARL W. KAUFMANN AND RICARDO C. THOMPSON
Smithsonian Tropical Research Institute, PO Box 2072, Balboa, Republic of Panama
Corresponding author: kaufmank@si.edu
ABSTRACT.?Bah?a Almirante is a shallow lagoon on the Caribbean coast of western Panama almost
entirely surrounded by land. Rainfall is most intense during the night and least intense in the late afternoon,
a pattern common in tropical coastal areas.Water temperatures are often elevated in the inshore waters
relative to surface temperatures immediately offshore, at times exceeding 30?C. Analysis of solar radiation,
wind speed, humidity and air temperature indicate that variations in solar radiation and wind speed were
responsible for much of the observed excursions from the offshore temperatures. Environmentally stressful
temperatures can result from a month or two of clear skies, and an equal period of cloudy skies can bring the
temperatures down again rapidly. Shallow water has the most extreme daily and annual ranges in tempera-
ture, but water up to 20 m shows a similar range in temperatures over periods of several years. Salinity at the
surface is usually 30 to 34 ppt, but can drop to as low as 20 ppt after heavy rain. Historical records of monthly
rainfall explain only 9.6% of the variation in monthly water temperature changes.There appears to be a
thermal gradient in the bay and adjacent areas across three sites with for which we have 4 to 6 years of hourly
temperature data. The innermost site, closest to the mainland, had the highest mean temperature and the
highest range in temperature. The two sites on the seaward side of the bay had less extreme temperatures.
KEYWORDS.?rainfall, water temperature, solar radiation, wind speed, Central America, temperature stress,
coral bleaching
INTRODUCTION
Bah?a Almirante is a shallow lagoon
about 20 by 30 km bounded by a large
coastal swamp and mangrove forest to the
northwest, the archipelago of Bocas del
Toro to the north and southeast, and the
mainland to the southwest. There are no
large marine passages to the Caribbean on
the north or to the Laguna de Chiriqu? to
the southeast. The mountains of the central
cordillera to the south and east are the
highest in Panama, but a low ridge about 5
km from the southwestern coast diverts
runoff from the mountains to either the La-
guna de Chiriqu? or to the Caribbean. Ac-
cordingly, the watershed of the bay is very
restricted, approximately the same size as
the bay and about half of that consists of
coastal wetlands. Chiriqu? Lagoon to the
southeast is similarly isolated from oceanic
influences, but has several major rivers
with sources in the central cordillera drain-
ing into it. Both lagoons have significantly
higher concentrations of nutrients, chloro-
phyll, and zooplankton biomass than the
open ocean, but Bah?a Almirante has sub-
stantially less (D?Croz et al. 2004) and in
addition has a thriving coral reef fauna not
found in much of the more turbid Chiriqu?
Lagoon (Guzm?n and Guevara 1998).
Offshore sea surface temperatures (SSTs)
in the Caribbean are affected by the annual
movements of the Intertropical Conver-
gence Zone (ITCZ) and on an intraseasonal
time scale, by the movement of tropical
waves westward across the Atlantic. Varia-
tions in insolation and other weather
related phenomena result in offshore
SST anomalies, and ultimately, offshore
temperatures are a major determinate of
inshore temperatures. However, local
weather conditions varies in a complex
way in Central America, as evidenced by
the great variation in rainfall patterns, and
the resulting differences in weather can be
expected to affect local water temperatures.
The Environmental Sciences Program
(ESP) at the Smithsonian Tropical Research
Institute (STRI) laboratory in Bocas del
Caribbean Journal of Science, Vol. 41, No. 3, 392-413, 2005
Copyright 2005 College of Arts and Sciences
University of Puerto Rico, Mayagu?ez
392
Toro has been collecting meteorological
and oceanographic data since 1999 and
other historical rainfall records are avail-
able going back to 1926. Our objective is to
describe the physical conditions affecting
the local fauna and flora, particularly those
related to temperature stress, and to de-
velop baseline values so that episodes of
extreme conditions will be recognized. We
describe the interannual to diurnal patterns
of change in meteorological parameters
and how they affect water temperatures in
Bah?a Almirante and nearby waters.
MATERIALS AND METHODS
Laboratory and instrument locations
The laboratory at Bocas del Toro is lo-
cated on a narrow land peninsula connect-
ing the town of Bocas del Toro with the
remainder of Isla Col?n, a low lying island
that forms part of the archipelago of Bocas
del Toro (Fig. 1). On the northwest side of
this neck of land is a bay opening to the
Caribbean Sea. On the southeast side is a
shallow bay, surrounded by mangroves,
opening to the Bah?a Almirante. Three per-
manent biological monitoring sites are lo-
cated near the laboratory (Guzm?n et al.
2005). A mangrove and a grassbed site are
several hundred meters northwest of the
pier used by the laboratory. A coral reef site
is about 500 m east just outside the bay. A
meteorological tower on a platform about
100 m east of the pier is used to collect me-
teorological and oceanographic param-
eters.
Water temperature
This is recorded hourly with individual
underwater temperature loggers (Hobo
StowAway TidbiT and Hobo Water Temp
Pro, Onset Computer Corp., Bourne, Mass.)
attached to the bottom at four sites, Col?n
Coral Reef, Col?n Grassbed, Cayo Rold?n,
and Cayo Agua (Fig. 1). Each logger is cali-
brated annually for several days in a water
bath at about 21?C against a glass NBS
traceable thermometer and all measure-
ments are adjusted by the calibration. The
accuracy of the data can be considered to be
at least ? 0.25?C. Data collection started
at the Col?n Coral Reef in March, 1999 at 1
m, 4 m, 10 m, and 20 m depth and at 2 m
the Col?n Grassbed. The 1 m site was dis-
continued in 2000. Water temperature data
has also been collected from 4 meters at
Cayo Agua and Cayo Rold?n starting in
December, 2000 and March, 1999, respec-
tively. Temperatures from March 25, 1999
to September 11, 2000 were also collected at
1 m, 10 m, and 20 m at Cayo Rold?n. Cayo
Agua is an island separating the Chiriqu?
Lagoon from the open ocean and the water
temperature site there is in a coral reef on
the south side of the island. Cayo Rold?n is
a mangrove island that forms part of a bar-
rier enclosing a smaller bay (9 km by 5 km)
on the south side of Bah?a Almirante and
bounded on its southwest side by the main-
land. Offshore water temperatures are
weekly means of a one degree quadrat cen-
tered at 81.5 W, 9.5 N. (IGOSS 2004). A
small part of the southwest corner of this
quadrat intersects the mouth of the
Chiriqu? Lagoon near the Cayo Agua site.
Rainfall
We have four sources of rainfall data.
The longest is a record of monthly rainfall
from January 1926 to October, 2004, col-
lected by the Chiriqu? Land Company in
Changuinola (Finca 8), R.P., about 33 km
northwest of Bocas del Toro (here identi-
fied as the Changuinola data). Changuinola
is 8 km from the coast at an elevation of 20
m and surrounded by banana plantations.
These data were collected daily by the
overseer of the plantation in a collector
similar to that used by the U.S. Forest Ser-
vice and phoned in to the main office every
day. The data were considered very impor-
tant since they were used to predict the size
of the banana crop, and it is likely that care
was taken in their collection (Clyde Ste-
phens pers. comm.). No extreme values or
other inconsistencies were found in the
data.
Starting in 1972, rainfall data were col-
lected at the airport in Bocas del Toro (here
identified as the Airport data) by the Insti-
tuto de Recursos Hidraulicos y Electrifi-
caci?n, the state owned electric company.
Monthly and annual summaries were pub-
PHYSICAL CONDITIONS OF BOCAS DEL TORO 393
lished in Estad?stica Paname?a (1973-1994).
In 1998, after the company was privatized,
data collection and distribution were taken
over by Empresa de Transmisi?n Electrica
S. A. We have daily airport rainfall data
supplied by them from 1989 to 2000 and
monthly data for 2001 to 2003. The airport
is less than 2 km from the STRI station.
The STRI ESP started collecting daily
rainfall data in a US Forest Service type rain
gauge on July 1, 2000 at the site of the main
laboratory in Bocas del Toro (here called
the Daily ESP data). Rainfall is recorded
each morning at about 8 AM to the nearest
0.2 mm. On May 16, 2002, the ESP started
collecting hourly data with a Vitel VRG-
TB3 Tipping bucket (here called the Tip-
ping Bucket data). Rainfall is recorded in
increments of 2.54 mm. The tipping bucket
is attached to the ESP meteorological tower
near the STRI pier.
