Quantitative attribution of the temperature associated with winter extreme cold events in China

Extreme cold events exert enormous impacts on the ecosystems and human health. A quantitative attribution analysis of the surface temperature anomalies associated with extreme cold events in China is conducted using an updated version of the coupled atmosphere-surface climate feedback response analysis method. For both extreme warm and cold years, the effects of anomalous surface albedo, water vapor, atmospheric dynamics, and aerosols are the major positive contributors, and the processes associated with surface dynamics provide a major negative effect. Speci�cally, the effect of surface albedo (water vapor) is directly linked to the change in local snow cover (precipitable water) over China. More local snow cover (precipitable water) provides a cooling (warming) effect and vice versa. The process associated with atmospheric dynamics exhibits an asymmetric contribution for the extreme warm and cold years. It provides a warming effect over most of China in the extreme warm years, which is possibly related to the nation-wide warm air advection. On the other hand, the effect of atmospheric dynamics displays large regional and year-to-year variations for the cold cases. In addition, a considerable impact of anthropogenic aerosols is detected over central-southern China, which is possibly related to the emission reduction policies in China since 2013. Although clouds show only a moderate impact on the national-averaged temperature anomalies, they provide a considerable effect over southern China.

The interannual variability of extreme temperature is regulated by both atmospheric internal variability and external forcing (Zhang et al. 2022).Low temperatures tend to appear in China when the East Asian winter monsoon is strong, and vice versa (Wang and Chen 2014).The Siberia high and the East Asia trough are the two main factors responsible for the occurrence of extreme cold events in East Asia (Zuo et al. 2015).The increase in the Ural blocking could enhance cold advection from the polar region and trigger downstream cold air intrusions, which is favorable for the occurrence of extreme cold events in Hong Kong (Cheung et al. 2015).It has been indicated that the cold events during La Niña winters were contributed by the Ural blocking owing to the loss of Arctic sea ice in the Barents Sea, whereas those during El Niño winters were closely related to the negative Arctic Oscillation (AO, Song and Wu 2022).In addition, Wang et al. (2023) found that, on interannual time scales, the AO and El Niño-Southern Oscillation (ENSO) could signi cantly modulate the rst and second modes of the winter extreme cold events in China before the 1980s, respectively.Reversed relationships of the two dominant modes with the AO and ENSO were observed after the 1980s (Wang et al. 2023).
A remarkable progress has been made in understanding the mechanisms responsible for the interannual variability of extreme cold events.Nevertheless, few studies have attempted to understand the quantitative contributions of atmosphere-surface coupled thermodynamic processes to extreme cold events.The coupled atmosphere-surface climate feedback-response analysis method (CFRAM) can directly isolate individual contributions of multiple physical and dynamical processes to surface temperature differences between two climate states (Cai and Lu 2009;Lu and Cai 2009;Zhang et al. 2022).It has been proven to be an e cient o ine diagnostic tool and successfully applied to quantify the relative contributions of various climate feedback processes to the surface and atmospheric temperature anomalies associated with ENSO (Deng et (Hu et al. 2020).The CFRAM has also been adapted to quantitatively examine the feedback attribution of surface temperature anomalies associated with the heatwaves over East Asia (Seo et al. 2021;Ha et al. 2022).
In addition to the dynamical processes, previous studies have found that the anthropogenic signal could be robustly detected in the changes in temperature extremes (Dong et al. 2018;Melbourne et al. 2019).In the CMIP5 multi-model warming projection, the anthropogenic greenhouse forcing and subsequent water vapor feedback processes are primarily responsible for the surface warming over East Asia (Fan et al. 2021).Recently, Zhang et al. (2022) conducted a quantitative attribution analysis of the surface temperature anomalies associated with a record-breaking warm winter (2016/2017) in China by utilizing an updated version of CFRAM in which the effects of aerosols and more species of trace gases were incorporated (CFRAM-A hereafter).It was reported that the overall warming pattern of extreme warm winter in 2016/17 might be largely attributed to the effects of anomalous water vapor, atmospheric dynamics, and aerosols (Zhang et al. 2022).Up to date, the relative roles of various physical processes associated with winter extreme cold events in China remain to be disclosed.
