1 Introduction

The global mean surface temperature was 0.99 °C higher in the first two decades of the twenty-first century (2001–2020) than in 1850–1900 (IPCC 2021), which has already caused multiple observed changes in the climate system. Especially, the magnitude and intensity of climate extremes have increased around the world in recent decades, affecting the natural and human systems (IPCC 2012; IPCC 2021). Therefore, the investigations about temperature extremes have received more and more attention worldwide due to their impacts (Byers et al. 2018). It has also been projected that more temperature extremes will increase mortality and morbidity in vulnerable groups (Dong et al. 2018), and will have an important influence on global agriculture (Vogel et al. 2019; Yan et al. 2021), vegetation phenology, and productivity (Crabbe et al. 2016). Hence, understanding the variations in temperature extremes is crucial to ascertain the magnitude and pattern of the risks posed by global warming.

It should be noted that significant changes in extreme temperatures have been observed on global and some regional scales within different datasets (Alexander et al. 2006; Dong et al. 2018; Zhou et al. 2016). Since the beginning of the twentieth century, the widespread significant variations in temperature extremes in the global are consistent with the warming trend of average temperatures (Alexander et al. 2006; Donat et al. 2013). These changes in the global are more pronounced for indices related to cold extremes than for indices related to warm extremes (Donat et al. 2013). In China, decreases in cold extremes and increases in warm extremes have also been found during 1961–2010 (Zhou et al. 2016).

It is important to understand the causes of the long-term trends of the observed temperature extremes and the possible influence of external forcing on the climate system (Dong et al. 2018), especially the El Niño-Southern Oscillation (ENSO), which is a climate signal from the oceans and can trigger pronounced changes in climate across the world (Sun et al. 2016). The ENSO plays a robust role in the climate of East Asia, which has been mainly ascribed to the interactions between the ENSO and the East Asian summer and winter monsoons (Ying et al. 2015; Miao et al. 2019). Another study investigated the relationship between ENSO and mean temperature peaks a few months before the monsoon (del Rio et al. 2013). Under the background of a warm ENSO event and the positive phase of the Tropical Indian Ocean dipole, the Eurasian teleconnection pattern dominated the midlatitude region across the Eurasia continent, providing the moisture conditions for the severe snowfall event in early winter 2018 on the Tibetan Plateau (TP) (Shen et al. 2021). On the other hand, the day-time and night-time temperatures are increasing over western Pacific and East Asia regions during the El Niño years (Nicholls et al. 2005). The variability of high-temperature extremes follows the ENSO cycle, which may be due to the intensification and persistence of ENSO activity, the warming of the tropical Indian Ocean, and changes in atmospheric teleconnection (Wang et al. 2014).

The above studies mainly focused on the impact of the average intensity of ENSO events, but the different types of ENSO were poorly considered. ENSO events can be divided into two types, Eastern Pacific (EP) ENSO events and Central Pacific (CP) ENSO events, which have different influences on the atmospheric circulation in East Asia (Larkin and Harrison 2005; Weng et al. 2009). It is noticed that CP ENSO events have been frequently observed in recent years (Wang et al. 2019), which may interfere with the robustness of climate predictions in East Asia. One of the studies found that two types of El Niño are associated with increased summer high-temperature extremes in different regions of China by influencing the location of subtropical high pressure in the western North Pacific and the East Asian jet stream (Gao et al. 2020). Overall, it is believed that the ENSO signals may be the source of changes in the spatial and temporal patterns of climate change (Miao et al. 2019). Thus, gaining a better understanding of the impact of different types of ENSO events on climate is necessary and would enable the identification of the key factors affecting extreme climate events

As the largest and highest plateau on earth, the TP is extremely sensitive to warming compared to surrounding areas (Duan and Xiao 2015), and an increasing number of climate extremes have occurred on the TP in recent decades (You et al. 2016; Zhou et al. 2016). Although some studies have investigated the variation in temperature extremes and its influence factors using the Coupled Model Intercomparison Project Phase 5 (CMIP5) models (Yin et al. 2019; You et al. 2018), these studies have focused on climate change and anthropogenic influences (Yin et al. 2019; Saleem et al. 2021). However, changes in extreme temperatures may be related to large-scale patterns of climate variability, such as ENSO (Saleem et al. 2021). Nevertheless, little attention has been paid to the response of temperature extremes on the TP to different ENSO types. Thus, it is of interest to gain a better understanding of the temporal and spatial variations in temperature extremes and their response characteristics to different ENSO types; then, we will be able to give an insight into the reasons behind the rapid changes in climate extremes and to develop an essential scientific basis for future projections of climate extremes and climate change policymaking.

