Variations of air temperature differences at different temporal scales over the Tibetan Plateau since 1961 and their possible causes

The air temperature difference is an essential indicator to measure weather and climate change. The Tibetan Plateau (TP) is a sensitive area to global climate change. The study on its temperature difference variations is of great significance for further understanding the evolution of weather and climate systems, and scientific disaster prevention and mitigation. In this study, the spatio‐temporal evolution characteristics of the temperature difference on daily, monthly, seasonal and annual scales and their possible causes are analysed based on the daily observations at 59 meteorological stations over the TP from 1961 to 2020 and the circulation indexes such as the East Asian summer monsoon, South Asian summer monsoon, midlatitude westerly and TP summer monsoon. The results indicate that the diurnal temperature range (DTR) on the TP tends to decrease from 1961 to 2020. Compared with the situation in the late 20th century, the annual average of the DTR in the eastern TP turns from a nonsignificant increasing trend into a pronounced decreasing trend, while that in the western TP maintains a decreasing trend. This phenomenon is primarily influenced by the minimum and maximum temperature, precipitation and atmospheric circulations. The standardized temperature difference at different temporal scales shows decreasing trends and is distributed in north–south and east–west distribution patterns. The annual temperature range is mainly influenced by the remarkable increase of the minimum temperature, especially in the northern and southeastern TP. Besides, the increases of maximum temperature and precipitation also affect the annual temperature range. In terms of the seasonal temperature range, it is dramatically influenced by different meteorological factors, especially in the northern TP. The monthly temperature range (Mon‐TR) is dramatically influenced by the maximum and minimum temperature in the first and second half years, respectively. Moreover, the circulation systems, such as the midlatitude westerly, South Asian summer monsoon and TP summer monsoon, are the main influencing factors of the temperature difference variations on the TP. The midlatitude westerly have a large influence on Mon‐TR in June and September, and the significantly influenced stations reaches 54.2% and 35.6%. The impact stations of the TP summer monsoon on Mon‐TR is the largest in August, and the significantly influenced stations is 39%. In July, the influenced stations of the TP summer monsoon and South Asian summer monsoon on the Mon‐TR reaches 30.5%–33.9%. The influenced area of the Mon‐TR in June is largest affected by single summer monsoon or westerly index. Previous studies have shown that the warming mechanisms and their influencing factors have changed in the TP since the 21st century. The study results might provide the foundation for a scientific understanding of climate warming on the TP.