In order to confirm the accuracy of the
various rainfall records, annual and
monthly totals were compared with linear
regressions where overlapping data exists.
The monthly total of the Daily ESP data
was very similar to the monthly total of the
Tipping Bucket data (N = 30, r2 = 0.980,
slope = 1.124, intercept set to 0, P < 0.0005).
That the Daily ESP data records about 12%
more rain than the Tipping Bucket data is
not unexpected since tipping buckets tend
to underestimate during periods of heavy
rainfall. The monthly totals for the Airport
FIG. 1. Smithsonian Tropical Research Institute laboratory at Bocas del Toro, Panama and surrounding area.
KARL W. KAUFMANN AND RICARDO C. THOMPSON394
data matched the Daily ESP data well (N =
45, r2 = 0.92, intercept = 1.7, slope = 0.881).
The monthly totals for the Changuinola
data also matched the monthly Airport
data from 1989 to June, 1998 well (N = 113,
r2 = 0.71, intercept = 10.5, slope = 0.813).
Two months from the Airport data and one
from the Daily ESP data were excluded
from the analysis based on the large residu-
als in the Airport/Daily ESP regression and
the lack of concordance with the Chan-
guinola data. Estimates for the two missing
Airport months were calculated from the
regression on the Daily ESP data and sub-
stituted before proceeding with further
analysis. A reduced major axis regression
on the yearly totals from Changuinola and
the Airport data (1973-2003, N = 28, r2 =
0.762, intercept = 943, slope = 0.835)
showed that the Airport station averaged
about 300 to 600 mm more than the Chan-
guinola station, with the largest differences
during drier years.
Air temperature and relative humidity
These were recorded hourly between Jan
1 and Dec 31, 2003 with a Viasala HMP45
relative humidity and temperature sensor
attached to the ESP meteorological tower
about 3 m above the water. Air tempera-
tures were compared to weekly measure-
ments taken with a digital thermometer at
the rainfall station and to hourly measure-
ments recorded in a nearby mangrove for-
est.
Wind speed and direction
These were recorded with an R. M.
Young Model 05103 anemometer (R. M.
Young Co., Traverse City, Michigan) at-
tached to the top of the ESP tower 7 m
above the water. Each hour, the scalar
mean wind speed and the wind direction at
that time were recorded. Data collection
started on August 20, 2002. Wind direction
and speed are checked on the ground sev-
eral times a year and compared to the in-
stantaneous reading from the anemometer
and the direction on a hand held compass.
This is to ensure that there hasn?t been any
major drift from the calibration at installa-
tion.
The anemometer is not sufficiently el-
evated to remove the interference from
trees around the lab or from the hills on Isla
Col?n from the NE to the NW quadrats.
Strong trade winds from the NE develop
further east along the coast from January to
March (Robertson et al. 1999) and are evi-
dent at Bocas del Toro on trips to more
open areas, but these winds are not well
recorded at the laboratory site. The data are
probably representative of wind conditions
on the lee side of the island, but not the
large scale wind pattern.
Solar radiation
This is recorded using a Licor LI-200 py-
ranometer (LI-COR, Inc., Lincoln, Ne-
braska). Data were recorded as the total ra-
diation for each hour in W/m2 based on
samples every 5 seconds. Two pyranom-
eters attached to arms sticking out from the
north and south sides of the tower are used.
To avoid shadows from the tower, data
from the south side are used from March to
September, and from the north side for the
remainder of the year, with the opposite
side then serving as a backup. The pyra-
nometer on the south side was not installed
until Jan 30, 2003. To ensure consistency
and to check against degradation of the
sensors, these pyranometers are calibrated
once a year against a set of Licor pyranom-
eters that we keep in the laboratory for the
purpose. All data are corrected for devia-
tion from the standards.
Salinity
It is measured every Wednesday be-
tween 1000 and 1200 EST at 0.5 m depth at
the grassbed adjacent to the laboratory and
at a nearby coral reef. Measurements are
either taken with an A & O Model 10419
temperature compensated hand held re-
fractometer or using a YSI Model 85 hand-
held sensor (YSI Inc., Yellow Springs,
Ohio). Both the refractometer and the YSI
were calibrated at least annually with dis-
tilled water and Copenhagen water or its
PHYSICAL CONDITIONS OF BOCAS DEL TORO 395
equivalent. The refractometer is accurate to
? 1 ppt, the YSI to ? 0.3 ppt.
Secchi disk
These readings were taken at the same
location and time as the salinity readings.
At the grassbed site, they are taken hori-
zontally at 0.5 m depth and the disk extinc-
tion distance is read in meters with a dive
mask. At the coral reef site, the measure-
ment is taken vertically and the extinction
is read in meters from above the water sur-
face.
Statistical Analyses
Analysis and statistics were done with
Microsoft Visual Foxpro 8.0 and Systat 9.0.
RESULTS
Interannual and regional patterns for rainfall
and water temperature
Annual total rainfall for the four data
sources is shown in Figure 2. The data
show maxima and minima during the same
years, but the Bocas del Toro totals are con-
sistently higher than Changuinola, demon-
strated above. The annual mean at Chan-
guinola is 2615 mm (N = 78, SD = 550 mm)
while that at Bocas del Toro is 3277 mm
(N = 28, SD = 461 mm).
Rainfall at Changuinola for 2002 was
4024 mm, the highest in 78 years. This
value is matched by a total of 4255 mm for
the Airport data, also the highest of the 28
years for which we have annual totals, and
4979 mm for the Daily ESP data.
Water temperatures at the three sites
were monitored from 1999 to 2004. Tem-
peratures at Isla Col?n, Cayo Rold?n, and
Cayo Agua, rose and fell together, even
when examined on a scale of a week or less
(Fig. 3A). Temperatures varied from an
hourly low of 26.2?C at Isla Col?n to a high
of 32.1?C at Cayo Rold?n. At the end of
each year, temperatures at all sites fell from
2 to as much as 5 degrees from what was
often the annual high in September or Oc-
tober to the annual low in December or the
first two months of the next year. In most
years, there was also a peak in May or June
followed by lows in July and August (Fig.
4). The offshore IGOSS temperatures fol-
lowed a pattern similar to the three sites we
monitored, but the inshore waters often
had higher highs and less commonly, lower
lows. The largest changes in mean daily
temperature at 4 m at Isla Col?n over ap-
proximately a month?s time were a drop of
2.8?C in 2001 over 30 days starting on No-
vember 5 and a rise of 2.8?C in 2002 over 27
days starting on May 8.
The mean temperatures from 1999 to
2004 at all sites and depths (Table 1) were
very similar, ranging from 28.5?C at the
Col?n Coral reef 10 m and 20 m sites to
FIG. 2. Annual rainfall records from Changuinola and Bocas del Toro. Annual total rainfall from four data
sources. Missing connectors between data points indicate missing data for a year. For description of data
sources, see methods section.
KARL W. KAUFMANN AND RICARDO C. THOMPSON396
29.1?C at the Rold?n 4 m site. The means of
the coldest two months of the year, January
and February, were all within 0.5?C of each
other (27.3?C to 27.9?C). For the warmest
two months of the year (September and Oc-
tober), the mean for Cayo Rold?n at 4 m
(30.5?C) was 0.9?C higher than that for Isla
Col?n at 4 m (29.6?C), not a great amount
FIG. 3. Regional water temperatures: Three sites and offshore, 1999 to 2004. A: Daily mean water temperatures
at 4 meters combined with 1 week smoothed offshore water temperatures (circles) from IGOSS database (81.5
W, 9.5 N). The horizontal line indicates the mean offshore temperature for 1999-2003 28.3?C. B. Difference in
mean monthly water temperature at Cayo Agua and Col?n for each month, 1999 to 2004. C. Same, for Rold?n
and Col?n. Line in B and C is LOWESS smoother.
PHYSICAL CONDITIONS OF BOCAS DEL TORO 397
but significant in view of the temperature
differences known to damage corals (be-
low). The mean for Cayo Agua (29.7?C)
was similar to Isla Col?n. The mean tem-
perature offshore was slightly colder (0.2 to
0.8?C) than all of the other sites. In the 23
year IGOSS record, the months with the
lowest and highest mean temperatures
were February (27.8?C) and September
(28.9?C), respectively; thus the two-month
means for the IGOSS data in Table 1 are a
good representation of the coldest and
warmest periods offshore. The January-
February IGOSS mean matches that of
Cayo Agua and is slightly warmer than the
remaining sites. The September-October
offshore temperature however, is just 1.1?C
warmer than the coldest months, while all
of the inshore means are over 2?C warmer
than their coldest months.