In this study, we examined the variations in temperature extremes across the TP using the meteorological observational dataset from 1980 to 2020 and extracted different ENSO types based on the national standard of China formulated. Then, we determined the influence of different ENSO types on the variation patterns of temperature extremes over the TP. The goals of this study were to accomplish the following: (1) analyze temporal and spatial patterns of temperature extremes on the TP, and (2) investigate the physical mechanism of ENSO events on the patterns of temperature extremes on the TP.

2 Data sources and methods

2.1 Study area

The TP is located in southwestern China, with a total area of 2.5 million km2, and is the most extensively elevated surface on Earth, with an average elevation of approximately 4500 m (Fig. S1) (Spicer et al. 2021). Due to the complex topography and high elevation of the TP, there is a corresponding gradient in the annual total precipitation from 16 mm in the northwest to 1764 mm in the southeast. In the coldest month, the average temperature is less than −5 °C, and it is less than 10 °C in the warmest month (Huang et al. 2016; Chen et al. 2015). There are approximately 130–140 frost-free days. Overall, the plateau mountain climate of the TP is dominated by the westerly jet, the East Asian and South Asian monsoons, with the spatial pattern exhibiting horizontal band differentiation from warm humid in the southeast to dry-cold in the northwest (Fig. S2) (Immerzeel and Bierkens 2012).

2.2 Data sources

2.2.1 Meteorological datasets

The meteorological station dataset was obtained from the China Meteorological Administration (http://data.cma.cn/), which includes the daily maximum temperature and minimum temperature data from 1980 to 2020. Data quality control is essential before the analysis of climate because erroneous outliers have an effect on temperature trends (Yan et al. 2015). We used the RClimDex package (http://etccdi.pacificclimate.org/) in R software (R Core Development Team, R Foundation for Statistical Computing, Vienna, Austria) to perform data quality control, which is considered to be an effective method for data quality control (Li et al. 2012; Tong et al. 2019). First, we replaced the missing temperature data and the inaccurate temperature data (such as minimum temperatures that exceeded the maximum temperatures) with −99.9. Then, the temperature data were examined for outliers. In this study, outliers are defined as daily values that are more than three times the standard deviation of the observed records and differ significantly from historical climatology extreme values and observation records of near stations at the same time. Finally, we use linear interpolation to replace the outliers. Our analysis is restricted to the meteorological stations that had no more than 25% missing data for any of the variables, and to apply a dataset that adequately captured the behavior of extremes, meteorological stations with a sufficient record length were required (Yong et al. 2021). Hence, 93 stations were maintained to study the variations in temperature extremes across the TP. The locations of the meteorological stations are shown in Fig. S1.

We use the meteorological station data after 1980 because of good quality in this period (Yin et al. 2019), but the meteorological stations have spatial discontinuities. In order to obtain high-resolution gridded climate data, spatial interpolation is usually required. In addition, the distribution of meteorological stations is uneven on the TP, with a higher concentration in the east of the TP and a sparse station density in the western, so we introduced some meteorological stations in the surrounding regions of the TP as a supplement for spatial interpolation (Sun et al. 2022). We utilized the local thin plate smoothing splines function of Anusplin 4.2 software (Australian National University, Centre for Resources and Environmental Research, Canberra), to process the spatial interpolation of temperature (Ye et al. 2020), which can introduce multiple factors as covariates in the processing procedure to obtain more accurate interpolation results (Hutchinson and Xu 2013). Previous studies have found that elevation also has an important effect on temperature distribution (Joly et al. 2018; Li et al. 2013). Hence, we introduced the digital elevation model as a covariate and aggregated the spatial resolution of temperature to 1 km from 1980 to 2020 (Sun et al. 2022).