| INTRODUCTION
The Sixth Assessment Report of the Intergovernmental Panel on Climate Change stated that the average global warming rate from 2001 to 2020 is 0.99 CÁdecade −1 , and the period from 2016 to 2020 is the warmest 5 years since 1850 (IPCC, 2021).The warming rate in China during 1961-2021 is higher than the global average in the same period.China is located in a sensitive area to global climate change, especially on TP, where the warming rate is the highest in the country (CCC, 2022).
DTR is the difference between the daily maximum (T max ) and minimum temperature (T min ) (Easterling et al., 1997;Li et al., 2022;New et al., 2000), which can reflect the variability characteristics of the interaction of the T max and T min (Roy & Balling Jr, 2005;Sun et al., 2019) and is an essential indicator of climate change.Simultaneously, the DTR also can affect the growth, development and productivity of plants and the climatic comfortableness for human beings (Cheng et al., 2014;Dhakhwa & Campbell, 1998;Hao et al., 2022;Ma et al., 2022;Peng et al., 2013).The research on DTR variations and their mechanisms began in the 1990s (Kong, 2020).The asymmetric warming of T max and T min on global and regional scales has resulted in a decreasing trend of global and regional average DTR and in annual and seasonal average DTR since the 1950s, and this decreasing trend slows down in most regions around 1980 (Alexander et al., 2006;Easterling et al., 1997;Hao et al., 2014;Karl et al., 1991Karl et al., , 1993;;Sun et al., 2019;Thorne et al., 2016;Vose et al., 2004Vose et al., , 2005)).The model simulation also shows that the global DTR will decrease by 0.30 C during 1900-2099 (Zhou et al., 2009b).
The TP is the plateau with the highest altitude and the largest area in the world, and it is also a sensitive area of global climate change (Liu & Chen, 2000).Its climate is affected by the synergistic effect of midlatitude westerly, plateau monsoon and South Asian monsoon.By changing the global atmospheric and hydrological cycles, the climate of TP can trigger the change of global environmental (Kang et al., 2010;Wu et al., 2007;Yang et al., 2014).Although the overall climate of TP shows a warming trend, the warming trend does not change consistently with time.The annual mean temperature can be divided into three stages, namely the warm period in the early 1960s, the cold period from the mid-1960s to the early 1980s (Cai et al., 2003) and the period of continuous temperature rise since the mid-late 1980s (Song et al., 2012;Wu et al., 2005).Note that different regions of the TP entered the period of continuous temperature rise at different times, and the southeast region entered the period of continuous temperature rise at the earliest time (Lin & Zhao, 1996).At the end of the 20th century, the average temperature, T max and T min on TP all showed an obvious increasing trend, among which T min had the most obvious increase (Song et al., 2012;Wu & Tang, 2017).
There have been many studies on the temperature variation and its characteristics over TP (Braganza et al., 2004;Duan & Wu, 2006;Easterling et al., 1997;Liu & Chen, 2000;Thorne et al., 2016;Vose et al., 2005;Zhou et al., 2009a), but there are few studies on DTR over TP (Kong, 2020).The annual and seasonal mean T max and T min of TP exhibit asymmetric warming characteristics which are consistent with the change of global mean series, thus resulting in a decrease in DTR and a decrease in annual mean DTR (Duan & Wu, 2006;Liu et al., 2006;You et al., 2016).Before the 21st century, the eastern TP was one of the regions in China where the DTR in spring and winter decreased significantly, and the decrease was stronger in spring than in winter (Hua et al., 2004).Meanwhile, the annual and summer DTR in most of the eastern TP showed an increasing trend (Ma & Li, 2003), while there is no obvious variation trend in the growing season (June-September) of TP from 2003 to 2016 (Zhu et al., 2020).In general, past studies have concluded that the DTR over TP shows a decreasing trend (You et al., 2017).However, limited by the length of data and sites in the study area, the variation amplitude of DTR over TP before and after the 21st century is not the same.The reduction of the DTR over TP (Dai et al., 1997;Karl et al., 1993;Thorne et al., 2016;Vose et al., 2005) is mainly controlled by surface energy and hydrologic balance (Lindvall & Svensson, 2015;Zhou et al., 2009a), and the main influencing factors include surface state, atmospheric boundary layer state, radiation effects of clouds and aerosols, atmospheric circulation state, and atmospheric composition (Lindvall & Svensson, 2015;Zhou et al., 2009a).Among them, the meteorological factors affecting DTR include precipitation, cloud cover, soil moisture and so forth.Precipitation and cloud cover are the main influencing factors, and precipitation is significantly and negatively correlated with DTR (Dai et al., 1997(Dai et al., , 1999;;Zhao et al., 2018;Zhou et al., 2009a).The increase of total cloud cover and the difference of diurnal cloud cover may lead to the increase of T min being greater than the increase of T max and the decrease of DTR (Duan & Wu, 2006;You et al., 2016).Some studies have shown that atmospheric circulation has a potential effect on the DTR in the TP region (Liu et al., 2018;Wu, 2010), but there are still few studies on the effect of atmospheric circulation on the DTR changes in the TP region (Duan & Wu, 2006;Liu et al., 2006;You et al., 2016You et al., , 2017)).In addition, the influence mechanism of atmospheric circulation modes on DTR changes remains unclear (Liu et al., 2018;Wu, 2010;You et al., 2016).Especially the relationship between the variation of DTR and westerly monsoon synergy effect over the TP has not been reported.
In view of the inconsistencies among the research conclusions on the variation range of DTR over the TP since the 21st century, and the insufficiencies of research on the relationship between the variation of DTR and westerly-monsoon synergy effect, this study attempts to analyse the variation characteristics and differences of the mean and extreme values of the temperature difference over the TP during 1961-2020 from different time scales and to explore the influencing factors of DTR and its relationship with the westerly-monsoon synergy effect.We hope this can improve the understanding of the climate and environment change over the TP and provide a basis for the sustainable development of TP and scientific response to climate change.The remainder of this paper is organized as follows.Section 2 describes the data and methods used in this study.Section 3 analyses in detail the temperature difference variations on the TP at different temporal scales and their possible causes.Finally, the main conclusions are shown in section 4.

