Introduction

In its Fourth Report, the Intergovernmental Panel on Climate Change (IPCC) provides an increase of close to 0.6 ± 0.2 °C in global surface temperature (^{IPCC, 2007}). There is a lot of research on the effects of climate change on a global scale, and some researchers have focused on determining the effect on regional and local scale, usually to study the climatic variables of temperature and daily precipitation, usually with average values. ^{Vázquez (2010}) mentioned that according to the World Meteorological Organization (WMO), one ofthe major consequences ofclimate change is perhaps the increased frequency ofextreme weather events.

An extreme weather event may occur within a few days, unlike the extreme weather events that occur for long periods. Among the definitions of extreme event, is understood as that event registered with the behaviour of a given variable, which has a low probability of occurrence and the value observed in that variable exceeds at a specified threshold. ^{Vincent et al. (2005)} determined that, the extremes of temperature change for South America, when find that the change to colder nights and warmer tropical nights tend to increase. ^{Gallant and Karoly (2010)} mentioned that in Australia, the duration of extremes hot and humid have increased, while the duration of fresh and dry extremes have decreased.

^{Sensoy et al. (2013)} concluded that for Turkey, the extremes such as summer days, tropical nights and warm nights have increased, while nights and cold days have decreased. These variations in weather extremes will have impacts in different sectors including agriculture, as indicated by the IPCC in its Fourth Assessment Report (^{IPCC, 2007}), considering that in temperate zones will increase in the yield of agricultural crops due to increased temperature, but in warmer areas the increase in temperature will affect the yield and will allow an increase in the presence of pests.

According to ^{Klein-Tank et al. (2009)} , in order to have a uniform view of changes in climate and weather extremes, the EPDMICC has defined a set of 27 indices describing particular features, including frequency, amplitude and persistence. Projected changes in these indices are indicative of future climate change on extreme weather parameters. Detailed information on these indexes can be found in ^{Alexander et al. (2006)} ; ^{Klein-Tank et al. (2009)}; ^{Zhang et al. (2011)} . Globally, ^{Alexander et al. (2006)} found that the index "number of warm nights per year" has increased near to 25 days since 1951, while the "number of cold nights per year" has decreased by 20 days for the same period.

Current policies for mitigation and adaptation to the effects of climate change, demand geneindexn information both locally and regionally. Mexico has a wide variety of areas ranging from very warm climate along the coast, dry in the central and north and, temperate in the mountains where cold polar was found in the highest peaks. Mexico also has a varied physiography. This diversity indicates that, the effects of climate change will be different in each region and therefore the mitigation and adaptation will be as well.

One of these regions is the Physiographic Province South Gulf Coastal Plains, which includes the States of Veracruz, Tabasco, Oaxaca, Campeche and Chiapas. The need to determine the behaviour of extreme weather related to temperature at the regional level will allow for valuable information for the development of adaptation measurements in the productive sectors. The aim of this study was to evaluate the tendencies ofthe climatic indices of weather extremes, based on maximum and minimum daily temperatures for the region South Gulf Coastal Plains.

Materials and methods

Study area. The region South Gulf Coastal Plains is one of the fifteen physiographic provinces in Mexico. It is located in the south-eastern part of Mexico (Figure 1), bounded to the west by the Transverse Volcanic Belt, to the south by the Sierra Madre del Sur and to the east the Yucatan Peninsula. Its average width varies between 125 and 150 km. Includes the coastal regions of southern Veracruz, covers virtually the entire State of Tabasco, some areas of northern Oaxaca and Chiapas, Campeche and southwest. Consisting of the sub-villages: Coastal Plain of Veracruz; Sierra de los Tuxtlas; Plains and Marshes of Tabasco (^{Medina et al., 2010}). The elevation in this region ranges from 0 meters above sea level on the border with the sea, 550 meters above sea level on the border with the Mountains; with sub-humid and humid climates, with annual average temperatures between 24 to 27 °C and annual rainfall between 1 500 and 2 250 mm (^{Medina et al., 2010}).

Figure 1 Location of weather stations analysed in the South Gulf Coastal Plains Region. 

Database. Featuring the database of daily precipitation, maximum dailytemperature and minimum daily temperature of the Environmental Information System of INIFAP, of the weather stations of the Monitoring Network of the National Water Commission for the States of Veracruz, Tabasco and Campeche, which was only used information from stations located within the South Gulf Coastal Plains region. Ofthese, 31 stations were selected according to the following criteria: 30 years or more information; 95% or more of data; elimination of year with 3 months or more without data; removing stations with many consecutive years without data. The Figure 1 shows the study area and weather stations analysed.