2.2.2 Large-scale atmospheric circulation, sea surface temperature, and Niño index data

The Extended Reconstructed Sea Surface Temperature (SST) V4 data were used in this study to estimate the monthly SST of the Pacific with a 2°×2° resolution. The data were acquired from the National Oceanic and Atmospheric Administration (NOAA) (https://psl.noaa.gov/data/gridded/) (Huang et al. 2015). The wind fields and geopotential height fields reanalysis datasets were obtained from ERA5 monthly averaged data on pressure levels from 1959 to the present, which is developed by European Centre for Medium-Range Weather Forecasts (ECMWF) (https://cds.climate.copernicus.eu) (Hersbach et al. 2020). The ERA5 reanalysis provides monthly data from January 1980 to December 2020 at a horizontal resolution of 0.25°×0.25° at different pressure levels. In addition, the monthly Niño3 index (regional mean SST anomaly in the eastern tropical Pacific (150°W–90°W, 5°S–5°N), In3), Niño4 index (regional mean SST anomaly in the central tropical Pacific (160°E–150°W, 5°S–5°N), In4), and Niño3.4 index (regional mean SST anomaly in the east-central tropical Pacific (170°W–120°W, 5°S–5°N), In3.4) for the period of December 1979 to January 2021 were obtained from Climate Indices: Monthly Atmospheric and Ocean Time-Series provided by the NOAA Physical Sciences Laboratory (https://psl.noaa.gov/data/climateindices/list/) (Yu et al. 2019).

2.3 Analytical methods

2.3.1 Calculation of extreme temperature indices

Four extreme temperature indices (the percentage of days when the daily minimum temperature (TNn) was < 10th percentile (TN10p); the percentage of days when the daily maximum temperature (TX) was > 90th percentile (TX90p); the minimum value of the daily minimum temperature (TNn); and the maximum value of the daily maximum temperature (TXx)) were applied to investigate the variations in temperature extremes. These selected indices were defined by the ETCCDI for the identification of extreme temperature events, and have been adopted by several studies to explore the dynamics of extreme events (IPCC 2012; Alexander et al. 2006). These four extreme temperature indices were calculated using RClimDex in R software with 1981–2020 as the base period. These indices were the calendar day 10th percentile centered on a 5-day window for the base period. Table S1 provides a detailed description of four extreme temperature indices.

2.3.2 Identification of different ENSO types

Traditionally, ENSO has been quantified with simple indices, such as SST anomaly in the Niño3 region and the Niño4 region, or a mixture of the two, in Niño 3.4 region. The fact that ENSO has different SST patterns makes the Niño3.4 index inadequate to serve as a common index to capture these different types of ENSO. Since the SST patterns of these two types of El Niño events are highly correlated, neither the traditional Niño 3 nor Niño 4 SST indices alone can effectively represent CP El Niño (Ren and Jin 2011) (Fig. S3). Niño3.4 index and Trans-Niño index (TNI, representing the difference between the normalized SST anomalies averaged in the Niño1+2 and Niño4 regions) are nearly uncorrelated at zero lag; these two indices were shown to be capable of describing the evolutions of ENSO and dramatic ENSO regime changes (Trenberth and Stepaniak 2001). In addition, the Niño3.4 index and TNI are nearly orthogonal (Kao and Yu 2009). However, the SST patterns of the warm-pool and cold-tongue El Niño are far from orthogonal and are very similar. Thus, these indices are not optimal to differentiate the warm-pool and cold-tongue types of ENSO.

Hence, we used the national standard of China formulated to identify different ENSO types (Wang et al. 2020, Ren; et al., 2017). First, all the ENSO events during the 1980–2020 period were chosen, and the ENSO events were identified when the 3-month smoothing average of the absolute value of the In3.4 reached and exceeded 0.5 °C for at least 5 months (In3.4 of ≥ 0.5 °C indicates an El Niño event, and a value of ≤ −0.5°C indicated a La Niña event); In3.4 was shown to be capable of describing the evolutions of ENSO and dramatic ENSO regime changes (Trenberth and Stepaniak 2001). Then, we constructed the EP and CP ENSO index (IEP and ICP) based on In3 and In4 (Ren and Jin 2011). The details of the calculations are described by Yu et al. (2019). IEP (ICP) was calculated as in Eq. 1.

$$\left\{\begin{array}{c}{I}_{\textrm{EP}}={I}_{\textrm{n}3}-\alpha \times {I}_{\textrm{n}4}\\ {}{I}_{\textrm{CP}}={I}_{\textrm{n}4}-\alpha \times {I}_{\textrm{n}3}\end{array}\right.,\kern1em \alpha =\left\{\begin{array}{c}0.4,\kern0.5em {I}_{\textrm{n}3}\times {I}_{\textrm{n}4}>0\\ {}0,\kern0.5em {I}_{\textrm{n}3}\times {I}_{\textrm{n}4}\le 0\end{array}\right..$$
(1)