| Data
The data used in this study are the observations from 59 meteorological observation stations on the TP from 1961 to 2020, provided by the Meteorological Information Center of the China Meteorological Administration, including the daily T max , T min , precipitation, sunshine duration and wind speed.The circulation data were monthly reanalysed data by NCEP/NCAR (including height field and wind field, respectively), and the time period was from 1961 to 2020.

| Index
The circulation factors include the East Asian summer monsoon, Indian summer monsoon, midlatitude westerly and TP summer monsoon (Figure 1).The East Asian summer monsoon index is proposed as below (Zhao et al., 2015), where Nor represents standardization and u is JJA mean 200-hPa zonal wind.When easterly anomalies appear around 20 N and westerly anomalies appear around 5 N and 35 N, the index is positive, and the EASM is stronger.The South Asian summer monsoon index (SASMI) is defined as an area-averaged seasonally (JAS) dynamical normalized seasonality (DNS) at 850 hPa within the South Asian domain (5 -22.5 N, 35 -97.5 E) (Li & Zeng, 2003), where V 1 and V are the January climatological wind vector and the mean of January and July climatological wind vectors, respectively, and V m,n is the monthly wind vector for the year n and month m.The norm A k k is defined as , where S denotes the domain of integration.At a given grid point (m, n), where V 1 is the latitude at the point (m, n) and Δs =aΔφΔλ=4, a is the mean radius of the earth and Δφ and Δλ (in radian) are the resolutions in the meridional and zonal directions, respectively.The A value of Equation (3) is subtracted in the right-hand side of the formula, because it is the critical value of significance of the quantity.The case when δ m,n >0 means that the prevailing winds direction shifts by at least 90 between winter and summer.
The westerly index is defined by the average of the 500 hPa geopotential height difference in the range of 70 -110 E from June to August between 35 N and 50 N (Li et al., 2008).The index is proposed as below, WI = 1 17 where N denotes the number of equidistant longitudes taken along latitude circles, with an interval of 2.5 longitudes.
The 600 hPa vorticity distribution is calculated by using the wind field.In summer, the main body of the TP is the centre of positive vorticity.Considering the stable establishment of positive vorticity, the jso-vorticity line of 5 units (1 unit =1 × 10 − 6 s − 1 ) is taken as the characteristic contour line, and the TP summer monsoon index (TP-SMI) represents the vorticity value of not less than 5 units in the area (30 -40 N, 70 -105 E; I V − area ) (Wang & Li, 2015).

