Increasing trend of extreme winter warm spells in China and the intra‐seasonal differences

Based on the daily temperature observation data at 1992 stations in China, this study investigates the intra‐seasonal variations and trends of extreme warm spells during 1981–2022. The results indicate that the nationwide extreme cold and warm spells have both increased rapidly since the 21st century. However, different from cold extremes which occur evenly in winter months, the nationwide super warm spells have distinct intra‐seasonal differences, with more occurrences in February. The number of nationwide extreme warm spells has increased from 8 during 1981–2001 to 14 during 2002–2022, and the intensity has also increased obviously. Due to the spatial differences, the study area is divided into three regions to reveal the detailed features. It is found that the warm spell frequencies in most of China increase from December to February while decrease in northeastern China. Results demonstrate that extreme warm spells are concentrated in late winter in most regions (stations). The significant increasing trends are widespread from northern China to most parts of southern China, and the areas with increasing trend exceeding +0.6 day/decade are concentrated in North China and the middle reaches of the Yangtze River. Linear trends indicate that the significantly increasing trends in late winter contribute a major part to the variation in the whole winter in China.


| INTRODUCTION
Climate warming has led to increasing trends of extreme temperatures in recent decades (IPCC, 2021).Numerous studies have revealed the multi-time scale variations of winter cold extremes or extreme cold spells, especially the significantly increasing frequency in the middle latitudes of the Northern Hemisphere, owing to the Arctic warming or the Arctic amplification effect (Cohen et al., 2020;Johnson et al., 2018;Zhang et al., 2012).Meanwhile, global warming also brings more warm winters and warm spells.In Europe, it is found that extreme warm spell events can greatly affect agriculture, the sustainability of ecosystems (Chapman et al., 2020), and the environment such as the glacier melt in the Alpine (Colucci et al., 2017).Extreme warm spells can even induce high-impact hazards by interacting with other weather and climate events, exerting much more severe and extended impacts on many aspects (Ridder et al., 2020).In addition, they can hugely increase the risk of secondary disasters, such as forest and grassland fires, snowmelt floods, and air pollution.The quick alternation of warm spells and cold spells can also cause more extreme temperature whiplash, making ecosystems, human activities, and environmental resources more vulnerable to adaptation (Casson et al., 2019).
In recent decades, China has suffered an enhancing influence from extreme winter warm spell events, especially in the North China Plain.For example, in the 2020-2021 winter, China experienced a temperature roller coaster.After several cold spells in the early winter, the temperature rebounded sharply in the late winter and led to the warmest February since 1951 (Hu & Dong, 2021), with a regional average maximum temperature spike exceeding 4 C for the first time in the North China Plain.The maximum temperature in Beijing-Tianjin-Hebei exceeded 20 C on February 19-21, 2021, which is close to the climatology in early May.This extreme warm spell not only prompted the warmest February in recent 72 years but also led to a 4-15 days advance of the greening season of winter wheat in the plain, which greatly weakened the ability to resist cold and frost in the following March.In early March of 2022, the strong temperature rise caused the rapid melting of outdoor snow at the Beijing 2022 Paralympic Games venue, seriously disturbing the planned schedules of the games (China Meteorological Administration, 2022).
In recent years, extreme temperature whiplash events occurred frequently in China, for example, in the winters of 2020/2021 and 2022/2023 (Wang et al., 2021;Yao et al., 2022;Yao et al., 2023;Zhang et al., 2022;Zheng et al., 2022).After the two cold extreme events in early winter, extreme warm spells are observed over China in late winter, and the average temperatures in February 2021 and 2023 are 2.9 C and 1.6 C higher than normal, respectively.However, studies on warm spells in China are much fewer than those on extreme cold spells.China has already issued national standards or meteorological standards for cold spells or cold surges (Wang et al., 2017;Wei et al., 2017;Zhou et al., 2017), but the monitoring of warm spells is seldom touched.For the 95th and 90th thresholds of temperature drops at each stations in China, Ding et al. (2021) provided a spatial distribution to illustrate the regional differences of the 95th and 90th thresholds of temperature drops at each station in China.In general, the amplitude of temperature rises is less than that of temperature drops for a certain station or region.However, the threshold distribution for extreme warm spells is still unclear.Besides, typical cases in different countries in Europe show that both the intensity and the frequency of such warm spell events are increasing (Bozkurt et al., 2019;Chapman et al., 2020;Young & Galvin, 2020).However, the trends of extreme warm spells in China are still not explored.
In this study, we will first compare the intra-seasonal differences between typical extreme warm and cold spell events from a nationwide range, and then explore the general and individual features for extreme warm spells in early winter and late winter based on three regions in China.A spatial distribution of the thresholds and trends for each station is also provided.

