Contribution of Surface Solar Radiation and Precipitation to Spatiotemporal 1 Patterns of Surface and Air Temperature Warming in China from 1960 to 2003 2

Contribution of Surface Solar Radiation and Precipitation to Spatiotemporal 1 Patterns of Surface and Air Temperature Warming in China from 1960 to 2003 2 Jizeng Du, Kaicun Wang, Jiankai Wang, Qian Ma 3 College of Global Change and Earth System Science, Beijing Normal University, 4 Beijing, 100875, China 5 Joint Center for Global Change Studies, Beijing 100875, China 6 Chinese Meteorological Administration, Beijing, 100081, China 7

This study shows an essential role of land energy budget in determining regional warming.

Introduction
With the rapid development of observational data and the simulation abilities of climate models, global warming has been regarded as undeniable (Hartmann et al., 2013).The increase in anthropogenic greenhouse gases and other anthropogenic impacts are believed to be the primary cause of global warming.However, there are significant spatial and temporal heterogeneities in climate warming, i.e., faster warming rates in semiarid regions and a "warming hole" in the central United States (Boyles and Raman, 2003;Huang et al., 2012), which represents a major barrier to the reliable detection and attribution of global warming (Tebaldi et al., 2005;Mahlstein and Knutti, 2010).Furthermore, the uncertainties in model simulations generally increase from the global to the regional scale because of uncertainty in regional climatic responses to global change (Hingray et al., 2007;Mariotti et al., 2011).Therefore, it is crucial to research not only the spatial and temporal patterns of regional climate changes but also regional climatic response mechanisms to global change.This approach can improve confidence in the detection and attribution of global climate change and prediction of future regional climate change.
The spatial heterogeneity of climate warming can be attributed to local climate factors and anthropogenic factors (Karl et al., 1991).For the former, local determining factors such as cloud amounts and precipitation can significantly influence regional warming speeds (Hegerl and Zwiers, 2007;Lauritsen and Rogers, 2012).Those spatial heterogeneities in climate-factor trends make important contributions to various changes in the land-surface energy balance.Existing studies have indicated that an increase in clouds can diminish downward shortwave solar radiation to the land surface, thus reducing the daytime temperature (Dai et al., 1997;Zhou et al., 2010;Taylor et al., 2011) while potentially increasing nighttime temperatures by intercepting outgoing longwave radiation (Shen et al., 2014;Campbell and VonderHaar, 1997).
Precipitation can alter the proportion of surface absorbed energy partitioned into sensible heat flux and latent heat flux and therefore has an inevitable impact on both land-surface and near-surface air temperatures (Wang and Dickinson, 2012;Wang and Zhou, 2015).In addition, precipitation plays a key role in the soil thermal inertia and surface vegetation, causing important feedback to regional and global warming (Wang and Dickinson, 2012;Seneviratne et al., 2010;Ait-Mesbah et al., 2015;Shen et al., 2015).
In addition to local climate factors, anthropogenic emissions of aerosols have a Atmos.Chem. Phys. Discuss., doi:10.5194/acp-2016-1022, 2016 Manuscript under review for journal Atmos.Chem.Phys.Published: 19 December 2016 c Author(s) 2016.CC-BY 3.0 License.significant effect on the regional climate system.Studies indicated that improving air quality in recent decades has led to brightening over North America and Europe (Wild, 2012;Vautard et al., 2009), whereas surface shortwave solar radiation (SSR) has declined in East Asia and India with increasing air pollution (Xia, 2010;Menon et al., 2002;Wang et al., 2012;Wang et al., 2015a).Consequently, the variation in SSR may have an impact on both local and global climate change (Wild et al., 2007;Wang and Dickinson, 2013b).
Land cover change can also alter the energy exchange between the land surface and the atmosphere; moreover, it has the potential to impact regional climate (Falge et al., 2005;Bounoua et al., 1999;Zhou et al., 2004).Previous studies have suggested that urbanization and other land-use changes contribute to promoting the warming effect caused by greenhouse gases (Kalnay and Cai, 2003;Lim et al., 2005;Chen et al., 2015).
Overall, the impacts of these factors on climate change may be very important on the regional scale, leading to a marked spatial difference in regional climate change, whereas they are usually omitted from the detection and attribution of climate change on the global scale (Karoly and Stott, 2006).
