To analyze the EmCEI, 0.5° × 0.5° latitude/longitude grids of CPC gauge-based high-resolution daily maximum temperature, minimum temperature, and precipitation from 1981 to 2021 are used. The CPC extracts surface synoptic weather observations from the Global Telecommunications System and obtains a set of quality-controlled highly quantitative input gauge and satellite data using a robust objective analysis technique with negligible bias (Xie et al., 2007; Chen et al., 2008). The geographical coverage of East Asia is presented with the red box (20°–50°N, 100°–150°E) in Figure 1. Extreme events are analyzed in terms of annual, semi-annual, seasonal, and subseasonal time scales. The semi-annual range is divided by the warm season from May to October (MJJASO) and the cold season from November to April (NDJFMA) of the following year. Thus, the cold season of 1981 is the 6 months from November 1981 to April 1982. In the seasonal range, spring (March–May, MAM), summer (June–August, JJA), fall (September–November, SON), and winter (December–February, DJF) are used. The subseasonal range is defined by monthly variation.

The EmCEI consists of five components (C1 to C5) with seven CEIs and measures the percentage of an area with values above or below the percentile thresholds for the considered variable. In a nonstationary climate, changes to the extremes of a climatic distribution can occur from a shift in the distribution, a change in variance of a distribution, a change in the skewness of a distribution, or a combination of all three possibilities (Allen et al., 2012; Seneviratne et al., 2012). The modified EmCEI considers extreme from both tails of the climatic distribution and can highlight the shifts in the climatic distribution or skewing toward a particular tail of the distribution (Gallant et al., 2014). The method of calculating the EmCEI is presented in Table 1 (see Dittus et al., 2015, for more details). Each component is calculated for subseasonal to annual periods at every grid point over East Asia.

The long-term trend in the most recent years (1981–2021) were computed using linear regression and are described in the units of percent per decade. The significance of these trends was determined using a modified nonparametric Mann–Kendall test that accounted for persistence in the time series (Hamed and Rao, 1998). The statistical significance was assessed at the 5% level (Kendall, 1975).

The extremes identified by the temperature components in EmCEI (C1 and C2) are based on the 90th and 10th percentiles of the daily maximum/minimum temperature. Meanwhile, the precipitation components in EmCEI (C3–C5) consider wet days within the subseasonal to annual time scales. The 90th percentile of this wet day precipitation has one threshold value at each grid and time range while the thresholds for the temperature have 365 daily values for each grid point.

Figure 2 shows the distributions of EmCEI thresholds used in this study for the daily maximum temperature (Tx), daily minimum temperature (Tn), and wet day precipitation at individual grid points from January to December. The spatial structures of thresholds for the annual, semi-annual, and seasonal time scales are analogous to the seasonal evolution of the monthly distributions (not shown). Each percentile value in Figure 2 has a larger amplitude toward the south and near the ocean, with obvious seasonality. The wet day extreme rainfall is high over countries in southeastern China near the Yangtze River basin, the Korean Peninsula, and western Japan. The amplitude is smaller in the cold season and less than 5 mm over North China during winter. The daily maximum temperature in summer covers the most area above 30°C at the 90th percentile while it drops to below at the 10th percentile. In winter, Tx10p, Tn90p, and Tn10p decrease to below zero over North and Northeast China and extend southward from November to next March (April for Tn10p). The highest and lowest threshold values from both daily maximum and minimum temperature exist in July–August and December–January, respectively with the temperature ranging from −40°C (Tn10p) to 38°C (Tx90p).

Figure 3 shows the annual trends of EmCEI and its subcomponents for East Asia from 1981 to 2021. Statistically significant trends are indicated by asterisks in the figure legend. In Figures 3B–F, each component (C1–C5) is indicated by a solid line (with gray shading) and has upper extreme (90th percentile, positive bar) and lower extreme (10th percentile, negative bar) values. The total EmCEI (3A, black solid line), which represents the average of the five components reveals a 9% decade⁻¹ increase in the percentage of the area experiencing extreme climate both contributed by the temperature and precipitation. The 1980s and 2010s have opposite phases, where the areas experiencing extreme climate by the contribution of lower extreme (cold and dry) temperature and precipitation in the 1980s have changed to the upper (90th percentile) extreme warm and wet condition in the 2010s.

As shown in Figures 3B–F, annual increases in the subcomponents are observed, where the increase in extreme temperature (C1: 12% decade⁻¹, C2: 12% decade⁻¹) is higher than the increase in extreme precipitation (C3: 9% decade⁻¹, C4: 5% decade⁻¹, C5: 5% decade⁻¹). The temperature extremes show a growing upper-tail prevalence in both maximum and minimum temperatures. That is, the East Asia region has experienced more extremely warm days and nights than extremely cool days and nights in the most recent decade. The precipitation component, C3, represents extreme total rainfall. In the 2010s, the area of the upper extreme mostly covered East Asia, and the lower extreme area covered less than 5%. The areal percentage of upper extreme heavy rainfall (C4) also contributed more than the lower extreme (10th percentile) in the 2010s. The area with extreme numbers of wet and dry days (C5) also exhibits a significant pattern, where the area experiencing the wet day extreme is larger than the dry day extreme in the 2010s.