| DATA AND METHODS
The daily maximum, mean, and minimum temperature data at 2513 weather stations during 1981-2022 are extracted from the Basic Meteorological Elements of China National Surface Weather Stations (V3.0), which is issued by the National Meteorological Information Center of China Meteorological Administration (Ren et al., 2012).This dataset has been strictly quality-controlled and widely used in operational meteorological services and scientific studies in China.In this study, 1992 stations with complete data in December-February from 1981 to 2022 are selected.The least-squares method is applied to fit the linear trend (Huang, 2004).As winter goes from 1 year to the next, it is mentioned in this study as the year for February, for example, winter 2022 means December 2021-February 2022.
In this study, a nationwide extreme warm (cold) spell event should last for at least two successive days, and meet one of the following two criteria.First, the daily maximum (minimum) temperature rise (drop) exceeds the 95th percentile of temperature rises (drops) on all days.Second, the daily temperature rise (drop) exceeds the 90th percentile on all days and at least one daily rise (drop) exceeds the 98th percentile.The daily temperature rise (drop) is calculated to detect the extreme warm (cold) spells, and this method can partly eliminate the warming trend since the consequent 2 days are in the same winter.

| RESULTS
Both extreme cold and warm events occurred frequently in recent years.This will increase the extreme cold-warm weather whiplash as explored by previous studies, for example, Lee (2022).However, the intra-seasonal features of extreme cold and warm spells may be different.The intra-seasonal variations of extreme cold and warm spells at the nationwide scale are analyzed.Figure 1 demonstrates 28 extreme cold spells and 22 extreme warm spells selected according to the criteria introduced above.Temperature variation in each day is also displayed by different percentile thresholds.It is obvious that warm events occur much more frequently in the second half of winter than in the first half.The frequency of warm events in February is 15, much larger than those in December (5) and January (2).However, there are no distinct monthly differences for the extreme cold spell events, with the corresponding frequencies being 12, 8, and 8 in December, January, and February, respectively.Remarkable differences in extreme temperature spell frequencies can also be found in 1981-2001 and 2002-2022.For the warm spell, the frequency almost doubled from 8 during 1981-2001 to 14 during 2002-2022.Similarly, the cold spell also increases from 10 to 18, which is consistent with the previous studies that the extreme cold surges have prevalently increased in China (Ding et al., 2021).The comparison among the four grades of temperature rises exceeding the 90th, 95th, 98th, and 99th percentile thresholds shows that the intensity of super warm spell events has also increased since the 21st century.There are only 4 (1) events with the national average maximum temperature exceeding the 98th (99th) percentile in 1981-2001, while the number jumps to 12 (6) in the late 21 winters.In the research period, the global surface temperature shows a significant increasing trend.But for the case selection, the warming trend effect is quite weak since the daily temperature change degree is used in the study instead of the temperature anomaly.
To identify the cold extremes, Ding et al. ( 2021) pointed out that a universal absolute threshold for the whole country is not appropriate.Due to the differences in latitudinal position and topography, remarkable discrepancies in daily temperature variations exist among different regions.Similar to Ding et al. (2021), the daily temperature rises exceeding the 90th and 95th percentile thresholds are calculated in this study.Figure 2 shows the distribution of the 90th and the 95th thresholds of daily mean temperature rises at the 1992 stations in China during the winter of 1981-2022.Different from the spatial pattern of temperature in China which increases from north to south, large values of the 90th percentile threshold are found in northern China, Tibet, and the south of the Yangtze River, with low values among them.Large thresholds of more than 4 C are found in the north of the Yellow River, with maximum centers of more than 6 C located in northern Xinjiang Province and Northeast China.Thresholds of 4-6 C are also found in eastern China and southern China.Affected by the low terrains, the thresholds over the middle reaches of the Yangtze River are relatively low with values less than 3 C, especially in Sichuan Basin.For the 95th percentile threshold, a similar geographical pattern but with higher values is observed (Figure 2b).Compared with Figure 2a, the threshold differences in the north of the Yellow River are more obvious, especially in Northeast China.The largest values exceeding 6 C are found in northern China and Tibet, and values exceeding 4 C are in southern China, with a relatively low-value region among them in the middle reaches of the Yangtze River.