To determine whether water tempera-
tures among the three sites were more dif-
ferent at some times of the year than others,
we calculated the mean temperature for
each month and examined the differences
among them. (Fig. 3B, C) .The monthly dif-
ferences between Cayo Agua and Isla
Col?n (Fig. 3B) were greatest during the be-
ginning and end of the year, when Cayo
Agua was as much as 0.5?C warmer. The
monthly differences between Cayo Rold?n
and Col?n were higher for the first 10
FIG. 4. Water temperatures at different depths: Isla Col?n, 1999 to 2004. A: Running one week mean water
temperatures at 4 m from Isla Col?n. B. Same, for 20 m depth. In the legend, the column to the right is the total
rain during the year from the Airport data. Heavier lines correspond to larger annual totals. C. Difference in
mean monthly water temperature between 2 m (grassbed) and 20 m (coral reef) depths at Isla Col?n. D. Same,
for 4 m minus 20 m differences. Line in C and D is LOWESS smoother (Systat 9.0 SPSS, Inc., Chicago)
KARL W. KAUFMANN AND RICARDO C. THOMPSON398
months of the year and lower for Novem-
ber and December, but the ANOVA was
not significant (Cayo Agua minus Col?n,
N = 38, r2 = 0.723, P < 0.001, Rold?n minus
Col?n, N = 33, r2 = 0.416, N.S.).
For all months combined, the mean dif-
ference in monthly means between Cayo
Agua and Isla Col?n was 0.16?C and the
maximum difference was 0.63?C in Decem-
ber, 2002 (t-test Cayo Agua minus Col?n,
N = 38, P < 0.0001). The mean difference
between Cayo Rold?n and Isla Col?n was
0.72?C and the maximum was 1.21?C in
May, 2001 (Rold?n minus Col?n t-test, N =
33, P < 0.0001).
To contrast interannual differences in
temperatures at different depths and relate
them to annual rainfall, 7 day smoothed
temperatures at 4 m and 20 m at Isla Col?n
were plotted on the same set of axes for
each depth (Fig. 4A, B). Temperatures at
20 m did not vary over a period of weeks or
even months as much as those at 4 m at Isla
Col?n, but over a period of several years,
sometimes reached the same highs and
lows (Table 1, Fig. 4). Years with high rain-
fall, 2002 and 2000, had generally lower
temperatures than those years with lesser
rainfall, 1999 and 2003, but there were nu-
merous exceptions. This pattern is more
distinct at 20 m. Note that four out of the
five years for which we have water tem-
perature data were years of above average
rainfall (Fig. 4).
Intraseasonal patterns
In 15 years of monthly Airport data
(1989-2004, Fig 5A), the month with the
least median monthly total rainfall was
September with less than half the rain of
the greatest month in December. The other
minimum and maximum of the year was
February and June, respectively. The
monthly pattern for 72 years of Chan-
guinola data (not shown) was similar. The
September to October minimum in rainfall
is very regular. In the Changuinola data, it
appears in box plots for every decade.
Water temperatures at the 2 m Isla Col?n
grassbed site (1999 to 2004 with some gaps,
Fig. 5B) can be characterized as a period of
rising temperatures from February to June,
followed by a dip from July to September
or October, and then a decrease reaching
annual lows in December and extending to
the following January and February.
As might be expected, the solar radiation
data (Fig. 5C) shows maxima where rain-
fall is minimal and minima where rainfall is
maximal, but the pattern is not as distinct.
Wind speed (Fig. 5D) was less variable but
because of the anemometer?s sheltered lo-
cation, the data do not show the strong NE
trade winds that are present from mid-
December to March each year. Air tempera-
ture (Fig. 5E) shows a pattern similar to
that of the water temperature, although we
only have one year of data. Relative humid-
TABLE 1. Water temperature means, maxima, and minima for each site and depth. The inshore data is for all
months for which we have at least 15 days of complete data, the IGOSS offshore data includes all data from April
7, 1999 to March 31, 2004 from quadrat noted in Fig. 3. Mean: Mean of 12 monthly means. Each of the 12 monthly
means was based on 4 to 6 means for a given month from 1999 to 2004, but different months were missing in
different years. Max, Min: The maximum and minimum of daily means found for each site. IGOSS offshore data
are smoothed over 2 weeks so Max and Min are not tabulated. Jan-Feb Mean: Same as Mean but only January
and February monthly means are used. Sept-Oct Mean: Same as Mean but only September and October means
are used. Temperatures are ?C.
Depth Mean Max Min
Jan-
Feb
Mean
Sept-
Oct
Mean
Cayo Agua 4 m 28.7 30.8 26.5 27.8 29.7
Cayo Rolda?n 4 m 29.1 31.6 25.8 27.9 30.5
Isla Colo?n Grassbed 2 m 28.6 31.4 25.6 27.5 29.7
Isla Colo?n Coral Reef 4 m 28.6 30.5 26.6 27.5 29.6
10 m 28.5 30.8 25.9 27.3 29.7
20 m 28.5 30.3 26.5 27.3 29.3
IGOSS offshore 0 m 28.3 ? ? 27.8 28.9
PHYSICAL CONDITIONS OF BOCAS DEL TORO 399
ity (Fig. 5F) is strongly affected by tem-
perature and is higher during months with
higher temperatures and vice versa.
Generally, during periods of rising water
temperatures, the 2 m site rose the most
rapidly and peaked earlier and at the high-
est temperature (Fig. 4). The temperatures
at the 4 m, 10 m and 20 m sites (not all
shown) rose at successively lower rates and
had lower and later peaks. In periods of
temperature decline, usually during the
peak rainfall months of August and De-
FIG. 5. Intraannual patterns 1. Box diagrams of summary statistics for each month. For each parameter, the
box diagrams show the median (horizontal line) and the upper and lower quartiles enclosing 50% of the data
(rectangle). The two lines above and below the rectangle extend to furthest points that are within a distance of
1.5 times the height of the rectangle. If there are any points outside of these boundaries, either above or below,
they are each marked on the diagram with a symbol. (Systat 9.0, SPSS, Inc., Chicago).
To the right of each set of monthly diagrams is a box diagram showing the summary for the entire period that
each monthly diagram covers. The length of the period is given at the top or bottom. A. Rainfall: Data are
monthly totals from Airport data from January, 1989 to August, 2004. B. Water temperature: Data are sample
water temperature for each hour from the Grassbed site (2m) from 4/25/99 to 4/24/04. C. Solar radiation: Data
are daily totals from 8/17/02 to 8/16/04. D. Wind speed: data are mean scalar wind speed for each hour from
8/17/02 to 8/16/04. E. Air temperature: Data are sample air temperature for each hour from 1/1/03 to
12/31/03. F. Relative humidity: Data are sample relative humidity for each hour from 1/1/03 to 12/31/03.
KARL W. KAUFMANN AND RICARDO C. THOMPSON400
cember, the 2 m and 4 m sites dropped
much faster and to a lower level than the 20
m site. The 10 m site usually followed the
two shallower sites but sometimes main-
tained a higher temperature similar to the
20 m site. Sometimes, 20 m temperatures
would continue to rise slowly for several
weeks after the shallower water started
dropping.
After periods of temperature decline at
10 m depth and above, temperatures at
20 m were often warmer. Presumably the
water remained stratified because the colder
surface water was less saline than the 20 m
water and still less dense despite the tem-
perature difference. This suggests that ver-
tical mixing was not occurring at these
times. Similar temperature reversals (not
shown) were observed at Cayo Rold?n, and
these periods of temperature reversals
would sometimes last for a month or more.
These episodes occurred most often from
June to August and October to January.
Salinities at both the Col?n coral reef and
the Col?n grassbed sites (Fig. 6A, B) were
high, around 34 ppt, during months of less
rain and lower, 30 to 31 ppt, during the
rainy months. At times the salinity fell to
around 25 ppt, and once, associated with
heavy rainfall, fell to 20 ppt in the grassbed
site. The secchi disk readings (Fig. 6C, D)
gradually decrease from high levels in Feb-
ruary until July and August and then in-
crease through October when there is re-
duced rainfall and higher salinity. The
horizontal readings in the grassbed were
generally lower than those at the coral reef,
taken vertically.