Note that the precipitation components in Figure 3 show both upper- and lower-tail extreme contributions, in contrast to the temperature components, where most years have a contribution from either the upper- or lower-tail extreme. Thus, the spatial coherence of the temperature extremes is much larger than that for precipitation extremes, as reported by Dittus et al. (2015). In their study, the EmCEI was first applied across five large-scale regions (the Northern Hemisphere, North America, Europe, Australia, and Asia); the areal extent of Asia (10°–70°N, 60°E–170°E) is much broader than that in our study, covering the tropics to high latitudes during 1951–2010. The annual EmCEI trend for Asia revealed a 5% decade⁻¹ increase, with a greater increase for the temperature components than for the precipitation components. Our result (Figure 3) is consistent with their study, however, there are several differences. First, the base period for the analysis includes up-to-date climate information. Second, we consider a more regional scale for the East Asia region. Third, various time scale changes are added and compared. Thus, several discrepancies appeared as follows. There is an abrupt increase in the temperature and precipitation extremes in the recent decade (the 2010s), which has led to an increasing trend of five components as well as the total EmCEI. The year 2021 is recorded to be the highest that experience extreme from a multivariable in more than 40% of East Asia. The amplitude of the year-to-year variability differs due to the difference in the base period. Because East Asia is significantly affected by the seasonal overturning of the monsoon flow, there are seasonal and subseasonal dependencies in extreme climate which will be explained in Figures 4–10.

The semi-annual trends of the EmCEI are shown in Figures 4, 5. In both warm and cold seasons, a significant increase in EmCEI is observed, where the increase is greater in the warm season (8% decade⁻¹) than in the cold season (6% decade⁻¹). The increases in C1 (11% decade⁻¹) and C2 (11% decade⁻¹) in the warm season are higher than those in the cold season (C1: 9%, C2: 8% decade⁻¹). The temperature component (C1 and C2) in the cold season show a large interannual to decadal variability during the late 1990s to early 2010s. The weaker trend may be related with the observed aerosol cooling in the 20th century (Arias et al., 2021) and the multidecadal variability (Miao and Wang, 2020), which counteracts the consistent increase in temperature over East Asia especially in the cold season. The contributions of the upper tail to extreme precipitation have a consistent trend with the temperature but a more robust increment in the recent decade during the both season and higher in warm season. Because East Asia has a distinct rainy season in summer, the large amount of rainfall affects the annual precipitation by processes with various time and spatial scales. The contribution of extremely heavy rainfall (C4) in the warm season is larger than in the cold season (5 and 2% decade⁻¹, respectively). The wet/dry day extreme (C5) in both seasons also has a significant trend and is highly correlated (R = 0.9 regardless of the trend) with the total rainfall extreme (C3).

In short, the extreme temperature shows a more distinct increasing trend in the warm season than in the cold season, as well as the upper-tail extreme of wet-day total rainfall, rain days, and heavy rainfall. In both seasons, EmCEI was highest in the most recent year of the analysis (2021 and 2020 for warm and cold seasons).

The EmCEI integrates several components that assess changes in temperature and precipitation extremes and returns a percentage value representing the fraction area of a region affected (Gallant et al., 2014). In Figure 6, we selected the years with the highest and lowest EmCEI in the annual and semi-annual trends and plotted the grid points that contribute to the extreme of each component. We can monitor the locations of temperature and precipitation extremes separately using this multivariate index. In the annual, warm, and cold seasons, the highest percentage of upper (lower) tail extremes occurred in 2021 (1984), 2021 (1992), and 2020 (1983), respectively. The highest percentage is mostly explained by either the upper or the lower tail extremes in the temperature while it is largely (or partly) explained by either side of the extreme in the precipitation. The upper (lower) tail extremes of C1 to C5 include warm (cool) days, warm (cool) nights, wet-day total rainfall, the contribution of heavy rainfall to total wet-day rainfall, and the number of wet (dry) days exceeding (below) the 90th (10th) percentile, respectively.