| Methods
The temperature difference is expressed as the difference between the T max and T min at different temporal scales (daily, monthly, seasonal and annual), including the DTR, Mon-TR, seasonal temperature range (Sea-TR) and annual temperature range (Ann-TR).Spring, summer, autumn and winter months are from March to May, June to August, September to November and December to the following February, respectively.The monthly (seasonal and annual) average DTR is the The linear trend estimation, M-K test and t test methods (Wei, 2007) are used to analyse the temporal variabilities of annual mean DTR, annual maximum DTR and annual minimum DTR on the TP and its abrupt changes.The MK trend test determines whether a trend exists for each time scale separately.This test consists of computing the MK statistic (S) (Equations ( 5) and ( 6)), where x j and x k represent the data points at times j and k (j > k), respectively, and n is the number of data points.
For n ≥ 10, the statistic S is approximately normally distributed with the mean and variance as follows (Equation ( 7)) (Douglas et al., 2000): where n is the number of data points, g is the number of tied groups (a tied group is a set of sample data with the same value) and t p is the number of data points in the pth group.
The standardized Z test statistic is computed by Equation ( 8), The Z statistic fits the standard normal distribution.A positive Z value indicates an upward trend, whereas a negative value indicates a downward trend (Neeti & Eastman, 2011).
The least squares estimation is adopted to obtain the linear regression coefficient of linear trend estimation.The data used in this research span 60 years.The critical values for linear trends passing the significance tests at the 0.05 and 0.01 significance levels are 0.254 and 0.330, respectively.

| Diurnal temperature range
From 1961 to 2020, the annual average, maximum and minimum values of the DTR, as well as the difference between these annual maximum and minimum values, show a decreasing trend and have noticeable regional differences in the whole TP, except for the local areas in the southern part (Figure 2).The annual average and minimum values of the DTR show a decreasing trend in most area of the TP, with the spatial distribution characteristics of lower in the northern part and higher in the central and eastern parts.The decreasing rates are larger in the southeastern part, with values of 0.06-0.12CÁdecade −1 .The difference in the DTR between the annual maximum and minimum is smaller in the northern part but higher in the central part, and the decreasing rates are the largest (0.12-0.18CÁdecade −1 ) in the central large-value area.
The intra-annual average, maximum and minimum values of the DTR on the TP are smaller during June-September, and these values are smaller in the warm season (April-September) than that in the cold season (January-March and October-December).The difference between the daily maximum and minimum values of the regional average DTR is the largest during February-March (Figure 3).
The standardized series of the monthly, seasonal and annual average DTR on the TP from 1961 to 2020 display decreasing trends (Table 1).Among them, the decreasing trends in April and June and those of seasonal and annual averages pass the significance test.The decreasing rate of the standardized DTR series is the smallest in November, while it is the largest for the annual average, with a value of −0.32 CÁdecade −1 .
The monthly average DTR during January-March and November-December exhibits decreasing trends in the northern TP and increasing trends in the southern and southeastern TP, and these trends pass the significance test at the 0.05 significance level in some areas.
The April DTR at 94.9% of the stations on the TP shows Figure 4 shows the spatial variations of the seasonal and annual average DTR.The average values of the DTR tend to decrease with time in most areas, except for the parts of the northeastern TP and the southern TP, where the trends are increasing at some periods.Specially, the average DTR at 54% of the stations shows significant decreasing trends in spring.The areas with remarkably decreasing trends of the summer average DTR are mainly located in the northern TP.The variation trend rates of the annual, autumn and winter average DTR all display a latitudinal pattern of increasing in the southern TP and decreasing in the northern TP.

| Monthly temperature range
The standardized series of the monthly Mon-TR on the TP from 1961 to 2020 show decreasing trends (Figure 5).Specifically, the decreasing trends are most obvious in April (−0.22 CÁdecade −1 ) and the weakest in October (−0.03CÁdecade −1 ).The remarkable decrease of the Mon-TR in April is attributed to the slight increase of the monthly T max in April (the only month with nonsignificant warming) and the remarkable increase of the monthly T min .The increasing trend rates of the monthly T max and T min in October are similar, ranging from 0.23 to 0.29 CÁdecade −1 .Therefore, the Mon-TR in October is more stable than in other months over the years, and its decreasing trend is the flattest.
The maximum value of the standardized Mon-TR series on the TP varies greatly in the first half year, and the historical maximum appears in June 1961.However, it remains relatively stable in the second half of the year.The minimum value of the standardized Mon-TR series shows the opposite variation characteristics from the maximum value, that is, slight changes in the first half of the year and large fluctuations in the second half of the year.Affected by the maximum and minimum values of the Mon-TR, the maximum amplitude of standardized Mon-TR is in April (5.3), followed by June (5.2),and the minimum is in March (4.0).Moreover, the difference of maximum amplitude of standardized Mon-TR from January to June can reach 1.3.However, the change amplitude of Mon-TR in the second half of the year is between 4.2 (November) and 5 (August), and the largest difference between the months is only 0.8.It can be seen that in the past 60 years, the standardized Mon-TR of the TP from January to June has a relatively large variation range (Figure 5).
Spatially, the standardized Mon-TR during January-March and December of the cold season (January-March and October-December) on the TP shows a distribution pattern of decreasing in the northern TP and increasing in the southern TP, and the decreasing trends are marked in most areas of the northern TP.From October to November, the standardized Mon-TR tends to decrease in