Daily patterns
More rain falls at night than during the
day, and the number of hours with mea-
surable rain was greater at night than dur-
ing the day (Fig. 7A). The least rainy part of
the day was the late afternoon, from 1400 to
1700 h. To examine the amount of rain fall-
ing as part of heavy downpours rather than
as light rain, the proportion of rain falling
during hours categorized by amount of
FIG. 6. Intraannual patterns 2. A, B. Salinity measured at 30 cm. at the coral and grassbed sites at Isla Colon.
C, D. Secchi disk readings at same sites. The coral reef data is from vertical readings, the grassbed data from
horizontal readings underwater near the surface. See Fig. 5 for explanation of box diagrams.
PHYSICAL CONDITIONS OF BOCAS DEL TORO 401
rain was tabulated (Table 2). Over 93% of
the hours had no measurable rain, and 48%
of the rain fell during 1% of the hours.
Rainfall is infrequent and very patchy.
The amount of solar radiation for a
mostly cloud free noon hour should be be-
tween 900 and 1040 W/m2, depending on
the time of year, (unpublished ESP data for
STRI Galeta station near Col?n). At the Bo-
cas station, 75% of the hours had less than
800 W/m2 of solar radiation for the noon
hour and 50% of those hours had less than
600 W/m2, which is a bit over half that of a
cloud free noon hour. This demonstrates
that it is frequently cloudy, in sharp con-
trast to the small amount of time that rain is
falling. The daily pattern (Fig. 7B) shows
more light at noon, but, in the absence of
clouds, the hours on either side of the noon
hour should have a symmetrical pattern,
since the local noon at the laboratory is
1230. However, the distribution of solar ra-
diation is slightly skewed to the right,
which is consistent with the number of
hours with rain being less in the afternoon
than in the morning. Air temperature
peaks, and humidity reaches a minimum
(Fig. 7C and D), at hour 15, about the same
time as the rainfall reaches its low. Wind
speed however (Fig. 7D), doesn?t reach a
minimum until hour 18, a time when the
rainfall has already started increasing.
Maximum wind speed during a typical
hour (Fig. 7E) is less than 18 km/hr 50% of
the time, but at times reaches 60km/hr.
The effect of weather on water temperature
Here, we analyzed the effect of solar ra-
diation, rain, wind, humidity, and air tem-
perature on water temperature at different
depths and the periods of time over which
high or low temperatures persisted.
Water temperature often fluctuates daily
in response to radiant solar heating, as well
as to other factors, with relative maxima
around noon and minima at night (Table 3).
At the 1 m site at Isla Col?n, the diurnal
fluctuations averaged 1.15?C but, the great-
est one day temperature change was over
3?C. In the grassbed at 2 m, the greatest
fluctuation was still 2.55?C. At 10 m and
20 m, there was often a daily signal
FIG. 7. Daily patterns. Box diagrams of summary
statistics for each hour of the day. Each figure has
the hour labeled as the time at the beginning of
the hour. A. Tipping bucket data from August 17,
2002 to August 16, 2004. Upper line: Percent of
hours with any amount of rain for each hour of
the day. Lower line: Percent of total rain for the 2
years that fell in the designated hour. B. Solar radia-
tion C. Air temperature D. Relative Humidity E. Wind
speed. Sites, starting and ending dates, and sum-
mary statistics for B to D are the same as correspond-
ing figure in Fig. 5. See Fig. 5 for explanation of box
diagrams.
KARL W. KAUFMANN AND RICARDO C. THOMPSON402
present, but at other times, one level or the
other would fluctuate several times a day
in a pattern not related to the sun, or to the
level immediately above or below it. The
mean fluctuation decreased to 0.20?C at 20
m, but the maximum daily fluctuation ob-
served was higher than the two levels
above it. At 20 m, the daily signal was very
weak (Fig. 4), but for brief periods each
year, the temperature would fluctuate by
up to a degree four or more times in the
course of a day, evidently as a result of ver-
tical mixing, and this is when the maximum
daily fluctuation of 1.4?C was recorded.
The occasional periods of rapid fluctua-
tions at 4 m and 10 m, with a random pe-
riod not related to the sun, suggested ver-
tical mixing. Often, fluctuations at one level
were not observed at a shallower level, so
the mechanism causing the fluctuations is
probably not from wave turbulence, but
may be from vertical density driven cur-
rents. Tidal currents cannot be discounted
although the mean tide range at Bocas del
Toro is only 24 cm (Tides&Currents for
Windows). Fluctuations at shallow depths
appeared more often than those at 20 m
and are not confined to the period of rapid
cooling at the end of the year (Fig. 4A). At
Cayo Rold?n, the daily fluctuations were
less than those at an equivalent depth at
Isla Col?n. Temperatures at Cayo Agua
show fluctuations similar to Isla Col?n,
both at 4 m.
Over periods longer than a day, the rela-
tive fluctuations of solar radiation presum-
ably result in increases or decreases in
mean daily water temperature, but other
parameters also have an effect. Wind and
humidity would affect evaporative cooling
of the water, air temperature, generally
lower than water temperature, should cool
the water, and rain mixing with the water
either directly or through runoff, should
also change the temperature. There would
also be complex interactions among these
TABLE 2. Frequency distribution of rainfall. The amount of rain during each hour, in mm, is categorized by
exponentially increasing size classes (Rainfall Category) from August 17, 2002 to August 16, 2004. The total rain
that fell in hours belonging to each size class (Total Rain), the percent of the hours with rainfall amounts in each
size class (Percent of Hours), and the percent of all of the rain falling in the two year period in each size class
(Percent of Rain) are tabulated. The first size class, 0-2, has an upper bound just below the minimum amount
recorded by the tipping bucket, 2.54 mm and so probably includes trace amounts of rain, but has 0 mm of rain
tabulated in column 2.
Rainfall
Category
(mm)
Total
Rain
(mm)
Number
of
Hours
Percent
of
Hours
Percent
of
Rain
0-2 0 16381 93.4 0
2-4 1781 701 4.0 26.5
4-8 1694 285 1.6 25.2
8-16 1288 104 .59 19.1
16-32 1331 59 .34 19.8
32-64 536 13 .074 8.0
64 81 1 .006 1.2
TABLE 3. Daily temperature fluctuations. The mean and maximum of the daily temperature ranges in ?C. Isla
Colo?n 1 m 3/99-8/01, 2 m, 4 m, 10 m, 20 m 3/99-7/04. Cayo Rolda?n 1 m, 10 m, 20 m 4/99-9/00; 4 m 4/99-1/04.
Cayo Agua 4 m 12/00-4/04. The 2 m site at Isla Colon is in the grassbed, the others at the coral reef.
Isla Colo?n Cayo Rolda?n Cayo Agua
Depth Mean Maximum Mean Maximum Mean Maximum
1m 1.15 3.07 0.80 2.81 ? ?
2m 1.06 2.55 ? ? ? ?
4m 0.38 1.29 0.65 2.36 .55 1.87
10m 0.34 1.29 0.47 1.54 ? ?
20m 0.20 1.40 0.31 .92 ? ?
PHYSICAL CONDITIONS OF BOCAS DEL TORO 403
factors. Wind, for example, could cause
vertical mixing, and the resultant change in
surface temperature would affect the rate
of radiative cooling. Here, however, we are
not interested in a complete model of the
effects of all of these factors. Solar radiation
is the major source of energy into the bay,
and we wanted to demonstrate the relation
between it, rainfall, and changes in water
temperature.
To evaluate the main effects of various
meteorological parameters on the 2 m
Col?n grassbed water temperature, we
used a multiple linear regression to com-
pare the 72 hour change in water tempera-
ture to various 72 hour summaries of solar
radiation, wind speed, rainfall, air tempera-
ture and humidity. The grassbed site was
used because it is expected to show a stron-
ger and more immediate response to me-
teorological parameters. A small stream en-
tering on the northwest side of the bay may
have provided more of a runoff effect than
at other, more open sites. Water tempera-
tures were smoothed over a 24 hour period
centered on each hour to remove diurnal
effects and then the 72 hour change in tem-
perature was calculated for each hour. For
rainfall and solar radiation, each smoothed
point was the sum over the previous 72
hours. Rainfall and solar radiation totals
were square root transformed to make the
relation with change in water temperature
linear. The solar radiation sum was divided
by 3 so it would be expressed as an average
daily total. For wind speed and air tem-
perature, we calculated the mean over the
previous 72 hours. Humidity was arcsine
transformed before calculating the 72 hour
mean. After transformations, the response
and predictor variables were generally lin-
ear with respect to each other except for
arcsine transformed relative humidity. At
humidity values less than the mean, water
temperature change was lower than ex-
pected based on the trend for values
greater than the mean. Solar radiation and
air temperature were similarly lower than
expected but rainfall and air temperature
were higher than the trend expected from
higher humidity values. Most, but not all of
these low humidity points were in the drier
parts of the year, January to March and
September to October.
