Temperature Variability over South America - USF School of...

Temperature Variability over South America

JENNIFER M. COLLINS

Department of Geography, University of South Florida, Tampa, Florida

ROSANE RODRIGUES CHAVES AND VALDO DA SILVA MARQUES

Laboratorio de Meteorologia, Universidade Estadual do Norte Fluminense, Campos Dos Goytacazes, Brazil

(Manuscript received 31 March 2008, in final form 8 June 2009)

ABSTRACT

The variation of air temperature at 2 m above the earth’s surface in South America (SA) between 1948 and

2007 is investigated primarily using the NCEP–NCAR reanalysis. In December–February (austral summer),

the majority of SA has a mean temperature between 218 and 248C during 1948–75, and for 1976–2007 the

mean temperature is above 248C. In June–August (austral winter), warmer temperatures are observed in the

tropical region in the recent period. The results indicate that Northeast Brazil (NEB) and central Brazil are

warmer in the more recent period. In the last seven years (2001–07) compared to the earlier periods, greater

warming is noted in the tropical SA region, mainly in NEB and over the North Atlantic Ocean, and cooling is

observed in part of the subtropical SA region. Supporting evidence for the warming in Brazil is given through

analyses of station data and observational data. The results presented here indicate that the climate change

over SA is likely not predominantly a result of variations in El Nino–Southern Oscillation (the most important

coupled ocean–atmosphere phenomenon to produce climate variability over SA). Instead, the climate

changes likely occur as a response to other natural variability of the climate and/or may be a result of human

activity. However, even without ascertaining the specific causes, the most important finding in this work is to

demonstrate that a change in the temperature patterns of SA occurred between 1948 and 2007.

1. Introduction

The Intergovernmental Panel on Climate Change

(IPCC) Fourth Assessment Report (AR4) (Bernstein

et al. 2007) considered that progress in understanding of

how climate is changing in space and time has been

gained through improvements and extensions of numer-

ous datasets and data analyses, broader geographical

coverage, better understanding of uncertainties, and

a wider variety of measurements. However, data cover-

age remains limited in some regions. The AR4 shows,

through the use of station data, that the warming of the

climate system is unequivocal, with 11 of the last 12 years

(1995–2006) ranking among the 11 warmest years in the

instrumental record of global surface temperatures since

1850. The report also notes that the total temperature

increase from the period 1850–99 to the period 2001–05 is

0.768C. The main AR4 conclusion is that most of the

observed increase in globally averaged temperatures

since the mid-twentieth century is very likely due to

the observed increase in anthropogenic greenhouse gas

emissions. Brown et al. (2008) also indicate that extreme

daily maximum and minimum temperatures have warmed

for most regions of the world since 1950 and the total area

exhibiting positive trends is significantly greater than what

can be attributed to natural variability.

Difficulties remain in reliably simulating and attrib-

uting observed temperature changes at smaller scales.

On these scales natural climate variability is relatively

larger, making it harder to distinguish changes expected

due to external forcings. Uncertainties in local forcings

and feedbacks also make it difficult to estimate the

contribution of greenhouse gas increases to observed

small-scale temperature changes (Bernstein et al. 2007).

Climate change has been studied over South America

(SA) (608S–108N, 908–208W) mainly in the context of

deforestation in the Amazon (Malhi et al. 2008; Nobre

et al. 1991). There have been few studies about climate

change that consider the whole of SA and focus on

Corresponding author address: Dr. Jennifer M. Collins,

Department of Geography, University of South Florida, 4202 East

Fowler Ave., Tampa, FL 33620.

E-mail: [email protected]

5854 J O U R N A L O F C L I M A T E VOLUME 22

DOI: 10.1175/2009JCLI2551.1

� 2009 American Meteorological Society

temperature. Part of the reason for this may be, as Frich

et al. (2002) note, that there are large regions in the world

where no digital data are readily available for analysis.

Caesar and Alexander (2006) also emphasize this point

regarding the sparse data coverage of freely available

station data for South America. Of the data available,

some do not have long climate records. Vincent et al.

(2005) note that this is particularly a problem for the

Amazon region of Brazil. They also note that there are

few published results available internationally. Of the

few, Victoria et al. (1998) examined mean temperatures

during 1913–95 (although there was missing data) in the

Amazon from 17 stations and found an increasing trend

of 0.568C century21. Vincent et al. (2005) examine trends

in the indices of daily temperature extremes over the

period 1960–2000. They were able to obtain their data

from a workshop held in Brazil in 2004 where scientists

from eight southern countries brought the daily clima-

tological data from their region. They found that there

were no consistent changes in the indices based on daily

maximum temperature. However, significant trends were

observed when considering minimum temperature. They

found significant increasing trends in the percentage of

warm nights and decreasing trends in the percentage of

cold nights at many stations. Dufek et al. (2008) suggests

that the National Centers for Environmental Prediction–

National Center for Atmospheric Research (NCEP–

NCAR) reanalysis can provide useful information re-

garding minimum temperature (which shows a warming

tendency) and consecutive dry-days indices at individual

grid cells in Brazil. Some regional differences in the

trends of the climate indices are observed when com-

paring the NCEP–NCAR reanalysis with observational

data; they suggest that discrepancies between reanalysis

and station data may be caused by model inefficiencies to

produce these data, such as surface processes represen-

tation, resolution, and others, as discussed in Cai and

Kalnay (2005). However, in their study they included

stations with up to 30% missing data and did not use the

Mann–Kendall test to determine trends for nonnormally

distributed data. Quintana-Gomez (1999), Rusticucci and

Barrucand (2004), and Marengo and Camargo (2008) are

other studies that examine temperature extremes in SA,

Quintana-Gomez (1999) focusing on Venezuela and

Colombia, Rusticucci and Barrucand (2004) focusing on

the area of Argentina, and Marengo and Camargo (2008)

focusing on southern Brazil.

In this work we investigate how the mean air tem-

perature at 2 m above the earth’s surface in SA varies, for

the most part utilizing the NCEP–NCAR reanalysis

(NNR) (Kalnay et al. 1996) between 1948 and 2007. Since

the mean temperature is derived from the minimum

and maximum temperature, it is important to consider

studies such as Vincent et al. (2005) and Alexander

et al. (2006) that note significant trends in the minimum

temperature but find that the changes in daily maximum

temperature are less marked, implying that our world

is becoming considerably less cold rather than consider-

ably hotter.