In China, the country is usually divided into several regions to explore the spatial differences in the climate (Ding et al., 2020).Figure 3 shows the ratio of extreme temperature rise days in late winter (from January 16 to February 28) to that in early winter (from December 1 to January 15) exceeding the 95th percentile thresholds in each station of China in the periods of 1981-2022.Considering the regional differences, three regions can be classified by the ratio distribution.In northwestern China (R1, in the west of 110 E, and north of 38 N), the stations with a ratio exceeding 1 or below 1 are quite scattered and thus are identified as a mixed region.In northeastern China (R2, in the east of 120 E, and north of 38 N), the ratio of extreme temperature rise days in late winter to that in early winter in most stations is below 1, and the ratio in the south of northeastern China is even below 0.8.Therefore, northeastern China is identified as a decreasing region for its extreme warm spells.In the rest part of China (marked by R3 in the figure), the ratio in most stations is above 1.2, indicating more F I G U R E 1 National-wide extreme warm spell (EWS; brown) and extreme cold spell (ECS; blue) events in winter 1981-2022.The colors from light to dark brown (blue) indicate the area-weighted daily maximum (minimum) temperature rise (drop) exceeding the thresholds of the 90th, the 95th, the 98th, and the 99th percentile, respectively.The upper panels show the total number of EWS and ECS events in 1981-2001and 2002-2022.extreme warm spells in late winter than in early winter.However, there are a few sites in the Sichuan Basin with a ratio below 1.0.
As mentioned before, the nationwide extreme warm spell has clear intra-seasonal differences.To demonstrate the detailed features at the regional scale, Figure 4 gives the total number of extreme warm spell days in each 10-day period in winter with daily temperature rise exceeding the 90th, 95th, and 98th percentiles in the above three regions during 1981-2022.The intraseasonal features of the days for each threshold are consistent in each region and consistent with event ratios explored in Figure 3.The extreme warm spell days in northwestern China are fewer in the first 20 days, and almost stable from late December to February.In northeastern China, the extreme warm spell days show a decreasing trend from December to February.The rest part of China demonstrates an increasing trend from early winter to late winter, and the warm spell frequency exceeding the 98th percentile in each 10-day in February is about twice more than those in December.Therefore, it could be considered that for most regions in China, extreme warm spells occur more frequently in late winter.

F I G U R E 3
The ratio of extreme temperature rise days in late winter to that in early winter exceeding the 95th percentile thresholds in each station of China in the periods of 1981-2022, respectively.R1, R2, and R3 are northwestern China, northeastern China and the rest of China, respectively.
The above analyses on frequency variations of extreme warm spells are conducted by separating the winter into early winter and late winter at each station.The distributions of linear trends of the days with the 90th percentile threshold in the two periods are given in Figure 5.In early winter, the significant decreasing trends are scattered in both the north of the Yellow River and the south of the Yangtze River, while the significant increasing trends are concentrated in the southern part of North China and the lower reaches of the Yangtze River (Figure 5a).Decreasing trends exceeding 0.6 day/decade are in the northern part of North China, parts of Southwest China and South China.In late winter, the trends of extreme warm spells show a quite different pattern from that in early winter.Significant increasing trends are widespread from northern China to most parts of southern China, while negative trends appear at only a few stations in the upper reaches of the Yellow River and Southwest China (Figure 5b).The stations with increasing trends exceeding 0.6 day/decade are concentrated in North China and the middle reaches of the Yangtze River.These results indicate that the occurrence of extreme warm spells has remarkable intraseasonal variations and trends.