If In3 × In4 > 0, the constant α is set as 0.4; if In3 × In4 ≤ 0, α is set as 0. If the absolute value of IEP (ICP) ≥ 0.5 °C for at least 3 months, an event is defined as an EP (CP) ENSO event. Note that a value of IEP (ICP) ≥ 0.5 °C for at least 3 months denotes an EP (CP) El Niño event, and a value of ≤ −0.5 °C for at least 3 months denotes an EP (CP) La Niña event, respectively. Table S2 shows the identified results of the different types and intensities of ENSO events. Due to the interannual oscillations of ENSO, we used a 9-year high-pass filter to remove the long-term trend and decadal variability of the ENSO index time series. The detrended ENSO index was used to analyze the correlation with the extreme temperature index.

2.3.3 Trend analysis

We used the nonparametric Mann–Kendall statistical test to determine the statistical significance of the trends in extreme temperature indices of each grid, and the magnitudes of the trends were estimated using Sen’s slope estimator method (Mann 1945; Sen 1968). The results of the nonparametric Mann–Kendall statistical test are not disturbed by outliers in the time series and are not affected by the distribution of the data. A positive slope indicates an upward trend; otherwise, the trend of the index is downward. However, the existence of positive autocorrelation in the data increases the probability of detecting trends when none exists, and vice versa. Autocorrelation in the data is often ignored (Hamed and Rao 1998). Here, trend analysis using Kendall’s slope estimator considered lag-1 autocorrelation in the time series residuals. When the time series of temperature extreme indices does not denote significant autocorrelation, the original Mann–Kendall statistical test was used to assess the statistical significance of trends (Saleem et al. 2021; Tangang et al. 2017). Regarding time series that exhibit serial correlation, we follow the research of Hamed and Rao (1998) to avoid the effect of autocorrelation on the variance of the Mann–Kendall trend test statistic (Tangang et al. 2017; Hamed and Rao 1998). We construct a modified nonparametric trend test applicable to autocorrelated data by correcting the variance value of the Mann–Kendall trend test statistic. In terms of its empirical significance level, the accuracy of the modified test was found to be superior to that of the original Mann–Kendall trend test (Hamed and Rao 1998). The 5% significance level was used in all of the significance tests (Miao et al. 2019; Poudel et al. 2020).

2.3.4 Composite analysis

To assess the role of climate variability (such as the ENSO), we used a composite analysis method to provide a direct representation of the possible influences of the different ENSO types on the spatial and temporal variations in temperature extremes. The method is similar to that used in previous studies (Zhang et al. 2010; Miao et al. 2019). In the study, we selected the years with the five highest and five lowest IEP and ICP values. We computed the means of the extreme temperature indices (Thigh and Tlow) for the high- and low-EP/CP ENSO index years from 1980 to 2020 and the means of extreme temperature indices (Tneu) during the neutral years. Then, we calculated the difference in these averages (Thigh–Tneu or Tlow–Tneu) to determine the influences of the different ENSO event types (Miao et al. 2019). Moreover, we used the two-sided Student’s t-test to determine the statistical significance of the composite differences. We looked into the possible influence of modes of large-scale variability on the spatial and temporal variations in temperature extremes. Following the composite analysis, the anomalies of wind and geopotential height fields were investigated by analyzing the differences between different types of ENSO years and neutral years to explain the mechanisms of the influence of different ENSO types (Wang et al. 2020; Yu et al. 2019).

3 Results

3.1 Temporal and spatial patterns of the temperature extremes

Figure 1a–d show the regional annual mean changes of the four extreme temperature indices (TN10p, TX90p, TNn, TXx) on the TP. Inspecting Fig. 1a, the mean annual TN10p significantly decreased (p < 0.01) at a rate of −0.41 day/year. However, the mean annual TX90p, TNn, and TXx display significantly increasing trends across the TP during the period of 1980–2020, with slopes of 0.32 day/year, 0.05 °C/year, and 0.03 °C/year, respectively (Fig. 1b, c, and d). Overall, the frequency of the extreme events related to low temperatures significantly decreased on the TP, while the frequency of the extreme events associated with high temperatures significantly increased.

