Spatiotemporal analysis of air temperature indices, aridity conditions, and precipitation in Iran

Global warming has become a major threat to life on the earth, and recognizing its impacts can definitely be useful in controlling and mitigating its adverse effects. In this study, time series variations in air temperature indices (frost days, Tmin, Tmax, Tmean, Tminmin, Tmaxmax, Tsoil-min), De Martonne aridity index (IDM), and total precipitation were investigated using a long-term meteorological data (1960–2019) of 31 synoptic stations throughout Iran. The results indicated that more than 94% of the stations had increasing trend in Tmean, in which about 70% were significant at the 0.05 level. The average increase in Tmin was calculated approximately 1.7 times higher than Tmax and also the increase in Tminmin was about 2.5 times higher than Tmaxmax. Our findings showed that abrupt changes in Tmin and Tmax mostly observed in the 1990s were upward in 87% of all the stations. Increase in annual Tmean at a rate of 0.3 °C per decade and reduction of 5 mm per decade in total annual precipitation led to decrease in the IDM aridity index by 0.35 per decade in Iran. The intensity of air temperature increase was higher in tropical regions than in cold regions. Trend analysis in the partial series before and after a change point showed that the trends in Tmean before the change point were negative, but turned to positive afterwards in some stations mostly located in the northwestern cold and mountainous regions of the country. Our results revealed that the climate in Iran, in general, has become warmer and drier in the past 60 years and continuation of the current global warming trend will exacerbate this problem in the future.