China has a vast territory and abundant types of climactic zones stretching from tropic to cold temperate, with a special alpine climate over the Tibet Plateau.In addition, Atmos.Chem.Phys. Discuss., doi:10.5194/acp-2016-1022, 2016 Manuscript under review for journal Atmos.Chem.Phys.Published: 19 December 2016 c Author(s) 2016.CC-BY 3.0 License.dramatic economic development and explosive population growth in recent decades has caused significant land cover change and serious air pollution, including frequent haze events (Yin et al., 2016;Cheng et al., 2014;Wang et al., 2016).The climatic diversity and intensive human activity in this region will likely lead to a unique response to global warming with obvious spatial differences in climate change.Karl et al. (1991) had analyzed the observational records for the period 1951-1989, finding that China's temperature warming trends were faster than those of the United States but slower than those of the former Soviet Union.Several studies had revealed that the warming rate in Northwest China had been approximately 0.33-0.39°C 10yr −1 during the second half of the last century (Li et al., 2012;Zhang et al., 2010), which was significantly higher than the average warming rate over China (0.25 °C 10yr −1 ) (Ren et al., 2005) or that on a global scale (0.13 °C 10yr −1 ) (Hegerl and Zwiers, 2007).
Air temperatures (Ta) over the Tibet Plateau have increased by 0.44 °C 10yr −1 over the last 30 years (Duan and Xiao, 2015), which was considerably faster than the overall warming rate in the Northern Hemisphere (0.23 °C 10yr −1 ) and worldwide (0.16 °C 10yr −1 ) (Hartmann et al., 2013).Understanding the characteristics and mechanisms of regional climate change is critical to advancing the knowledge and predication of future climate change.
Ta is a common metric for judging climate change on the global or regional scales.
However, land surface temperature (Ts) is beginning to play an increasingly important role in climate change research because it has the distinct advantage of being directly related to the land surface energy budget.Previously, Ts values used in regional climate research are primarily derived from satellite retrievals or reanalysis datasets (Weng et al., 2004;Peng et al., 2014), both of which have good global coverage but questionable accuracy and integrity.Furthermore, satellite-derived Ts values are only available under clear sky conditions, limiting their application to climate change studies.
In China, Ts has been measured as a conventional meteorological observation item by nearly all weather stations, as is Ta.This study found that observations of Ts have a good relationship with Ta in terms of spatial-temporal patterns and can equally accurately reflect the characteristics of climate change.More importantly, Ts is more sensitive to the local land surface energy budget, particularly surface solar radiation (SSR) and precipitation.
From the perspective of energy, both SSR and precipitation are key factors controlling the land surface energy budget; therefore, their changes most likely cause regional differences in the warming rate of Ta (Wild, 2012;Manara et al., 2015;Hartmann et al., 1986).For the first time, this study analyzed the relationship between Atmos.Chem.Phys. Discuss., doi:10.5194/acp-2016-1022, 2016 Manuscript under review for journal Atmos.This paper is organized as follows: Section 2 introduces the data and method used in the study.Section 3 includes three parts: the first part describes the spatial and temporal pattern of climate warming over China; the second part analyzes the impact of the variation in SSR and precipitation on Ta and Ts; and the third part illustrates the spatial and temporal pattern of the warming trend of Ta and Ts after adjusting for the impact of SSR and precipitation.The adjustment removed impact of land-atmosphere interaction on the warming, leaving impact of large scale warming caused by the elevated atmospheric greenhouse gases substantially.Our results show that adjustment substantially reduced the spatial contrast of warming trends of Ta and Ts in China, which is agree with the expectation of global warming.A summary and discussion are presented in Section 4.

Data
The meteorological observational data used in this study are recently released daily Atmos.Chem. Phys. Discuss., doi:10.5194/acp-2016-1022, 2016 Manuscript under review for journal Atmos.Chem.Phys.Published: December 2016 c Author(s) 2016.CC-BY 3.0 License.meteorological datasets, including the China National Stations' Fundamental Elements Datasets V3.0 (CNSFED V3.0), which can be downloaded from the China's National Meteorological Information Center (http://data.cma.gov.cn/data)(Cao et al., 2016).This dataset includes Ts, Ta, the barometric pressure, relative humidity, and sunshine duration.All of the observational records of the climate variables include quality control and homogenization of the processes of data acquisition and compilation.