Some studies of the reanalysis skill and trends have

been performed in the Southern Hemisphere, and strong

spurious trends have been found associated with changes

in the observing system. The NNR has been constructed

with a model and data assimilation system kept un-

changed, but it is affected by changes in the observing

systems, especially the introduction of the satellite ob-

serving system in 1979 (Kalnay and Cai 2003). According

to Kalnay and Cai and Nunez et al. (2008) there is a gap

between the NNR estimate of surface temperature and

the station observations. There is a large negative jump

in the NNR between the 1970s and 1980s before and after

the use of satellite data associated with the impact of the

change in observing systems on the reanalysis climatol-

ogy, which is much larger in the Southern than in the

Northern Hemisphere. However, Kalnay and Cai show

that this gap could be associated with changes in land use

(including urbanization and agricultural practices such as

irrigation), and the NNR reflects the changes associated

with atmospheric temperature changes due to changes in

circulation and greenhouse warming.

However, even before 1979 there is also good agree-

ment between the two datasets (NNR and observations)

but problems are observed over regions with topography

(Kalnay and Cai 2003; Rusticucci and Kousky 2002;

Nunez et al. 2008). Rusticucci and Kousky do a com-

parative study between the NNR and station data con-

sidering the maximum and minimum temperature over

Argentina and they show that there is good agreement

between the station data and reanalysis gridpoint data

for low elevations in central and eastern Argentina.

Poor results are observed closer to the Andes (due to the

differences between model topography and real topog-

raphy) and all low latitudes (due to physical parame-

terizations of cloudiness and convections).

It is important to emphasize that the El Nino–Southern

Oscillation (ENSO) is the most important coupled ocean–

atmosphere phenomenon to produce climate variability

over SA on an interannual time scale (Ropelewski and

Halpert 1987; Zhou and Lau 2001; Barros et al. 2002). In

the tropical region, climate anomalies associated with

ENSO are caused by a perturbation in the Walker and

Hadley circulation over the east Pacific. The extratropical

response to ENSO occurs indirectly through alteration of

the Rossby wave train patterns and teleconnections. The

precipitation and latent heat anomalies caused by tropical

sea surface temperature (SST) anomalies alter the forcing

15 NOVEMBER 2009 C O L L I N S E T A L . 5855

of large-scale Rossby waves that propagate into the ex-

tratropics, producing, for instance, changes in extra-

tropical storm tracks and wave patterns with smaller

frequency (Trenberth et al. 1998). Dettinger et al. (2000)

note that storminess appears to be diverted northward

from the midlatitudes to about 308S with less storminess

indicated over southern Argentina. Ropelewski and

Halpert (1987) also find ENSO-related precipitation for

northeastern SA and southeastern SA. There also ap-

pears to be a relationship between the ENSO and rainfall

in the South Atlantic convergence zone (SACZ) clima-

tology region (e.g., Carvalho et al. 2002).

Dettinger et al. (2000) note that El Ninos are associated

with warm temperatures in much of tropical SA, Para-

guay, Uruguay, Patagonia, and southeasternmost Brazil.

They also note that El Ninos bring cooler temperatures to

the eastern Amazon region. Dettinger et al. show a no-

table symmetry around the equator between the Pacific–

North America (PNA)-type pattern over North America

and a similar pattern over SA. The PNA pattern/wave

train associated with El Nino brings a great wavy me-

ridional excursion of atmospheric circulations over North

America that brings cool high-latitude air down into the

southeast United States, causing the cooler temperatures

there. It is possible, therefore, that similar processes from

the corresponding wave train are occurring in SA (M. D.

Dettinger 2009, personal communication). Marengo and

Camargo (2008) consider extreme temperatures and

note that ENSO explains the steep increase in the air

temperature extreme trends. In this paper we consider

therefore if ENSO significantly affects mean temperature

variability over SA.

On time scales longer than interannual, the dominant

climate phenomenon in the Pacific Ocean and overlying

atmosphere has an ENSO-like spatial distribution (Zhang

et al. 1997) of surface temperatures and atmospheric

circulations and interhemispheric climate effects on the

Americas. It should be noted that many authors treat the

two terms—decadal ENSO-like and Pacific decadal os-

cillation (PDO) (Mantua et al. 1997)—synonymously,

but Vimont (2005) notes that they are only related with

the prominent spatial features of decadal ENSO-like

variability being generated by physical mechanisms that

operate through the interannual ENSO cycle. The PDO

was originally intended to reflect the decadal variability

of the North Pacific but it has a substantial expression in

the South Pacific (Dettinger et al. 2000). Between nearly

1948 and 1976 (1977 until around 1998) the PDO index

was negative (positive). From 1998 the PDO index os-

cillated between negative and positive. Zhang et al.

(1997) also note a further climate shift further back in

the 1940s, which they also associate with ENSO-like

decadal variability. Because of the length of the NNR, it

is not possible to examine this shift in the 1940s de-

termined by Zhang et al. In this paper, we also discuss

the temperature shifts in relation to the PDO.

2. Methodology

The temperature variability at 2 m above the earth’s

surface in the NNR is investigated in SA and in the

west part of the South Atlantic Ocean between 1948 and

2007. We consider the whole of SA, tropical SA (TSA)

(208S–108N, 808–358W), and subtropical SA (SSA) (608–

208S, 758–508W) as shown in Fig. 1. Hereafter, air tem-

perature at 2 m above the surface will be referred to only

as ‘‘temperature.’’ Temperature is classified as a B cate-

gory variable in the NNR. This designation indicates that,

although there are observational data that directly affects

the value of the variable, the model also has a very strong

influence on the analysis value (Kalnay et al. 1996).

To consider the reliability of the results from the NNR,

we obtained station data from the Institute of Astronomy

and Geophysics/University of Sao Paulo (IAG/USP)

and the Brazilian National Institute of Meteorology

(INMET) for some other Brazilian locations. The IAG/

USP Sao Paulo data had long enough temperature re-

cords for the same period of time that we examined in the

FIG. 1. Map of South America showing tropical (TSA) (208S–

108N, 808–358W) and subtropical (SSA) (608–208S, 758–508W)

South America.