Fig. 1
figure 1

Linear regression trends for regional annual extreme temperature indices (a–d) and spatial patterns of the annual trends (e–h) on Tp during 1980–2020. The blue point is the annual series of the considered index, and the red line is the linear trend. The light red shadings correspond to 95% confidence intervals. The gray dot indicates the statistically significant at the 95% confidence level

Full size image

Figure 1 e–h illustrate the spatial patterns of slopes for TN10p, TX90p, TNn, and TXx on the TP during 1980–2020. The TN10p trend is spatially consistent across the TP, with significant decreasing trends over most of the TP. The annual changes ranged from −0.83 to 0.01 day/year, with the strongest decreases occurring on the northwestern TP (Fig. 1e). In addition, TNn significantly increased in most regions of TP (Fig. 1f). Figure 1g also shows that the TX90p statistically significantly increased over the entire TP. TXx varied from approximately −0.07 to 0.1 day/year, and the western and eastern parts of the TP had the highest rate of increase (Fig. 1h).

3.2 Response of temperature extremes to different ENSO types

Figure 2 a, b, and c show the time series of the monthly In3.4, IEP, and ICP from 1980 to 2020. The time series indicate that the SST anomalies caused by the CP ENSO events were weaker than those caused by the EP ENSO events. The detailed results are presented in Table S2. The frequency of EP El Niño events (eight events) was greater than that of CP El Niño events (three events) in a 41-year study period. However, the frequency of EP La Niña events (seven events) was similar to that of CP La Niña events (four events). Furthermore, these events also reflect distinct seasonal characteristics; the ENSO events tended to peak intensity in winter (November, December, January, and February) (Table S2) and gradually decline in the following year (Fig. 2a, b, and c). Thus, we divided the ENSO phases into the developing and decaying phases. Based on the peak time of IEP and ICP, the year before the peak time was considered the developing phases, and the year after the peak time was considered the decaying phases.

Fig. 2
figure 2

a–b Time series of the monthly ENSO index from 1980 to 2020 and d–f correlation coefficient of ENSO index with extreme temperature index. (a) Time series of In3.4; (b) time series of IEP, and (c) time series of ICP. (d) Correlation of In3.4 with extreme temperature index; (e) correlation of IEP with extreme temperature index; (f) correlation of ICP with extreme temperature index; the orange lines are the time series of the ENSO index. The slate gray lines are the 9-year high-pass filtering of the ENSO index. * denotes the statistical significance at 0.05 level

Full size image

We used the Pearson correlation coefficient between the monthly regional average of extreme temperature indices and the monthly IEP (ICP) to explore the response characteristics of temperature extremes to the different ENSO types. In neutral years, the IEP is positively correlated with the intensity of extreme temperature, and the correlations between ICP and extreme temperature intensity are not significant (Fig. S4). Figure 2 d, e, and f show the correlations between the monthly regional mean of extreme temperature indices and the monthly IEP (ICP and In3.4) during the different ENSO types. During the whole study period, we revealed that only IEP and In3.4 were significantly and positively correlated with the frequency of cold extremes. During La Niña years, the correlation coefficients between ICP years and TNn (TXx) were statistically significant at the 95% level in La Niña years (r = 0.46 (0.49) in CP La Niña years and 0.36 (0.31) in EP La Niña years). For the EP El Niño event, IEP has a significant negative correlation with the intensity of temperature extremes (r = −0.28 and −0.33), while a significant positive correlation with TN10p (r = 0.29). These results are suggestive of an asymmetric response characteristic of temperature extremes over the TP to the different ENSO types.

3.3 Non-uniform influence of the different ENSO phases

We applied a composite analysis method to assess the seasonal response characteristics of temperature extremes to the different ENSO types. Fig. S5 shows the composite differences between the developing phases of ENSO years and the neutral period. Compared to the neutral period, TX90p in CP E1 Nino years tended to be larger over large areas in the western and central parts of the TP (difference of more than 2 days) and showed a significant increase in the central part of the TP (Fig. S5c). The TN10p difference values between EP E1 Niño years and neutral years were statistically significant at the 95% confidence level in the southern and eastern parts of the TP (Fig. S5i). Compared with the period of EP La Niña years, the frequency of warm extremes (TX90p) in E1 Niño years was relatively higher (over 6.8 days) in most of the grids on the TP, except for the northeastern regions. However, TXx was greater in EP La Niña years than in the neutral period of the western TP (Fig. S5o and p).