Figure 1 shows that the number of stations used in this study (1977 selected stations from a total of 2479 stations) is abundant and significantly greater than in previous studies (i.e., 57-852 stations) (Kukla and Karl, 1993;Shen and Varis, 2001;Liu et al., 2004;Li et al., 2015); therefore, the observational data have better spatial coverage and higher confidence of detecting regional climate change (Fig. 1).Our study is the first to use the observations of Ts for research into regional climate change.
Observations of Ts at weather stations are different from data retrieved via other approaches, such as satellite data and reanalysis.All of the observational fields of Ts are 4 m × 2 m square bare land plots in a weather station.The surface of the observational field must be kept loose, grassless, flat, and at the same level as the ground of the weather station.Three thermometers are placed on the surface of the observational field, including a surface thermometer, a surface maximum thermometer, Atmos.Chem.Phys. Discuss., doi:10.5194/acp-2016-1022, 2016 Manuscript under review for journal Atmos.Chem.Phys.Published: December 2016 c Author(s) 2016.CC-BY 3.0 License.and a surface minimum thermometer.The thermometers are deposited on the surface of the observational field horizontally: half of each thermometer is embedded in the soil and the other half is exposed to the air.When the observational field is covered by snow, the thermometers are removed from the snow and placed on the snow surface.In addition, the exposed parts of the thermometers must be kept clean from dust and dew.
To verify the reliability of the Ts observational records, we analyzed the relationship between Ta and Ts in the observed records for 1960-2003.As shown in Figures.S1, the mean Pearson Correlation Coefficients between Ts-max and Ta-max calculated from the monthly anomalies were 0.775, 0.843, and 0.806 for the annual, warm, and cold seasonal scales, respectively, and were statistically significant (99% confidence) for all stations.The mean correlation coefficients between Ts-min and Ta-min were 0.861, 0.842, and 0.865 for the annual, warm, and cold seasonal scales, respectively, and were statistically significant (99% confidence) for all stations.The numerous than those for other climatic variables, i.e., only 85 sites were used for SSR observation in Liu et al. (2004); 90 sites were used in Li et al. (2015).
More importantly, it was found that sensitivity drifting of the instruments used for the SSR observations led to a faster dimming rate before 1990 and that instrument replacements from 1990 to 1993 had resulted in a falsely sharp increase in SSR (Wang, 2014;Wang et al., 2015a).The sparse distribution and low quality of SSR observations make it difficult to quantify the variation in SSR and detect its impact on climate change.
We therefore used sunshine duration-derived SSR in this study, which is based on an effective hybrid model developed by Yang et al. (2006).This model has subsequently been improved (Wang et al., 2015a;Wang, 2014) and has proved to be performed well in regional and global applications (Tang et al., 2011;Wang et al., 2012).Sunshine duration-derived solar radiation not only can accurately reflect the impact of clouds and aerosols on the SSR but also can more exactly reveal long-term SSR trends (Wang et al., 2015a;Wang, 2014).Sunshine duration has a better correlation with the satellitederived SSR, reanalysis, and climate model simulations of SSR than the observed SSR in China (Wang et al., 2015a).
There are 2,474 meteorological stations reporting data; however, the lengths of the effective observation records for the stations are different.In addition, only a small The monthly anomaly relative to the 1961-1990 climatology was calculated based on a monthly mean value of the daily observation value, and if a month has more than 7 daily missing values, it was classified as a missing value (Sun et al., 2016;Li et al., 2015).The annual anomalies are the average of the monthly anomalies for the entire year.The anomalies in the warm seasons are the averages of the monthly anomalies from May to October, and the anomalies in the cold seasons are the averages of the monthly anomalies from November to the next April.

Method
As shown in Fig. 1, the spatial distribution of the weather stations over Mainland China is extraordinarily asymmetric and the density of weather stations in East China is far greater than in West China.We used the area-weight average method to reduce these biases when calculating the national mean.First, we divided the study region into 1° × 1° grids (see Fig. S2); there are 953 grids over China.Second, we assigned all Atmos.Chem.Phys. Discuss., doi:10.5194/acp-2016-1022, 2016 Manuscript under review for journal Atmos.Chem.Phys.Published: December 2016 c Author(s) 2016.CC-BY 3.0 License.selected stations to the grids; there are 627 grids with stations, accounting for 65.79% of the total.Finally, the grid box value is taken to be the average of all of the stations on the grid, and the national mean is the area-weight average of all of the effective grids (Jones and Moberg, 2003).