NNR, 1948–2007. The other locations have data over a

shorter time period beginning in 1961 and, while some of

the stations, such as Brasilia, Curitiba, Goiana, Manaus,

Porto Alegre, and Recife, have complete records during

this period, others, such as Belem, Rio Branco, Fortaleza,

Florianopolis, Maceio, Natal, Rio de Janeiro, and Sal-

vador, do not, with Palmas having only 14 years of data

from 1961 to 2007. Palmas was omitted from the analysis

herein and it is acknowledged that a shorter period of

time and missing data is a serious limitation to the use of

the other station data for such climate change analysis,

but this data will be used to supplement evidence pro-

vided by the reanalysis.

Because of the lack of freely available and easily ac-

cessible data in South America (as noted earlier), other

individual station data are not considered in this paper.

However, for consideration of trends, we also use the

Climate Research Unit (CRU) temperature dataset,

CRUTEM3, which consists of land air temperature anom-

alies on a 5 3 5 gridbox basis (Brohan et al. 2006). The

global dataset is derived from a collection of homoge-

nized, quality controlled, monthly average temperatures

for 4349 stations. It is based on 1961–90 average condi-

tions. Uses and limitations of this dataset are discussed by

Brohan et al. (2006).

Trends are examined by the best-fit linear trend from an

ordinary least squares (OLS) regression and confirmed by

the Mann–Kendall test (described by Onoz and Bayazit

2003). We consider the significance of these trends. Some

of the other studies mentioned earlier that examine tem-

perature in SA used the Mann–Kendall test for trend

detection. Vincent et al. (2005) consider the estimator of

the slope proposed by Sen (1968) and they use the Kendall

rank correlation. This method of trend estimation is ap-

plicable to their work, which examined extremes of tem-

perature (minimum and maximum temperature) where

the data has a skewed distribution. The Mann–Kendall

test can be used for any distribution. Our study of the

reanalysis examines the best-fit linear trend from an OLS

regression as well since we only consider the mean tem-

peratures in SA where the residuals examined are nor-

mally distributed. In addition, the coefficients of skewness

are all less than 1 for the South American reanalysis data

used, and the kurtosis values are also less than 1 following

the Z distribution for normally distributed data. However,

some of the station data is nonnormally distributed and,

therefore, the Mann–Kendall test alone is necessary to

determine trends and their significance.

In addition, the means and standard deviations of the

NNR temperature are determined for all months of

the year for the periods 1948–75 and 1976–2007, and

the results presented in this paper are the average of

December–February (DJF austral summer) and June–

August (JJA: austral winter). September–November

(SON: austral spring) and March–May (MAM: austral

autumn) are also considered. To evaluate the temper-

ature variability during the second period only (1976–

2007), this period is further subdivided into 1976–91

and 1992–2007, and a similar analysis is conducted. This

division into the periods 1948–75 and 1976–2007 is

based on the work of Obregon and Nobre (2003), who

verify the occurrence of a climate shift in the mid-1970s

from station precipitation data in SA. This is related to

a SST change mainly in the Pacific Ocean. This abrupt

climate shift affected almost all of SA except Northeast

Brazil (NEB). We also consider the temperature

for the last seven years (2001–07) since in the recent

period there is a marked temperature increase begin-

ning in 2001 in the whole of SA and globally. The sta-

tistical significance of the difference between the

composites, for the different periods considered in this

work, is calculated from the OLS regression applying

a Student’s t test, assuming that the population vari-

ances are not equal. With 27 (15) degrees of freedom,

the values of t are 1.70, 2.05, and 2.77 (1.75, 2.13, 2.97)

for 10%, 5%, and 1% statistical significance, re-

spectively.

The results shown in this paper could be associated

with natural and anthropogenic variability, thus we

verify the occurrence of all ENSO events based on the

Nino-3.4 index (Trenberth and Stepaniak 2001) during

the period 1948–2007. In the first period (1948–75),

there are five La Nina events (1950–51, 1957, and 1973–

75 are strong; 1963–64 and 1965–66 are moderate) and

five El Nino events (1955–56 and 1972–73 are strong,

1964–65 and 1968–69 are moderate, and 1950–51 is

weak). In the second period (1976–2007), we observe

five La Nina events (1975–76 and 1988–89 are strong,

1998–2001 is moderate, and 1984–85 and 1995–96 are

weak) and ten El Nino events (1982–83, 1991–92, and

1997–98 are strong; 1987–88, 1994–95, and 2002–03 are

moderate; and 1976–77, 1977–78, 2004–05, and 2006–07

are weak). To examine the ENSO effect further, in

addition to examining the periods 1948–75 and 1976–

2007, we next examine the differences in temperature

between these periods after the ENSO years identified

above have been removed.

We also perform an analysis involving the four

strongest El Nino and the three strongest La Nina events

observed between 1976 and 2007 for the DJF season. We

consider the La Nina 1975–76, 1988–89, and 1998–2000

events and the El Nino 1982–83, 1986–87, 1991–92,

and 1997–98 events. We do not evaluate this difference

in JJA because only one La Nina (1988) event and two

El Nino events (1991 and 1997) occurred having a Nino-

3.4 index value above one during the period 1976–2007.


3. Results

a. Annual mean temperature: NNR

Figure 2 compares the global, whole of SA, TSA, and

SSA annual mean air temperature at 2 m above the

earth’s surface (8C) between 1948 and 2007. The period

between 2000 and 2006 is considered as the globally

warmest of the last 100 years (Bernstein et al. 2007). In

the dataset used here, in the recent period the temper-

ature increase began in 2001 in all of SA and globally.

The year 2007 was the least warmest of the past seven

years for the majority of SA (Fig. 2), and the global value

for 2007 also dropped compared to the immediate pre-

ceding years, although it is still the fifth warmest in the

128-year period of record (Lawrimore 2008).