In the annual EmCEI (Figures 6A,B), C1/C2 in 1984 and 2021 cover approximately 50% of East Asia, with opposite extreme conditions. The year 2021 was the eighth of the top 10 warmest years on record that occurred in the last decade. East Asia was also exposed to one of the hottest summers in China, Korea, and Japan. Most areas exposed to the lower tail extreme in 1984 are shifted to the upper tail extreme in 2021. For C3 to C5, the area experiencing wet extreme are mostly over North and Southcentral China in 2021. In Japan, dry extreme conditions prevail all around the country in 1984 while the opposite wet extreme occur over western Japan in 2021. The area exposed to the dry extreme in C3 to C5 mostly overlaps the wet extreme with more extended regions. As shown in warm season EmCEI (Figures 6C,D), the cool night extreme (C2) in 1992 occur in a larger area than the cool day extreme (C1). During the 2021 warm season, C1 and C2 concurrently reveal a 30% area of warm day and night extreme over the south and eastern coast and Northeast China. The areas with lower-tail precipitation extremes in C3 and C5 occur over southern China and western Japan, whereas small areas in C4 are scattered across East Asia. The upper-tail precipitation extreme in the warm season is consistent with the annual range, showing the contribution of extreme precipitation is mostly from the warm season. In the cold season EmCEI (Figures 6E,F), cold and dry extremes prevailed over the continent adjacent to the ocean and central-north China, respectively in 1983. The Korean Peninsula and Japan experienced extreme cold. While in winter 2020, simultaneous but opposite warm and wet extremes with the area more extended are observed. C3 and C5 reveal consistent wet extremes in North China which means the frequent wet days contributed to the extreme rainfall. In C4, strong heavy rainfall dominates over Korean Peninsula and Japan.

Based on the long-term trends and the information in Figure 6, we can identify the land area affected by the extreme climate in a specific time period, which component makes the largest contribution, and which locations have the greatest exposure to various extreme conditions. The spatial distribution of extreme conditions clearly showed multivariate features on an equal basis over the East Asia region. From the two highest records by upper- and lower-tail extreme temperature, the area of cold extreme at the lowest year are changed to be exposed to the warm extreme at the highest year.

Long-term trends of EmCEI and subcomponents from spring to winter are shown in Figure 7. In each season, indices from subseasonal months (Figure 8) are plotted to evaluate their contribution to the season.

In spring (Figure 7A), there are some years with marked spatial coverage that has experienced warm extremes in the recent decade. East Asia was exposed to cold and dry conditions in the 1980s–1990s and changed to the opposite conditions (warm and wet) afterward except for 2–3 years of opposite variations. The areas of extreme total rainfall are gradually increased and more frequent days of precipitation is found in the recent decade. In the case of heavy rainfall extreme, only a small area is exposed. Among the seasons, the increasing trend of maximum temperature (C1: 9% per decade) and total rainfall (C3: 7% per decade) is the highest in spring with the largest contribution from March and May, respectively. There is a large interannual variability from the late 1990s to the 2000s during April (Figure 8A).

In summer (Figure 7B), the contrasting tendency before and after the 2000s in the temperature extremes is obvious. The contributions from the minimum temperature are larger and have the highest increment among the seasons with the largest contribution from decreasing trend of the cool night in June (shown in Figure 9C). Because of the distinct rainy season over East Asia, smaller but significant trends are found in all three components of the wet day precipitation (C3 to C5). In the recent decade, there is an abrupt increase in the regions with extremely heavy and frequent rainfall causing the extreme total rainfall in East Asia. On the subseasonal scale, June to August shows the most steady and consistent trends and less interannual to decadal variability in comparison with other seasons.

In fall, changes toward the warm and wet extreme conditions are also obvious. Compared to other seasons, the enhanced wet day extreme is the largest. There are large monthly discrepancies (especially in temperature). In November, low trends and large interannual variability prevail through the analysis period. In spite of the monthly discrepancies, a warm and wet shift toward the recent decade is clearly found in seasonal (Figure 7C) and subseasonal (Figure 8C) EmCEI.

In winter, long-term trends in the minimum temperature (C2) and wet/dry day extremes are not significant and there is the largest interannual to multi-decadal variability, resulting in the lowest magnitude in trend among all the seasons. In the context of global warming, East Asian winter air temperature exhibits pronounced and complicated subseasonal variations (Yang and Fan 2022). In Figures 7D, 8D, winter shows the largest interannual variability in the spatial coverage of extreme temperature during the analysis period as well as the highest subseasonal variability among the seasons. In addition, the largest interannual variability occurs in the late 1990s to early 2010s in C1 and C2 (Figure 7D). During winter 2020/21, there were extreme cold waves that invaded China from December 2020 (Dai et al., 2021), then the record-breaking warm event in February 2021 (Zhang et al., 2021) was caused by Arctic sea ice melting and stratospheric sudden warming. As such, the EmCEI in December 2020 and February 2021 is revealed as the opposite extreme in Figure 8D.

In short, the seasonal long-term trend in the extreme maximum (minimum) temperature was strongest during spring (summer) and weakest during winter. The seasonal long-term trend in the extreme total rainfall, heavy rainfall, and wet days were highest in spring, summer, and fall, respectively. The tendencies in the EmCEI and five components represent an increasing proportion of East Asia experiencing warm and wet extremes in the 2010s of the recent decade except for weak signal in the winter due to large subseasonal and interannual variability.