The linear trends reported in this study were calculated by a linear regression based on the least square method.Based on the anomalies of grids, there are two common ways to calculate the national mean trends of the variables in China.The first method (Method I) calculates the national mean monthly anomalies by taking the area-weight of every grid first and then calculates the national mean trend based on the time series of the national average anomalies.The second method (Method II) calculates the trend at every grid first and then the national mean trend over China is the area-weighted average value of the trends on all of the grids.In our study, we calculated the national mean trends of the temperatures using both methods as both methods are widely used in the existing studies (Gettelman and Fu, 2008).Same results are derived from those two methods if time series of all grids is integral and have no missing data (Zhou et al., 2009).However, as noted, we selected 1,977 stations (see Fig. 1) that the valid data of observation records are longer than 30 years during the period 1960-2003, which is a reasonable compromise between the integrity of the observation records and the spatial coverage.The missing data in the time series for some grids results in a little difference between the results of these two methods.To avoid misunderstanding, the trends derived from Method I was discussed in the main text, but results from two methods were shown in Table 1.
In this study, a multiple linear regression (see Eq. ( 1)) was used to calculate the sensitivity and impact of changes of SSR and precipitation on the temperatures (Roy and Haigh, 2011).This can be expressed as where z represents the monthly anomalies of Ts-max, Ts-min, Ta-max, and Ta-min; x and y are the monthly anomalies of the SSR and precipitation, respectively; a and b are the corresponding sensitivities of the temperatures to SSR and precipitation, respectively; c is constant term; and  indicates the residuals of the equation.
To adjust for the impact of SSR and precipitation on temperatures, we took x as a time series of SSR and y as a time series of precipitation, while a and b are the sensitivities of the climate variables to changes in SSR and precipitation, respectively.
The method of adjusting for the impact of SSR and precipitation is expressed as ).In the annual records, all of the temperatures showed an obvious warming trend over China (Figs. 2a and Figs. 3a).As shown in Table 1, the national mean warming rate for Ts-max was 0.227 °C 10yr −1 and the rate for Ta-max was 0.167 °C 10yr −1 from 1960 to 2003.The warming rate of Ta-max based on the 1,977 stations in this paper was a little higher than both that of the global average (0.141 °C 10yr −1 ) from 1950 to 2004 (Vose et al., 2005) and that of a previous analysis of China (0.127 °C 10yr −1 ) from 1955 to 2000 based on 305 stations (Liu et al., 2004).
The seasonal contrasts of warming of Ta-max and Ts-max are important.Ts-max had an with previous studies on global or regional scales (Li et al., 2015;Liu et al., 2004;Easterling et al., 1997).For the daily contrast in warming rates, there are still apparent uncertainties in its causes and physical mechanism (Hartmann et al., 2013).Some previous studies hold an opinion that the microclimate (e.g.urban heat island) has a larger impact on minimum temperatures due to the lower and more stable boundary layer at night (Zhou and Ren, 2011;Christy et al., 2009), while many investigators argued that the variabilities of surface solar radiation are the main reason for this daily contrast in warming rates (Sanchez-Lorenzo and Wild, 2012;Makowski et al., 2009).
For seasonal contrast in warming rates of temperatures, previous studies attributed that the most precipitation are concentrated in warm season because of monsoonal climate in China.The higher relative humidity in warm seasons will suppress changes of temperatures (Liu et al., 2004;Karl et al., 1993).In this paper, our results had partly explained the causes of those daily or seasonal contrast in warming rates of temperatures.
However, there remain slight differences between our results and previous studies with respect to the temperature warming rates, which might have several causes.The number of stations used in our study is much greater than in previous studies, which has led to better spatial coverage and a better representation of our analytical result.In addition, we used the area-weight average method both to reduce the impact of uneven station densities and to improve the representation of West China, which has a sparse station distribution (see Fig. 1), when calculating the national mean trend.Previous studies had indicated that the Tibet Plateau and the northwestern arid and semiarid regions were experiencing significantly more rapid warming trends than other regions in China (You et al., 2016), which may lead to our results being slightly higher than those of previous studies.In addition, the study periods of the various studies are not exactly the same.In addition, differences in the method used to calculate the national mean trend also may result in significant differences.As shown in Table 1, the absolute value of difference between Method I and Method II ranges from 0.011 to 0.033 °C 10yr −1 , account for 3.4% to 14.3% of trends (taking the results of Method I as reference).been reported in multiple previous studies (Liu et al., 2004;Li et al., 2015).