With regard to the response of temperature extremes to the decaying phases of ENSO years, it was observed that the composite differences of TXx are significant on the TP in the decaying phase. The lower intensity regions of warm extremes were located in the western and northern TP, while the significant warming areas occurred in the southern and eastern (Fig. S6d, h, l, and p). Compared to TXx, TX90p showed similar spatial patterns in the decaying phase of CP E1 Niño and CP La Niña years (Fig. S6c and g). On the contrary, TN10p was greater in the decaying phases of CP E1 Niño, CP La Niña, and EP E1 Niño years, but these changes were not statistically significant (Fig. S6a, e, and i).

Figure 3 summarizes these results of composite analysis. From Fig. 3a, we found that the composite differences of TN10p in the developing phases are larger than the decaying phases. In addition, the median of TN10p is the largest in the developing phases EP E1 Niño years (6.59 days). Nevertheless, the frequency of warm extremes (TX90p) featured the largest difference occurred in the developing phases EP La Niña years (median = 7.31day, Fig. 3c). Furthermore, it was observed that TNn tended to decrease in the EP E1 Niño and CP La Niña years (Fig. 3b).

Fig. 3
figure 3

Box plots showing the distributions of temperature extreme indices at the developing (decaying) phases of CP E1 Niño years, CP La Niña years, EP E1 Niño years, and EP La Niña years. The blue box represents the developing phase, and the orange box represents the decaying phases. The black lines within the boxes denote the median; the upper and lower bounds are the 25th and 75th percentiles, respectively

Full size image

3.4 Effects of large-scale atmospheric circulation

Figure 4 depicts the anomalies in the geopotential height and wind at 500 hPa caused by the developing and decaying phases of the different ENSO events. During the developing phases, EP El Niño events resulted in positive geopotential height anomalies at 500 hPa over northern China and negative geopotential height anomalies over the western TP (Fig. 4a). However, positive geopotential height anomalies centered on Japan and northeastern and western China. The negative geopotential height anomalies centered in the western Pacific and southern China can be detected. These geopotential height anomalies favored the prevalence of easterly winds over the TP (Fig. 4b). Moreover, the 850 hPa level pressure showed a similar change pattern, even though the magnitude of change was lower than 500 hPa (Fig. S7b). Next, Fig. 4g indicates opposite geopotential height patterns between EP El Niño events and EP La Niña events, and that the TP was dominated by westerly winds (Fig. S7c). Furthermore, the geopotential height anomalies during the CP La Niña events were different from those during the El Niño events. A negative geopotential height anomaly and a cyclonic circulation anomaly occurred on the northeastern TP during the CP La Niña events (Fig. 4d). In terms of the decaying phases, the TP was dominated by northwesterly winds (Fig. 4e–h). The positive geopotential height anomalies at 500 hPa are over the TP in the EP El Niño years (Fig. 4e), while the negative geopotential height anomalies over the TP occurred in the period of CP La Niña years (Fig. 4h). Figure 4f features opposite geopotential height patterns between the decaying and developing phases; the negative geopotential height patterns were distributed in the western TP. For the EP La Niña years, a positive and a negative geopotential height could be found over the southwestern TP and northern China (Fig. 4g). Nevertheless, TP was dominated by the negative geopotential height in 850 hPa pressure levels during the La Niña years (Fig. S8c and d).

Fig. 4.
figure 4

Changes in the average wind speed (arrows, unit: m/s) and geopotential height (shading, unit: gpm) at 500 hPa. a–d The average wind speed and geopotential height during the developing phases of different ENSO years (EP El Niño, CP El Niño, EP La Niña, CP La Niña) minus those during the neutral year. e–h The average wind speed and geopotential height during the decaying phases of different ENSO years (EP El Niño, CP El Niño, EP La Niña, CP La Niña) minus those during the neutral year. The lengths of the arrows indicate the changes in the wind speed. The black dot indicates locations where the difference of geopotential height was statistically significant at the 95% confidence level according to the Student’s t-test. The red arrows illustrated the significant test results of the winds at the 90% confidence level according to Student’s t-test

Full size image

4 Discussion

4.1 Variability of extreme temperatures on the TP

This paper presents an analysis of the spatial and temporal variations in temperature extremes across the TP during 1980–2020. The response characteristics of temperature extremes to different ENSO types and the related atmospheric circulation variations during years with different ENSO types were examined to explore the underlying mechanisms driving these changes.