Figure 2 shows that in the period 1948–75 the global

mean temperature is around 13.88C; for the period 1976–

96 it oscillated around 148C; and for 1997–2007 the mean

annual temperature is above 148C, with maximum values

in 1998 and 2005. The maximum value of temperature in

1998 is associated with the strong ENSO event of 1997–

98. The year 2005 is the warmest in the period 1948–2007,

with a mean temperature of 14.58C. In this year normal

ENSO conditions are observed over the equatorial Pa-

cific for most months except January to March. From

2001 to 2007 a warming trend is observed globally, al-

though the earth saw a cooling in 2007 compared to the

preceding years. Over the period 1948–2007, there is

a significant (to the 1% level) upward linear trend noted

with an estimate of 0.098C decade21 based on the results

of the regression. The Mann–Kendall test also showed

a significant increasing trend (the Z value was 5.57).

The curve of annual mean temperature for SA is

similar to the curve for the globe in the period 1948–75,

with values around 18.58C; from 1976 the annual mean

temperature is above 18.58C except in 1999 and 2000. In

FIG. 2. Global, whole (SA) (608S–108N, 908–208W), tropical (TSA) (208S–108N, 808–358W),

and subtropical (SSA) (608–208S, 758–508W) annual mean air temperature at 2 m above the

earth’s surface (8C) between 1948 and 2007.


SA there is also a warming trend during 2001–06; how-

ever, in 2007 the temperature decreased. For SA the

warmest years are 1983 and 1998. Over the period 1948–

2007, there is a significant (to the 1% level) upward

linear trend noted with an estimate of 0.088C decade21

based on the results of the regression. The Mann–Kendall

test also showed a significant increasing trend (the Z

value was 3.92).

The temporal series of annual mean temperature over

TSA is similar to the series for the whole of SA and for

the globe with the maximum values in 1998, associated

with the ENSO phenomenon. In the period 1948–75 the

annual mean temperature over TSA is slightly above

238C; from 1976 to 2001 it is around 23.58C. In the pe-

riod 2001–07 a warming trend also occurs in TSA, with

warmest temperatures in 2005 with values of nearly

248C. Over the period 1948–2007, there is a significant

(to the 1% level) upward linear trend noted with an

estimate of 0.088C decade21 based on the results of the

regression. The Mann–Kendall test also showed a sig-

nificant increasing trend (the Z value was 4.20).

The temporal series of the annual mean temperature

over SSA has some similarities to the other series de-

scribed above for the period 1948–2007. In the period

1948–56 the annual mean temperature is around 128C,

between 1957 and 1976 it is around 12.58C, and from 1977

until 1998 it is nearly 138C. From 1999 the values of an-

nual mean temperature decreases and oscillates around

12.38C, with 2007 registering the lowest annual mean

temperature (11.78C) for all periods considered here.

This result is consistent with Lawrimore (2008) who

presents a summary of global temperature and notes that

FIG. 3. Mean air temperatures (8C) at 2 m above the earth’s surface in DJF and JJA during (a),(d) 1948–75 and (b),(e) 1976–2007; (c),(f)

the difference of the mean between these two periods. In (c) and (f) shaded areas represent statistical significance at the 5% level.


although the temperatures were warmer in 2007 for the

earth’s land temperatures, the globe actually experienced

a cooling when the ocean temperatures are also consid-

ered. He notes that the only widespread land areas with

below-average annual temperature occurred in SSA, with

the 2007 winter being unusually cold there. The results

from the OLS regression show a trend with an estimate of

0.088C decade21 significant to the 3% level. The Mann–

Kendall test also showed a significant increasing trend

(the Z value was 2.65).

We observed also that the temporal series noted above

(Fig. 2) corresponds well with the sign of the PDO. Be-

tween nearly 1948 and 1976 (1977 until around 1998) the

PDO index was negative (positive), which coincided with

lower (higher) temperatures over the globe and SA.

From 1998 the PDO index oscillated between negative

and positive.

b. Summer: NNR

The spatial patterns of the summertime temperature

in SA are nearly similar in both periods, 1948–75 and

1976–2007 (Figs. 3a,b). However, there are some nota-

ble differences. Over the majority of SA, the mean

temperature is between 218 and 248C during the period

1948–75. In this period, we observe temperatures be-

tween 188 and 218C over the mountain regions of Rio

Grande Sul (Serra Gaucha) and over the ‘‘Serra da

Mantiqueira,’’ located within the states of Sao Paulo,

Minas Gerais, and Rio de Janeiro. These mountains are

winter tourism places in Brazil. Temperatures above

248C are observed in the northern part of the Amazon

region, east and north of NEB and east of the Andes

cordillera.

In the period 1976–2007, we observe a change in the

temperature pattern in SA, with southward (in TSA and

SSA) and eastward displacement of the 248C isotherm

(Fig. 3b). Thus, in this period the temperature over most

of the continent is above 248C and the area with tem-

peratures below 248C decreased in size. In the ‘‘Serra

Gaucha’’ and ‘‘Serra da Mantiqueira’’ the temperature

is above 218C. Over most of the continent and the west-

ern part of the South Atlantic Ocean the temperature is

warmer in this second period.

The temperature difference between the periods 1948–

75 and 1976–2007, in the majority of SA, shows statistical

significance to at least the 5% level (Fig. 3c). These values

are significant particularly over Brazil, Argentina, and

part of Chile, with temperature differences above 0.68

and 1.28C, respectively. The last value is higher than the

increase in the total global temperatures (0.768C in-

crease) from 1850–99 to 2001–05, estimated by Bernstein

et al. (2007). We also observe a warming above 0.68C

over the South Atlantic Ocean, with a pattern that re-

sembles the SACZ (Chaves and Nobre 2004; Chaves

and Ambrizzi 2005).

To evaluate temperature variability in the second

period only (1976–2007), this period is further sub-

divided into 1976–91 and 1992–2007 (Fig. 4). The spatial

patterns of the temperature are similar in both periods

(figure not shown). However, in the period 1976–91 the

278C isotherm is located in the north of Argentina, while

it is found along the SA north coast during 1992–2007.

FIG. 4. Difference of the mean air temperatures (8C) at 2 m above the earth’s surface in (a) DJF and

(b) JJA for 1992–2007 and 1976–91. Shaded areas represent statistical significance at the 5% level.


Thus, we observe cooling in SSA and warming in TSA

in the period 1992–2007 when compared with 1976–91

(Fig. 4a). Considering a similar number of ENSO events

between the periods 1976–91 and 1992–2007, the results

above suggest a non-ENSO influence and is indicative of

other natural influences and/or a human influence on the

South American climate.