Introduction
Air temperature plays an important role in the interaction of the Earth's climate system. Surface air temperature has risen over the past century due to human activities via excessive emissions of greenhouse gases into the atmosphere (IPCC 2013). Global warming has also led to changes in other climatic variables, which has brought about fundamental concerns toward environment as well as anthropogenic issues. The regional advent of global climate variation does not often conform to the behaviour of global mean temperature (Pokorná et al. 2018). In other words, the effects of climate change have been shown to differ from region to region (IPCC 2014). It is then important to identify the cause of such events as well as searching the existence of some signs or particular periodicities in data of interest.
An increase of 1°C in global temperature leads to approximately 1.5 to 2% increase in the global mean precipitation and about 4% increase in the global surface runoff (Labat et al. 2004;Salzmann 2016). In the same context, each 1°C further increase in temperature may, on the average, cause to decrease global wheat production by 6% (Asseng et al. 2014). Similarly, Sinha et al. (2020) demonstrated that a rise of 2°C in the global mean surface air temperature will confront an additional 15% of the global population under water scarcity condition. Lenton et al. (2017) showed increasing monthly temperature variability and autocorrelation in large parts of northern hemisphere. Estrada et al. (2017) discovered that since 2000, the accumulated total costs caused by the impacts of global and local climate change were about 2.6 times greater than the costs without urban-weather-related factors in every city all over the world.
In the second half of the twentieth century, warming in the Middle East regions was greater than the global warming. These regions are, therefore, propounded as potential hotspots due to climate change (Mostafa et al. 2019;van Oldenborgh et al. 2009). Having an area of 1.648 million km 2 and a population of about 82 million, Iran is one of the largest countries in the Middle East ranked as the second most populous country in the region after Egypt, along with Turkey. The vulnerability to the climate change impacts is more pronounced in developing countries, such as Iran, which are heavily dependent on their natural resources and communities with high social diversities.
Since relating the variability of regional surface variables to the large-scale atmospheric circulations have been a common practice in last several decades, Kahya (2011) reported that North Atlantic Oscillation (NAO), a dominant mode of Atlantic sector climate variability, has the far-field influence on interannual to decadal variations in main climate and hydrological variables in the eastern Mediterranean region. For example, Dezfuli et al. (2010) studied the regional rainfall teleconnections to the Southern Oscillation (SO) and NAO at the seasonal time scale to search regional drought predictors across the southwest Iran. They found that the NAO is negatively correlated with autumn rainfall such that it is least likely for an extreme autumn drought to occur when June-August NAO is negative. Dezfuli et al. (2010) concluded that similar to droughts, the wet winter seasons were not significantly in association with either SO or NAO. Vazifehkhah and Kahya (2018) focused on the influences of winter NAO and Arctic Oscillation (AO) extreme phases on hydrological drought using a standardized streamflow index (SSFI) over Turkey and northern Iran. The outputs of SSFI for the positive extreme phases of NAO and AO appeared to notably different in Turkey and Iran in such a way that multiple drought events were detected in Turkey for all timescales as, in most cases, opposed to those in Iran. The regions in the western and eastern Turkey were exposed to drought conditions in various magnitudes during the positive extreme phases of NAO, whereas fewer droughts were observed in Iran around the Caspian Sea during the negative NAO and AO extreme phases.
In a close geography,  investigated hydrological and agricultural droughts and their characteristics using the Standardized Streamflow Index (SSFI) and Standardized Soil Moisture Index (SSI) across the Konya Closed Basin in Turkey, respectively. They also analysed the teleconnections of various large-scale climate indices (e.g., NAO, AO) over 3-month SSFI and SSI series applying cross wavelet transform and pointed out an increase in drought duration and severity in the basin since 1999. In a proceeding study , they were concerned with non-parametric multivariate standardized drought index (NMSDI) based on precipitation and soil moisture data in conjunction with copula functions in the same basin. The bivariate return period analysis of the NMSDI using the two typical drought characteristics (duration and severity) resulted in a high risk for southeastern and southwestern parts of the basin for the 3-month NMSDI series while north to northwestern parts could be exposed to high risk for the 6-month series.
Precipitation and air temperature are two most influential and crucial variables in climatic and hydrological studies. Changes in precipitation and air temperature sharply alter the hydrological cycle; therefore, trend and abrupt change detection analysis in rainfall and temperature patterns is important in understanding climate dynamics. A rigorous precipitation time series analysis provides us a better understanding of changes in the aridity index. In this study, we aimed to examine air temperature indices and I DM aridity index time series in Iran using non-parametric methods to achieve the following objectives: (i) To analyse air temperature time series variables including average daily minimum air temperature (T min ), average daily maximum air temperature (T max ), mean daily air temperature (T mean ), absolute minimum air temperature (T minmin ), absolute maximum air temperature (T maxmax ), mean daily minimum soil surface temperature (T soil-min ), and finally frost days (FD: defined as the number of days with minimum daily temperature below 0°C) in association with the influences of climate change; (ii) To explore changes in aridity conditions using modified De Martonne aridity index; As far as is known, the determination of changes in climate is a prerequisite for a better understanding of the climate and developing adaptation and mitigation measures at regional and local scales (Gebrechorkos et al. 2019). Upon an overall review on previous studies, it is said that none has been focused on monotonic trends in climate variables during a long-term period like more than 50 years in Iran. The potential of more than one break points in a climate time series, at the first time, has been of primary interest in this study. In addition, the trends before and after the breakpoints has been also considered for accurate pertinent inferences in this study. Our results can be applicable in managements of water resources and agriculture.

Study area and data
Iran is located in the Middle East region between 25°0 0′ to 39°47′ N and 44°02′ to 63°20′ E, and covers almost 1.1% of the world's land area ( Fig. 1; Table 1). Studies concerning climate change require long-term data, which are usually limited in Iran. The absence of recorded data in the country influences climate change studies, and it is also a main limiting factor for impact assessments and policy-making processes. In this research, climate data were collected from Iran Meteorological Organization, IRIMO (https://data.irimo. ir/). Many synoptic stations in Iran are newly established and do not have a long-term record. We first selected a set of the stations having observations since 1960 among more than 300 synoptic stations, and later excluded those with more than 10% missing observations among the remaining stations. Finally, we identified 31 synoptic stations that had a long-term recorded data for the last 60-year period

Data quality test
The quality of each synoptic station's data is assured by IRIMO before being presented to users. Suspicious data are strictly checked out and subjected to WMO procedures before use (Balling et al. 2016;Shirvani 2015;WMO 2003). In this regard, Grubbs test (Grubbs and Beck 1972) was applied to identify outlier data point. In this procedure, when one or more data points in one station were detected as an outlier, the data points at the same year from one or more neighbour stations had to be investigated. If these data points did not match the outlier, the outlier was identified for sure and removed from the time series; otherwise, it is considered as a valid data point. This test was applied to all the data in this study.