The mean of the significant trends was −0.41 day/year for the TN10p. However, the opposite trends were observed for the TX90p, TNn, and TXx from 1980 to 2020 (Fig. 1a–d). These results indicate that the frequency and intensity of the warm extremes on the TP increased during the 41-year study period, while the intensity and frequency of the cold extremes decreased. Previous studies have reported similar variations in the temperature extremes in China, Asia, and the global (Dong et al. 2018; Zhou et al. 2016; Alexander et al. 2006), and all of the variations in temperature extremes on the TP are larger than those in the other regions of China (Yin et al. 2019). The main reason for this difference is that due to the surface albedo reduction caused by the reduced snow cover under global warming, the surface albedo feedback can warm the surface air by absorbing more solar radiation and then emitting more longwave radiation upward (Ding et al. 2018; Kang et al. 2010). Moreover, the radiative forcing effect of greenhouse gases can also cause significant air warming (Aguilar et al. 2009).

Our results indicate that the variations in the four extreme temperature indices were not spatially uniform on the TP. The TN10p shows significant decreasing trends over most of the TP, while the TX90p displays statistically significant increasing trends (Fig. 1e). These results reflect the decreasing frequency of the extreme cold events and the increasing frequency of the extreme warm events over the entire TP, which is consistent with the mean values of the significant changes in the TN10p (TX90p) on the TP (Fig. 1a and c). Furthermore, the intensity of temperature extremes shows significant regional variations, and TNn significantly increased across the TP (Fig. 1e–h). The possible reasons for this are related to the pattern of the amounts of clouds above the TP. The variations in the cloud amount play an essential role in the energy balance and thus the temperature distribution (Ding et al. 2018).

4.2 Non-uniform response of extreme temperature to different ENSO types

Both tropical and midlatitude ocean signals can affect the thermal conditions on the TP (Liu et al. 2020). In particular, ENSO events are a major factor controlling the variability of climate on a global scale because they modulate the variability in global atmospheric circulation (Miao et al. 2019). Our results demonstrate that there is an asymmetric relationship between temperature extremes over the TP and the different ENSO types. Compared to the neutral years, the EP El Niño episodes result in more extreme cold events (Fig. 2). The factors responsible for the impacts of EP El Niño on cold extremes are shown in Fig. S9c. The related SST induces a double-cell anomaly of Walker circulation in the tropics, with significant ascending motion anomalies in the tropics of Central-Eastern Pacific and Indian Oceans and downward motion anomalies over the Maritime Continent (Zhao et al. 2022), which contributes to generation of the northwesterly anomalies on the TP, inducing more extreme cold events over the TP. In addition, the relationships between ICP during the CP La Niña episodes and the intensity of temperature extremes are significantly positively correlated on the TP (Fig. 2). The La Niña episodes related to SST induce a single-cell anomaly of Walker circulation in the tropics, with ascending motion anomalies over tropical west Pacific and descending motion anomalies over tropical Central Pacific; ascending motion anomalies over the Indian Ocean are weak and thus cannot trigger an atmospheric wave train to lead to the generation of the anticyclonic anomaly in Northeast Asia (Tian and Fan 2020), whereas a cyclonic anomaly occurred over the northeastern TP, triggering significant westerlies and cooling across the TP (Fig. S9b and d). During the CP El Niño episodes, the SST anomalies over the Indian Ocean are weak and thus cannot trigger an atmospheric wave train to lead to the generation of the anticyclonic anomaly (Fig. S9a). Therefore, the influence of the CP El Niño episodes on the temperature extreme over the TP is weak.

Notably, the different phases of ENSO also have different influences on the changes in temperature extremes. During the developing phase, the frequency of the extreme cold events on the western TP increased significantly during the EP E1 Niño years (Fig. S5i). The characteristic may be attributed to the circulation anomalies in western TP, which result in the local extreme cold events (Fig. 4a). Moreover, the concurrence of a higher water vapor flux from Indian Ocean and changes of the subtropical high during EP E1 Niño years could enhance the differences in temperature variability between the southeastern and western TP. Next, the extreme warm events during the CP E1 Niño years tended to occur more frequently on the central of TP (Fig. S5c). The large-scale atmospheric circulation revealed that the CP El Niño events resulted in a positive geopotential height anomaly centered to Japan and northeastern China, and negative geopotential height anomalies centered over the southwest China. These geopotential height anomalies favored the prevalence of the easterly winds over the TP (Fig. 4b). The negative geopotential height over southwest China and the easterly winds can induce warm humid flows, which offset the cold conditions across the majority of the TP (Liu et al. 2020; Wang et al. 2020). By contrast, in EP La Niña years, ascending motion anomalies over the Indian Ocean are weak and thus cannot provide sufficient water vapor to trigger precipitation on the TP (Gao et al. 2018). Here, less precipitation leads to reduced evapotranspiration, and warm extremes cannot become cooler during the precipitation day (Yong et al. 2021), which may contribute to the more warm extreme events on the TP. Moreover, we also found that the changes in the temperature extremes were stronger during the EP La Niña years than during the EP E1 Niño years, which is consistent with the results of Saleem et al. (2021).