Analysis of the quality and homogeneity of the series. Quality control data (^{Klein-Tank et al, 2009}) was performed using the "RClimDex" module v.1.0 (^{Zhang and Feng, 2004}), implemented in R v.3.0.1, making the correction or deletion of data when compared with climatic parameters on the same dates from a nearby station. For the analysis of homogeneity and the detection of jumps or breaks in the series of daily minimum temperature and máximum temperature, the F test of maximum penalty was used for changes in mean data not documented in accordance with the recommendations of^{Wang (2008)} through the RHTests module v. 3.0 (^{Wang and Feng, 2010}) implemented in R v. 3.0.1 which includes the process of homogenization of the series with adjustment per quantile (QM), in case they are not homogeneous (^{Wang et al, 2010}). With this series we made a descriptive statistics analysis in order to identify anomalies in the thermal variables of daily maximum temperature, daily minimum temperature and daily thermal oscillation with the climatic normal for a 30 year period according to the WMO.

Determination of indices of extremes. We selected 13 of the 27 indices of climate extremes set by the "Expert Panel on Detection, Monitoring and Indices of Climate Change" (EPDMICC) (^{Vázquez, 2010}; ^{Zhang et al.,2011}), related to temperature and climatic parameters, average maximum temperature and average minimum temperature (Table 1). These values were obtained using the RClimDex module developed by the Climate Research Department, Meteorological Service ofCanada (^{Zhang and Feng, 2004}) and implemented in the R software version 3.0.1 that calculates these indices on a monthly or yearly basis, for which the following were considered as threshold temperatures: 25 °C upper threshold of maximum daily temperature and 20 °C upper threshold of minimum daily temperature, these values were selected to have results of the South Gulf Coastal Plains region which can be compared with results from other regions as well (^{Zhang et al, 2011}).

Table 1 Indices of climate change analysed in this study. 

After obtaining the indices, we proceeded to perform a correlation analysis with the values of the slopes of the extremes in each climatic index of the 31 stations obtained in the process using the RClimDex module, since it was considered that these are values of the same type for all variables so were not standardized at all. Subsequently we conducted a major analysis and the values of the new variables and weighted components was performed in a cluster analysis using the "Cluster Analysis" module "Minitab" software in order to identify groups of stations with similar extremes in the set of indices of climate extremes.

Regional trend. In order to evaluate the tendency of these climate indices in the region studied, the values of each index for all stations were averaged, obtaining regional values annually. These values developed trend charts ofthe indices, showing higher significance, including the average trend line moving for 10-year periods. Also, to identify the stational behaviour of the extreme climate indices we developed graphs of frequency indices of cold nights and hot days frequency for the four seasons in the South Gulf Coastal Plains region.

Results and discussion

For the study area, the weather normals for the 30 year period from 1978 to 2007 the values: 31.7 °C maximum annual temperature, 20.4 °C minimum annual temperature and 11.3 °C annual temperature variation. For the period studied (1961-2007), in the Gulf Coastal Plains region of Mexico, the anomalies based on moving averages of 10 years, 0.33 °C and 0.86 °C of maximum annual temperature and minimum annual temperature were obtained respectively, while the temperature variation showed a negative trend with an anomaly of -0.53 °C indicating that the maximum and minimum temperatures have come as time elapses.

Indices of thermal extremes. The Table 2 shows the extremes found for each of the studied climate indices summarized.

Table 2 Number of stations with significant negative trend, no significant trend, positive significant trend for each climate index. Bold numbers indicate that more than 40% of the stations showed a significant trend. 

It is observed that, the number of stations with significant positive trends is significantly higher than for the negative trend. For climate indices: tropical nights, extreme minimum temperature, frequency ofwarm nights, hot days frequency, duration of warm periods and diurnal temperature range, 45% or more ofthe stations were significant positive; that is, a tendency to increase these climatic parameters. Regarding the frequency ofclimatic indices ofcold nights and cold days, 45% or more of the stations showed significant negative trends, indicating a decrease in the number of days in the year when the nights and days are cold.

