Observed Changes in Return Values of Annual Temperature Extremes over Argentina

Matilde Rusticucci Departamento de Ciencias de la Atmósfera y los Océanos, Facultad de Ciencias Exactas y Naturales, Universidad de Buenos Aires, and Consejo Nacional de Investigaciones Científicas y Técnicas, Buenos Aires, Argentina

Search for other papers by Matilde Rusticucci in
Current site
Google Scholar
PubMed
Close
and
Bárbara Tencer Departamento de Ciencias de la Atmósfera y los Océanos, Facultad de Ciencias Exactas y Naturales, Universidad de Buenos Aires, and Consejo Nacional de Investigaciones Científicas y Técnicas, Buenos Aires, Argentina

Search for other papers by Bárbara Tencer in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

Extreme temperature events are one of the most studied extreme events since their occurrence has a huge impact on society. In this study, the frequency of occurrence of absolute extreme temperature events in Argentina is analyzed. Four annual extremes are defined based on minimum and maximum daily data: the highest maximum (minimum) temperature of the year, and the lowest maximum (minimum) temperature of the year. Applying the extreme value theory (EVT), a generalized extreme value (GEV) distribution is fitted to these extreme indices and return values are calculated for the period 1956–2003. Its spatial distribution indicates that, for warm extremes, maximum temperature (Tx) is expected to be greater than 32°C at least once every 100 yr throughout the country (reaching values even higher than 46°C in the central region), while minimum temperature (Tn) is expected to exceed 16°C (reaching 30°C in the central and northern regions). Cold annual extremes show larger gradients across the country, with Tx being lower than 8°C at least once every 100 yr, and Tn lower than 0°C every 2 yr, with values even less than −10°C in the southwestern part of the country.

However, the frequency of occurrence of climatic extremes has changed throughout the globe during the twentieth century. Changes in return values of annual temperature extremes due to the 1976–77 climatic shift at six long-term datasets are then analyzed. The lowest Tx of the year is the variable in which the 1976–77 shift is less noticeable. At all the stations studied there is a decrease in the probability of occurrence of the highest Tx if the study is based on more recent records, while the frequency of occurrence of the highest Tn increases at some stations and decreases at others. This implies that in the “present climate” (after 1977) there is a greater frequency of occurrence of high values of Tn at Observatorio Central Buenos Aires and Río Gallegos together with a lower frequency of occurrence of high values of Tx, leading to a decrease in the annual temperature range.

The most noticeable change in return values due to the 1976–77 shift is seen in Patagonia where the 10-yr return value for the highest Tn increases from 13.7°C before 1976 to 18.6°C after 1977. That is, values of the highest Tn that occurred at least once every 10 yr in the “past climate” (before 1976) now happened more than once every 2 yr.

Corresponding author address: Matilde Rusticucci, Departamento de Ciencias de la Atmósfera y los Océanos, FCEN, Universidad de Buenos Aires, Ciudad Universitaria Pab II (1428), Buenos Aires, Argentina. Email: mati@at.fcen.uba.ar

Abstract

Extreme temperature events are one of the most studied extreme events since their occurrence has a huge impact on society. In this study, the frequency of occurrence of absolute extreme temperature events in Argentina is analyzed. Four annual extremes are defined based on minimum and maximum daily data: the highest maximum (minimum) temperature of the year, and the lowest maximum (minimum) temperature of the year. Applying the extreme value theory (EVT), a generalized extreme value (GEV) distribution is fitted to these extreme indices and return values are calculated for the period 1956–2003. Its spatial distribution indicates that, for warm extremes, maximum temperature (Tx) is expected to be greater than 32°C at least once every 100 yr throughout the country (reaching values even higher than 46°C in the central region), while minimum temperature (Tn) is expected to exceed 16°C (reaching 30°C in the central and northern regions). Cold annual extremes show larger gradients across the country, with Tx being lower than 8°C at least once every 100 yr, and Tn lower than 0°C every 2 yr, with values even less than −10°C in the southwestern part of the country.

However, the frequency of occurrence of climatic extremes has changed throughout the globe during the twentieth century. Changes in return values of annual temperature extremes due to the 1976–77 climatic shift at six long-term datasets are then analyzed. The lowest Tx of the year is the variable in which the 1976–77 shift is less noticeable. At all the stations studied there is a decrease in the probability of occurrence of the highest Tx if the study is based on more recent records, while the frequency of occurrence of the highest Tn increases at some stations and decreases at others. This implies that in the “present climate” (after 1977) there is a greater frequency of occurrence of high values of Tn at Observatorio Central Buenos Aires and Río Gallegos together with a lower frequency of occurrence of high values of Tx, leading to a decrease in the annual temperature range.

The most noticeable change in return values due to the 1976–77 shift is seen in Patagonia where the 10-yr return value for the highest Tn increases from 13.7°C before 1976 to 18.6°C after 1977. That is, values of the highest Tn that occurred at least once every 10 yr in the “past climate” (before 1976) now happened more than once every 2 yr.

Corresponding author address: Matilde Rusticucci, Departamento de Ciencias de la Atmósfera y los Océanos, FCEN, Universidad de Buenos Aires, Ciudad Universitaria Pab II (1428), Buenos Aires, Argentina. Email: mati@at.fcen.uba.ar

1. Introduction

Extreme events have a huge impact on society and the ecosystem. Particularly, high and low temperatures are one of the most studied extreme events, since their occurrence severely influences agriculture (many crops are affected by the number of frost days or the amount of hot days per year), human health (the heat wave that affected Europe in 2003 caused between 22 000 and 35 000 deaths according to Schär and Jendritzky 2004; see Schär et al. 2004 for a statistical analysis of the heat wave), demand for energy, water resources, and the availability of drinkable water, among others. Under a climatic change, a small variation in the mean values of temperature can be associated with large changes in the frequency of extreme events (Katz and Brown 1992).

