The thermal condition of cities may be increased over time by global warming and by the population and urban growth known as the Urban Heat Island (UHI) effect. Studying the causes of these variations in temperature help to understand the behavior of climate in general, and is also useful for evaluating the increasing lack of comfort of the inhabitants, to determine the increased power consumption needed to mitigate the deleterious ef- fects of thermal change.

The temperatures in a town are often higher than that in its rural surroundings. London was the first urban heat island to be documented [1] , but since then many cities have been studied [2] -[4] . The UHI effect can be at- tributed to many physical differences between urban and rural areas, including absorption of sunlight, increased heat storage by manmade surfaces, obstruction of re-radiation by buildings, absence of plant transpiration, dif- ferences in air circulation, and other phenomena [5] [6] .

Weather stations located in cities are an important part of the meteorological network, and their data are used to assess climate change, so that the city’s influence on global temperature trends is constantly studied to per- form adequate data assimilation [7] [8] . While the first studies determined a nonlinear relationship between pop- ulation growth and the UHI [9] , findings over the past two decades have underlined that UHIs are more diverse than originally suspected. For this reason, the most reliable way to investigate possible urbanization bias is to analyze the climate behavior at each particular urban weather station [3] .

Urban meteorological information used for this study was obtained from Córdoba Observatory (BO) weather station, which has been making observations of weather conditions since its installation in the late 19th century. The BO weather station is located near the center of Córdoba city, Argentina, which currently has more than 1 million inhabitants, in an area of the southern hemisphere where there are few weather stations. Population density in the region is also very low (see http://www.fao.org/ag/AGP/AGPC/doc/Counprof/Argentina/argentina.htm). Studies about UHI for Argentina have been performed by [10] -[13] , with particular attention to Buenos Aires city.

In the context of climate change postulated by IPCC [14] [15] , the distribution of seasonal mean temperature anomalies has shifted toward higher temperatures and the range of anomalies has increased [16] . An important change is the emergence of a category of extremely hot summertime outliers, more than three standard devia- tions (3σ) warmer than the climatology of 1951-1980 base period. This hot extreme, covering much less than 1% of the Earth’s surface during the base period, now typically covers about 10% of the land area [17] .

In the past, most analysis of surface air temperature observations focused on mean values. In recent years, the analysis of extreme temperatures has highlighted changes in high frequency extreme temperature events [16] [18] , in the number of days exceeding various temperature thresholds and percentile extremes [19] -[21] , in re- gional trends of daily maximum (T_max) and minimum (T_min) temperatures [22] [23] , and in the diurnal tempera- ture range (DTR) [24] . T_max and T_min trends have also been studied and, in general, minimum temperatures were seen to have increased at a higher rate than maximum temperatures [16] . Also, [23] studied the statistical distri- bution of minimum and maximum temperatures from long series of more than a century of 24 European stations, and verify that there are several distinct stable climatic regions in Europe, and each of them have a different thermal behavior. While Eastern European stations present an increase of about 2˚C of the maximum tempera- tures in the last 30 years with respect to the preceding time spans, in Central Europe there is an increase of about 1˚C of the maximum temperatures at some stations, but no discernible increase at the others.

As [25] point out, one important aspect which may influence trend estimates in different periods is the exis- tence of very low frequency variability in the climate, such as the Pacific Decadal Oscillation (PDO), which has a period of 50 - 60 years [26] . This effect was also noted in a previous work [27] which, using the Tome-Mi- randa method [28] , found that the variation over time of annual rainfall in Córdoba Province, Argentina, does not follow a linear trend but has notable breakpoints, with an important influence of the southern Pacific Decad- al Oscillation (PDO).

The objective of this work was to analyze the variation of air temperature at Córdoba Observatory (BO) wea- ther station, which has been making observations of weather conditions always at the same place since its in- stallation in the late 19th century. The study was performed for the 1960-2010 period considering meteorologi- cal information from rural and suburban stations near Córdoba city as a reference of regional thermal conditions. Also it was studied the influence of these changes on diurnal temperature range which have been proposed as an indicator of climate change.

The temperature data analyzed were obtained at 3 weather stations operated by the “Servicio Meteorológico Na- cional” (SMN) of Argentina in Córdoba Province. These are Córdoba Observatory (BO), Pilar Observatory (PI), Pajas Blancas Airport (PB), which were always located as shown in Table 1 and Figure 1. The observational regime is similar in the three weather stations, as they integrate the official weather stations network, by using maximum and minimum conventional thermometers, placed into the standard shelter over a field covered with grass (garden stations).

The population of Cordoba city has grown steadily in the last century, and currently has 1,300,000 inhabitants. Pajas Blancas Airport is located on the outskirts of the city, 11.5 km from BO (see Figure 1), so that could be considered as a suburban station. Pilar Observatory (PI) is located near the town of the same name, which has a population of only 13,000, and is 41.4 km from BO in a predominantly rural area, so that urban influence has minimal impact (rural station). Although there are small differences in altitude between stations, the relief be- tween Córdoba and Pilar cities is flat, so there are no weather factors that modify the expression of the data ex- cept the climate dynamics.