Modified De Martonne aridity index
An aridity index is a numerical indicator of the degree of dryness in the climate at a selected location. In this study, where P mean is average of precipitation (mm) T mean is average of air temperature (°C) I DM is modified De Martonne aridity index

Time series monotonic analysis
We analysed time series of interest by (i) Mann-Kendall test for trend analysis, (ii) Sen's slope estimator for calculating magnitude of the trend line, and (iii) Pettitt test for detecting abrupt changes in time series.
Mann-Kendall trend test The Mann-Kendall (M-K) trend test (Kendall 1975;Mann 1945) is a rank-based non-parametric test in determining statistically significant trend in a time series. The correlations between the rank order of the observed values and their chronological order are considered in this test. We herein carried out the M-K trend test in each station at α = It is important to note that positive serial correlation in a time series tends to increase the chance of being significant in the M-K test. In other words, when serial correlation in data is significant, the positive or negative serial correlation will alter Var (S) value to be less or more than real value. Hamed and Rao (1998) proposed the modified M-K trend test to remove the effects of serial correlation. The details of the method can be found elsewhere (e.g., Daneshvar Vousoughi et al. 2013;Dinpashoh et al. 2011;Jhajharia et al. 2014;Partal and Kahya 2006).
Sen's slope estimator Sen's slope estimator (Sen 1968) is a non-parametric linear regression slope that is a better alternative to simple linear regression. In this method, the effects of outlier data points on the results of the trend are negligible which can be considered as the major advantage over other parametric methods. The details of the method can be found elsewhere (e.g., Daneshvar Vousoughi et al. 2013;Dinpashoh et al. 2011;Jhajharia et al. 2014;Partal and Kahya 2006).
Pettitt change point test The Pettitt test (Pettitt 1979) was utilized in identifying potential abrupt changes in a time series for each of the considered indices and calculates their statistical significance (Bickici Arikan and Kahya 2019). This test is non-parametric rank-based and distribution-free so that it is insensitive to outliers. In this study, we performed this test at 95% and 99% confidence intervals (α = 0.05 and α = 0.01) for detecting abrupt change. Aftermath, the average amount of changes can be calculated for post-breakpoint against prebreakpoint. Trends for pre-and post-breakpoint series were obtained using simple linear regression (performed only for series with a significant Pettitt test at the 0.01 level). One of our innovations in this research study was using the Pettitt test for detection of multiple abrupt changes in the time series. For those series that had one significant breakpoint (at a 0.01 level), the time series was divided into two sub-series at the detected point. Then the Pettitt test was applied again for the both sub-series. This is repeated if there was a significant change point in any sub-series.

Spatial analysis
Cokriging geostatistical technique was employed for the interpolation of the analysis indices over Iran using 96 stations to extract high-resolution maps for the period 2009-2018. The details of the method can be found elsewhere (e.g., Rostami et al. 2020;Yang et al. 2021)

Return periods
Return periods can be used to develop plans and policies for reducing risk and loss caused by extreme climate events in both short-and long-term planning. Using EasyFit5.5 software, we performed a comprehensive frequency analysis considering 65 different distributions and determined then the best-fit probability distribution through the Kolmogorov-Smirnov test at each station. Through the cumulative distribution of the selected probability distribution, we computed magnitudes of the variable under consideration with respect to the following return periods: 2, 10, 25, 50, and 100 years. This was repeated for each climate variable in the rest of analysis.

Results and discussion
It is evident in Fig. 2 that the highest differences between T min and T max changes were observed in the eastern and southeastern Iran as well as at the stations located in the Zagros mountain range, whereas the lowest differences were found in the coastal area in the Caspian Sea.