Identifying trends in climatic extremes implies an extra demand on data quantity and quality than the study of changes in mean values. This is because even a relatively small amount of missing data raises the possibility that an extreme event has not been recorded, especially when absolute extreme events are being studied. Also, when investigating trends at the extreme ends of a climatic distribution, the quality of data can also influence the results since an outlier can be incorrectly considered as a true extreme value or vice versa (a genuine extreme may be rejected as an outlier).

During the last few years new datasets of daily temperature in different parts of the globe have become available allowing the study of trends in extreme temperature events. The majority of the findings revealed that the persistence and intensity of extreme temperature values has changed—globally, but not homogeneously—throughout the period of observation (Trenberth et al. 2007). Frich et al. (2002) used extreme temperature indices based on daily minimum and maximum temperature series to investigate possible global changes in the frequency and/or severity of climatic extremes during the second half of the twentieth century. They found an increase in warm summer nights, a decrease in the number of frost days and a reduction of intra-annual extreme temperature range. Alexander et al. (2006) reported that most of the global land area sampled in their study showed a significant decrease in the annual occurrence of cold nights and a significant increase in the annual occurrence of warm nights, which implies a positive shift in the distribution of daily maximum temperature throughout the globe.

Over Australia, the study of trends in extreme temperature indices revealed that the frequency of warm temperature extreme events has generally increased over the period 1957–96, while the number of extremely cool temperature events has decreased, consistently with increasing trends in mean minimum and maximum temperatures during that period (Collins et al. 2000). In Southeast Asia and the South Pacific, Manton et al. (2001) also found significant increases in the annual number of hot days and warm nights and significant decreases in the annual number of cool days and cold nights in the period 1961–98. These trends were stronger for indices based on minimum temperature than for those based on maximum temperature. In Europe, Klein Tank and Können (2003) observed a symmetric warming of the cold and warm tails of the daily minimum and maximum temperature distributions during 1946–99. However, this symmetry disappeared when trends were analyzed in two consecutive periods: for the 1946–75 subperiod (an episode of slight cooling), the annual number of warm extremes decreased, but the annual number of cold extremes did not increase; for the 1976–99 subperiod (an episode of pronounced warming), the annual number of warm extremes increased two times faster than the decrease in the number of cold extremes.

Significant increasing trends in the percentage of warm nights and decreasing trends in the number of cold nights were found at many stations in South America according to Vincent et al. (2005). The authors believe that this warming is mostly due to more warm nights and fewer cold nights during the summer (December–February) and fall (March–May). They also found that trends in indices based on daily maximum temperature were not consistent and that the stations with significant trends were located closer to the west and east coasts of the continent.

In Argentina, Rusticucci and Barrucand (2004) showed that the strongest (positive) changes over time occurred in mean summer minimum temperature, while its standard deviation decreased. On the other hand, mean maximum summer temperatures mostly decreased over time in northern Argentina, but increased in Patagonia (southern Argentina). Overall, negative trends were obtained for the number of cold nights and warm days per summer, while the number of warm nights and cold days has increased at certain locations. They also studied the relationship between seasonal mean temperature and the frequency of occurrence of extremely warm and cold days. Results indicate that an increase in the summer mean temperature is mostly due to an increase in the frequency of warm events in the central and northern part of the country, and to a decrease in the number of cold events in the southern region (Barrucand and Rusticucci 2001).

All the studies mentioned above use the empirical distribution function to study extreme events. Nonetheless, very small discrepancies in the estimate of the empirical distribution function can lead to substantial discrepancies in the distribution of extreme events (Coles 2001). Therefore, an alternative approach consists of applying the extreme value theory (EVT) to study extreme events and extrapolate information to unobserved levels. Based on the annual and daily rainfall data on the central coast of Venezuela and using different modeling strategies and inference approaches, Coles et al. (2003) showed that the 1999 rainfall, which caused the worst environmentally related tragedy in Venezuelan history, was extreme, but not implausible given the historical evidence.

Statistical modeling of extreme values has been applied to daily data throughout the globe in order to extrapolate the probability of occurrence of events that are more extreme than any other observed. In this context, Zwiers and Kharin (1998) applied the statistical modeling of extremes to observed daily records of minimum and maximum temperature and they compared it to the extremes of climate simulated in an equilibrium doubled CO2 experiment conducted with a general circulation model (GCM). Bonsal et al. (2001) fitted a generalized extreme value (GEV) distribution to the annual extremes of daily minimum and maximum temperature in Canada. They found that 20-yr return values of annual extremes of daily minimum temperature increase approximately 4°C from the period 1900–49 to 1950–98 over southern Canada, while for maximum temperature return values decrease by 1°C.

In this paper, we use a theoretical distribution function to model extreme temperature events observed during the second half of the twentieth century in Argentina. A GEV distribution is fitted to annual extremes of daily minimum and maximum temperature at different stations in the country for the period 1956–2003. Then, return levels of annual extremes are estimated and their spatial distribution is analyzed.

However, the frequency of occurrence of climatic means and extremes has changed throughout the globe during the twentieth century. Several studies have analyzed the presence of climatic changes in different atmospheric and oceanic variables. Trenberth (1990) found a different regime after 1976 evident both in the time series of Pacific mean sea level pressure and in the temperature anomalies. In the Southern Hemisphere, Seidel et al. (2004) found a warming of a few tenths of degree in different upper-air temperature datasets while Huang et al. (2005) observed a decrease in January–May precipitation over the Amazon region and the southern tip of South America together with an increase of rainfall in the sub-Amazonian South America.