The spatial patterns in the variabilities for the temperatures
For Ts-max and Ta-max, the warming rates of South China and the North China Plain in the warm seasons were considerably lower than those in the cold seasons, resulting in a more obvious spatial heterogeneity in the warm seasons (Figs.4b and 4h).However, the warming rates of both Ts-max and Ta-max in the Sichuan Basin and the Pearl River Delta were lower in the cold seasons than in the warm seasons.Despite of the spatial and seasonal patterns of Ta-max were clearly similar to those of Ts-max.For Ta-max, both the seasonal asymmetry and the spatial heterogeneity of the warming trend were less than those of Ts-max.
For Ts-min and Ta-min, the warming rates were highest in North China and generally decreased from north to south (Figs.4d and 4j).The average warming rates of Ts-min and Ta-min in the cold seasons (Figs.4f and 4l) were faster than those in the warm seasons (Figs.4e and 4k).This variation of warming rate with latitudes have been attributed to dynamics amplification (Wallace et al., 2012;Ding et al., 2014).In this study, we focus on the spatial heterogeneity of the warming rates at similar latitudes and diurnal contrast of the warming rates.
By contrasting the annual variation and spatial pattern of trends, we found that Ts and Ta had an extremely significant correlation with each other.Based on the time series Ts-max and Ta-max were 0.877, 0.799, and 0.921 on the annual, warm, and cold seasonal scales, respectively.The correlations between Ts-min and Ta-min were 0.976, 0.969, and 0.977 on the annual, warm, and cold seasonal scales, respectively.In the spatial pattern of the trends (Figs. 4), the correlations between Ts-max and Ta-max were 0.488, 0.465, and 0.522 on the annual, warm, and cold seasonal scales, respectively.Those between Tsmin and Ta-min were 0.638, 0.670, and 0.594 on the annual, warm, and cold seasonal scales, respectively.
In summary, Ts had a significant correlation with Ta both in annual variation (Figs.
3) and in long-term trends (Figs. 4), indicating that Ts observational records are reliable for climate change research.However, the correlation between Ts-min and Ta-min was significantly higher than that between Ts-max and Ta-max.Ts-min is closely related to the land-atmosphere longwave wave radiation balance during the nighttime, which is closely related to the atmospheric greenhouse effect (Dai et al., 1999).During the daytime, Ts is directly determined by the land surface energy balance, i.e., the incoming energy including SSR and atmospheric longwave radiation (Wang and Dickinson, 2013a), and its partitions into latent and sensible heat fluxes (Zhou and Wang, 2016).Despite its dependence on the land-atmosphere sensible heat flux, Ta is also impacted by local and/or large-scale circulation.So, the changes of land surface Ta-max than that between Ts-min and Ta-min.

Impact of surface solar radiation
Figs. S4 shows that SSR had an important relationship with Ts-max and Ta-max but not with Ts-min and Ta-min.The national mean of the partial correlation coefficients between SSR and Ts-max is 0.552 and 98.9% of the stations are statistically significant at the 1% level.Meanwhile, the national mean of the partial correlation coefficients between SSR and Ta-max is 0.441, and 95.4% of the stations are statistically significant at the 1% level.This relationship is stronger in South China and on the North China Plain, i.e., it reaches 0.810 for Ts-max and 0.765 for Ta-max.
Ts-max is primarily determined by the land surface energy budget, whereas Ta-max is influenced by the land surface energy flux and other factors, including both large-scale circulation and anthropogenic heat flux.Therefore, the correlation between Ts-max and SSR is higher than that between Ta-max and SSR (Figs.Spatially, overall, the partial correlation coefficients between Ts-max and Ta-max and SSR are higher in South China than in North China (see Figs. S4a-c and Figs.S4g-i).