Trend analysis
An overall evaluation of the results given Table 3 and Fig. 3 points out that 50% of the stations had a significant decreasing FD trend (at the 0.05 level). In fact, global warming has led to significant reduction in frost days in half of all the stations over the past 60 years. The lowest trend slope was observed in the northeast, in which FD decreased approximately 6 days per decade. Nevertheless, 10% of stations have also experienced a significant increasing trend in their FD time series, whereby FD has risen up 6 days per decade in Shahrekord. Similar indication was documented earlier by Caloiero (2017) who found positive FD trends in some parts of New Zealand. In general, global warming is expected to cause decrease in FD; however, we discovered opposite conditions so that FD trend has been increasing or close to zero at some stations mostly located in cold and mountainous regions. Along the coastal areas in the Caspian Sea, FD has dramatically reduced; however, conflicting results with other stations were observed  in Gorgan, which is located in the eastern side of the Caspian Sea and has a drier climate than the rest of the coastal area. It is worthwhile to mention that no frost days have been recorded in Bandarabbas in the last 60 years. It is located in one of the most tropical parts of Iran, so that annual T mean of that station was calculated as 27°C being the highest value among all the stations. The results of Shi et al. (2018) showed that frost days decreased with a slope of 2.9 days per decade in China during the period 1961-2015. Our findings similarly indicated that frost days were decreased 2 days per decade on average during the last 60 years in Iran. The analysis of T min (T max ) time series revealed that 71% (68%) of all the stations had a significant positive trend at the 0.05 level. The Shahrekord station, which has the most elevated location with an altitude of more than 2000 m, showed contrasting results to other stations in such a way that T min demonstrated statistically significant negative trend. Some stations have experienced fairly strong increasing trends as the both T min and T max trends were significant at 61% and 48% of the stations at the 0.001 level, respectively. One of our striking findings in this study is that the increase in T min values was sharper than in T max values. Although many researchers have acknowledged this fact, some opposite results have been also reported (e.g., Caloiero (2017) for New Zealand). Wang et al. (2018b) stated that T min contributed greatly to the overall T mean increases in northern China. They have claimed that warm air temperature indices appeared to be increasing since late 1980s and early 1990s. Our results concerning the increase in T max , which was often started during the mentioned period in Iran, confirmed their indications. In addition, Yosef et al. (2019) concluded similar results in Israel.
The results of T mean showed that the stations located in the Zagros mountainous regions are less warmer than other regions so that the Shahrekord station has even experienced significant negative trend. Moreover, Shirvani (2015) came to similar conclusion that a decreasing trend was one of characteristic behaviours in the T mean time series at Shahrekord. His findings also demonstrated that the increasing trend of T mean was sharper in the southwest of the country. Pokorná et al. (2018) presented that the surface air temperature trends are not geographically monotonous in different regions of the world. Anyhow, the picture in this study has been consistent with the common expectation that T mean has risen up 0.3°C per decade on average over the past 60 years in Iran, indicating that the increase in air temperature in Iran was higher than the global average. In a recent study conducted in Iraq (Iran's western neighbour), Salman et al. (2017) suggested that air temperature in the country had been increasing 2 to 7 times faster than global mean surface air temperature rise.
In the case of T minmin analysis, our results showed that 65% of all the stations experienced a significant positive trend while 6% of the stations had a significant negative trend at the 0.05 level. The increase in T minmin in the northeastern part of the country was so sharp that the slope of trend based on the Sen's method occurred as an increase of 1.2°C per decade. A similar pattern appeared in its counterpart variable so that significant positive trends in T maxmax values were observed in 55% of stations.
In the case of the soil related temperature variable, the analysis of T soil-min time series revealed that almost half of all the stations had increasing trends, whereas the remaining half had decreasing trends. Out of all the stations, we detected significant positive trends in 19% stations as well as significant negative trends in another 19% of the stations at the 0.05 level. Our findings also revealed that no station has been noted to have an increasing T soil-min trend in the Zagros Mountains. Most of the stations located in the northwest part of the country, where similar geologic conditions alike the Zagros Mountains exist, have also demonstrated a declining behaviour in T soil-min values, but none of them have been statistically significant. Qian et al. (2011) reported that a significant negative trend was observed for T soil in eastern Canada. Furthermore, the increasing trend of T soil-min was observed in the northeastern stations of Iran as well as in the stations located in the coastal regions of the Caspian Sea (except the Gorgan station). In a recent study focusing on the north   Araghi et al. (2017) showed that trend of soil surface temperature was positive in warmer months while negative in colder months. We also found that mean annual T soil-min and mean annual T min differently behaved and this distinction between the two varies between −1.2 and −3.5°C in various locations in Iran. This feature could be interpreted by precipitation changes and snow cover as well as solar radiation and other meteorological indices such as humidity and wind speed. Zhang et al. (2005) has obtained similar results for Canada.
In a summary presented in Fig. 3 and Table 3, about 75% of all the stations have exhibited negative trends (p > 0.05) or such tendencies in the I DM aridity index and total annual precipitation. More specifically, nearly 20% of all the stations showed a statistically significant negative trend in both precipitation and I DM values (p < 0.05). A negative trend in precipitation was previously reported in some regions of Iran by Dinpashoh et al. (2014) and Zamani et al. (2017). Şarlak and Mahmood Agha (2018) has also reported a sharp declining trend in I DM aridity index in Iraq. In the Zagros mountain ranges and the northwestern regions, the rainfall showed declining trend in all stations and its average value was decreased by an average rate of 15 and 10 mm per decade, respectively. The east of the Caspian Sea (Gorgan) and Zagros mountainous region has become drier more than all other areas. In the east of the Caspian Sea, the increase in air temperature was not statistically significant; therefore, the significant decrease of rainfall was thought to be the main reason for the dramatic decrease in I DM aridity index in the area. In general, the increasing air temperature on one side and declining precipitation on the other side have caused a noticeable decrease in the I DM aridity index in Iran. The results of Tabari