| Seasonal and annual temperature range
The standardized Sea-TR on the TP displays significant decreasing trends in all seasons, with the largest decreasing rate in spring (−0.31CÁdecade −1 ) and the lowest rate in winter (−0.28 CÁdecade −1 ).The spring Sea-TR displays decreasing trends in about 85.4% of the TP stations, among which the decreasing trends are significant in 30.5% of the TP stations.The standardized Sea-TR in summer and autumn shows similar variation trends in spatial distribution, namely decreasing trends in most TP and increasing trends in some parts of the southern TP.The standardized Sea-TR in winter significantly decreases in the northern TP but increases in the central and southern TP (Figure 6).
The standardized average Ann-TR on the TP considerably decreases at the rate of −0.19 CÁdecade −1 , and the spatial differences in its variation trends are large.Specifically, the standardized average Ann-TR in the southeastern TP and most of the northern TP show decreasing trends with rates between 0.10 and 0.40 CÁdecade −1 .The larger decreasing rates

| Comparative analysis of the temperature difference at different temporal scales
A period of 30 years is taken as the time window, and the data in the window is averaged.Then the time window is moved to obtain the sliding estimation sequence.The 30-year windows is chosen because the World Meteorological Organization usually uses 30-year to calculate a standard climate state.To investigate the variation trend of the DTR climatology, we adopt the 30-point moving average to analyse the variation trends of the annual average, maximum and minimum DTR on the TP (Figure 7a).The results indicate that the decreasing rate of the annual minimum DTR reaches the minimum during 1974-2003, and then it increases noticeably.The annual maximum DTR during 1974-2003 shifts from a decreasing trend to an increasing trend.In addition, the annual average DTR turns from a decreasing trend to an increasing trend during 1973-2002.The MK test data of annual mean DTR, annual maximum DTR and annual minimum DTR shows that the abrupt change years are 1974, 1973and 2004, respectively (Figure 7b-d), respectively (Figure 7b-d).Meanwhile, the mean annual DTR and maximum annual DTR show an obvious increasing trend in the early 21st century.It can be seen that for annual mean DTR, annual maximum DTR and annual minimum DTR, the relatively stable climate has a relatively obvious change with the change in the early 21st century.
Comparing the average values (monthly, seasonal and annual average DTR) and the extreme values (Mon-TR, Sea-TR and Ann-TR) of the temperature difference on different temporal scales, we found that the average and extreme values of the Sea-TR, Ann-TR and April temperature difference show significant decreasing trends.In particular, the decreasing rates of the April Ann-TR and annual average DTR are the largest.For the extreme values of the October Mon-TR and the average values in November, the decreasing rates are the smallest (Figure 8).  the temperature difference extreme values in most of the TP are larger than those of the temperature difference average values.However, the temperature difference average values tend to increase in most semi-arid regions during autumn, but the temperature difference extreme values tend to decrease.In winter, the variation trends of the temperature difference average values in most of the TP are consistent with those of the temperature difference extreme values, but their variation trends are opposite in the eastern parts of the arid and semi-humid zones (Figure 9).