South of 35° N, the national mean of the partial correlation coefficients between Ts-max (Ta-max) and SSR is 0.654 (0.552), whereas that between Ts-max (Ta-max) and SSR is just 0.417 in north of 35° N.During daytime, Ts and Ta is largely determined by how much energy is used to evapotranspiration (Shen et al., 2014).In south China where soil moisture is high, the energy used for evapotranspiration is near linearly related to SSR (Wang and Dickinson, 2013b;Zhou et al., 2007).However, energy used for evapotranspiration is more dependent on precipitation in the northwest China where the soil is dry during most time of a year.As a result, the energy available for heating surface and air temperature is not so closely related SSR.Therefore, the correlation coefficients between SST and Ts-max (or Ta-max) were stronger in south China.The correlation between Ta-max and SSR along the coast was significantly lower than inland (see Figs.To quantify the impact of SSR on temperature, the sensitivity of temperatures to changes in SSR has been calculated (Eq.( 2)).As Figs.S5 shows, Ts-max was the most sensitive to SSR, followed by Ta-max, and their national means were 0.092 °C (W m −2 ) −1 and 0.035 °C (W m −2 ) −1 , respectively.Ts-min and Ta-min were insignificantly sensitive to SSR because they primarily depend on atmospheric longwave radiation during the nighttime.
Based on the above analysis, we calculated the impact of changes in SSR on temperature (see the Method Section).From 1960-2003, the national mean decreasing rate of SSR was −1.502 W m −2 10yr −1 , as calculated from monthly anomalies at 1,977 stations, and the trend was significant in most regions over China (see Figs. S6).Our results are considerably less than the global average dimming rate (−2.3 ~ −5.1 W m −2 10yr −1 ) between the 1960s and the 1990s (Gilgen et al., 1998;Liepert, 2002;Stanhill and Cohen, 2001;Ohmura, 2006) and the national mean dimming rate across China (−2.9 ~ −5.2 W m −2 10yr −1 ) between the 1960s and the 2000s based on radiation station observations (Che et al., 2005;Liang and Xia, 2005;Shi et al., 2008;Wang et al., 2015a).
As noted in data section, the sensitivity drifting and replacement of the instruments used for the SSR observations results in a significant homogenization in stations observation records (Wang, 2014;Wang et al., 2015a), which causes a great uncertainty Due to the decreasing trend in SSR, the warming trend of Ts-max and Ta-max decreased by 0.139 °C 10yr −1 and 0.053 °C 10yr −1 , respectively, in the national mean.
Spatially, the decreasing rate of SSR in South China and the North China Plain was significantly higher than in other regions, especially in the warm seasons (Figs.5b).
Therefore, the cooling effect of decreasing SSR on Ts-max and Ta-max was more significant in South China and the China North Plain, resulting in significantly lower warming rates of Ts-max and Ta-max there than in other regions (see Figs. 4).The spatial consistency between decreasing SSR and the warming slowdown of Ts-max (Ta-max) implies that variation in SSR is the primary reason for the spatial heterogeneity of the warming rate in Ts-max (Ta-max).

Impact of Precipitation
Figs. S7a shows that there is a significant negative correlation between Ts-max and precipitation; the national mean of the partial correlation coefficients is −0.323, and 99.3% of the stations are statistically significant at the 1% level.Seasonally, the correlation is stronger in the warm seasons (regional mean: −0.405) than in the cold seasons (regional mean: −0.276).In warm seasons, the correlation in North China (regional mean: −0.459) is clearly stronger than in South China (regional mean: −0.365).
In cold seasons, the correlation is highest on the Southwestern Yunnan-Guizhou Plateau and in most regions of North China (regional mean: −0.305) (Figs.S7b and Figs.S7c), whereas it is relatively weak in Southeastern China, the Tibet Plateau, Dzungaria, the Tarim Basin, and some regions of Northeastern China (regional mean: −0.117).The correlations between Ta-max and precipitation have similar spatial and seasonal patterns (Figs.S7g-i) too, and 35.4% of the stations are statistically significant at the 1% level; these are primarily concentrated in arid and semiarid regions of China (regional mean: −0.167) (Figs.S7e-f and Figs.S7j-l).
Precipitation has a negative relationship with temperature because precipitation can reduce temperatures by increasing surface evaporative cooling (Dai et al., 1997;Wang et al., 2006).The impact of precipitation on temperature is higher in the warm seasons over China, which is consistent with seasonal changes in the correlation between Ts-max and Ta-max and precipitation (see Figs. S7b-c and Figs.S7h-i).
The national mean sensitivities of Ts-max and Ta-max to precipitation were −0.321 °C previous studies (Zhai et al., 2005;Wang et al., 2015b).