These results show an increase in the values of indexes related with increasing daily maximum daily temperatures and low temperatures, unlike ^{Vincent et al. (2005)} , who found that most stations with no significant changes relate to the increase in maximum and minimum temperature indices, although their study covers a considerably larger area than the one analysed in this article. The variation between stations, of two of the thirteen studied indices are shown in Figure 2. Climatic indices, cold nights and warm day's frequency are shown, i.e., periods of extremes heat and in Table 2, showed more trian 40% of stations with significant extremes for these indices.

Figure 2 Observed tendencies on the weather stations analysed in the period from 1961 to 2007, for the indices (a) frequency of cold nights; and (b) frequency of hot days. Black triangles indicate significant downward trend, grey triangles indicate significant upward trend mild white circles indicate no significant change. 

In the calculation of these two indices, four weather stations were incomplete due to the lack of information in some months of the series, so it was decided not to include it into this figure. It is observed that most of the stations have a tendency to decrease the frequency of cold nights (Figure 2a) while the frequency index shows increases in hot days in most of the weather stations (Figure 2b). These two indices show variation between the analysed stations because there are stations with opposite extremes in the two stations with equal indices and the extremes, as well as nearby stations with very different extremes for the same index are also presented. ^{Klein-Tank et al. (2009)} indicated that in order to reduce this variability is convenient to work with regional averages rates, softening the variations found.

The values found for the 13 thermal extremes indices and parameters of minimum daily temperature and maximum daily temperature for the stations studied are summarized in the Table 3. In the State of Campeche only four stations were analysed, which is reflected in the larger standard error in all indexes. For the period studied, we had summer days almost the whole year and tropical nights are presented in larger amounts in the southwestern part of Campeche. The maximum and minimum temperature extremes are found in the southern part of Veracruz. The frequency of cold nights and cold days, while the frequency of hot days and warm nights is similar throughout the region. The duration of warm periods ranging from 30 to 50 days per year, while the duration of cold periods is similar to the southeast of Campeche and Tabasco, but not for southern Veracruz in which last up 73 days, which may relate to the fact that in this area there are weather stations at elevations between 100-345 meters.

Table 3 Values of the temperature extremes and two climatic variables, avarage of the located stations by State in the South Gulf Coastal Plains region for the period 1961-2007. The average value and the standard error.  

The principal component analysis resulted that 86.7% ofthe variability in climate indices in the region, was sized in the first two principal components in which the greatest load is on the indexes: tropical nights, length of cold periods, duration of warm periods, summer days, warm nights frequency, frequency of cold nights, warm days frequency and frequency of cold days.

The Figure 3 presents these indices and the trend line with moving averages per decade for the South Gulf Coastal Plains region. The greatest load brought about by these indices to the variation in the study area indicates the great influence they have on the changing extremes for the study area. At the regional level the number of tropical nights has increased 30 days in 47 years, similar to the increases found for the southern part of Brazil by ^{Vincent et al. (2005)} and lower than the increase reported by ^{Sensoy et al. (2013)} for Turkey. Summer days show a slight increase, although we believe that the study area has warmer weather, this index does not have a strong impact on the observed variation in the South Gulf Coastal Plains region.

Figure 3 Tendencies of climatic indices, on a regional scale, with higher load on the principal component analysis: a) tropical nights; b) summer days; c) duration of cold periods; d) duration of warm periods; e) frequency of cold nights; f) frequency of cold days; g) frequency of warm nights; and h) the frequency of hot days. 

The duration of cold periods show a small increase (12 days), while the duration of warm periods has increased 26 days, value within the range estimated in more temperate regions mentioned by^{Alexander et al. (2006)} . Regarding the frequency indices ofnights and cold days, it has a negligible increase (1% and 3.8% respectively), while the frequency of hot days and warm nights have a higher increase (12% and 10% respectively); 40 days more in the year these conditions are presented in comparison with the beginning of the period studied. ^{Vincent et al. (2005)} found similar values of frequency indexes on warm nights and hot days frequency for Perú and the southern part ofArgentina, while ^{Caesar et al. (2011)} found globally lower values than those obtained in this paper.

These graphics show the tendencies at regional level as a result of averaging the values ofthe indices for each station as mentioned by ^{Klein-Tank et al. (2009)} . The cluster analysis applied to the balanced variable values by the two main components, showed that 31 weather stations are included in four groups with a level of similarity of at least 85%. The grouping found is unresponsive to the spatial location of the weather stations studied, but rather to the general trend obtained by averaging all climatic indices. Trying to understand this classification, we found that it responds to the tendencies of the climatic parameters annual mean daily maximum temperature and annual mean daily minimum temperature, variables that originate these indices of extremes.