When we consider the difference between 1976–91

compared with 1948–75 and 1992–2007 compared with

1948–75 (Figs. 5a,b), we observe warming over almost the

entire SA continent in both periods. Although this

warming is greater over TSA in the period 1992–2007,

with values above 0.98C over NEB. In this period cooling

is observed only in the central Andes cordillera. How-

ever, it should be noted that the NNR model uses

a spectral representation of the orography; thus, the re-

sults in the Andes may have been influenced by the Gibbs

effect (Navarra et al. 1994). Already, in the period 1976–

91 the warming is greater in SSA, with values above 1.28C

in Argentina and Chile. Assimilation of the satellite data

in the more recent period may be a consideration when

analyzing these results; however, the fact that we see

FIG. 5. Difference of the mean temperatures (8C) in DJF and JJA between periods (a),(c) 1976–91

and 1948–75 and (b),(d) 1992–2007 and 1948–75. Shaded areas represent statistical significance at the

5% level.


similar trends with observational data (discussed later in

section 3e) gives us some confidence in the results.

The mean temperature for DJF for the period 2001–

07 is shown in Fig. 6. In this period a warming over al-

most the whole of SA is evident (cf. Fig. 6a with Figs.

3a,b), with an increase of the area where temperatures

are above 248C over the continent. We also observe an

increase of the area with temperatures above 278C in the

North Atlantic Ocean near the northern coastline of SA.

However, over the central Andes cordillera the area

with lower temperatures gets cooler, with temperatures

below 68C for the period 2001–07 (again it is important

to consider the Gibbs effect when interpreting the re-

sults in mountainous regions).

The temperature differences in DJF between 2001–07

and 1948–75, 1976–2007 and 1948–2007 are shown in

Fig. 7. We observe warming in TSA mainly in NEB and

over the equatorial North Atlantic Ocean. These tem-

perature differences are greater when 2001–07 is com-

pared with the 1948–75 period, with values above 1.48C

in NEB (Fig. 7a). In part of the SSA region the warming

is observed with values above 1.28C only when com-

pared with the period 1948–75. We observe cooling over

part of the SSA region, mainly over the Andes cordil-

lera, in the period 2001–07 when compared with the

periods 1976–2007 and 1948–2007 (Figs. 7b,c).

Figures 8a–c shows the analysis of the period 1948–75

compared to 1976–2007, as in Figs. 3a–c, but now with

significant ENSO events removed. It can be seen when

comparing Fig. 3 and Fig. 8 that similar results emerge;

that is, there is a significant increase in temperatures

for the majority of SA in the more recent period. This

suggests that ENSO does not account for the noted

warming in most of SA. Figure 9 shows the mean tem-

perature difference between the four strongest La Nina

and El Nino events in the period 1976–2007. We observe

warming in SSA, mainly over the central part of Ar-

gentina, with values above 0.68C, and cooling in TSA,

with values around 20.68C corresponding to La Nina

events. This difference has statistical significance at

greater than the 10% level. Comparing values of the

mean temperature difference between ENSO extremes

with the values of the mean temperature difference

between the periods above (Figs. 3c, 4a, and 5a,b),

we see that the values are slightly greater between the

periods than those associated with the difference be-

tween extreme ENSO events. The values and signifi-

cance of the difference between positive and negative

phases of ENSO should be higher (if this phenomenon is

the main cause for the temperature changes) when

compared with the two long periods considered. Thus,

these results indicate that the ENSO phenomenon

cannot be largely responsible for these temperature

differences between the different periods. This is in

agreement with Vincent et al. (2005), who note that

ENSO has not affected the series of cold days and cold

nights.

The interannual temperature variability in DJF, de-

termined by the standard deviation, is not intense over

the majority of SA for all periods 1948–75, 1976–2007,

and 2001–07 (Fig. 10). In the TSA region the standard

deviations are around 0.58C and in SSA they are between

FIG. 6. Mean air temperature (8C) at 2 m above the earth’s surface in (a) DJF and (b) JJA for 2001–07.


18 and 3.08C. During the period 2001–07 values above 18C

are observed only over the Andes cordillera (Fig. 10c).

This result may be associated with a smaller number of,

and weaker, El Nino events occurring in this period.

The temperature variability over SSA is higher in the

period 1992–2007 than the period 1976–91 (not shown).

Thus, the temperature variability observed in the whole

1976–2007 period (Fig. 10b) is due to the greater tem-

perature variability in the period 1992–2007.

c. Winter: NNR

The wintertime temperature patterns in SA show tem-

peratures above 248C over part of the Amazon region

and in the northern part of NEB in periods 1948–75 and

1976–2007 (Figs. 3d,e). However, in the period 1976–

2007, the area with temperatures above 248C increases

in size. Thus, we observe warmer temperatures in TSA

during 1976–2007 compared to 1948–75. The differences

in wintertime mean temperature between these two

periods are more evident in TSA (Fig. 3f). The position

of the 218 and 248C isotherms over NEB indicates that

this region was cooler in the past (1948–75) than in the

more recent time period (1976–2007). In the first period

(1948–75) the 218C isotherm is around Salvador

(;138S); however, in the second period this isotherm

shifts to the south.

Over the North Atlantic Ocean, the temperatures are

also greater in the period 1976–2007 with values above

278C along the northern coastline of SA (Figs. 3d,e). Over

the western part of the South Atlantic Ocean it is also

warmer during 1976–2007 with southward displacement of

FIG. 7. Difference of the mean temperatures (8C) in DJF and JJA between periods (a),(d) 2001–07 and 1948–75; (b),(e) 2001–07 and

1976–2007; and (c),(f) 2001–07 and 1948–2007. Shaded areas represent statistical significance at the 5% level.


the 248C isotherm. We observe a warming of 0.68C in

central Brazil and NEB in the period 1976–2007 compared

to the period 1948–75 (Fig. 3f).

Similar to the analysis of the summer season, Figs. 8d–f

show the analysis of the period 1948–75 compared to

1976–2007, as in Figs. 3d–f, but now with significant

ENSO events removed. It can be seen when comparing

Fig. 3 and Fig. 8 that, again, similar results emerge; that

is, there is a significant increase in temperatures for the

majority of SA in the more recent period—thus sup-

porting the suggestion that ENSO does not account for

the noted warming in most of SA.