This study provides a quantitative attribution of individual physical and dynamical processes for the surface temperature anomalies associated with extreme cold events in China, utilizing the CFRAM-A.The rest of this paper is organized as follows.The datasets and method applied are described in section 2. The quantitative attribution of individual physical and dynamical processes for the surface temperature anomalies in the extreme warm and cold winters are discussed in section 3. Finally, conclusions of the study and further discussions are provided in section 4.

Data And Analysis Methods
The quality-controlled observational station data of daily 2-m air temperature in 1964 meteorological stations across China during 1962-2019 is collected from the China National Stations' Fundamental Elements Datasets V3.0 (Cao et al. 2016;Ren et al. 2012).The method of calculating extreme cold events is the same as that of Wang et al. (2023), but for daily mean temperature.Speci cally, an extreme cold day is identi ed when the daily mean temperature is below the 10th percentile of daily mean temperature during the period 1962-2019 (10th percentile for a total of 870 days; 58 × 15; the 15 days correspond to 7 days on either side of the target date).The number of extreme cold events is de ned as the cumulative number of extreme cold days during each winter (Wang et al. 2023).The winter of a speci c year is de ned as the December of the preceding year and the January and February of the current year.The Empirical Orthogonal Function (EOF) analysis is applied to extract the dominant modes of extreme cold events in China (North et al. 1982;Dawson 2016;Wang et al. 2023).
Monthly mean geopotential height, winds, solar irradiance at the top of the atmosphere (TOA), ozone mixing ratio, air temperature, surface temperature, speci c humidity, cloud amount, cloud liquid and ice water content, black carbon mixing ratio, organic carbon mixing ratio, sulfate aerosol mixing ratio, sea salt mixing ratio, dust mixing ratio, surface albedo, surface latent ux, and surface sensible heat ux from the MERRA-2 reanalysis product are utilized (Gelaro et al. 2017).The MERRA-2 has a horizontal resolution of 0.5° × 0.625° and 42 pressure levels ranging from 1000 hPa to 0.1 hPa.Monthly trace gases mixing ratio (CO 2 and CH 4 ) are obtained from https://ads.atmosphere.copernicus.eu/cdsapp#!/dataset/cams-global-greenhouse-gas-inversion.The CO 2 concentration has a horizontal resolution of 1.9° × 3.75° and 39 model levels from surface to TOA for 1979-2021.The CH 4 concentration has a horizontal resolution of 2° × 3° and 34 model levels from surface to TOA for 1990-2020.
Daily mean geopotential height, winds, air temperatures, and vertical velocity from the fth generation European Centre for Medium-range Weather Forecasts reanalysis (ERA5; accessed from the Copernicus Climate Change Service Climate Data Store) with a horizontal resolution of 1°×1° are applied to diagnose the thermal process by the following the thermodynamic energy equation (Hersbach et al. 2020): where is the horizontal temperature advection, is the adiabatic heating due to vertical motion, and is the diabatic heating.Air temperature (T), the horizontal wind (V), and omega (W) are extracted from the ERA5 reanalysis datasets, whereas , the potential temperature, is derived from T using the Poisson's equation (Cheung et al. 2013). is the diabatic heating rate per unit mass, and is the speci c heat at constant pressure (= 1,004.07According to Wang et al. (2023), the dominant modes of the extreme cold events in China exhibited an interdecadal shift around the 1990s.In addition, more accurate datasets and a rather "clean" aerosol background without major volcanic eruptions could be obtained after 1996 (Zhang et al. 2022).Therefore, we consider the averaged winter state of 1996-2019 as the base state, and the warm (cold) winter averaged state as the contrast to the base state.