Therefore, this paper also aims to study the effects of the 1976–77 climate shift on return levels of annual temperature extremes. For this reason, the longest annual extreme series available in Argentina are divided into two consecutive subperiods: before 1976 (pre-1976) and thereafter (post-1976), and the GEV distribution is fitted to each subperiod. Annual extreme return levels are then recomputed and possible changes in the frequency of occurrence of extreme temperatures due to the regime shift are analyzed.

The outline for the remainder of this paper is as follows. The data and methodology used are described in section 2. Results obtained for the period 1956–2003 are shown in section 3 and the influence of the 1976–77 shift is analyzed in section 4. A summary and conclusions are presented in section 5.

2. Data and methodology

Daily data of maximum and minimum temperature (Tx and Tn, respectively) for 43 stations over the period 1956–2003, provided by the Servicio Meteorológico Nacional (National Weather Service), were used to estimate the spatial distribution of fixed annual extreme return levels and periods in Argentina (Fig. 1). An exhaustive quality control of these data was performed by Barrucand and Rusticucci (2001).

Another longer-term dataset was used to assess changes in annual extreme return levels throughout the observational period. This dataset consists of six stations from Argentina with record lengths between 67 and 110 yr: San Miguel de Tucumán, Tucumán (TUC, 26.80°S, 65.20°W, 1891–2000); Observatorio Central Buenos Aires, Buenos Aires (OCBA, 34.58°S, 58.48°W, 1906–2004); Pergamino, Buenos Aires (PGM, 33.93°S, 60.55°W, 1931–2004); Pilar, Córdoba (PIL, 31.66°S, 63.88°W, 1931–2004); Santa Rosa, La Pampa (SRS, 36.56°S, 64.26°W, 1937–2003); Río Gallegos, Santa Cruz (RGA, 51.61°S, 69.28°W, 1896–2004). A supplementary internal consistency analysis was made prior to return level calculations, by searching the time series for outliers and indisputable erroneous data (such as Tn greater than Tx). At these stations, changes in return levels due to the 1976–77 climate shift were studied by comparing the distribution of annual temperature extreme return levels based on the observational period pre-1976 (beginning of the time series–1976) and post-1976 (1977–end of the time series).

Four different annual temperature extremes based on daily data were calculated for each station and each period studied: the highest maximum (minimum) temperature of the year called HTx (HTn), and the lowest maximum (minimum) temperature of the year, called LTx (LTn). For warm extremes (HTx, HTn), the year was considered as the period of time from 1 July to 30 June next year so that austral summer was not broken. For cold extremes (LTx, LTn), the calendar year was used.

An extensive analysis of missing data was conducted at each station record individually. The year was considered missing for the series of warm (cold) annual extremes when 20 (or more) nonconsecutive days, or 10 (or more) consecutive days were missing in the period from December to February (June to August).

Extreme value analysis was performed in this study by fitting the GEV distribution to the sample of annual extremes defined above at each station using the method of maximum likelihood estimation (MLE). Following Coles (2001), the GEV distribution for warm annual extremes is described by
i1520-0442-21-21-5455-e1
defined for {z:1 + ξ(zμ)/σ > 0}, −∞ < μ < ∞, σ > 0, and −∞ < ξ < ∞, where μ, σ, and ξ are the adjustable parameters of the distribution that determine the location, scale, and shape of the distribution, respectively. The GEV distribution function for cold annual extremes is
i1520-0442-21-21-5455-e2
defined for {z:1 − ξ(zμ̃)/σ > 0}, −∞ < μ̃ < ∞, σ > 0, and −∞ < ξ < ∞, where again μ̃, σ, and ξ are the location, scale, and shape parameters of the distribution. The P-yr return level is defined as the value that is exceeded by an annual extreme at least once every P years, and is obtained by inverting the fitted GEV distribution. In this study, 2-, 10-, and 100-yr return levels are calculated at each station both for warm and cold annual extremes.

Different goodness-of-fit (GOF) tests were used to examine whether the GEV distributions fit the empirical distributions of annual extremes. As suggested by Coles (2001), we firstly analyzed the probability plot (PP plot) and the quantile plot (QQ plot), which allow visual comparison between the theoretical and the empirical distributions for the annual extremes. Since linearity represents a perfect fit, the level of agreement between both distributions can also be measured with the linear correlation coefficient.

Furthermore, we applied a standard Kolmogorov–Smirnov (KS) GOF test that measures the overall difference between two cumulative distribution functions. The KS statistic D is defined as the maximum absolute difference between two distribution functions:
i1520-0442-21-21-5455-eq1
where F(x) is the fitted distribution function (GEV) and SN(x) is the empirical distribution function estimated from a sample of N observations as the proportion of data values less than or equal to x. The null hypothesis that the annual extremes are drawn from a GEV distribution is rejected when the value of D is greater than a critical value. This critical value is determined by the parametric bootstrapping technique (Kharin and Zwiers 2000). In this procedure, 1000 samples of annual extremes are generated from the original sample by random resampling with replacement. These “new” samples are the same size as the original and are used to reestimate the GEV distribution parameters and the corresponding D value. The limit of the 90% confidence interval of the sample of D is then used as the critical value for the rejection of the null hypothesis at the 10% significance level.

3. Spatial distribution of annual extremes in the period 1956–2003

In this section, results based on the period in which most of the stations have available information, 1956–2003, are shown. First, the presence of linear trends in the four annual temperature extremes defined in the previous section is analyzed. Then, the spatial distributions of return levels and return periods are shown.