Again, for the summer season we also divide the

second period (1976–2007) in this analysis of winter

temperatures in two parts: 1976–91 and 1992–2007.

There is no accentuated difference in the temperature

pattern in SA in both periods for winter (figure not

shown). However, we observe warming during 1992–

2007 in the central Amazon region, and the equatorial

tropical North Atlantic Ocean near SA, with values

above 0.68 and 0.38C, respectively and cooling in part of

the SSA region, in the Andes cordillera, and Chile with

values below 21.58C (Fig. 4b). The reader should note

the Gibbs effect referred to earlier.

The area with warming increases in the central part of

SA, NEB, and the North Atlantic Ocean near SA in the

period 1992–2007 more than in 1976–91 when compared

with the temperatures in the period 1948–75 (Figs. 5c,d).

In the period 1976–91, a cooling is observed in the central

Amazon region.

FIG. 8. Mean air temperatures (8C) at 2 m above the earth’s surface in DJF and JJA during (a),(d) 1948–75 and (b),(e) 1976–2007; (c),(f)

the difference of the mean between these two periods. Note in each period significant ENSOs have been removed. In (c) and (f) shaded

areas represent statistical significance at the 5% level.


When we compare the mean temperature in 2001–07

(Fig. 6b) with the mean temperature of other periods

(Figs. 7d–f), we observe that the warming is greater in

the recent period in the TSA region and over the North

Atlantic Ocean, with values above 0.68C. However,

we observe cooling with values below 22.78C over parts

of Chile, Peru, and Argentina (again the Gibbs effect

relevant for the Andes should be noted for these areas).

As seen with the DJF data, the interannual tempera-

ture variability in JJA is not strong over the majority of

SA (Figs. 10d–f), except over SSA, where the values of

the standard deviation are between 18 and 28C in the

periods 1948–75 and 1976–2007. In the period 2001–07,

the standard deviation is above 1.58C, which is lower

than those in the summer season considering the same

period. However, we observe a greater area of the SSA

region where the variability in temperature increased

during the period 2001–07.

d. Autumn and spring: NNR

The spatial patterns of autumn and spring are shown

in Figs. 11 and 12, respectively, for the periods 1948–75

and 1976–2007. Similar to what we observed in summer

and winter, the temperature patterns in the spring and

autumn season also changed when we compared the

period 1976–2007 with 1948–75. In 1976–2007 com-

pared to 1948–75, over central SA in autumn the area

with temperatures between 218 and 248C decreases and

for temperatures greater than 248C the area increases

in size, with southward displacement of the 248C iso-

therm. Over the equatorial Atlantic we observe tem-

peratures above 278C. The temperature difference

between 1948–75 and 1976–2007 is significant over al-

most all of SA, with warming between 0.38 and 0.68C

(Fig. 11c). The increasing temperature changes from

the period 1948–75 to 1976–2007 are more notable in

autumn than in the spring. For spring (Fig. 12) we ob-

served warming over NEB, subtropical SA, and the

equatorial Atlantic and cooling over the Amazon re-

gion. Figure 13 shows that, when the period 2001–07 is

considered, the temperature patterns in autumn and

spring are different compared to earlier periods. This is

similar to what is observed in winter and summer, with

greater temperatures over SA in the period 2001–07.

The patterns of the four seasons also show that the

NNR performs well in capturing the intraseasonal

variability.

e. Station data and CRUTEM3 observational data

An examination of trends for the station data obtained

from INMET and IAG/USP (Table 1) shows general

agreement with that found with the reanalysis. The slope

and significance is only shown for stations where the

data were determined to be normally distributed (as

noted by a skewness and kurtosis value of less than 1).

To test for all stations (including ones that were non-

normally distributed), as noted earlier, the Mann–Kendall

test is used. Hence, the Z value from the Mann–Kendall

test is noted in addition to the direction of the trend. A

Z value greater than 1.96 indicates significance at the

5% level. It can be seen that there is good agreement

with the results from the NNR in that all of the stations

having a significant trend showed one in which the tem-

peratures were increasing. An examination (not shown)

of the variability of the NNR compared to station data

shows generally good agreement, despite an offset in the

actual values (noted earlier by Kalnay and Cai 2003;

Nunez et al. 2008).

An examination of trends for the annual mean tem-

perature in the CRUTEM3 dataset revealed similar

trends (and significance) to that observed with the

NNR (section 3a). With CRUTEM3, all areas exam-

ined by an OLS regression (global, SA, TSA, and SSA)

revealed increasing temperature trends (significant to

the 1% level). The Mann–Kendall test for trend de-

tection confirmed these positive significant trends with

a Z value of 4.60, 4.93, 4.31, and 3.13 for global, SA,

TSA, and SSA, respectively. These results provide us

with greater confidence in the results obtained from

the NNR.

FIG. 9. Temperature differences (8C) between the four strongest

La Nina and El Nino events observed between 1976 and 2007.


4. Summary and conclusions

This work shows how the air temperature at 2 m

above the earth’s surface in SA varies in the NNR

dataset between 1948 and 2007. There are significant

increasing temperature trends observed in the Brazilian

station data examined and the CRUTEM3 dataset of

observed values, in addition to the trends observed in

the NNR. Finding these trends also with the CRUTEM3

and station datasets provides greater confidence in the

results from the NNR that these are real changes rather

than due to changes in observing systems. Considering

the NNR, in DJF during 1948–75 over the majority of

SA, the mean temperature is between 218 and 248C.

However, in the period 1976–2007, the mean tempera-

ture over most of SA is above 248C with warming in the

TSA and SSA regions.

The analysis from the NNR also shows cooling

(warming) in the SSA (TSA) latitudes in the period 1992–

2007 compared with the period 1976–91. Since both pe-

riods have a major El Nino event, this suggests that the

warming in TSA may not be due to the El Nino phe-

nomenon but rather some other component of natural

variability of the climate and/or may be a result of human

activity. Indeed, when major ENSO events were removed

from the data series, the warming was still observed.