Results
The in-situ observations of climatological extreme cold events and the rst EOF mode of the cold days in boreal winter are presented in Figs.1a-b for 1996-2019.The winter cold extremes occur around 4-8 days per year and are concentrated mostly over central-southwestern China (Fig. 1a).The rst EOF mode of cold days shows a national-coherent distribution with a maximum of interannual variability over northern and central-southwestern China (Fig. 1b), similar to the climatological structure.These features are consistent with the result of Wang et al. (2023) for the period of 1993-2017.The time series of nationalaveraged cold days and the corresponding principal of the rst EOF mode exhibit similar interannual variations (Fig. 1c).We de ne an extreme cold winter when the number of national-averaged cold days is above 0.8 standardized deviations, and an extreme warm year when it is below − 0.8 standard deviations.(Fig. 1c).The extreme cold winters exhibit nation-wide negative surface temperature anomalies over China and the maximum anomaly center appears over northern and central-southwestern China (Fig. 2a).The extreme warm winters also present a national-coherent structure with maximum positive values over northern China (Fig. 2b).These spatial patterns of surface temperature anomalies observed in the extreme cold and warm winters, including the magnitude and maximum centers, can be well captured by the MERRA-2 (Fig. 2).Thus, the MERRA-2 data is utilized for the CFRAM-A calculations to quantitatively attribute the surface temperature anomalies in the extreme warm and cold winters.
3.1 Quantitative attribution of surface temperature anomalies in extreme cold winters The result from CFRAM-A calculations for extreme cold winters is shown in Fig. 3.The MERRA-2 surface temperature differences and the sum of partial surface temperature differences between the cold winters and the 1996-2019 base state derived from the CFRAM-A analysis are presented in Figs.3a-b.Apparently, the CFRAM-A calculations well reproduce the spatial structure and the magnitude of surface temperature anomalies in the MERRA-2 (Figs. 3a-b).
The area-averaged sum of partial surface temperature differences over China is about − 0.85 K from the CFRAM-A calculations, which is quite close to the number of − 0.86 K in the MERRA-2 (Figs. 3a-b).The errors associated with the o ine calculation and linearization process are quite small at the surface ( gure not shown), as validated by previous studies (Chen et  The quantitative contribution of individual physical and dynamical processes to the surface temperature anomalies in extreme cold winters is presented in Figs.3c-l.The processes associated with surface albedo contribute to an overall cooling of − 0.46 K in China, the effect that shows a similar structure to the total temperature differences (Fig. 3g).The anomalies of water vapor provide almost a uniform cooling effect, which induce − 0.26 K overall cooling (Fig. 3h).
In addition, the processes associated with atmospheric dynamics and aerosols exhibit area-averaged partial surface temperature differences of − 0.11 K and − 0.07 K, respectively (Figs. 3j and l).Speci cally, the processes associated with aerosols provide a considerable cooling effect over central-eastern China but a warming effect in the southwestern region (Fig. 3i).The atmospheric dynamics contribute to cooling effects over most parts and warming over northeastern China (Fig. 3l).Although the effect of cloud is negligible for the national-averaged temperature differences, it provides a considerable cooling over southern China (Fig. 3i).The effect of surface dynamics is the only negative contributor, which shows a remarkable warming in eastern China and a relatively weaker cooling in western China (Fig. 3k).It provides an overall warming of 0.09 K for the extreme cold winters.On the other hand, the changes in solar irradiance, ozone, CO 2, and CH 4 exert a negligible impact on the surface temperature difference in extreme warm winters (Figs.3c-f).Overall, the remarkable cooling in northeastern China is primarily attributed to the changes in albedo and water vapor, whereas the changes in albedo, water vapor, atmospheric dynamics, and aerosols jointly contribute to the cooling in central-southern China.
We where is the total surface temperature anomalies with respect to the winter base state of 1996-2019.and are the spatial PAP coe cient and partial surface temperature anomalies related to the ith process.and denotes the calculation of area average.The sum of all equals to the area average of the total surface temperature anomalies.