Since we needed detrended series to fit the stationary GEV distribution, we used the method of least squares in order to detect the presence of linear trends in the series of annual extremes. Figure 2 shows the sign and magnitude of the linear trends found in HTx, HTn, LTx, and LTn for 43 stations in the period 1956–2003. Significant values at the 5% level are marked with filled symbols. It can be seen that HTx has negative significant trends in the central-eastern part of the country. This is consistent with Rusticucci and Barrucand (2004), who found negative trends in the maximum mean summer temperature and in the frequency of occurrence of warm days (based on the 95th percentile) in Argentina in the period 1959–98. Collins et al. (2000) found negative significant trends in hot days (frequency of daily Tx greater or equal than 35°C) in southeast and southwest Australia. The number of hot days has been less than normal in eastern China since the 1970s (Zhai et al. 1999). And a small downward trend was observed in the frequency of occurrence of warm days in the United States in the period 1910–98 (Easterling et al. 2000). The central-western region shows positive significant trends in HTx. HTn trends are negative in the central part of the country and positive at the rest of the stations. However, the significance level is only reached at a few stations and the pattern is not as clearer as for HTx. It is noticeable that extremes have less significant trends than mean values as it can be seen in Rusticucci and Barrucand (2004). Alexander et al. (2006) found positive trends in the warm nights in southern South America, based on gridded data, with the regions of significant change restricted to northern Argentina and southern Brazil. The LTx trend is positive at the majority of stations, but only significant in the central-eastern region. LTn has positive trends almost all over the country, only significant in the western and the central-eastern regions. In Buenos Aires, the largest city in the country, maximum temperature has a negative nonsignificant trend of 0.179°C/10 yr in its highest value of the year and a positive significant trend of 0.301°C/10 yr in the lowest. The highest and lowest minimum temperatures increase significantly at a rate of 0.342° and 0.542°C every 10 yr, respectively. In the central-western region, trends are positive both for warm and cold extremes. However, in the central part of the country, HTx decreases while LTx increases leading to a drop in the maximum temperature annual range.

After filtering the corresponding linear trends, we proceeded to fit the GEV distribution to the four annual temperature extremes at each station. Figures 3 and 4 show the QQ plots and the PP plots for the GEV distribution fitted to the series of annual extremes at OCBA. From the QQ plots, HTx seems to be the extreme that is worst represented by the GEV distribution since empirical values are lower than theoretical values when HTx is greater than 40°C. From the PP plots, LTn is the one that shows less agreement with the theoretical distribution, since dispersion is greater. However, linear correlation coefficients are greater than 0.98 for all cases.

We performed this goodness-of-fit analysis for every station and we also applied the Kolmogorov–Smirnov test in order to measure the level of agreement between the empirical and the GEV distributions. Since both the graphical and the nonparametric tests were consistent and brought a 10% level of confidence in that annual temperature extremes are drawn from a GEV distribution at all the stations studied, we proceeded to analyze the spatial distribution of return levels in Argentina in the period 1956–2003. The gridding method used for interpolation of return values was the kriging method (Wackernagel 2003).

Figure 5 shows 2-, 10-, and 100-yr return level spatial distributions for warm annual extremes. Both HTx and HTn have similar patterns to those of mean daily values (not shown). According to the GEV distribution fitted to the highest maximum temperature in the year (HTx), maximum temperature is expected to be greater than 32°C at least once every 100 yr throughout all the country, reaching even 46°C in the central region, where mean maximum temperature reaches the highest values of the country. The 2-yr return level distribution shows that maximum temperature will be greater than 30°C in almost any part of the country with probability 0.5. The theoretical distribution for HTn shows that minimum temperature is expected to exceed a value of 14°C in any one single year with probability 0.5, and a value of 16°C with probability 0.01, reaching values even higher than 30°C in the central and northern part of the country. As it can be seen, the region of greatest return levels of HTn is located further north than HTx.

In Fig. 6, 2-, 10-, and 100-yr return level spatial distributions for cold annual extremes show larger gradients across the country, especially in the southern part of Argentina. Maximum temperature is predicted to be lower than 12°C in any part of the country once every 2 yr and lower than 8°C every 100 yr. Minimum temperature, on the other hand, will be lower than 0°C every 2 yr in any part of the country, with values even less than −10°C in the southwestern part of the country. In the Patagonia region, minimum temperatures are expected to be less than −18°C at least once every 100 yr.

There are other applications—hydrology, oceanography, wind engineering, insurance industry, and risk assessment on financial markets, among others—where a threshold value might be important. Therefore, we performed the inverse analysis: fixing the return value of an annual temperature extreme, we studied the spatial distribution of its return periods in Argentina. Consistent with Fig. 5, a 32°C return level in HTx (not shown) results in return periods of 1 yr almost all over the country, except in Patagonia where return periods between 2 and 10 yr are expected. Increasing the HTx return value to 40°C (Fig. 7), the central region of the country shows return periods between 1 and 5 yr and the eastern region, between 5 and 50 yr. There are three stations (Salta Aero, Malargüe Aero, Río Gallegos Aero; gray crosses in Fig. 7) that increase their return periods to values much greater than 100 yr, showing that this temperature is not likely to occur. For HTn, a return value of 20°C (not shown) gives return periods of less than 2 yr throughout the country, while a return value of 25°C is associated with return periods of less than 5 yr in the northern region and between 5 and 50 in the central region and northern Patagonia. There are five stations (Jujuy Aero, Dolores Aero, Río Gallegos Aero, Colonel Suárez Aero, Malargüe Aero; gray crosses in Fig. 7) with return periods greater than 100 yr.