The first group covers two weather stations: one in Campeche and another one in Veracruz, show positive trend in both parameters; the second group is formed with three stations: one from Campeche and two from Veracruz, with a negative trend in both parameters; the third group, the largest one, includes 25 weather stations: two in Campeche, seven in Tabasco and 16 from Veracruz, this group shows a negative trend for annual mean daily maximum temperature and positive for annual mean daily minimum temperature; and the last group contains only one station from Veracruz, shows a positive trend for maximum temperature and negative trend for minimum temperature, this station contains several data lost, considered to be the cause to be separated from the others.

Seasonal tendencies. In order to determine the seasonal behaviour oftemperature extremes, we developed graphics of the decadal means mobiles for the indices frequency of cold nights, warm nights frequency and frequency of hot days for the spring (March-April- May), summer (June-July-August), autumn (September-October-November) and winter (December-January-February), as shown in Figure 4. In the three graphics there are relatively stable tendencies of each index in the four seasons of the year, up to the year 1990. After this year the tendencies between the seasons had a larger variation.

Figure 4 Seasonal tendencies of the indices a) frequency of cold nights; b) frequency of warm nights; and c) frequency of hot days. 

During spring, the frequency of cold nights has remained almost unchanged over the period 1961-2007, but not for the frequency ofwarm nights and hot days, showing an increase of 6% and 9%, respectively, which is equivalent to 5 and 8 days with minimum and maximum temperatures above 90 percentile. During summer, the highest frequency of cold nights, warm nights and hot days are manifested, increasing from 10 to 16% for cold nights, 11 to 26% for warm nights and 10-26% for hot days, which is equivalent to having 14 cold nights, 23 warm nights and 23 hot days during the summer, in the last decade analysed.

In autumn it is observed that frequency ofcold nights remains stable throughout the period with a value of10%, while the frequency of warm nights and hot days have a significant increase of 16% (9% in the first decade and 25% in the last decade) for warm nights and, 13% (8% in the first decade, 21% in the last decade) for hot days, nights, this is an increase of almost 14 nights with minimum temperature higher than 90 percentile and 12 days with maximum temperature above the 90 percentile for the three months of autumn. Winter cold evenings showed decreased but increase in the number ofnights and warm days with values of15% and 16% respectively at the end ofthe decade for the period 1961-2007 which corresponds to 13 nights and 14 days significantly hot during this season of the year.

By comparing the tendencies of the frequencies of warm nights with hot days, for the two seasons ofthe year in which the highest increase was observed, summer and autumn, warm nights increased from 11 to 26% in summer and 9-25% in the autumn, while hot days show an increase of 10-26% in summer and 8 to 21% in autumn. These results indicated that the frequency of nights with night-temperatures above the 90^th percentile has increased, so that the number of warm nights equals the number of hot days in summer and are superior in autumn, that is, during this season there are more warm nights than hot days as shown in the Figure 5.

Figure 5 Tendencies of the indices: frequency of hot days (black lines) and frequency of warm nights (grey lines) for summer and fall. 

These results confirm the results in the cluster analysis, in which the group with most stations show tendencies stable to slightly positive for the daily maximum temperature and significantly positive tendencies for the daily minimum temperature, resulting in the decrease of the diurnal temperature range. In the South Gulf Coastal Plains region the slight increase in the frequency of cold nights is not consistent with that reported by ^{Alexander et al. (2006)} for the same area, this may be due that the study just mentioned was made in a global level and in ours was made at regional level. In contrast, the increase in night temperatures and daytime temperaturas occur in summer (Jun-Aug) and (Oct-Jan), agreeing White those reported by ^{Vincent et al, (2005)} in South America for the summer (December to February) and autumn ( Mar-May) as well.

Conclusions

In the South Gulf Coastal Plains region, the indices of extremes related to daily minimum and maximum temperature show significant changes, for decreasing as well as increasing for the study area in the period 1961-2007, being the most notorious: summer days, tropical nights, warm periods duration, frequency of warm nights and frequency of hot days.

The management of average index values of the extremes softens the variation found in the study area and allows a broader view of regional changes.

For the GulfCoastal Plains region of Mexico, the increase in warm nights and hot days occur more during the summer and autumn, seasons that need preventive measurements due to the effects that these increases might have on living beings.

The frequency of warm nights has increased in such a way that is similar to the frequency of hot days in summer and higher in autumn. This implies the largest presence of the highest night temperatures.