Therefore, for Ta-max and Ts-max, the reduction in precipitation aggravated the warming trend in the North China Plain, Sichuan Basin, and parts of Northeastern China, whereas the increase in precipitation primarily slowed the warming trend in Northwestern China and on the Mongolian Plateau (Figs. 6d).On national average, the impact of increasing precipitation resulted in the warming trends of Ts-max and Ta-max being decreased by −0.007 °C 10yr −1 and −0.002 °C 10yr −1 , respectively.However, compared to SSR, the impact of precipitation on Ts-max was smaller by approximately an order of magnitude.For Ts-min and Ta-min, the impact of changes in precipitation was insignificant.

Trends of surface and air temperature after adjusting for the effect of SSR and precipitation
Based on the above analysis of the impact of SSR and precipitation on temperatures, we found that the variation of SSR and precipitation had little effect on Ts-min and Ta-min.Therefore, we only analyzed their impact on Ts-max and Ta-max.After adjusting for the impact of SSR and precipitation (Figs.7), the warming rates of Ts-max and Ta-max increased by 0.146 °C 10yr −1 (64.3%) and 0.055 °C 10yr For Ta-max, the warming rate increased by 0.069 °C 10yr −1 (76.4%) in the warm seasons and 0.034 °C 10yr −1 (11.7%) in the cold seasons.
After adjusting for the impact of SSR and precipitation, the difference in warming rates between Ta-max and Ta-min changed from 0.190 to 0.134 °C 10yr −1 , a decrease of 29.1%, and the difference between Ts-max and Ts-min changed from 0.088 to 0.058 °C 10yr −1 , a decrease of 33.0%.
More importantly, after adjusting for the impact of SSR and precipitation, the spatial coherence of the warming rates of Ts-max and Ta-max in South and North China clearly improved (Figs. 8).The regional difference between the North China Plain, South China, and other regions in China shrank significantly due to the increase in the warming rates in South China and the North China Plain.In addition, the warming trend of Ts-max and Ta-max became more statistical significant in North China Plain and South China (see Figs. S9).
To further prove this, we selected two regions in China for further investigation: R1 primarily includes the North China Plain and R2 primarily includes the Loess In addition to SSR and precipitation, temperatures' warming rates may be affected by many other factors, such as land cover change, that have not been discussed in this study due to lack of data, i.e., land cover and land use (Liu et al., 2005;Zhang et al., 2016).After adjusting for the impact of SSR and precipitation changes, spatial differences in the warming trends clearly decreased; however, some regional differences remain.The warming rate of Ts-max in the Sichuan Basin remained significantly lower than in other regions after adjusting for these impacts.In addition, the north-south difference in the warming rates of Ts-min and Ta-min cannot be explained by the impacts of SSR and precipitation.Further study is needed.Table 1.The warming rates (units: °C 10yr −1 ) of the temperatures on annual, warm, and cold seasonal scales.Raw and Adjusted represent the warming rates calculated for the data before and after adjusting for the impact of solar radiation and precipitation.
Method I represents the first method, which calculates the national mean anomalies first and then calculates the national mean trend based on this time series; Method II represents second method, which calculates the trend of every grids first and then calculates the national mean value of the trends of all grids using the area-weight average method.We calculated the national mean trends of the temperatures using both methods.
Chem.Phys.Published: December 2016 c Author(s) 2016.CC-BY 3.0 License.SSR (and precipitation) and Ta or Ts in terms of their spatial-temporal patterns and further quantified the impact of the variations of SSR and precipitation on Ta and Ts in China for the period of 1960-2003.
high correlation between Ta and Ts indicates that the observations of Ts are reliable for detecting climate change.SSR is the most fundamental energy resource for Ts and Ta.Most previous studies had used the observed SSR to analyze the relationship between the variation in SSR and Ta over Mainland China.However, sites for SSR observation were far less Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2016-1022,2016 Manuscript under review for journal Atmos.Chem.Phys.Published: December 2016 c Author(s) 2016.CC-BY 3.0 License.
Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2016-1022,2016   Manuscript under review for journal Atmos.Chem.Phys.Published: December 2016 c Author(s) 2016.CC-BY 3.0 License.number of stations existed prior to 1960 and the observational records of Ts at many stations became significantly abnormal after 2003 because of automation.Therefore, in our analysis, we selected 1,977 meteorological stations (see Fig.1) that the valid data of observation record must be longer than 30 years during the period of 43 years between 1960 and 2003.