Bernstein et al. (2007) note that only models including

anthropogenic forcing can simulate the observed patterns

of warming. They also note ‘‘no coupled global climate

model that has used natural forcing only has reproduced

FIG. 10. Standard deviation of the air temperature (8C) at 2 m above the earth’s surface in DJF and JJA in (a),(d) 1948–75;

(b),(e) 1976–2007; and (c),(f) 2001–07. Shaded areas represent values above 18C.


the continental mean warming trends in individual con-

tinents (except Antarctica) over the second half of the

20th century’’ (Bernstein et al. 2007, p. 39).

The differences in wintertime mean temperature be-

tween the periods 1948–75 and 1976–2007 are more evi-

dent in the TSA region. The results indicate that NEB

was cooler in the past (1948–75) than in the more recent

time period (1976–2007). Over the North Atlantic Ocean

the temperatures are also greater in the period 1976–

2007, with values above 278C along the northern coastline

of SA. We also observe warming in central Brazil, NEB,

and north of SA in the period 1976–2007 compared to

the period 1948–75. We note a warming in the central

Amazon region and over the equatorial tropical North

Atlantic Ocean near SA and cooling in SSA in the period

1992–2007 compared with 1976–91. When we evaluate

the mean temperature for 2001–07 compared with all

others periods, the warming is greater in the TSA region

and over the North Atlantic Ocean, and cooling is ob-

served over part of SSA. Increasing mean temperature

trends can be seen more clearly in summer than winter for

a larger portion of SA. This is consistent with Vincent

et al. (2005), who note that warming observed in extreme

nighttime temperatures is mostly due to warm nights and

fewer cold nights during summer and fall.

The results presented here indicate that the climate

change over SA is not due to ENSO alone (the most

important coupled ocean–atmosphere phenomenon to

FIG. 11. Mean air temperatures (8C) at 2 m above the earth’s surface in autumn (MAM) during (a) 1948–75 and (b) 1976–2007; (c) the

difference of the mean temperature between these two periods. In (c) shaded areas represent statistical significance at the 5% level.

FIG. 12. As in Fig. 11, but in spring (SON).


produce climate variability over SA on an interannual

time scale) and is likely due to other natural variability

of the climate and/or the result of human activity. In fact,

when considering a more modest increase on the global

scale, some authors (Wigley and Barnett 1992; Hulme

and Jones 1994) state that natural processes alone can-

not explain this increase, thus making the argument for

anthropogenic warming for SA stronger. Indeed, there

has been more land use change and greater carbon di-

oxide release during the most recent period (Kalnay and

Cai 2003). Only global and regional climate models with

anthropogenic forcing can, perhaps, explain tempera-

ture warming due to human activities. It is possible that

changes in the observational networks could influence

the results; however, to verify this hypothesis it would be

necessary to examine complete datasets from meteoro-

logical stations over a long period of time.

Acknowledgments. J. Collins would like to thank

the University Corporation for Atmospheric Research

(UCAR) for the support to attend the Latin American

Data Workshop (21–23 August 2008). R. Chaves

is grateful to the FAPERJ (APQ1171.225/2006 and

APQ170.339/2006), CNPq (PQ-301591/2005 and EU-

472832/2006-9), TECNORTE, and FINEP (CICLONES

and PREVRIO) for financial support and to the USF

Department of Geography for support through the use of

facilities during the preparation of this paper. The au-

thors thank the Institute of Astronomy and Geophysics/

University of Sao Paulo (IAG/USP) and the Brazilian

National Institute of Meteorology (INMET) for access to

some station data and we thank the U. K. Climate Re-

search Unit (CRU) for making gridded observational

data available on the internet. We would also like to

thank Thomas Boucher (Plymouth State University) and

George Douglas Lunsford (University of South Florida)

for review and discussion of our statistical techniques. In

addition we thank four anonymous reviewers for com-

ments and suggestions on the manuscript.

FIG. 13. Mean air temperatures (8C) at 2 m above the earth’s surface in (a) autumn (MAM) and

(b) spring (SON) during 2001–07.

TABLE 1. Temperature trends from station data 1961–2007 (slope

and significance obtained from a linear regression; for those stations

where the data is normally distributed, * indicates nonnormally

distributed data) and Z value and trend noted from the Mann–

Kendall test (note the value is significant to the 5% level if the Z

value is greater than the absolute value of 1.96). Data obtained from

INMET, except for Sao Paulo, which is obtained from IAG.

Station Slope Significance Z Trend

Belem 0.029 0.000 5.89 Increasing

Brasilia * * 4.33 Increasing

Curitiba 0.034 0.000 4.73 Increasing

Florianopolis * * 3.83 Increasing

Fortaleza 0.013 0.005 2.84 Increasing

Goiana 0.044 0.000 6.46 Increasing

Manaus 0.012 0.002 2.84 Increasing

Maceio * * 21.55 No trend

Natal * * 2.04 Increasing

Porto Alegre * * 0.79 No trend

Rio Branco * * 2.10 Increasing

Rio de Janeiro 0.038 0.000 2.67 Increasing

Salvador * * 3.87 Increasing

Recife 0.025 0.000 6.46 Increasing

Sao Paulo 0.041 0.000 6.76 Increasing


REFERENCES

Alexander, L. V., and Coauthors, 2006: Global observed changes

in daily climate extremes of temperature and precipitation.

J. Geophys. Res., 111, D05109, doi:10.1029/2005JD006290.

Barros, V. R., A. M. Grimm, and M. E. Doyle, 2002: Relationship

between temperature and circulation in southeastern South

America and its influence from El Nino and La Nina events.

J. Meteor. Soc. Japan, 80, 21–32.

Bernstein, L., and Coauthors, 2007: Climate Change 2007: Synthesis

Report. IPCC, 104 pp.

Brohan, P., J. J. Kennedy, I. Harris, S. F. B. Tett, and P. D. Jones,

2006: Uncertainty estimates in regional and global observed

temperature changes: A new dataset from 1850. J. Geophys.

Res., 111, D12106, doi:10.1029/2005JD006548.

Brown, S. J., J. Caesar, and C. A. T. Ferro, 2008: Global changes in

extreme daily temperature since 1950. J. Geophys. Res., 113,

D05115, doi:10.1029/2006JD008091.

Caesar, J., and L. Alexander, 2006: Large-scale changes in ob-

served daily maximum and minimum temperatures and anal-

ysis of a new gridded data set. J. Geophys. Res., 111, D05101,

doi:10.1029/2005JD006280.