We then picked the hour from 1200 to
1300 on two days each week, 3 to 4 days
apart (Sundays and Wednesdays) and cal-
culated two separate multiple linear regres-
sions for water temperature change against
the meteorological parameters, one for be-
low average humidity, and the other for
above average humidity. Both regressions
were significant, the first explained 23.5%
of the variance, the second 44.5%. The effect
of solar radiation was significant in both (P
= 0.002 and P < 0.0005 with Bonferroni ad-
justments, N = 69 and 61). None of the re-
maining effects, including humidity, were
significant. The humidity and air tempera-
ture variables exhibited collinearity.
Initially, we used data from mid 2002 to
the end of 2003, the period for which we
had data on all 5 meteorological param-
eters. Since temperature and humidity did
not have significant effects and we lacked
data for an additional year from these two,
we removed them from the analysis. We
had data for the remaining parameters up
to November, 2004. The second regression
analysis, using solar radiation, wind speed,
and rainfall, explained 40.7% of the vari-
ance and was highly significant (N = 207,
P < 0.0005). Solar radiation showed the
strongest effect on change in water tem-
perature with a standardized coefficient of
0.500 and the transformed wind speed was
next with a standardized coefficient of
?0.184. Both effects were significant (P <
0.0005 and P = 0.008, respectively) The dif-
ference in signs indicates that solar radia-
tion increases water temperature and wind
decreases it, and the greater magnitude of
the standardized coefficient for solar radia-
tion indicates a greater effect from varia-
tions in solar radiation than from wind. The
effect of wind is probably a combination of
evaporative cooling mediated by air tem-
perature and humidity and vertical mixing.
It is likely to have a stronger effect less shel-
tered sites. The effect of rainfall was not
significant (P = 0.642). When the regression
used hour 8 instead of 12, a time when the
effect of the most recent solar radiation is
less compared to wind speed, the results
are similar but the standardized coeffi-
KARL W. KAUFMANN AND RICARDO C. THOMPSON404
cients for solar radiation and wind speed
are 0.331 and ?0.227 and the regression ac-
counts for only 28.4% of the variation.
A similar regression against the 72 hour
change in water temperature was done for
solar radiation alone at depths of 2 m, 4 m,
10 m, and 20 m at the Isla Col?n grassbed
and coral reef sites (Fig. 8). (r2: 0.308, 0.347,
0.341, and 0.084. Regression slopes: 0.209,
0.162, 0.155, and 0.037, units are ?C /kW/m2/
day, N = 196, 197, 173, and 197, P < 0.001 in
all cases). For the top three levels, there was
a strong effect on water temperature from
solar radiation, decreasing slightly with
depth. At 20 m, the effect of solar radiation
was very weak, but the regression was still
significant. At all four depths, the point at
which the regression line crossed from a
positive to negative effect was a daily av-
erage of about 4 kW/m2/day. This is ap-
proximately the median value for annual
solar radiation (Fig. 5C). At 2 m, three days
of solar radiation averaging 6 kW/m2/day,
or 3 kW/m2/day above the median, in-
creased the daily mean water temperature by
an average of 0.62?C (= 0.209 ? 3).
There are much longer records available
for rainfall than for water temperatures,
and any correlation found would be useful
in estimating historical water temperature
ranges. Although rainfall did not appear to
affect water temperature over 72 hours, we
looked for an effect of monthly total rainfall
on monthly water temperature change. Be-
cause rain is accompanied by clouds and
hence reduced solar radiation, and also by
stronger winds, the cumulative effect of a
month of rain would be expected to reduce
water temperature. Elevated humidity ac-
companying rainfall might be expected re-
duce evaporation and partially offset the
cooling effects of wind. In fact, lower water
temperatures in rainier years had already
been observed (Fig. 4). To measure the ef-
fect of rainfall, which implicitly includes ef-
fects from wind, runoff, and humidity, we
compared the monthly change in water
temperature with monthly totals from the
Airport data up to the end of 2001 and after
that with the Daily ESP data corrected with
Airport/Daily ESP regression parameters.
Changes in mean daily water temperature
at 4 m from the beginning to end of each
month at Isla Col?n were used as the de-
pendent variable in a regression against the
monthly Airport/Daily ESP data. The re-
gression was significant (N = 48, P = 0.032),
but the regression only explained 9.6% of
the variance. To explain some of the rea-
sons for the low correlation between rain-
fall and temperature change, two more re-
gressions were done, one to check the effect
of rain on solar radiation and another to
check the effect of solar radiation on water
temperature. Using monthly rain as the in-
dependent variable and mean daily total
solar radiation as the dependent variable,
the regression was significant (N = 22, P =
0.002) and the regression explained 38.7%
of the variance. Using the mean daily total
solar radiation as the independent variable
and the water temperature as the depen-
FIG. 8. Relation between solar radiation and change in water temperature. Regression between 72 hr tem-
perature change and 72 hr total solar radiation (divided by 3 to adjust to a daily value). Temperature was first
smoothed with a 24 hr running mean. Values at noon on Sundays and Wednesdays were used to avoid overlap.
Data was from the Isla Col?n 2 m grassbed, and the 4, 10 and 20 m coral reef sites.
PHYSICAL CONDITIONS OF BOCAS DEL TORO 405
dent variable, again the regression was sig-
nificant (N = 22, P = 0.006), and the regres-
sion explained 31.9% of the variance,
similar to the results obtained with the 72
hour regression above. The difference in N
between the three regressions is because we
have 48 months of combined water tem-
perature and rainfall data starting in 1999
but we have only 22 months of combined
solar radiation, rainfall and water tempera-
ture data, starting in 2002.
One year of meteorological and
marine changes
To demonstrate the effect of meteorologi-
cal conditions on the water temperature, a
graph was prepared of the parameters over
an entire year, 2003 (Fig. 9). The bottom
four graphs in the figure show three day
running means and totals. The horizontal
line through the wind speed graph is at 6.5
km/hr, the median annual value wind
speed (Fig. 5D). Note that peaks in wind
speed are often associated with peaks in
rainfall but that the relative height of the
rainfall peaks are not necessarily correlated
with the height of the wind peaks. Since the
anemometer is in a protected area, it does
not accurately record wind from the NW to
NE. Strong NW winds are found in other
parts of Panama from late December to
mid-April (Cubit 1989; Robertson et al.
1999; Windsor 1990) which would often be
associated with clear skies and lack of rain.
These winds are present in Bocas del Toro
in more exposed areas and probably affect
water conditions in these areas more than
they do in our coral reef and grassbed sites
on Isla Col?n. However, the anemometer
FIG. 9. Effect of rain, wind, solar radiation, and air temperature on water temperature. Data for the year 2003
are shown. Data for each day at noon are plotted. The following are running totals and means are for the 72
hours preceding the plotted point. Water temperatures are means for the 24 hours centered on noon. From the
bottom up: Rain-total in mm. Wind?mean scalar wind speed in km/hr. Solar radiation: mean total daily solar
radiation (divided by 3) in kW/m2/day. The values are plotted as values above and below 4 kW/m2/d. Air
temperature: mean in ?C. Water temperature: 24 hour running means centered on noon in ?C. 2 m grassbed, 4
m coral reef, as labeled. Maximum daily temperature range at 20 m, 4 m, and 2 m: taken from the coral reef and
grassbed sites.
KARL W. KAUFMANN AND RICARDO C. THOMPSON406
probably provides a good indication of lo-
cal wind conditions at the STRI lab and in
other places along the south sides of islands
in the archipelago, and so one could con-
clude that moderate winds are most often
associated with rain there. The daily total
solar radiation was plotted as anomalies
from the annual median daily solar radia-
tion, about 4 kW/m2/day (Fig. 5C).
Daily mean air temperatures and water
temperatures at 2 m rise and fall together
and the highs and lows are correlated with
those in rain, wind, and solar radiation.
Water temperature at 20 m behaved quite
differently. It was much more stable, with-
out the 1 to 2 degree changes over the
course of a month, except in late November
and December. At times, it didn?t change at
all (within the 0.18?C sensitivity of the ther-
mometers) for over a day at a time. There
were three periods during the year (mid-
May, July to August, and mid-October to
December) when the 20 m temperature was
as much as a degree warmer than the 2 m
temperature.