Figure 4a presents the PAP coe cients associated with the ten partial temperature differences shown in Figs.3c-l, and their sums over China.Consistent with the above analysis, albedo and water vapor act as the main positive contributors to the cooling in China for the cold cases, and atmospheric dynamics and aerosols play a secondary positive role (Fig. 4a).Those positive effects are compensated by the negative contribution of surface dynamics to some extent (Fig. 4a).The relative roles of the physical processes vary in different regions.Speci cally, the cooling in northeastern China is mainly attributed to the processes associated with albedo, whereas the effects of atmospheric dynamics play the dominant roles in central-eastern China cooling (Fig. 4c and d).The processes of atmospheric dynamics play an opposite role in northeastern China compared to other regions (Fig. 4b-d), resulting in a weaker overall effect of atmospheric dynamics for the national-averaged cooling.
We further investigate the anomalies of 925-hPa geopotential height and winds, precipitable water, snow cover, and aerosol optical depth (AOD) to explore the possible causes of partial temperature changes in the extreme cold winters (Fig. 5).At the lower troposphere, anomalous negative geopotential height appears over the high latitudes, but anomalous positive geopotential height dominates over the mid-latitudes (Fig. 5a).The geopotential height shows positive and negative anomalies at lower levels over northern and southern East Asia, respectively, enhancing the north-south pressure gradient (Fig. 5a).An anomalous negative geopotential height and a cyclonic circulation anomaly appear at the mid-upper levels over the mid-latitudes of East Asia ( gure not shown).These features are similar to those for the negative phase of the AO, which favor the development of the East Asian trough, leading to cold advection from the high latitudes (Figs.5a-b).On the other hand, the quasi-barotropic positive geopotential height anomalies over northeastern China contribute to the local warming effect (Figs. 3l and 5a).
The spatial pattern of snow cover anomalies is similar to the contribution of surface albedo for surface temperature changes (Figs.3g and 5d).The local snow cover anomalies are dominant in the change of surface albedo.Speci cally, the positive anomalies of snow cover increase the surface albedo, re ecting more incident solar radiation and leading to cooling over China, in particular for northeastern China and north of Xinjiang Uygur Autonomous Region (Figs. 3b,   3g and 5d).This result demonstrates that the contribution of surface albedo to surface temperature changes is primarily related to the change in local snow cover.Water vapor act as a greenhouse gas, and more water vapor provides a warming effect and vice versa.The cooling effect of water vapor is clearly linked to the below-normal precipitable water (Fig. 3h and 5c).
Moreover, the effect of aerosols is related to the AOD change in the extreme cold winters (Figs.6a-f).By scattering or absorption of incoming shortwave solar radiation, an increase (a decrease) in aerosol concentration basically contributes to cooling (warming) in surface temperature.The cooling and warming effects of aerosols are associated with the positive anomalies of AOD over most of China and the negative anomalies of AOD over the southwestern edge of China, respectively (Figs. 3j, 5e and 6a).In fact, the cooling effect of aerosols in central China is mainly due to the change in sulfate, whereas the warming effect in southwestern China is linked to the changes in organic carbon (Figs.6a-e).In northwestern China, the cooling effect of aerosols is dominated by the change in dust (Figs.6a and 6f).
The total cloud cover presents positive anomalies over most parts of China and negative anomalies in most of south China and Xinjiang Uygur Autonomous Region in the extreme cold winters (Fig. 7a), which seems to be inconsistent with the surface temperature differences induced by clouds (Fig. 3i).We further decompose the total temperature changes related to clouds into partial temperature changes associated with the long-wave and short-wave effects of clouds (Figs.7b-c).In general, an increase in total cloud cover causes less incident solar radiation to reach the surface (i.e., short-wave effect), resulting in surface cooling and vice versa (Figs.7a and 7c).In addition, more clouds add downward long-wave radiation, exerting a warming effect on surface temperature changes (Figs.7a-b).Although total cloud cover shows only a slightly positive anomaly over southern China (except for the Guangdong Province), the shortwave related cloud effect contributes to a remarkable surface cooling (Figs.7a and 7c).On the other hand, the long-wave related cloud effect is stronger than the short-wave related cloud effect in northern China (Figs. 7c-d and 3i).The cloud effects are dominated by its short-wave radiation effect in south China and long-wave radiation effect in the rest of China.