In terms of the decaying phases, our results indicate that the frequency and intensity of cold extremes over the western TP have experienced increases in EP La Niña years, while the intensity of warm extremes decreased (Fig. S6). Based on the large-scale atmospheric circulation in the decaying phases of the EP La Niña years, we observed that a negative geopotential height anomaly centered to the western TP, and the strengthening westerly corresponded to more extreme cold events over the western TP (Fig. 4f). The weakened wind in conjunction with the positive geopotential height was largely responsible for the above atmospheric anomalies tied to more warm extremes in the decaying phases of EP E1 Niño years. Dry conditions at positive geopotential heights favor more insolation, enhancing incoming solar radiation and thus increasing the likelihood of warm extreme events (Gao et al. 2020). Nevertheless, the resulting large-scale atmospheric circulation in the decaying phases of CP La Niña years demonstrates that a negative geopotential height anomaly and a cyclonic circulation anomaly occurred over the northeastern TP, triggering weakened easterly winds (Fig. S6g and h). This result suggests the existence of stronger mid-tropospheric cold air activities over these areas (Wang et al. 2020). During the decaying phases of CP E1 Niño years, warm extremes increased in the eastern TP, but decreased in the western part. Compared to EP E1 Niño years, a positive geopotential height shifts eastward, while a negative geopotential height controls cooling in the western part of the TP.

4.3 Importance and uncertainties

In this study, we divided the ENSO types based on the values of the EP (CP) ENSO indices and analyzed the different response characteristics of temperature extremes on the TP to different ENSO events and the large-scale atmospheric circulation anomalies associated with them. Our results show that there is an asymmetric characteristic between temperature extremes on TP and different ENSO types. On the other hand, the development and decay phases of ENSO have different effects on the temperature extremes over the TP, which are mainly attributed to the El Niño-related SSTs inducing Walker circulation anomalies in the tropical Pacific and upward motion over the tropical Indian Ocean, which induce non-uniform temperature extremes over the TP. However, it should be noted that uncertainties may occur if only the different ENSO events are used to conduct climate change forecasts. Previous investigations have reported that other factors can affect local microclimates and the changes in temperature extremes, for example, precipitation (Bao et al. 2017; Yong et al. 2021), elevation (Liu et al. 2009), topographic heterogeneity (Sun and Zhang 2016), and anthropogenic aerosols (Seong et al. 2021). However, the underlying mechanisms controlling the variations in the temperature extremes caused by complex factors are still ambiguous, and future researches are required to analyze the interactions between the different influence factors.

5 Conclusions

This study focused on the spatial and temporal variations in temperature extremes across the TP and its response to ENSO. Specifically, the results of this study indicate that more intense and more frequent warm extremes and less intense and less frequent cold extremes have occurred on the TP since the 1980s, which reflected that TP is extremely sensitive to warming. Next, the different response characteristics of temperature extremes to different ENSO events were analyzed. Our results suggest that there is an asymmetric relationship between temperature extremes over the TP and the different ENSO types. Compared to the neutral years, the EP El Niño episodes result in more extreme cold events, whereas the influence of the CP El Niño episodes on the temperature extreme over the TP is weak. Moreover, the relationships between ICP and the intensity of temperature extremes are significantly positively correlated during CP La Niña episodes. Then, the developing and decaying phases of El Niño have different effects on the temperature extreme of the TP, which was largely attributed to the cause of the atmospheric anomalies. ENSO-related SST induces Walker circulation anomaly in the tropical Pacific and the ascending motion over the tropical Indian Ocean, which contributes to the generation of the anomalies of wind and geopotential height around the TP, inducing non-uniform temperature extremes over the TP. This study highlights the non-uniform variations of temperature extremes over the TP during years with different ENSO types. The benefits of improved regional climate prediction are enormous for the sensitive ecosystems and agriculture on the TP. In the future, incorporating the differences in the SST anomaly patterns and other influence factors will be helpful for producing robust temperature extreme predictions for the TP.