Figure 7 also shows the spatial distribution of return periods for fixed cold extreme return levels. It can be seen that a return value of 6°C in LTx gives return periods between 1 and 5 yr in the Patagonia and the central region, between 5 and 50 yr in the central region and over 100 yr on the coast of Buenos Aires and the northeastern region. A return level of 10°C (not shown) gives return periods of 1 yr at all the stations studied, except in the northeastern region, meaning that maximum temperatures of less than 10°C are likely to happen at least once a year all over the country, except in the northeast. Minimum temperatures less than 0°C (not shown) have return periods between 2 and 10 yr in the northeast and 1 yr in the rest of the country. However, the analysis of the −5°C return level for LTn gives return periods greater than 100 yr in the northeast, and between 1 and 5 yr in almost all the rest of the country.

4. 1976–77 climate shift

With the purpose of analyzing the influence of the 1976–77 shift in the frequency of occurrence of annual extremes, we study the presence of changes in the annual temperature extreme return values drawn from a GEV distribution fitted to two consecutive base periods: before 1976 and after 1977.

Time series (not shown) for the annual temperature extremes defined above for the six long-term stations used in this study: TUC, OCBA, PGM, PIL, SRS, and RGA show that even though these extremes refer to only one day of the year, they do not represent isolated extremes. For five out of six stations, year 1995 shows a relative maximum in HTx, while 1967 displays a relative minimum in LTn. Minimum temperature has less interannual variability than maximum temperature. Also, 11-yr running means denote the important interdecadal variability that these variables present. At Río Gallegos cold extremes show greater variability than warm extremes. It is also evident that HTx has decreased during the period under study at most of the stations, while LTn has increased, especially at OCBA and PIL.

Prior to the GEV fit, a Student’s t test has been carried out in order to evaluate the presence of significant changes between the two periods in the series of annual extremes. All the stations analyzed in this section show significant changes in at least one of the annual extremes, with LTx the variable that showed less significant changes. Since the GEV distribution needs stationary series to be fitted, linear trends have been removed after separating the annual extreme series into two subperiods (pre- and post-1976). This removal enlarges the existing shift, and sometimes it even changes its significance, leading to a bigger shift between the two consecutive subperiods of the detrended series of annual extremes. As can be seen in Table 1, HTx shows negative trends both in the complete and the pre-1976 periods at all stations, except at RGA where positive trends are found, while the last subperiod presents different signs for different stations. Warm and cold extremes in minimum temperature (HTn and LTn) tend to increase in all of the subperiods analyzed, except at SRS where trends are negative, but only significant for HTn in the pre-1976 period. LTx is the extreme variable with the lowest values of trends.

Figures 8 –11 show return levels drawn from the GEV distribution as a function of return periods for each variable, station and subperiod studied. As can be seen, LTx is the variable in which the 1976–77 shift is less noticeable (Fig. 10). In general, the first subperiod (pre-1976) shows return values almost equal to those obtained for the complete period (not shown) of each station. This implies that the behavior of return values based on the complete period is dominated by the first subperiod.

At all the stations, HTx return values diminish from the first to the second subperiod, that is, return periods for fixed return values increase (Fig. 8). Therefore, there is a decrease in the probability of occurrence of maximum temperature warm extremes if we base our study on more recent records. At Río Gallegos, this statement is only true for return periods greater than 50 yr. At OCBA and Río Gallegos, HTn return values increase basing our calculations on the second subperiod; that is, the frequency of occurrence of minimum temperature warm extremes increases (Fig. 9). However, return values at Tucumán and Pilar have the opposite behavior since minimum temperature warm extremes become less frequent. At Santa Rosa, the shape parameter of the GEV distribution changes its sign from negative in the first subperiod to positive in the following one. This change of sign implies a crossing of the two return value distributions.

Although LTx is the variable that shows least change, return values are slightly higher for the GEV distribution fitted to the last subperiod (Fig. 10). This means that, though not significantly, maximum temperature cold extremes tend to occur less frequently during the last part of the century. It should be noticed here the case of Río Gallegos where, regardless of the base period, return values abruptly decrease with the increase of return periods. LTn return values are higher for the GEV distribution fitted to the last subperiod at all the stations (Fig. 11). At OCBA, the 1976–77 shift implies a significant change in return values since the confidence intervals of the two distributions do not overlap, while at Santa Rosa and Río Gallegos this change is not significant because of the complete overlapping of the confidence intervals. At these two stations, return values are greater for the second subperiod only for small return periods because of the change of sign in the shape parameter of the GEV distribution.

The most noticeable change in return values due to the 1976–77 shift is seen in HTn at Río Gallegos where, for example, the 10-yr return value increases from 13.7°C in the pre-1976 period to 18.6°C in the post-1976 period. At Tucumán, return values for HTx decrease almost 4°C from one period to the other, while for LTn they increase between 2° and 5°C, leading to a decrease in the annual temperature range. At OCBA, changes in maximum temperature return values are not significant, but the differences in minimum temperature return values show a reduction in the frequency of occurrence of cold extremes and an increase in the occurrence of warm extremes.

5. Conclusions

Return levels of extreme minimum and maximum temperatures have been estimated by fitting a GEV distribution to blocks of annual maxima. Four different extremes based on minimum and maximum daily data have been defined: the highest maximum (minimum) temperature of the year called HTx (HTn), and the lowest maximum (minimum) temperature of the year, called LTx (LTn). As shown by the different goodness-of-fit tests performed in this study, the GEV distribution fits the four annual extremes studied with a significance level of 10%. This analysis of extreme events allows the estimation of return values for periods longer than the available records for its application in different fields, like hydrology, oceanography, wind engineering, insurance industry, and risk assessment on financial markets, among others.