Quantitative attribution of surface temperature anomalies in extreme warm winters
For the warm cases, the partial surface temperature differences derived from the CFRAM-A are presented in Figs.8c-l.In the extreme warm winters, the areaaveraged surface temperature difference over China is approximately 0.91 K, and that derived from the CFRAM-A calculations is 0.89 K (Figs. 8a-b).The spatial structure and magnitude of the surface temperature anomalies in the MERRA-2 are also well reproduced by the CFRAM-A calculations (Figs.8a-b).Similar to the cold cases, the feedback processes in solar irradiance, ozone, CO 2 , and CH 4 exert negligible impacts on the surface temperature anomalies in the extreme warm winters (Figs.8c-f).The atmospheric dynamics inducing 0.43 K area-averaged warmings play a major role in the temperature anomalies over most parts of China for the extreme warm winters (Fig. 8l).The processes associated with water vapor, surface albedo, and aerosols also make considerable contributions (0.17 K, 0.13 K, and 0.07 K area-averaged warming, respectively; Figs.8g, 8h, and 8j).Speci cally, the process associated with water vapor exerts a considerable warming effect over most of China but a cooling effect over southeastern China (Fig. 8h).Surface albedo exhibits a cooling effect in northernmost China and almost a warming effect in the rest of China (Fig. 8g).The spatial pattern of the aerosol effect is depicted by a dipole structure, with cooling in southern China and warming in central China (Fig. 8j).The effects of cloud and surface dynamics present large regional features and make only moderate contributions to the overall warming (0.02 K and 0.01 K, respectively; Figs.8i and 8k), consistent with Zhang et al. (2022).In particular, the cloud effect is characterized by a remarkable warming in southern China (Fig. 8i).
Overall, atmospheric dynamics act as a main positive contributor to the nation-wide warming in China for extreme warm winters, and the processes associated with albedo and water vapor play a secondary role (Fig. 4e).Speci cally, the warming in northeastern China is mainly attributed to the processes associated with atmospheric dynamics, water vapor, and albedo, whereas the effects of atmospheric dynamics, aerosols, and albedo play the dominant roles in central-eastern China warming (Fig. 4f and g).In southern China, the warming is primarily attributed to the changes in clouds, compensated by the cooling effects of water vapor and aerosols (Fig. 4h).
The anomalies of 925-hPa geopotential height, winds, precipitable water, snow cover, and AOD in the extreme warm winters are presented in Fig. 9.At the lower troposphere, anomalous negative geopotential height appears over the high latitudes, but anomalous positive geopotential height dominates over the mid-latitudes (Fig. 9a).The upper and middle troposphere displays a similar structure of geopotential height anomalies ( gure not shown).Anomalous positive geopotential height and anticyclonic circulations occupy major parts of China.These features signify a positive phase of the AO.According to Wang et al. (2023), these circulation anomalies associated with the positive AO are unfavorable for deepening the East Asian trough, and hence result in anomalous warm temperature advection and diabatic heating over most of China (Fig. 9b), contributing to the warming effect of atmospheric dynamics in the extreme warm winters (Fig. 8l).
In the extreme warm winters, below-normal snow cover appears over most of China, leading to reduce in the surface albedo which re ects fewer incident solar radiation and induces surface warming over China (Figs. 8b, 8g and 9d).The positive anomalies of snow cover increase the surface albedo, re ecting more incident solar radiation and leading to cooling over the northernmost of China (Figs. 8g and 9d).The warming (cooling) effect of water vapor over major parts (southeast) of China is clearly linked to above-normal (below-normal) precipitable water (Fig. 8h and 9c).The cooling and warming effects of aerosols are associated with the negative anomalies of AOD over most of China and the positive anomalies of AOD over the southwestern edge of China, respectively (Figs. 8j and 9e).In fact, the warming effect of aerosols in central China is mainly due to the change in sulfate, whereas the cooling effect in southern China is linked to the changes in black carbon, organic carbon, and sea salt (Figs.6g-k).In northwestern China, the warming effect of aerosols is dominated by the change in dust (Figs.6g and 6l).