The top three graphs in Fig. 9 show tem-
perature ranges, each vertical bar showing
the range for one day. Temperature
changes are relatively high and uniform
throughout the year. At 4 m, the changes
are less. Note the period of high fluctua-
tions at the end of the year at 20 m. This
was a period of heavy rainfall, high winds,
low solar radiation, when the 2 m tempera-
tures were lower than the 20 m tempera-
tures. The pattern of relatively large (>1?C)
daily fluctuation at 20 m over a short pe-
riod of descending temperatures in No-
vember or December is also present in each
of the four years (1999, 2001-2003) for
which we have November and December
data. At 4 m there are several periods of
higher daily fluctuations, particularly in
May and July. At these times, the fluctua-
tions happen irregularly and are not related
to changes in solar radiation.
DISCUSSION
Bocas del Toro rainfall
Rainfall in Central America is related to
the ENSO (El Ni?o/Southern Oscillation),
but not in a simple way. The sea surface
temperature anomalies in the Tropical
North Atlantic and in the eastern tropical
Pacific play more important roles, and
sometimes they are correlated with one an-
other and with El Ni?o (sea surface tem-
perature anomalies in the eastern Pacific),
and sometimes not (Enfield and Alfaro
1999; Alfaro 2003).
There are two patterns of annual rainfall
in Central America (Maga?a et al. 1999; Al-
faro 2002). The most widespread consists of
a dry season from the middle of November
to the middle of May with a rainy season
for the remaining months, mid-spring to
late fall. There is a mid-summer reduction
in rain in July and August, called variously
the veranillo, canicula or Mid Summer
Drought. In Panama it is called the Vera-
nillo de San Juan (Espinosa 1998). The
summer rains in this cycle are a result of
convective activity from the northern mi-
gration of the ITCZ as well as easterly
waves, atmospheric disturbances moving
westwards from April to November in the
tropical North Atlantic (Maga?a 2003; En-
field and Alfaro 1999). In the winter, when
the ITCZ moves south with the sun, the
rains stop, strong northeast winds develop,
and there are high levels of solar radiation,
even though the sun at noon is below the
equator. Rainfall patterns in Central
America are however rather variable and
they are not entirely governed by maritime
air masses. The mountains and the dynam-
ics of the Central American atmosphere can
cause local and seasonal variations, so
much so that the second pattern is quite
different from the first (Alfaro 2002).
Alfaro (2002) described the second pat-
tern as having a nearly homogenous period
from January to the middle of October, fol-
lowed by a period of heavier rain at the end
of the year, although he notes that this ho-
mogenous period has a minimum in March
and a relative maximum in July. This pat-
tern matched his stations along the Car-
ibbean coast of Honduras, Costa Rica and
Panama.
Bocas del Toro has this Caribbean pat-
tern, but with some substantial differences
(Fig. 5A). While there is no dry season,
there is relatively dry period in February
PHYSICAL CONDITIONS OF BOCAS DEL TORO 407
and March and another, slightly drier, in
September and October. The heaviest rain
comes in December, but is rivaled by a July
high. This September low is found both in
the 72 year record from Changuinola and
the 15 year Airport record. In other words,
the slight pattern in Alfaro?s ?nearly ho-
mogenous? period is accentuated in Bocas
del Toro. Some of these differences may be
attributed to Bocas del Toro?s location in
the far south of the range of Alfaro?s sites.
Sites to the south had earlier starting points
for the wet season of the more common
pattern, related to the earlier arrival of the
ITCZ and perhaps this is related to Bocas
del Toro?s more extreme form of the Car-
ibbean pattern. The heavy winter rain can
be attributed to frontal systems from the
north moving into the tropics (Maga?a et
al. 2003; Empresa de Transmision Electrica
S. A.).
The rainfall pattern at the STRI lab at
Galeta, near Col?n on the Caribbean side,
follows the more common rainfall pattern
with a distinct dry season during the boreal
winter, but the timing of the onset of the
dry season is different with heavy rain usu-
ally lasting into the middle of December
(Cubit 1989). On Barro Colorado Island,
Panama, the dry season is generally from
January to March, and the daily pattern is
the opposite of that in Bocas del Toro, with
maximum rainfall occurring in the early af-
ternoon (Windsor 1990).
The diurnal cycle of low precipitation in
the late afternoon and a high early in the
morning (Fig. 7A) is typical of northern
hemisphere and tropical oceans adjacent to
large land masses and is the opposite of the
pattern normally found on land. According
to one theory, on land late afternoon thun-
dershowers develop because of solar heat-
ing of the land and produces convective
currents of rising moist air. At sea, this ef-
fect is not as strong and air tends to flow
towards continents to replace the rising air.
At night winds flow in the opposite direc-
tion as the land cools faster than the sea,
and the warm air flowing out to sea pro-
duces nighttime precipitation (Dai 2001).
A competing theory (Mapes et. al. 2003b)
explains such a diurnal cycle along the
coast of Columbia as the result of gravity
waves developing from land heating and
propagating toward the sea. The waves op-
erate at a higher altitude than sea breeze/
land breeze theory above and do not in-
volve horizontal translation of moisture but
rather a compression and decompression of
the air. During the day these waves result
in a warm anomaly over the ocean which
effectively cap any convective activity and
reduce the amount of rain. At night, they
result in a cool anomaly and enhance con-
vective activity. The nightly seaward
breeze of the first theory, according Mapes
et al. (2003b) is not strong enough to pro-
duce the observed offshore convection.
An interesting element of the model used
by Mapes et al. (2003a, 2003b) and observed
in data from that area, is a very sharp gra-
dient in rainfall along the coast, with more
rain at sea and less over the land. They at-
tribute this to a thin layer of cool near-
surface air, below the air producing the
gravity waves, which develops over the
land at night and inhibits convective pro-
cesses on the landward side of the coast.
This would explain the lower annual mean
rainfall at Changuinola, 8 km inland from
the coast, 2615 mm, compared to the airport
data at Bocas del Toro, 3277 mm.
Factors affecting water temperature
The relationship observed between rain-
fall, solar radiation, and water temperature
can explain many elements of the annual
cycle of temperature extremes and nutrient
mixing in Bah?a Almirante, as exemplified
by Fig. 9. At the Col?n site in Bah?a Almi-
rante and over periods of three days, solar
radiation is the most important determi-
nant of water temperature changes, fol-
lowed by wind speed (Fig. 8). Rainfall does
not show a strong correlation. Over periods
of a month, there is a correlation between
rainfall and water temperature, but as men-
tioned, the percent of the variation in water
temperature explained by rainfall is only
9.6% while rainfall explains 38.7% of the
variation in solar radiation, and solar radia-
tion explains 31.9% of the variation in wa-
ter temperature. If the factors affecting the
correlation between rainfall and solar ra-
diation are different from those affecting
KARL W. KAUFMANN AND RICARDO C. THOMPSON408
the correlation between solar radiation and
water temperature, then the product of the
two (0.387 x 0.319 = 0.123) should approxi-
mate the relatively low proportion of the
water temperature change explained by
rainfall, which it does. Since it rains for
only a small proportion of the time (Fig.
7A) and mostly at night, and when it does
rain, most of the rain falls in an even
smaller proportion of the hours (Table 2), it
is not unreasonable to expect that the fac-
tors affecting the production of rain from
convection clouds in the atmosphere would
not be closely related to factors affecting
vertical mixing in the water column. Wind,
which accompanies rain and increases cool-
ing by evaporation, is a possible exception,
but it is not always closely correlated with
rainfall.
Increased rainfall then can be expected to
be related to lower water temperatures and
vice versa, but the effect will not be a strong
one and would probably only be obvious
over periods longer than a month. In fact,
years with heavy rainfall had generally
lower water temperatures over a year than
those years with less rainfall (Fig. 4), but
this did not result in persistent temperature
differences from the offshore temperatures
into subsequent years. Note that 2002 had
more rain than any year in the 78 year
Changuinola record. Temperatures at 20 m
for that year were generally lower than for
all of the other years except for those at the
end of 1999 and the beginning of 2000.
On the other hand, the 4 m temperatures
in June and October of 2002 are among the
highest observed, reaching a 30.7?C in early
October. Even in an otherwise rainy year
then, a month of clear skies can result in
elevated temperatures. Of the five years for
which we have nearly complete water tem-
perature data, four had above average rain-
fall, and the fifth, 2003, was not appreciably
below average. It is possible that a dry year,
with even longer periods of above average
solar radiation than observed, could have
higher temperatures. But, because of the
loose coupling of rainfall and water tem-
perature, rainfall records may not necessar-
ily indicate periods of elevated or reduced
temperatures.