Results based on the period in which most of the stations have available and adequate information, 1956–2003, show that the highest annual maximum temperature (HTx) decreases while the lowest annual maximum temperature (LTx) increases in central Argentina leading to a drop in the maximum temperature annual range. In the central-western region trends are positive both for warm and cold extremes. It was also found that maximum temperatures greater than 32°C are expected at least once a year all over the country, except in the Patagonia, where this value has a return period of 10 yr. More extreme events indicate that maximum temperatures greater than 40°C may occuronce a year or every five years in the central and northern regions, where maximum temperatures reach the highest values of the country. Minimum temperatures below 0°C happen every 2–10 yr in the northeast and once a year in the rest of country.

Another aim of this study was to establish how the 1976–77 shift influenced the frequency of occurrence of annual temperature extremes. By fitting the GEV distribution to annual extremes in the period pre- and post-1976, we found that GEV distributions based on the complete period are generally very similar to the ones based on the period pre-1976. Nevertheless, the 1976–77 shift led to significant changes in return values. For example, assuming that the period post-1976 can be considered as the “present climate,” at Río Gallegos the highest minimum temperature of the year expected once every 10 yr increased from 13.7°C in the past to 18.6°C in the present. At OCBA, where the urban heat island is more intense, an increase of 3.3°C was found after 1976 in the lowest minimum temperature of the year that occurs every 10 yr and 4.1°C in the one that happens every 100 yr.

Acknowledgments

This work has been partly funded by the following projects: UBACYT X135, ANPCYT BID 1728/OC-AR PICT 38273, and the European Project of the Sixth Framework Programme CLARIS-A Europe–South America Network for Climate Change Assessment and Impact Studies (GOCE-CT-2003-001454). We also thank Eric Gilleland and Richard W. Katz for the extRemes toolkit used in this analysis (http://www.isse.ucar.edu/extremevalues/evtk.html) and the anonymous reviewers who helped to improve the manuscript.

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.

    • Search Google Scholar
    • Export Citation
  • Barrucand, M., and M. Rusticucci, 2001: Climatología de temperaturas extremas en la Argentina. Variabilidad temporal y regional. Meteorológica, 26 , 85101.

    • Search Google Scholar
    • Export Citation
  • Bonsal, B. R., X. Zhang, L. A. Vincent, and W. D. Hogg, 2001: Characteristics of daily and extreme temperatures over Canada. J. Climate, 14 , 19591976.

    • Search Google Scholar
    • Export Citation
  • Coles, S., 2001: An Introduction to Statistical Modeling of Extreme Values. Springer, 208 pp.

  • Coles, S., L. R. Pericchi, and S. Sisson, 2003: A fully probabilistic approach to extreme rainfall modeling. J. Hydrol., 273 , 3550.

  • Collins, D., P. Della-Marta, N. Plummer, and B. Trewin, 2000: Trends in annual frequencies of extreme temperature events in Australia. Aust. Meteor. Mag., 49 , 277292.

    • Search Google Scholar
    • Export Citation
  • Easterling, D. R., J. L. Evans, P. Ya Groisman, T. R. Karl, K. E. Kunkel, and P. Ambenje, 2000: Observed variability and trends in extreme climate events: A brief review. Bull. Amer. Meteor. Soc., 81 , 417425.

    • Search Google Scholar
    • Export Citation
  • Frich, P., P. Alexander, P. Della-Marta, B. Gleason, M. Haylock, A. Klein Tank, and T. Peterson, 2002: Observed coherent changes in climatic extremes during the second half of the twentieth century. Climate Res., 19 , 193212.

    • Search Google Scholar
    • Export Citation
  • Huang, H-P., R. Seager, and Y. Kushnir, 2005: The 1976/77 transition in precipitation over the Americas and the influence of tropical sea surface temperature. Climate Dyn., 24 , 721740.

    • Search Google Scholar
    • Export Citation
  • Katz, R. W., and B. G. Brown, 1992: Extreme events in a changing climate: Variability is more important than averages. Climatic Change, 21 , 289302.

    • Search Google Scholar
    • Export Citation
  • Kharin, V. V., and F. W. Zwiers, 2000: Changes in the extremes in an ensemble of transient climate simulations with a coupled atmosphere–ocean GCM. J. Climate, 13 , 37603788.

    • Search Google Scholar
    • Export Citation
  • Klein Tank, A. M. G., and G. P. Können, 2003: Trends in indices of daily temperature and precipitation extremes in Europe, 1946–99. J. Climate, 16 , 36653680.

    • Search Google Scholar
    • Export Citation
  • Manton, M. J., and Coauthors, 2001: Trends in extreme daily rainfall and temperature in Southeast Asia and the South Pacific: 1961-1998. Int. J. Climatol., 21 , 269284.

    • Search Google Scholar
    • Export Citation
  • Rusticucci, M., and M. Barrucand, 2004: Changes in temperature extremes over Argentina. J. Climate, 17 , 40994107.

  • Schär, C., and G. Jendritzky, 2004: Hot news from summer 2003. Nature, 432 , 559560.

  • Schär, C., P. L. Vidale, D. Lüthi, C. Frei, C. Häberli, M. A. Liniger, and C. Appenzeller, 2004: The role of increasing temperature variability in European summer heatwaves. Nature, 427 , 332336. doi:10.1038/nature02300.