The total cloud cover presents negative anomalies over most of China and slightly positive anomalies over the northernmost parts in the extreme warm winters (Fig. 7d).Corresponding to the remarkable reduction of total cloud cover in southern China, the short-wave related cloud effect contributes to a maximum warming center whereas the long-wave related cloud effect shows a relatively weaker cooling in southern China (Figs. 7e-f).The cloud effects are dominated by its short-wave radiation effect in southern China and long-wave radiation effect in the rest of China, consistent with the cloud effects in the extreme cold winters.

Year-to-year variations among extreme warm and cold winters
The PAP coe cients associated with the partial surface temperature differences due to the changes in different processes in each warm and cold cases are shown in Fig. 10a for verifying whether the conclusions drawn from the composite analysis are applied to the individual cases.The processes associated with surface albedo and water vapor provide large positive contributions (cooling effects in the cold cases and vice versa) to the surface temperature changes for both extreme warm and cold years, and the effect of anomalous surface dynamics is a major negative contributor (Fig. 10a).The contributions provided by these three drivers are basically proportional to the total temperature anomalies.On the other hand, the in uences of atmospheric dynamics display a larger year-to-year variation in the extreme cold years than in the extreme warm years, exhibiting asymmetric contributions to the warm and cold cases (Fig. 10a).Speci cally, the atmospheric dynamics induce considerable warmings for the warm cases except in 2015.They show a dominant warming effect for 2017.On the other hand, for the extreme cold years, considerable effects of the atmospheric dynamics for overall cooling only appeared in 2005 and 2008 (Fig. 10a).
This seemingly smaller overall effect is possibly because there exist opposite contributions of the atmospheric dynamics between northeastern China and other places (Fig. 3l).This spatially heterogeneous feature for the contribution of atmospheric dynamics to surface temperature change can also be seen from the individual extreme cold years (e.g., Fig. 11).
Aerosols provided considerable warming effects in the winters of 1999, 2000, 2002, and 2017, and cooling effects in the winters of 2008 and 2011 (Fig. 10a).
It seems that the effect of aerosols does not entirely depend on the warm and cold cases.Instead, the contribution follows the changes in AOD which show negative anomalies before 2004 and after 2015 but positive anomalies in between (Fig. 10b).Given the dominant role of anthropogenic aerosols in the total aerosol effect as discussed in the previous sections, the case-by-case effect of aerosols is possibly related to industrialization development and the emission reduction policies in China since 2013 (Zhang et al. 2022).

Summary And Discussion
A quantitative attribution analysis of the surface temperature anomalies associated with winter extreme cold events in China is conducted in this study.The temperature anomalies exhibit almost a coherent spatial pattern, warming in extreme warm winters and cooling in extreme cold winters, with the maximum anomaly center over northern China.Using the CFRAM-A analysis method, we decompose the surface temperature anomalies into partial surface temperature changes due to the processes of solar irradiance, ozone, carbon dioxide (CO 2 ), methane (CH 4 ), surface albedo, water vapor, clouds, aerosols, surface dynamics, and atmospheric dynamics.
The effects of anomalous surface albedo, water vapor, atmospheric dynamics, and aerosols are the major positive contributors, and the processes associated with surface dynamics provide a major negative effect for both extreme warm and cold years.Speci cally, the effect of surface albedo is directly linked to the change in local snow cover over China, which contributes to overall warming and cooling patterns for the extreme warm and cold years, respectively.The effect of water vapor is basically linked to precipitable water anomalies.The effect of aerosols may directly depend on the abnormal winter AOD, and considerable impacts of anthropogenic aerosols are detected over central-southern China, related possibly to industrialization development and the emission reduction policies in China since 2013.In addition, the cloud effect presents a large regional dependent feature, which is dominated by its short-wave radiation effect in southern China and long-wave radiation effect in the rest of China.The decrease in total cloud cover in the extreme warm winters leads to remarkable warming in southern China, and vice versa for the extreme cold winters.