Change point detection
The results illustrated in Figs. 3 and 4 show that an abrupt upward change, which was mostly observed in the 1990s, occurred at the 90% of all the T mean time series. In more detail, most of T min and T max change points detected in the first and second half of the decade of 1990, respectively. Abrupt changes of I DM and precipitation were also downward in 26% of all the stations. It seems that abrupt downward changes in the I DM series in the western and northwestern Iran were greater than those in other regions. It is important to emphasize that the areas in the eastern parts of the country (i.e., Kerman and Birjand) were categorized as arid areas before I DM abrupt downward change; however, after the shift in 1997, they have transformed to hyper-arid areas. Abrupt downward changes in the eastern Caspian Sea (Gorgan) and a station in the Zagros mountain range (Sanandaj) have also caused these areas to change their phase from Mediterranean to semi-arid.
As shown in Fig. 4, 60% of the stations revealed an abrupt downward change in FD mostly during the 1990s. Nevertheless, 20% of the stations experienced an abrupt upward change mainly during the 1980s or 1970s, but none occurred during the 1990s. It is important to note that no significant increasing air temperature trend was observed in all the stations experiencing an abrupt upward change in their FD time series. The highest decrease in FD was observed in Mashhad, located in the northeastern Iran, where the mean value of FD reduced about 30 days after the shift. The pattern of T soil-min changes seemed to be contradictory to the other temperature indices, since it had an abrupt downward change at 50% of all the stations that mostly occurred in the 1980s, whereas it had opposite change at 30% of all the stations.