Another factor affecting water tempera-
ture and for which we have no direct mea-
sure is the influence of runoff from rivers.
However, changes in water temperature
happen uniformly over the entire bay, and
this uniformity even extends to Cayo Agua
at the edge of the Chiriqu? Lagoon which
has several large rivers draining into it (Fig.
3) and would be expected to show a differ-
ent response to runoff than Bah?a Almi-
rante. This large scale uniformity in tem-
perature response suggests that local
effects, such as freshwater runoff, are not a
major factor in determining water tempera-
ture. The influence of marine currents on
water temperature changes has not been di-
rectly addressed in this study. We sug-
gested above that even though the tidal
range is small, subsurface currents may
have been responsible for some of the rapid
fluctuations found at some depths while
leaving water just above and below un-
changed. The strong northeast winds that
are known to occur from late December to
March may bring in some offshore water,
particularly to Cayo Agua, but most of the
cooling observed is during the period of
heavy rain just before these winds develop.
Most of the time, inshore water is either
warmer or colder than offshore water and
our argument is that local solar radiation,
and other associated parameters, creates
substantial differences from offshore wa-
ters, regardless of the amount of mixing.
The reason that Bah?a Almirante and the
outer part of Chiriqui Lagoon at Cayo
Aqua has 4m temperatures higher than the
surface temperatures offshore may be be-
cause the inshore waters receive relatively
more solar radiation than offshore. An ex-
planation for this could be a thin layer of
cold air developing over land at night,
mentioned above, which inhibits convec-
tion landward of the coastline (Mapes et al.
2003b). If the effect of this layer extended
seaward, perhaps aided with a cold ther-
mally driven slope breeze from the 3000 m
mountains immediately southwest of Bah?a
Almirante, convection might be partially
inhibited during the morning hours and
the increased solar radiation would heat
the water relative to the offshore water. The
effect would be stronger at Cayo Rold?n
PHYSICAL CONDITIONS OF BOCAS DEL TORO 409
which is more landlocked and closer to the
mountains than the other two sites.
Temperature inversions and vertical mixing
Mixing of the water column, and in par-
ticular the mixing of nutrients brought in
by runoff (D?Croz 2005) depends largely
on the density of the surface water relative
to deeper water, and there are two oppos-
ing processes affecting this. Normally one
would expect the top layers of water to be
warmer than the lower layers since warm
water is less dense. Less saline water is also
less dense, but when it is added to the bay
during rainy periods, the rain is accompa-
nied by clouds and wind, which cool the
water. If it cools too much, it will sink, but
if not, it could float on top and still be
colder than the underlying water. The pe-
riods when lower salinity, and presumably
nutrient rich, surface water floats on top of
warmer bottom water are not evenly dis-
tributed throughout the year, but are most
often found during the rainier months of
May to August and October to December
(Fig. 4C, D). Periods without temperature
inversions, February to May, and Septem-
ber, are associated with months having less
rain, when surface warming from increased
solar radiation maintains a more normal
thermocline.
Presumably, the rapid daily temperature
fluctuations observed annually at 20 m at
the end of the year indicate vertical mixing.
That the fluctuations occur without simul-
taneous fluctuations at more shallow
depths suggest that these were the result of
local subsidence events because of the drop
in water temperature rather than turbulent
mixing from wind. Normally the rain, and
associated runoff, would be expected to de-
crease the density of the surface water and
inhibit mixing, but the decreased tempera-
ture reduces the density enough for it to
sink. The increased wind may contribute to
mixing, but subsidence from increased
density seems more likely since the fluctua-
tions of the 4 m and 10 m water were often
not present when the 20 m water was un-
dergoing fluctuations. This suggests that
turbulent mixing was not happening since
it would have affected all depths at once.
Throughout the year, there were several
times when the 2 m water, as well as the 4
m (Fig. 9) and even the 10 m water (not
shown), were colder than the 20 m water
for two or more weeks at a time. If they
didn?t mix with the underlying water even
when colder, it suggests that there was very
little vertical mixing with the 20 m water
occurring throughout most of the year,
along the lee side of Isla Col?n. The diurnal
stability of the 20 m water for most of 2003,
except November and December and a few
days in July confirms this. We don?t have
sufficient data to determine if this stability
at 20 m extends to other parts of Bah?a
Almirante.
Temperature stress differences among sites
and depths
Temperature increases of 1 degree above
the summer mean maxima are known to
damage or kill corals, (Glynn and D?Croz
1990; Jokiel and Brown 2004), so the small
differences in temperature noted at differ-
ent depths and at the three different sites
can have a substantial impact on the corals
and other organisms in the Bocas del Toro
area. Of equal importance is the length of
time that water temperatures stay elevated
as well as the presence of other factors such
as lowered salinity, turbidity, and amount
of mixing of the water column.
Cayo Rold?n often has 4 m temperatures
more than a degree higher than Isla Col?n,
and these usually occur during the middle
of the year when temperatures are already
high. Its lowest temperatures are some-
times as low as Isla Col?n (Table 1) and as
the offshore surface temperatures (Fig. 3A,
C), so overall, Cayo Rold?n has a higher
annual temperature range and higher over-
all temperatures. Cayo Agua often has tem-
peratures about 0.5 degrees warmer than at
Isla Col?n, but these often occur during the
beginning and end of the year when the
water is colder (Fig. 3A, B). During the
warm part of the year, temperature differ-
ences are less. In other words, the tempera-
ture at Cayo Agua does not exhibit excur-
sions from the offshore water as extreme as
at Cayo Rold?n and has slightly lower ex-
cursions than at Isla Col?n. This is probably
KARL W. KAUFMANN AND RICARDO C. THOMPSON410
because it is more exposed to the open
ocean and not surrounded by as much
land.
Although the annual range of tempera-
tures is similar at all depths, there is a sub-
stantial difference in the rate and frequency
of temperature changes at each depth. Ob-
served as a seven day running mean (Fig.
4), 4 m temperatures at Cayo Rold?n
changed more rapidly and frequently than
at 20 m (Fig. 4). Mean daily temperature
fluctuations were as much as five times
greater at 2 m than 20 m at Isla Col?n
(1.06?C vs. 0.20?C). Although the differ-
ence in maximum fluctuation values were
not as extreme, large temperature fluctua-
tions in shallow water were a daily event at
2 m, but only occurred a few times a year at
4 m and once a year at 20 m (Fig. 9). At Isla
Col?n, the maximum fluctuations in 1 m to
2 m of water ranged from 2.6?C to 3.1?C
and were about half this at 4 m to 20 m. At
Cayo Rold?n, the mean daily fluctuations
at 4 m and deeper were larger than those at
Isla Col?n, while the 1 m fluctuation was less.
The Cayo Rold?n site then, is potentially
the most stressful for corals and other or-
ganisms, based on its higher mean tem-
perature (about 0.9?C ) and its even higher
(up to 1.3?C) differences from Isla Col?n
temperatures at 4 m. Cayo Agua is the least
stressful based on its mean temperature be-
ing slightly lower than Isla Col?n and its
tending to be slightly warmer than Isla
Col?n during the colder parts of the year,
thus reducing its overall temperature
range. At all sites, temperature fluctuations
are greater and more frequent at shallower
depths, but even at 20 m, it is possible for
animals to be subjected to short term fluc-
tuations of over 1?C. Over several years,
temperature ranges at 20 m are approxi-
mately the same as the range in daily mean
temperatures at 1 m to 2 m. Besides tem-
perature stress, shallow sites may be sub-
jected to salinities of up to 20 ppt, and
higher turbidity, particularly in small shal-
low bays like that at Isla Col?n. All three
sites were more stressful for temperature
sensitive organisms than the offshore wa-
ters, which had a range of just 1.1?C be-
tween the means for the coldest and warm-
est months of the year.
Water temperatures can rise substan-
tially in with just a few weeks of high solar
radiation, even at 20 m. At 4 m, mean daily
water temperature was observed to rise
2.8?C in just 30 days at Isla Col?n. The data
for this study only encompassed five years
of water temperature data and one would
expect that a longer record might have in-
cluded periods with unusually long
stretches of clear weather and subsequent
higher temperatures. This could happen
even if offshore temperatures were normal,
just because of the vagaries of the local
weather. If there were such a period of el-
evated temperatures, the temperature
stress would probably be most extreme at
the more inshore sites. Similar inshore ex-
cursions from offshore temperatures have
been documented in Hawaii (Jokiel and
Brown 2004).