The processes associated with atmospheric dynamics exhibit asymmetric contributions between the extreme warm and cold years.They provide a coherent and dominant warming effect over most of China in the warm cases, which is likely associated with nation-wide warm air advection.On the other hand, the effects of atmospheric dynamics display large regional and year-to-year variations for the cold cases.that the anthropogenic greenhouse forcing and subsequent water vapor feedback processes were primarily responsible for the surface warming over East Asia.In addition, a considerable effect of warming associated with aerosols over southeastern China had been revealed by the CFRAM-A analysis in 2016/17 winter, and it may ultimately be attributed to the effectiveness of emission reduction policies implemented in China since 2013 (Zhang et al. 2022).The warming effects due to the substantial decrease in aerosol concentration in China may be anticipated.Nevertheless, how the policies of mitigating pollution and reducing anthropogenic aerosols contribute to the surface temperature changes associated with extreme cold events warrants further investigations.
Recently, the Aerosol Chemistry Model Intercomparison Project has been endorsed by the Coupled-Model Intercomparison Project 6 and designed to quantify the climate impacts of aerosols (Collins et al. 2017).This provides an opportunity for our next work to investigate the regional climate impacts of future aerosol emission reduction policies in extreme temperature events.

Declarations
Figures Same as in Fig. 3, but for extreme warm winters.
Same as in Fig. 5, but for extreme warm winters.The number on the upper right of each panel denotes the area-averaged value over China.
received increased research interest during the recent decades, given their devastating effects on human health and the ecosystems (McMichael 2013; Ebi and Bowen 2016; Cui et al. 2017; Harris et al. 2018; Raymond et al. 2020; Guan et al. 2022).In the context of global warming, the extreme cold events in China have increased remarkably after 2006 (Fu and Ding 2021).For example, extraordinarily frequent and long-lasting snowstorms hit China in January 2008, with excessive snow amount, persistent low temperatures, and severe icing conditions over central and southern China (Wen et al. 2009; Zhou et al. 2009).Severely extreme cold events also occurred over China in the winters of 2011, 2016, and 2021 (Gong et al. 2014; Song and Wu 2017; Yu et al. 2022).Surface air temperature was observed more than 10°C lower than the historical average in some parts of China during 20-25 January 2016 (Song and Wu 2017), and the new record-breaking low temperatures reached − 19.6°C and − 19.9°C in Beijing and Tianjin on 7 January 2021, respectively (Zhang et al. 2021).
further calculate the spatial 'pattern-amplitude' (PAP) coe cient, following Park et al. (2012) and Deng et al. (2013), to quantify the overall contributions of individual processes to surface temperature changes.The PAP coe cient is de ned as:

Figure 1 (
Figure 1 (a) Climatological cold days and (b) the rst EOF mode of cold days in boreal winter (DJF) during 1996-2019 from in-situ observations.(c) Detrended time series of China winter cold days (thick blue line) and the corresponding PC1 (dashed orange line).The dashed horizontal lines represent the 0.8 standardized deviations of the cold days.

Figure 3 (
Figure 3 (a) Surface temperature differences (K) in winters between cold cases and 1996-2019 base state in MERRA-2.(b) Sum of partial surface temperature differences (K) derived from the CFRAM-A, and partial surface temperature differences (K) due to the changes in (c) solar irradiance, (d) ozone, (e) CO 2 , (f) CH 4 , (g) surface albedo, (h) water vapor, (i) clouds, (j) aerosols, (k) surface dynamics, and (l) atmospheric dynamics.Stippling in (a) denotes the values exceeding the 0.5 standard deviations of 1996-2019 inter-annual variation.The number on the upper right of each panel denotes the area-averaged value over China.