Time series trends before and after the abrupt change
Four patterns are distinguishable in the trend patterns of our data set. For the sake of brevity, our demonstration charts are limited to six of the selected stations in each type. The first type includes the stations that experienced a decreasing trend pre-and post-breakpoint series (Fig. 5a). In contrast to the first one, the second type corresponds to the stations that had an increasing trend in pre-and post-breakpoint series (Fig. 5b). The third type is dedicated to the stations that, the negative trend turned to positive due to the shift in such a way that the series contains a declining (increasing) trend before (after) the breakpoint (Fig. 5c). The fourth type is opposite to the third; that is to say, the series experiences an increasing trend before breakpoint and decreasing afterwards (Fig. 5d). Our analysis revealed that FD trends belonging to Type 1 were noted in 40% of the stations in the form of decreasing pre-and postbreakpoint. In nearly 30% of the stations, FD trends followed Type 4 in the form of increasing before breakpoint and decreasing afterwards.
Analysis of T min time series revealed that the trends occurred as increasing before and after the breakpoint in the majority of the stations. In some stations mostly located in the cold and mountainous regions, the trend in T min series was negative before the shift and positive afterwards. Unlike T min series, a considerable number of stations followed Type 3 pattern regarding T max . At the majority of all the stations located in the northwest parts of the country where cold and mountainous areas take place, the trend in T mean series was decreasing before the change point and increasing afterwards (Type 3), while the trend was positive over the entire period (Type 2) in the central, southeastern and northeastern regions of Iran. In the same context, the results of Reeves et al. (2007) showed that the trend in the annual T mean time series was decreasing during the 1960-1985 period in Libby, Montana, Fig. 4 The average amount of the temperature indices changes for post-breakpoint compared to prebreakpoint. Labels show abrupt change year. Background coloured maps represent the spatial distribution of each index between 2009 and 2018. Up and down pointing triangles denote upward and downward abrupt changes, respectively (p < 0.05). Also, solid circles demonstrate non-significant abrupt changes (p > 0.05) in the USA, but an abrupt upward change caused the series to have increasing trends. Moreover, Fioravanti et al. (2016) presented similar results for Italy. In this study, we found that the trends in T minmin and T maxmax series after change point were increasing in more than 70% of the stations. A careful look on the T soil-min results showed that the trend was increasing after the breakpoint in the majority of the stations that experienced an abrupt shift at the 0.01 significance level.
About half of the stations had an abrupt downward change and their T soil-min values before the shift were higher than the latter values; therefore, the resulted overall trend in these stations was decreasing. In these stations, the trend in T soil-min was rising after the breakpoint in accordance with other temperature indices due to global warming, but the reason behind why most of these stations experienced an abrupt downward change in the 1980s can have various causes which will be addressed later. In a relevant study, Wang et al. (2018a) reported that increasing trend of T soil-min in China is at least doubled since 1998.
For I DM and precipitation time series, Gorgan (located in the east of the Caspian Sea) is a good example; the trend of precipitation in this station was increasing at a rate of 24 mm per decade until 1982, but an abrupt downward change led to a decreasing trend in the series at a rate of 10 mm per decade after the change point. Since 1982, decreasing precipitation along with temperature growth at a rate of 0.3°C per decade at this station has led to a reduction in I DM at a rate of 0.5 per decade, while I DM was increasing at a rate of 0.2 per decade before the change point. In other words, the climate conditions in Gorgan used to be humid at a low rate, but after an abrupt shift that occurred in 1982, it started to be drier. Therefore, we can speculate that the aridity conditions in this region has been shifting from Mediterranean to semi-arid due to global warming and climate change.

Multiple abrupt changes in time series
Over a long period of time, climatic patterns can possibly shift multiple times (Yu and Ruggieri (2019). To provide a better analysis of the results, five types were defined as below: (i) Time series that experienced two abrupt upward changes is categorized as Type 1 (Fig. 6a). (ii) Type 2 is defined as the time series having two abrupt downward changes (Fig. 6b). (iii) Type 3 presents time series that experienced an initial abrupt change in upward direction and downward afterwards (Fig. 6c). (iv) Type 4 represents the time series in which the first abrupt change was increasing one afterwards, and d an increasing trend before abrupt changes and a decreasing one afterwards downward as the second one was upward (Fig. 6d). Finally, (v) Type 5 is defined as the time series containing three abrupt changes (Fig. 6e). The indices of I DM , precipitation, and T maxmax did not have more than one abrupt shift in their time series in any station. In the case of T min time series, 14 stations were identified to have two abrupt upward changes. Four stations (Ahvaz, Bushehr, Kermanshah, and Yazd) mostly located in the tropical regions experienced three abrupt upward changes in T min time series. The modified Pettitt test analysis for T max revealed that multiple abrupt changes were observed only in five stations, while in all of them, the first abrupt shift was downward and the second one was upward (Type 4). In addition, Reeves et al. (2007) showed that the annual T mean time series in Libby, Montana, in the USA experienced two change points between 1960 and 2000, so that the first abrupt change was downward and the second one was upward. In most stations, two abrupt upward changes were detected in T mean time series, and the second change point occurred in a shorter time interval than the first change point.
As previously discussed, some of the studied indices (especially T max ) had an abrupt downward change and a decreasing trend between 1970 and 1980 in some stations. A detailed review in earlier studies revealed that results of Easterling et al. (1997) are consistent with the outcomes of our research concerning T max in the southeastern Asia. Moreover, Yu and Ruggieri (2019) claimed that global surface temperature experienced an abrupt downward change between 1963 and 1976. They explained that the cause of this event could be World War I and II (1914-1918 and 1939-1945), particularly the Great Depression (1929)(1930)(1931)(1932)(1933) and other recessions (such as 1938, 1945, and 1958) that subsequently induced a reduction to greenhouse gas emissions due to the downfall of economic activities and closure of many factories. Greenhouse gas molecules remain in the atmosphere for decades, so the planet's temperature may not respond until the greenhouse gases begin to dissipate. In another relevant study, Pokorná et al. (2018) demonstrated that global warming has never been ubiquitous so that here are seasons, regions, and time periods with negligible or even negative air temperature trends (frequently referred to as warming holes). Therefore, we speculate that the abrupt downward change in T max or negative trend in T soil-min in some stations might be occurred due to aforementioned causes.