| Causes of temperature difference variations
The analysis of the temperature difference at different temporal scales on the TP from 1961 to 2020 indicates that April and November are the periods with the anomalous variations of the temperature difference average and extreme values and also the periods of the advancing and retreating of the circulation systems such as the westerly and the East Asian summer monsoon.Therefore, in this section, we investigate the relationships of the F I G U R E 9 Variation trends of the average and extreme values of the (a) spring, (b) summer, (c) autumn, (d) winter and (e) annual temperature difference, especially in the arid (orange areas), semi-arid (yellow areas), semi-humid (green areas) and humid (blue areas) zones.The dots indicate the temperature difference variation trends, "2" denotes that the variation rates of the temperature difference extreme values are larger than those of the temperature difference average values, "1" represents that the variation rates of the temperature difference extreme values are smaller than those of the temperature difference average value, "−2" indicates that the temperature difference extreme values tend to increase and the temperature difference average values tend to decrease, and "−1" donates that temperature difference extreme values tend to decrease and the temperature difference average values tend to increase

| Meteorological factors
The standardized annual average T max on the TP shows a remarkable upward trend.Especially since 1997, the warming phenomenon is noticeable.The standardized seasonal average T max also tends to increase, with the strongest trend in summer and the weakest trend in spring.Both the standardized annual average T min and annual average precipitation display distinct increasing trends.The standardized average T min and precipitation have significant increasing trends on the seasonal scale, except for the standardized average precipitation in summer.The standardized wind speed presents a considerable decreasing trend in each season, where the decreasing rate is the largest in spring and the lowest in winter.The standardized average sunshine duration shows marked decreasing trends in both summer and autumn.The variation trends of the monthly T max , T min and precipitation are increasing, and the monthly average wind speed and sunshine duration tend to decrease.
From the correlations of the temperature difference with the standardized meteorological factors (Figure 10), it can be found that the Ann-TR is significantly positively correlated with the annual T max , significantly negatively correlated with the annual T min and precipitation, and weakly positively correlated with the annual average wind speed and sunshine duration.Moreover, the significantly affected areas are mainly concentrated in the semi-arid and semi-humid zones.
On the seasonal scale, the spring Sea-TR is significantly positively correlated with the T max and significantly negatively correlated with the T min and precipitation.For the summer, autumn and winter Sea-TR, its correlation with the T min is negative.The standardized monthly Mon-TR is significantly positively correlated with the standardized monthly T max and significantly negatively correlated with the standardized monthly T min .Note that the correlation coefficient between the monthly Mon-TR and T max (0.7159) is the largest in April, and the monthly Mon-TR is significantly negatively correlated with the monthly T min and precipitation.The negative correlation between the Mon-TR and the T min in November is the largest (−0.7401) among all months.Additionally, the Mon-TR is also significantly affected by the monthly T max and wind speed.
The partial correlation analysis between the Mon-TR and different meteorological factors on the TP indicates that there are 20% of areas or above where the Mon-TR is significantly affected by the monthly T max and T min , and the significantly influenced stations even reaches 70% above during July-September.The significantly influenced stations of the monthly average wind speed on the Mon-TR is large from June to August and from October to November, reaching 15%-45%.The significantly influenced stations of the monthly average sunshine duration on the Mon-TR is also large during June-August, namely 70%-92%.From April to September, there are 40%-90% of areas on the TP where the Mon-TR is significantly affected by 3-4 meteorological factors.Specifically, in April, the Mon-TR is affected by 3-4 meteorological factors, and there are 90% of areas (larger than the areas in the other months) where the Mon-TR is significantly impacted by the monthly T max , T min and precipitation (Figure 11).