Figure 6
Figure 6 Partial surface temperature differences (K) due to the changes in the (a) total aerosols, (b) black carbon, (c) organic carbon, (d) sulfate, (e) sea salt, and (f) dust in the cold winters.(g-l) Same as in (a-f), but for the warm winters.The number on the upper right of each panel denotes the area-averaged value over China.

Figure 7 Anomalies
Figure 7 Anomalies of (a) total cloud cover (%) in extreme cold winters, and partial surface temperature differences with respect to the 1996-2019 base state (K) due to the changes in (b) long-wave and (c) short-wave effects of clouds.(d-f) Same as in (a-c), but for extreme warm winters.

Figure 10 (
Figure 10 (a) The PAP coe cients (K) as a function of extreme cold (light blue background) and warm (light red background) years associated with the partial surface temperature differences (K) due to the changes in solar irradiance, ozone, CO 2 , CH 4 , surface albedo, water vapor, clouds, aerosols, surface dynamics, and atmospheric dynamics.Different feedback processes are represented by different colors using the color scheme on the bottom.Abbreviations 'SR', 'O 3 ', 'CO 2 ', 'CH 4 ', 'AL', 'WV', 'CLD', 'AER', 'SUR', 'ATM', and 'SUM' stand for the processes of solar irradiance, ozone, carbon dioxide, methane, surface albedo, water vapor, cloud, aerosols, surface dynamics, atmospheric dynamics, and the sum, respectively.(b) Anomalies of winter AOD at 550 nm from 1996 to 2019.Blue (red) dots indicate extreme cold (warm) winters.
al. 2012; Park et al. 2012; Hu et al. 2016), the Northern Annular Mode (Deng et al. 2013), global warming (Chen et al. 2017, 2018; Hu et al. 2018, 2021; Kong et al. 2022), and the inter-model warming spread (Cai and Lu 2009;Lu and Cai 2009) energy balance in surface-atmosphere column, in order to isolate local partial temperature changes due to various individual radiative and dynamical processes(Cai and Lu 2009;Lu and Cai 2009).FollowingZhang et al. (2022), an updated version of the CFRAM that incorporates the effect of aerosols and more species of trace gases (namely the CFRAM-A) is applied in this study.The total temperature change can be decomposed as: 3 represent solar irradiance, cloud, surface albedo, water vapor, and ozone, respectively.GHG refers to greenhouse gases, including methane (CH 4 ) and carbon dioxide (CO 2 ).AER stands for aerosols including black carbon, organic carbon, sulfate, sea salt, and dust.and are the convergence of total non-radiative energy due to the changes in atmospheric dynamics and surface dynamics, estimated as a residual from the radiative energy perturbations.For more details, refer to Zhang et al. (2022).
al. 2017; Park et al. 2012).Therefore, the CFRAM-A is reliable by using the time-mean atmospheric and surface elds in the o ine radiative transfer calculations and the linearization of radiative energy perturbations (Seo et al. 2021; Zhang et al. 2022).
Dong et al. 20183)f atmospheric dynamics to the surface temperature anomalies is out of phase in northeastern China compared to the other places.According toWang et al. (2023), the effect of AO on extreme cold events is limited over northeastern China after the 1990s whereas ENSO could modulate the extreme cold events over southern China.Whether and to what extent the AO and ENSO contribute to the asymmetric contributions of atmospheric dynamics warrant further investigations.The effects of greenhouse gases (e.g., CO 2 and CH 4 ) on interannual timescale are negligible in this work; however, they have been clearly detected as the dominant anthropogenic external forcings on the changes in extreme temperature under global warming or on the long-term scale (e.g.,Dong et al. 2018).Numerous previous studies have been focused on the anthropogenic greenhouse forcing in the future climate projections(Fischer andKnutti 2015; Bador et al. 2016; Alexander and Arblaster 2017; Lewis and King 2017; Di Luca et al. 2020; Fan et al. 2021; van der Wiel and Bintanja 2021).Fan et al. (2021) suggested