    • Search Google Scholar
    • Export Citation
  • Seidel, D. J., and Coauthors, 2004: Uncertainty in signals of large-scale climate variations in radiosonde and satellite upper-air temperature datasets. J. Climate, 17 , 22252240.

    • Search Google Scholar
    • Export Citation
  • Trenberth, K. E., 1990: Recent observed interdecadal climate changes in the Northern Hemisphere. Bull. Amer. Meteor. Soc., 71 , 988993.

    • Search Google Scholar
    • Export Citation
  • Trenberth, K. E., and Coauthors, 2007: Observations: Surface and atmospheric climate change. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 235–336.

    • Search Google Scholar
    • Export Citation
  • Vincent, L. A., and Coauthors, 2005: Observed trends in indices of daily temperature extremes in South America 1960–2000. J. Climate, 18 , 50115023.

    • Search Google Scholar
    • Export Citation
  • Wackernagel, H., 2003: Multivariate Geostatistics: An Introduction with Applications. 3rd ed. Springer, 387 pp.

  • Zhai, P., A. Sun, F. Ren, X. Liu, B. Gao, and Q. Zhang, 1999: Changes of climate extremes in China. Climatic Change, 42 , 203218.

  • Zwiers, F. W., and V. V. Kharin, 1998: Changes in the extremes of the climate simulated by CCC GCM2 under CO2 doubling. J. Climate, 11 , 22002222.

    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

Stations used in this study. Stars indicate stations with longer-term data records.

Citation: Journal of Climate 21, 21; 10.1175/2008JCLI2190.1

Fig. 2.
Fig. 2.

Spatial distribution of linear trends (in °C 10 yr−1) for the four annual temperature extremes (HTx, HTn, LTx, LTn) in the period 1956–2003. Significant values at the 5% level in filled symbols.

Citation: Journal of Climate 21, 21; 10.1175/2008JCLI2190.1

Fig. 3.
Fig. 3.

“Quantile plots” for the GEV distribution fitted to the four annual temperature extremes (HTx, HTn, LTx, LTn) at station OCBA (1906–2004).

Citation: Journal of Climate 21, 21; 10.1175/2008JCLI2190.1

Fig. 4.
Fig. 4.

“Probability plots” for the GEV distribution fitted to the four annual temperature extremes (HTx, HTn, LTx, LTn) at station OCBA (1906–2004).

Citation: Journal of Climate 21, 21; 10.1175/2008JCLI2190.1

Fig. 5.
Fig. 5.

Spatial distribution of return values (in °C) of warm extremes: (top) HTx and (bottom) HTn for 2-, 10- and 100-yr return periods, drawn from a GEV distribution fitted to annual extremes in the period 1956–2003.

Citation: Journal of Climate 21, 21; 10.1175/2008JCLI2190.1

Fig. 6.
Fig. 6.

As in Fig. 5, for cold extremes: (top) LTx and (bottom) LTn.

Citation: Journal of Climate 21, 21; 10.1175/2008JCLI2190.1

Fig. 7.
Fig. 7.

Spatial distribution of return periods (in years) of annual temperature extremes for fixed return values (indicated on top of each map), drawn from a GEV distribution fitted to annual extremes in the period 1956–2003.

Citation: Journal of Climate 21, 21; 10.1175/2008JCLI2190.1

Fig. 8.
Fig. 8.

Return values (in °C) vs return periods (in years) for HTx for (top to bottom) TUC, OCBA, PGM, PIL, SRS, RGA. Return values (solid lines) are drawn from a GEV distribution fitted at each station to the first subperiod, beginning–1976 (gray lines), and second subperiod, 1977–end (black lines). 95% confidence intervals indicated in dotted lines.

Citation: Journal of Climate 21, 21; 10.1175/2008JCLI2190.1

Fig. 9.
Fig. 9.

As in Fig. 8, for HTn.

Citation: Journal of Climate 21, 21; 10.1175/2008JCLI2190.1

Fig. 10.
Fig. 10.

As in Fig. 8, for LTx.

Citation: Journal of Climate 21, 21; 10.1175/2008JCLI2190.1

Fig. 11.
Fig. 11.

As in Fig. 8, for LTn.

Citation: Journal of Climate 21, 21; 10.1175/2008JCLI2190.1

Table 1.

Linear trends (in °C 10 yr−1) for the four annual temperature extremes and the different periods studied at six stations in Argentina: TUC, OCBA, PGM, PIL, SRS, and RGA. Significant values at the 5% level in bold.

Table 1.
Save
  • 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.

    • Search Google Scholar
    • Export Citation
  • Barrucand, M., and M. Rusticucci, 2001: Climatología de temperaturas extremas en la Argentina. Variabilidad temporal y regional. Meteorológica, 26 , 85101.

    • Search Google Scholar
    • Export Citation
  • Bonsal, B. R., X. Zhang, L. A. Vincent, and W. D. Hogg, 2001: Characteristics of daily and extreme temperatures over Canada. J. Climate, 14 , 19591976.

    • Search Google Scholar
    • Export Citation
  • Coles, S., 2001: An Introduction to Statistical Modeling of Extreme Values. Springer, 208 pp.

  • Coles, S., L. R. Pericchi, and S. Sisson, 2003: A fully probabilistic approach to extreme rainfall modeling. J. Hydrol., 273 , 3550.

  • Collins, D., P. Della-Marta, N. Plummer, and B. Trewin, 2000: Trends in annual frequencies of extreme temperature events in Australia. Aust. Meteor. Mag., 49 , 277292.