| Circulation systems
The climate on the TP is mainly influenced by the circulation systems such as the East Asian summer monsoon, South Asian summer monsoon, midlatitude westerly and TP summer monsoon.The analysis shows that the interannual variation of EASMI has a weak decreasing trend, while SASMI and WI have a significant decreasing trend, both of which have passed the 0.01 significance test.The TP-SMI shows a weak increasing trend.Since the 21st century, both monsoon and westerly wind indexes have shown obvious changes, especially the SASMI and WI, which are mainly positive before the 21st century and negative after the 21st century.At the beginning of the 21st century, the trend has changed from increasing to decreasing.
Analysis of the relationship between different circulation factors and Mon-TR shows that in June and September, the WI has a greater influence on Mon-TR over the TP, followed by January and March.From June to August, the influence of summer monsoon and the westerlies on Mon-TR is slightly different.In June, the stations of the westerlies significantly affecting Mon-TR is the largest, reaching 54.2%.In July, the stations that the SASM and TP-SMI significantly affects Mon-TR reaches 33.9% and 30.5%, respectively.In August, TP-SMI has a significant effect on Mon-TR, affecting 39% of the area.Among them, the impact stations of TP-SMI on Mon-TR is relatively large in each month, ranging from 18.6% to 39% (Figure 12).
In summer, the area affected by summer monsoon or westerly is the largest in June, followed by August and July.In summer, the Mon-TR variation over the TP is mainly influenced by a single circulation factor.In June, the ratio of the area under the joint and significant influence of two circulation factors is 14%, while the area under the influence of three or more circulation factors is small, which mainly occurs in July (6%) and August (1%) (Figure 13).

| CONCLUSIONS
From 1961 to 2020, the DTR on the TP and its variation amplitude are smaller in the warm season than those in the cold season.The annual maximum and minimum values of the DTR show decreasing trends, especially in southeastern TP.The significant decreasing rate of the standardized Ann-TR is −0.19 CÁdecade −1 , with great spatial differences.Compared with the end of the 20th century, the annual mean DTR in eastern China turns from an nonsignificant increasing trend to a significant decreasing trend, while in western China it still maintains a decreasing trend.The trends and area of the decrease of the standardized Sea-TR are the largest in spring, and the variation trends in the other months present a distribution pattern of weakening in the north and strengthening in the south.Compared with the end of the 20th century, the normalized series of Sea-TR in spring, summer and winter changes from a weak decreasing trend to a significant decreasing trend, while it still shows a significant decreasing trend in autumn.Compared with the previous 21st century, the seasonal variation trend of normalized Ann-TR also shows obvious changes from a downward trend in the west and an upward trend in the east to a decreasing-increasingdecreasing trend from northwest to southeast.
The average and extreme values of temperature difference in most areas of the TP tend to decrease, and the decreasing rate of the temperature difference extreme value is higher.The regions with significant differences between the trends of the average and extreme values of temperature difference are concentrated in semi-arid and semi-humid areas.Since the 21st century, the Ann-TR in eastern China has turned from an increasing trend to a decreasing trend (Hua et al., 2004).In June, the Mon-TR changes from a decreasing trend to an increasing trend (0.43/10a), whereas in September the Mon-TR variation rate decreases to 0.7/10a.The decreasing trend of Mon-TR in April is weakened, but the decreasing trend of standardized series of Sea-TR in each season is strengthened.This may be the result of changes in meteorological factors associated with the circulation system.By analysing the influence of meteorological factors and circulation system on the temperature difference, it is found that the meteorological factors have significant influence on the temperature difference.The influence degree of different meteorological factors on the Ann-TR and Sea-TR in spring is as follows: T min > T max > precipitation, and the Sea-TR in summer, autumn and winter has a significant negative correlation with T min .Mon-TR variation is positively correlated with T max , monthly mean wind speed and monthly sunshine duration, but negatively correlated with T min and monthly precipitation.The Mon-TR variation is significantly affected by T min and T max in July-September, and the affected stations is more than 70%.In addition, average monthly wind speed and sunshine duration are affected.
In June and September, the westerly wind in the middle latitudes affects the Mon-TR over the TP.In June, the Mon-TR variation in nearly half (47%) of the stations is significantly influenced by single and two circulation factors.In July, the Mon-TR variation in 35% stations is affected by three factors.In August, about 33% of the stations is significantly affected by three factors.
It should be pointed out that this study only discusses the possible causes of temperature difference changes through the correlation between temperature difference, circulation index and meteorological elements.In the future, it is necessary to reveal the driving mechanism of temperature difference from the mechanism of atmospheric circulation change, and to analyse the causes of mean and extreme temperature difference changes since the 21st century.