Return period
After a consecutive analysis procedure applied for the calculation of return periods in each of 31 stations, the magnitudes with respect to the five return periods at 9 selected stations are plotted in Fig. 7. It is evident that the highest increases in FD return periods' values were frequently observed at the Time series of selected stations with a two abrupt upward changes, b two abrupt downward changes, c the first abrupt change is upward and the second one is downward, d the first abrupt change is downward and the second one is upward, and e three abrupt changes stations located in the coastal area across the Caspian Sea. Due to limited space, we can only briefly refer here to the highlights of our return period computations of all the stations. The highest magnitudes of T mean with respect to the return periods under consideration (i.e., 10, 25, 50, and 100 years) were observed in the northwest region of Iran. In particular, increase (hereafter defined as a deviation from the average value of the study period) in T mean at Khoy and Zanjan stations appeared to be highest among 31 stations for all the return periods; for example, it valued as 3.1°C (approximately %25) for the 100 years. A group of 5 stations in the northwest region including these two stations exhibited a value of 2.7°C as the mean increase in magnitude for the 100-year return period. A group of 7 stations in the west and northwest region deviated from the rest of stations by having high increase value in T soil-min for all the return periods. As noted in T mean case, highest increases in T soil-min among all stations for the 100year return period were computed with a value of 4.1, 4.5, and 4.9°C at Khoy, Khorramabad, and Zanjan stations, respectively. The mean increase in magnitude of this group is 3.96°C for the 100-year return period. For the case rainfall, Yazd, Bandarabbas, and Bushehr stations, which are located in a hyper-arid area having high decrease amount in I DM , exhibited highest increases in rainfall for all the return periods. This case might refer to a flooding potential in these areas. The lowest magnitudes of T minmin with respect to the 50-and 100-year return periods was computed as -37.9°C and -35.0°C at Shahrekord station, respectively. The highest values of T maxmax with respect to the 100-year return period were found in Abadan and Ahvaz stations (located in the southwest of the country) having a respective magnitude of 53.1°C and 52.9°C.

Conclusions
In general, the temperature has increased with less intensity in the cold regions in comparison to the tropical regions. A coupled effect of increased air temperature and diminished rainfall have resulted in decrease in I DM aridity index by 0.35 per decade on the average across the country. The Zagros mountainous region has exhibited drier conditions than other areas as its I DM decreased at a rate of 0.7 per decade on the average. Similarly, the northwest Iran has become drier due to reduced rainfall and increased air temperature. It is noteworthy to mention that Gorgan has noticeably dried up having significant decrease in rainfall amounts as the rate of rainfall reduction in Gorgan reached 33 mm per decade. An abrupt upward change in T mean series, in which most of these change points occurred during the 1990s was detected in more than 90% of the stations. A modified Pettitt test was utilized to determine multiple change points. Our results showed that four stations located mainly in the tropical regions of the country experienced three abrupt upward changes in the T min time series. The results of T soil-min showed that about half of all the stations had negative trend, that is opposite to the behaviours of the other temperature indices. However, it is evident that the trend in the T soil-min series was positive after the change point in most stations with an abrupt shift. In other words, the In general, we herein speculate that global warming has caused Iran to have warmer and drier climate with more extreme air temperature conditions. As long as the policy of sustainable control of greenhouse gas emissions remains excluded from the agenda of all countries around the world, temperature related increasing trends due to global warming will keep growing at a higher intensity. One of the most effective ways to prevent this crisis is to take advantage of using renewable energy instead of fossil fuels. As an alternative solution, Gorjian et al. (2019) indicated that Iran has a high potential in solar energy as the annual rate of solar energy in Iran is estimated at a range of 4.5-5.5 kWh/m 2 . The outcomes of our study can be useful in revealing the impacts of global warming and climatic change as well as providing a more comprehensive viewpoint for continuing researchers on the subject.