CONCLUSIONS
Organisms inhabiting inshore waters ex-
perience a substantially different environ-
ment than those exposed to offshore wa-
ters, and these differences are often closely
related to local weather phenomena. Tem-
peratures of offshore surface water are af-
fected by complex global weather patterns
which generally also affect inshore water.
But land, and in particular mountains and
the shape and location of shorelines can al-
ter these patterns. The wide variation in
rainfall over short distances in Central
America testifies to this.
Bah?a Almirante is a small bay almost en-
tirely surrounded by land, with a very lim-
ited watershed and relatively little circula-
tion from the Caribbean Sea adjacent to it.
The annual rainfall pattern is similar to
other Caribbean coastal sites in that it does
not have the pronounced dry season dur-
ing the winter that most of the Pacific side
and interior parts of Central America have.
However, unlike other more northern Car-
ibbean coastal sites, it has two distinct pe-
riods with lesser amounts of rain instead of
just one. The water in the bay has near ma-
rine salinities and supports a wide variety
of corals and other reef building organisms.
Elevated temperatures in the bay are cor-
PHYSICAL CONDITIONS OF BOCAS DEL TORO 411
related with periods of high solar radiation,
and the converse is also true, so the pri-
mary determinate of water temperature, or
more precisely its excursions from offshore
water temperatures, is the amount of con-
vective activity over the bay. When atmo-
spheric conditions enhance convection,
clouds form, sometimes accompanied by
rain and wind, solar radiation during the
day decreases, and the water cools. Ex-
tended periods with few clouds can result
in temperatures of over 30?C. Water closer
to the mainland usually has temperatures
even greater than water closer to the ocean.
Elevated water temperatures, particu-
larly those that persist for extended periods
of weeks to months, cause stress to many
organisms, particularly corals. Because the
temperature stresses at Bocas del Toro oc-
cur as a result of differences in insolation,
they are more intense at shallower depths
than at 20 m, although extended elevated
temperatures at 20 m also occur. Stresses
from reduced salinity and turbidity accom-
panying rain are also present in Bah?a
Almirante, even though it has a relatively
small watershed.
Extended cloudy periods cause tempera-
tures at 10 m and above to drop below 20 m
temperatures for a month or more. This
temperature inversion is maintained by the
slightly lowered salinity from rain accom-
panying the clouds and indicates that little
vertical mixing is occurring. At the end of
the year, decreased solar radiation, often
accompanied by heavy rain, lowers tem-
peratures to their annual lows and the 20 m
water is then mixed with the upper layers.
There is evidence that there is more fre-
quent turbulent mixing within the upper
10 m of water.
Since water temperature in Bocas del
Toro is strongly influenced by insolation
and hence the amount of clouds, one might
expect historical rainfall records to be a
good predictor of past temperatures
stresses. Unfortunately, monthly rainfall to-
tals only explain 9.6% of the variation in
monthly water temperature changes. Even
during a particularly rainy year there can
be several months of relatively clear skies
that result in elevated temperatures. The
year 2002 was the rainiest since 1926 and
yet water temperatures for June and Sep-
tember were among the highest recorded.
The annual range in water temperature
in Bah?a Almirante is more than twice that
in the offshore waters, and the frequency of
rapid temperature changes over periods of
several weeks is also higher. Local meteo-
rological conditions then have intensified
the seasonal offshore temperature ranges
and have even created a stress gradient
from the outer to the inner part of the bay.
Cold thermally driven breezes flowing
over the bay from the mountains to the
southwest may be causing this by inhibit-
ing convection during the morning hours.
The effect is mostly to increase tempera-
ture, but evaporative cooling during storms
can also reduce temperatures below those
offshore. Periods of clear skies for a month
can result in elevated water temperatures
at 4 m of almost 3?C. Other tropical embay-
ments with restricted circulation might be
subject to similar temperature increases
above adjacent open ocean water.
Acknowledgments.?Empresa de Trans-
misi?n El?ctrica S. A. supplied rainfall data
for their station at Bocas del Toro. Clyde
Stephens and Ogier Rodriguez supplied
rainfall data from the Chiriqu? Land Com-
pany. Jessica Riggs volunteered her time in
helping to maintain instruments and enter
much of the manually collected data. Arca-
dio Castillo, William Pomaire and Gabriel
J?come collected the weekly and daily data
at the laboratory. Reviewers Robert Stallard
and Steve Paton offered valuable com-
ments.
LITERATURE CITED
Cubit, J. D., R. C. Thompson, H. M. Caffey, and D. M.
Windsor. 1989. Meteorology and hydrography of a
shoaling reef flat on the Caribbean Coast of
Panama. Coral Reefs 8(2):59-66.
Dai, A. 2001. Global precipitation and thunderstorm
frequencies. Part II: Diurnal variations. J. Climate
14:1112-1128.
D?Croz, L., J. B. Del Rosario, and P. G?ndola. 2005. The
effect of fresh water runoff on the distribution of
dissolved inorganic nutrients and plankton in the
Bocas del Toro Archipelago, Caribbean Panama.
Caribb. J. Sci. Special Publication: Marine Fauna
and Environments of Bocas del Toro, Panama.
41:414-429.
KARL W. KAUFMANN AND RICARDO C. THOMPSON412
Espinosa, Jorge A. 1998. Veranillo de San Juan within
the Panama Canal Watershed. Meteorology Sec-
tion, Panama Canal Commission.
Empresa de Transmisi?n Electrica S. A. http://
www.hidromet.com.pa/
Enfield, D. B., and E. J. Alfaro. 1999. The dependence
of Caribbean rainfall on the interaction of the tropi-
cal Atlantic and Pacific Oceans. J. Climate 12:2719-
2733.
Estad?stica Paname?a (1973-1994). Situaci?n Fisica,
Meteorolog?a. Contralor?a General de la Rep?blica,
Direcci?n de Estad?stica y Censo. Panama City,
Panama.
Guzm?n, H., P. A. G. Barnes, C. E. Lovelock, and I. C.
Feller. 2005. A site description of the CARICOMP
mangrove, seagrass and coral reef sites in Bocas del
Toro, Panama Caribb. J. Sci. Special Publication:
Marine Fauna and Environments of Bocas del
Toro, Panama. 44:430-440.
Guzm?n, H. M., and C. Guevara. 1998. Arrecifes cor-
alinos de Bocas del Toro, Panam?: I. Distribuci?n,
estructura y estado de conservaci?n de los arrecifes
continentales de la Laguna de Chiriqu? y la Bah?a
de Almirante. Rev. Biol. Trop. 46:601-623.
Glynn, P. W., and L. D?Croz. 1990. Experimental evi-
dence for high temperature stress as the cause of El
Nino-coincident coral mortality. Coral Reefs 8:181-
191.
IGOSS. 2004. Integrated Global Ocean Services System
at the International Research Institute for Climate
and Prediction web site. nmc Reyn_SmithOIv1
weekly sst data set, November 8, 1981 to August
25, 2004. Lamont-Doherty Earth Observatory Cli-
mate Data Library. http://ingrid.ldeo.columbia.edu/.
Jokiel, P. L., and Eric K. Brown. 2004. Global warming,
regional trends and inshore environmental condi-
tions influence coral bleaching in Hawaii. Global
Change Biol. 10:1627-1641.
Maga?a, V. O., J. L. V?zquez, J. L. P?rez, and J. B.
P?rez. 2003 Impact of El Ni?o on precipitation in
Mexico. 2003. Geophysica Internac. 42(3):313-330.
Mapes, B. E., T. T. Warner, M. Xu, and A. J. Negri.
2003a. Diurnal patterns of rainfall in northwestern
South America. Part I: Observations and context.
Mon. Wea. Rev. 131:799-811.
Mapes, B. E., T. T. Warner, and M. Xu. 2003b. Diurnal
patterns of rainfall in northwestern South America.
Part III: Diurnal gravity waves and nocturnal con-
vection offshore. Mon. Wea. Rev. 131:830-844.
Robertson, D. Ross, S. E. Swearer, K. Kaufmann, and
E. B. Brothers. 1999. Settlement vs. environmental
dynamics in a pelagic-spawning reef fish at Car-
ibbean Panama. Ecol. Monogr. 69(2):195-218.
Tides&Currents Pro for Windows v. 2.5b. 1997. Nau-
tical Software, Inc.
Windsor, D. M. 1990. Climate and moisture variability
in a tropical forest: Long-term records from Barro
Colorado Island, Panama. Smiths. Contrib. Earth
Sci. No. 29, 1-145.
PHYSICAL CONDITIONS OF BOCAS DEL TORO 413