| CONCLUSION AND DISCUSSION
Global warming is causing a much warmer Arctic and an increasing trend of cold extremes in the middle latitudes of the Northern Hemisphere.Meanwhile, more extreme warm spells are found in some regions, such as Europe.In China, the warm extremes in the recent several years have also severely affected many industries.However, studies on warm spells in China are much fewer than those on extreme cold spells.The basic features of warm spells, such as the spatial and multi-time scale distribution, are still unclear.In this study, the daily temperature observation data at 1992 stations during the 1981-2022 winters are used to investigate the intra-seasonal variations and trends of warm spells in China.
The results indicate that both the nationwide extreme cold and warm spells have increased rapidly since the 21st century.From 2002 to 2022, the frequency of super warm spells is 14, almost double that from 1981 to 2001 (8).The super cold spell frequency also jumps from 10 to 18.However, different from cold extremes, the nationwide extreme warm spells enhance obviously during 2002-2022 and concentrated mainly in late winter.
The 90th and 95th percentile threshold distributions are displayed to illustrate the spatial difference of warm spells in China.Large values are found in northern China, Tibet, and the south of the Yangtze River, with low gaps among them.Due to the differences in latitudinal position and topography, there are remarkable discrepancies in daily temperature variations among different regions, which indicates that a universal absolute threshold for the whole country is not appropriate.China is divided into three regions based on the ratio of extreme warm spell days in late winter to that in early winter: the mixed region (northwestern China), the decreasing region (northeastern China), and the increasing region (the rest part of China).In most regions, the number of extreme warm spell days demonstrates an increasing trend from early winter to late winter.Distributions of linear trends of the warm spell days show quite different changes.In early winter, the stations with significant decreasing and increasing trends are rather scattered, while in late winter, significant increasing trends are widespread over most parts of China.Therefore, both the increasing trends and the enhancing intensity of warm spells in late winter contribute a major part to the trends over China in the whole winter.
In recent years, the effects of atmospheric teleconnection and wave train on the intra-seasonal variability of winter warm and cold events have attracted much attention.Wei et al. (2014Wei et al. ( , 2020) ) have revealed different roles of middle-troposphere teleconnection pattern in the in-phase and out-of-phase temperature evolutions in early and late winter.A similar wave train is also important in the rapid transition of temperature in North China (Ma & Zhu, 2020).The above analyses are obtained from a longer timescale.For the extreme warm spell, it belongs to the typical weather process and the major influencing circulations are synoptic systems.Therefore, the in-depth mechanism should be explored by circulation evolution from typical case studies and multi-case composition in the future.

F
I G U R E 4 The total number of days (ordinate) in each 10 days of winter (abscissa) with daily temperature rise exceeding the 90th, 95th, and 98th percentile in three regions of China during 1981-2022: (a) R1 (northwestern China), (b) R2 (northeastern China), and (c) R3 (the rest of China).

F
I G U R E 2 Distribution of (a) the 90th and (b) the 95th thresholds of daily mean temperature rises ( C) during the winter of 1981-2022 in China.In (a), main regions and rivers mentioned in the paper are marked.

F
I G U R E 5 Distributions of linear trends of the days with daily temperature exceeding the 90th percentile in (a) early winter and (b) late winter during 1981-2022.Only the stations with trends significant at 0.05 (big dots) and 0.10 (small dots) levels are shown.