Cai, M., and E. Kalnay, 2005: Can reanalysis have anthropogenic

climate trends without model forcing? J. Climate, 18, 1844–

1849.

Carvalho, L. M. V., C. Jones, and B. Liebmann, 2002: Extreme

precipitation events in southeastern South America and large-

scale convective patterns in the South Atlantic convergence

zone. J. Climate, 15, 2377–2394.

Chaves, R. R., and P. Nobre, 2004: Interactions between the sea

surface temperature over the South Atlantic Ocean and the

South Atlantic convergence zone. Geophys. Res. Lett., 31,

L03204, doi:10.1029/2003GL018647.

——, and T. Ambrizzi, 2005: Atmospheric response for two con-

vection schemes in sensitivity experiments using SST anoma-

lies over the South Atlantic Ocean. CLIVAR Exchanges, No.

33, International CLIVAR Project Office, Southampton,

United Kingdom, 25–27.

Dettinger, M. D., D. S. Battisti, R. D. Garreaud, G. J. McCabe Jr.,

and C. M. Bitz, 2000: Interhemispheric effects of interannual

and decadal ENSO-like climate variations on the Americas.

Present and Past Interhemispheric Climate Linkages in the

Americas and Their Societal Effects, V. Markgraf, Ed., Cam-

bridge University Press, 1–16.

Dufek, A. S., T. Ambrizzi, and R. P. da Rocha, 2008: Are reanalysis

data useful for calculating climate indices over South Amer-

ica? Trends and directions in climate research Ann. N. Y.

Acad. Sci., 1146, 87–104.

Frich, P., V. Alexander, P. Della-Marta, B. Gleason, M. Haylock,

A. M. G. Klein Tank, and T. Peterson, 2002: Observed co-

herent change in climatic extremes during the second half of

the twentieth century. Climate Res., 19, 193–212.

Hulme, M., and P. D. Jones, 1994: Global climate change in the

instrumental period. Environ. Pollut., 83, 23–26.

Kalnay, E., and M. Cai, 2003: Impact of urbanization and land-use

change on climate. Nature, 423, 528–531.

——, and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis

Project. Bull. Amer. Meteor. Soc., 77, 437–471.

Lawrimore, J., 2008: Global temperature highlights of 2007.

Weatherwise, 61, 18.

Malhi, Y., J. T. Roberts, R. A. Betts, T. J. Killeen, W. Li, and

C. A. Nobre, 2008: Climate change deforestation and the fate

of the Amazon. Science, 319, 169–172.

Mantua, N. J., S. R. Hare, Y. Zhang, J. M. Wallace, and

R. C. Francis, 1997: A Pacific interdecadal climate oscillation

with impacts on salmon production. Bull. Amer. Meteor. Soc.,

78, 1069–1079.

Marengo, J. A., and C. C. Camargo, 2008: Surface air temperature

trends in southern Brazil for 1960–2002. Int. J. Climatol., 28,

893–904.

Navarra, A., W. F. Stern, and K. Miyakoda, 1994: Reduction of the

Gibbs oscillation in spectral model simulations. J. Climate, 7,1169–1183.

Nobre, C. A., P. Sellers, and J. Shukla, 1991: Amazonian de-

forestation and regional climate change. J. Climate, 4, 957–988.

Nunez, M. N., H. H. Ciapessoni, A. Rolla, E. Kalnay, and M. Cai,

2008: Impact of land use and precipitation changes on surface

temperature trends in Argentina. J. Geophys. Res., 113,

D06111, doi:10.1029/2007JD008638.

Obregon, G., and C. A. Nobre, 2003: Rainfall trends in Brazil. Bull.

Amer. Meteor. Soc., 84, 1008–1009.

Onoz, B., and M. Bayazit, 2003: The power of statistical tests for

trend detection. Turk. J. Eng. Environ. Sci., 27, 247–251.

Quintana-Gomez, R., 1999: Trends of maximum and minimum tem-

perature in northern South America. J. Climate, 12, 2104–2112.

Ropelewski, C. F., and M. S. Halpert, 1987: Global and regional

scale precipitation patterns associated with the El Nino/

Southern Oscillation. Mon. Wea. Rev., 115, 1606–1626.

Rusticucci, M. M., and V. E. Kousky, 2002: A comparative study

of maximum and minimum temperatures over Argentina:

NCEP–NCAR reanalysis versus station data. J. Climate, 15,

2089–2101.

——, and M. Barrucand, 2004: Observed trends and changes in

temperature extremes over Argentina. J. Climate, 17, 4099–4107.

Sen, P. K., 1968: Estimates of the regression coefficient based on

Kendall’s tau. J. Amer. Stat. Assoc., 63, 1379–1389.

Trenberth, K. E., and D. P. Stepaniak, 2001: Indices of El Nino

evolution. J. Climate, 14, 1697–1701.

——, G. W. Branstator, D. Karoly, A. Kumar, N.-C. Lau, and

C. Ropelewski, 1998: Progress during TOGA in understanding

and modeling global teleconnections associated with tropical

sea surface temperatures. J. Geophys. Res., 103, 14 291–14 324.

Victoria, R. L., L. A. Martinelli, J. M. Moraes, M. V. Ballester, and

A. V. Krushche, 1998: Surface air temperature variations in

the Amazon region and its borders during this century.

J. Climate, 11, 1105–1110.

Vimont, D. J., 2005: The contribution of the interannual ENSO

cycle to the spatial pattern of decadal ENSO-like variability.

J. Climate, 18, 2080–2092.

Vincent, L. A., and Coauthors, 2005: Observed trends in indices of

daily temperature extremes in South America 1960–2000.

J Climate, 18, 5011–5023.

Wigley, T. M. L., and T. P. Barnett, 1992: Cambio Climatico,

Evaluacion Cientifica del IPCC. Organizacion Meteorologica

Mundial, 382 pp.

Zhang, Y., J. M. Wallace, and D. S. Battisti, 1997: ENSO-like

interdecadal variability. J. Climate, 10, 1004–1020.

Zhou, J., and K. M. Lau, 2001: Principal modes of interannual and

decadal variability of summer rainfall over South America.

Int. J. Climatol., 21, 1623–1644.


Temperature Variability over South America - USF School of...

Documents

Transcript of Temperature Variability over South America - USF School of...