In general, the climatic conditions of the three stations are very similar. In fact, in the 2001-2010 period, in all of them was observed that 40% of the winds are from the Are NE, while S winds represent 20% annual frequency.

The average wind speed is between 8 and 17 km/h, reaching maximum values between August and October months. In addition, in PI and BO stations calm periods account for about 30% of cases. In relation to rain, the three stations show precipitation values of about 800 mm in 80 days per year, mainly concentrated in the six months from October to March. Finally, cloudiness in all these stations is between 3 and 4 eighths per year.

The thermal regime variation was analyzed from records of daily maximum (T_max) and minimum temperatures (T_min) for 1970-2009 periods, during which observations are available for the three weather stations. Tests per- formed reveal the homogeneity condition of these series. The criterion of using daily extreme values was adopted to avoid the influence of the calculation procedure on the mean temperature, as well as because these have been identified as particularly sensitive indicators of climate change [29] . Maximum and minimum tem- peratures, while presenting their own uncertainties are not subject to any calculation procedure [30] .

With the daily data of extreme temperatures, T_max and T_min, the diurnal temperature range (DTR) was calcu- lated as:

DTR = T_max – T_min

and the annual amplitude for each year was calculated as the mean value of DTR.

To assess the partial trends of the study variables and see if there are changes in the period analyzed the Tomé-Miranda (TM) method was employed [28] . TM method, explained in detail in [27] , obtains the square difference between the observed values of the time series and the values evaluated from a set of partial trends. These partial trends are obtained subject to the condition that the interval between breakpoints must exceed a certain value, called Minix, and also imposing restrictions on the difference between two successive trends. Then the algorithm determines the best combination of continuous segments, i.e. those that minimize the sum-of- square error, SS. In the present work, the Minix values adopted were 15 and 20 years, as shorter periods of time could hardly be regarded as climate change. Also, the algorithm was used considering that two consecutive trends must be of different signs in order to analyze the time series by cyclical behavior.

To evaluate the level of significance of the results obtained by the algorithm, the Mann-Kendall test [31] was used. The Mann-Kendall statistic S is calculated using the formula:

where x_j and x_k are the annual temperature values in years j and k, j > k, respectively, and

The S values and variance of S (VAR(S)) are used to compute the test statistic Z as follows:

Since Z has a normal distribution, the presence of a statistically significant trend is evaluated using the Z value. To test for either an upward or downward monotone trend, a two-tailed test at α level of significance (α = 0.001 (^***), 0.01 (^**), 0.05 (^*) and 0.1 (+)) was used.

This work studied the thermal condition along 1960-2010 period of Córdoba city, Argentina, measured at the Córdoba Observatory (BO), located at the same place since 1873, in relation to the meteorological data of Pajas Blancas Airport (PB) and Pilar Observatory (PI) as suburban and rural reference stations, respectively.

^**Significant at 99% confidence level; ^*Significant at 95% confidence level.

The results about the urban heat island effect showed that:

1) The average minimum and maximum temperatures of the rural stations are approximately the same (the dif- ference is less than 0.5˚C), while the differences between urban and rural temperatures are about 1˚C.

2) Frequency analysis shows that all categories of daily minimum temperatures below 15˚C present more fre- quency in rural areas, showing that nighttime cooling are consistently lower in the rural context than in the town. The frequency of maximum temperatures in the urban area for the categories below 30˚C is lower than those of the rural areas. The thermal ranges of highest frequency are higher in BO, indicating that thermal conditions in the city during the day are warmer.

3) In comparative terms, the relationship between the mean annual minimum temperatures in PB and PI (sub- urban and rural stations) and BO (urban station), as well as the relationship between PI and PB, show no sig- nificant trends throughout the period analyzed, i.e. there is no objective evidence that “urban contamination” may have changed the thermal regime during the period analyzed, even though the city’s population has dou- bled in the same period.

With regard to climate change observed in the analyzed data it was noted that:

1) The increase in extreme high temperatures (greater than 3σ) reported by [8] is not present here.

2) Annual mean values of T_min and DTR throughout the period analyzed show trends to increase and decrease, respectively. However, T_min, T_max and DTR all exhibit a breakpoint around 1990. This breakpoint coincides with a change in the Pacific Decadal Oscillation (PDO), which had a warm phase during the 1977-1990 pe- riod and a cold phase from there on.

3) The change of thermal amplitude is inversely correlated with variation in precipitation that has occurred in the region. The increase of water vapor in the atmosphere may be responsible for the increase seen in mini- mum temperature.

We wish to thank A. R. Tome and P. M. A. Miranda of the University of Lisbon, Portugal, who freely provided the source code for obtaining the breakpoint parameters. In addition, we are grateful to Mr. José Barcelona for technical support, to the Servicio Meteorológico Nacional for providing the data and to SeCyT-UNC and CONI- CET for financial support. Also we would like to express our appreciation to Prof. B. J. Gonzalez Kriegel for collecting the hourly temperature data.