F
I G U R E 1 Study region and circulation systems on the Tibetan Plateau (TP) monthly (seasonal and annual) average value of the DTR at each station during the study period.

F
I G U R E 4 Spatial distribution of the variation trends of the standardized seasonal and annual average DTR over the TP from 1961 to 2020.The purple dots denote that the linear trends pass the significance test at the 0.05 significance level, and the colour spots indicate the DTR linear trends decreasing trends, and these trends at 50.9% of the stations pass the significance test at the 0.05 significance level.The variation trends of the monthly DTR from May to September present a remarkable zonal distribution, that is, the trends are obviously decreasing in the north and southeast while increasing in the central part.However, the increasing trends fail to pass the significance test (figure omitted).

F
I G U R E 5 The maximum and minimum values of the standardized monthly temperature range (Mon-TR) on the TP from 1961 to 2020, and their variation amplitudes and climate tendencies most parts of the TP, except for the southwestern TP (increasing trend).In the warm season (April-September), the standardized Mon-TR in most parts of the TP shows decreasing trends, especially in most of the northern TP from June to September (figure omitted).

F
I G U R E 6 Same as Figure4, but for the standardized seasonal (Sea-TR) and annual temperature range (Ann-TR) on the TP from 1961 to 2020 appear in Golmud (−0.40 CÁdecade −1 ) and Delingha (−0.33 CÁdecade −1 ) of northern Qinghai Province.In parts of southern and central TP, there are increasing trends, and the increasing rates range from 0.10 to 0.30 CÁdecade −1 (Henan County, Qinghai Province), as shown in Figure6e.

F
I G U R E 7 30-point moving average values and the MK test plots of the annual average, maximum and minimum DTR Further analysis of the multiyear variation trends of the average and extreme values of the monthly temperature difference suggests that the variation trends of the temperature difference extreme values in most areas of the arid zones are the same as those of the temperature difference average values.The variation rates of the temperature difference extreme values in January, March, April, September and November are larger in most areas.In semi-arid and semi-humid areas, the variation trends of the temperature difference extreme values are more obvious than those of the temperature difference average values, and the variation trends of the temperature difference mean and extreme values are opposite, especially in May and October.Moreover, the variation trends of the temperature difference average and extreme values are the same in most months in the humid zone (figure omitted).In most areas of the TP, the Sea-TR average values follow the same trend as the Sea-TR extreme values.In the eastern parts of the semi-arid regions, the variation trends of the Sea-TR extreme values in spring and summer are increasing, opposite to those of the Sea-TR average values.During autumn, the variation rates of

F
I G U R E 8 Average and extreme values of the temperature difference at different temporal scales.Shaded area denotes that the correlation coefficients pass the significance test at the 0.05 significance level 1 0 Correlations of the Ann-TR with (a) annual maximum temperature (T max ), (b) annual minimum temperature (T min ), (c) annual precipitation (Pre), (d) annual average wind speed (V) and (e) annual sunshine duration (SSD).The purple dots denote that the correlations pass the significance test at the 0.05 significance level, and the colour spots indicate the correlations F I G U R E 1 1 The stations ranges (%) of the Ann-TR and Mon-TR significantly influenced by different meteorological factors in each month temperature difference variations on the TP with the meteorological factors (T max , T min , precipitation, wind speed and sunshine duration) and the atmospheric circulations.

F
I G U R E 1 2 The stations ranges (%) of the Mon-TR significantly influenced by different circulation systems in each month

F
I G U R E 1 3 Significant influence range of different circulation factors on Mon-TR in summer (%)