    • Search Google Scholar
    • Export Citation
  • Easterling, D. R., J. L. Evans, P. Ya Groisman, T. R. Karl, K. E. Kunkel, and P. Ambenje, 2000: Observed variability and trends in extreme climate events: A brief review. Bull. Amer. Meteor. Soc., 81 , 417425.

    • Search Google Scholar
    • Export Citation
  • Frich, P., P. Alexander, P. Della-Marta, B. Gleason, M. Haylock, A. Klein Tank, and T. Peterson, 2002: Observed coherent changes in climatic extremes during the second half of the twentieth century. Climate Res., 19 , 193212.

    • Search Google Scholar
    • Export Citation
  • Huang, H-P., R. Seager, and Y. Kushnir, 2005: The 1976/77 transition in precipitation over the Americas and the influence of tropical sea surface temperature. Climate Dyn., 24 , 721740.

    • Search Google Scholar
    • Export Citation
  • Katz, R. W., and B. G. Brown, 1992: Extreme events in a changing climate: Variability is more important than averages. Climatic Change, 21 , 289302.

    • Search Google Scholar
    • Export Citation
  • Kharin, V. V., and F. W. Zwiers, 2000: Changes in the extremes in an ensemble of transient climate simulations with a coupled atmosphere–ocean GCM. J. Climate, 13 , 37603788.

    • Search Google Scholar
    • Export Citation
  • Klein Tank, A. M. G., and G. P. Können, 2003: Trends in indices of daily temperature and precipitation extremes in Europe, 1946–99. J. Climate, 16 , 36653680.

    • Search Google Scholar
    • Export Citation
  • Manton, M. J., and Coauthors, 2001: Trends in extreme daily rainfall and temperature in Southeast Asia and the South Pacific: 1961-1998. Int. J. Climatol., 21 , 269284.

    • Search Google Scholar
    • Export Citation
  • Rusticucci, M., and M. Barrucand, 2004: Changes in temperature extremes over Argentina. J. Climate, 17 , 40994107.

  • Schär, C., and G. Jendritzky, 2004: Hot news from summer 2003. Nature, 432 , 559560.

  • Schär, C., P. L. Vidale, D. Lüthi, C. Frei, C. Häberli, M. A. Liniger, and C. Appenzeller, 2004: The role of increasing temperature variability in European summer heatwaves. Nature, 427 , 332336. doi:10.1038/nature02300.

    • Search Google Scholar
    • Export Citation
  • Seidel, D. J., and Coauthors, 2004: Uncertainty in signals of large-scale climate variations in radiosonde and satellite upper-air temperature datasets. J. Climate, 17 , 22252240.

    • Search Google Scholar
    • Export Citation
  • Trenberth, K. E., 1990: Recent observed interdecadal climate changes in the Northern Hemisphere. Bull. Amer. Meteor. Soc., 71 , 988993.

    • Search Google Scholar
    • Export Citation
  • Trenberth, K. E., and Coauthors, 2007: Observations: Surface and atmospheric climate change. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 235–336.

    • Search Google Scholar
    • Export Citation
  • Vincent, L. A., and Coauthors, 2005: Observed trends in indices of daily temperature extremes in South America 1960–2000. J. Climate, 18 , 50115023.

    • Search Google Scholar
    • Export Citation
  • Wackernagel, H., 2003: Multivariate Geostatistics: An Introduction with Applications. 3rd ed. Springer, 387 pp.

  • Zhai, P., A. Sun, F. Ren, X. Liu, B. Gao, and Q. Zhang, 1999: Changes of climate extremes in China. Climatic Change, 42 , 203218.

  • Zwiers, F. W., and V. V. Kharin, 1998: Changes in the extremes of the climate simulated by CCC GCM2 under CO2 doubling. J. Climate, 11 , 22002222.

    • Search Google Scholar
    • Export Citation
  • Fig. 1.

    Stations used in this study. Stars indicate stations with longer-term data records.

  • Fig. 2.

    Spatial distribution of linear trends (in °C 10 yr−1) for the four annual temperature extremes (HTx, HTn, LTx, LTn) in the period 1956–2003. Significant values at the 5% level in filled symbols.

  • Fig. 3.

    “Quantile plots” for the GEV distribution fitted to the four annual temperature extremes (HTx, HTn, LTx, LTn) at station OCBA (1906–2004).

  • Fig. 4.

    “Probability plots” for the GEV distribution fitted to the four annual temperature extremes (HTx, HTn, LTx, LTn) at station OCBA (1906–2004).

  • Fig. 5.

    Spatial distribution of return values (in °C) of warm extremes: (top) HTx and (bottom) HTn for 2-, 10- and 100-yr return periods, drawn from a GEV distribution fitted to annual extremes in the period 1956–2003.

  • Fig. 6.

    As in Fig. 5, for cold extremes: (top) LTx and (bottom) LTn.

  • Fig. 7.

    Spatial distribution of return periods (in years) of annual temperature extremes for fixed return values (indicated on top of each map), drawn from a GEV distribution fitted to annual extremes in the period 1956–2003.

  • Fig. 8.

    Return values (in °C) vs return periods (in years) for HTx for (top to bottom) TUC, OCBA, PGM, PIL, SRS, RGA. Return values (solid lines) are drawn from a GEV distribution fitted at each station to the first subperiod, beginning–1976 (gray lines), and second subperiod, 1977–end (black lines). 95% confidence intervals indicated in dotted lines.

  • Fig. 9.

    As in Fig. 8, for HTn.

  • Fig. 10.

    As in Fig. 8, for LTx.

  • Fig. 11.

    As in Fig. 8, for LTn.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 548 112 8
PDF Downloads 302 78 6