Figs. 2 and Figs. 3  show the long-term changes in Ts-max and Ta-max, Ts-min and Ta- Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2016-1022,2016   Manuscript under review for journal Atmos.Chem.Phys.Published: December 2016 c Author(s) 2016.CC-BY 3.0 License.average rate of 0.172 °C 10yr −1 in the warm seasons and 0.354 °C 10yr −1 in the cold seasons.For Ta-max, it was 0.091 °C 10yr −1 and 0.294 °C 10yr −1 in the warm and cold seasons, respectively.The increases in Ts-max and Ta-max in the cold seasons were much larger than those in the warm seasons, which is consistent with previous studies of China and other regions(Shen et al., 2014;Vose et al., 2005;Ren et al., 2005).Similarly, the warming rates of Ts-min and Ta-min in the warm seasons were clearly lower than those in the cold seasons too.As shown in Figs. 3, Ts-min increased by 0.315 °C 10yr −1 and Ta-min increased by 0.356 °C 10yr −1 (see Figs. 3a) from 1960 to 2003.The warming trend of Ta-min is generally consistent with earlier studies(Shen et al., 2014;Li et al., 2015;Liu et al., 2004); however, it is considerably larger than that reported for the global average (0.204 °C 10yr −1 )(Vose et al., 2005).Ts-min increased at a rate of 0.221 °C 10yr −1 in the warm seasons and 0.447 °C 10yr −1 in the cold seasons from 1960 to 2003.Ta-min increased at rates of 0.245 °C 10yr −1 and 0.505 °C 10yr −1 in the warm and cold seasons, respectively.On a national average scale, all temperatures increased from 1960 to 2003.The warming rate of Ts-min (Ta-min) was significantly faster than that of Ts-max (Ta-max) and the warming rates of all temperatures in cold seasons were generally higher than those in warm seasons.These basic characteristics of the temperature changes are consistent Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2016-1022,2016   Manuscript under review for journal Atmos.Chem.Phys.Published: December 2016 c Author(s) 2016.CC-BY 3.0 License.

Figs. 4
Figs. 4 demonstrates a clear spatial heterogeneity in the warming rates for Ts-max Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2016-1022,2016 Manuscript under review for journal Atmos.Chem.Phys.Published: December 2016 c Author(s) 2016.CC-BY 3.0 License. of national mean yearly anomalies (see Figs. 2 and Figs.3), the correlations between Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2016-1022,2016   Manuscript under review for journal Atmos.Chem.Phys.Published: December 2016 c Author(s) 2016.CC-BY 3.0 License.energy balance caused by SSR and precipitation have different levels of effect on Ts and Ta during the day, which most likely causes a lower correlation between Ts-max and S4a and Figs.S4g).On the seasonal scales, the partial correlation between Ts-max (Ta-max) and SSR in warm seasons Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2016-1022,2016   Manuscript under review for journal Atmos.Chem.Phys.Published: December 2016 c Author(s) 2016.CC-BY 3.0 License. is higher than that in cold seasons, and the national mean partial correlation coefficients for the warm and cold seasons are 0.579 and 0.498 for Ts-max and 0.544 and 0.386 for Ta-max, respectively, consisting with the seasonal cycle of SSR intensity over China.
Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2016-1022,2016 Manuscript under review for journal Atmos.Chem.Phys.Published: December 2016 c Author(s) 2016.CC-BY 3.0 License. in trend estimation.Tang et al. (2011) used quality-controlled observational data from 72 stations and two radiation models based on 479 stations to determine both that the dimming rate over China is −2.1 ~ −2.3 W m −2 10yr −1 during 1961-2000 and that the SSR has been essentially unchanged since 2000; this finding is generally consistent with our results.
Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2016-1022,2016   Manuscript under review for journal Atmos.Chem.Phys.Published: December 2016 c Author(s) 2016.CC-BY 3.0 License. 10 mm −1 and −0.064 °C 10 mm −1 , respectively.As shown in Figs.S8, there were apparent seasonal and spatial changes in the sensitivity of Ts-max and Ta-max to precipitation (Figs.S8a-cand Figs.S8g-i).In warm seasons, these sensitivities were highest in the Tibet Plateau, the Loess Plateau, the Inter Mongolia Plateau, Dzungaria, and the Tarim Basin(Figs.S8b and Figs.S8h).In cold seasons, the distribution of regions with high sensitivity extended to all of North China and Southwest China (Figs.S8c and Figs.S8i).Overall, the sensitivities of Ts-max (Ta-max) were significantly higher in arid regions (dry seasons) than in humidity regions (rainy seasons)(Wang and Dickinson, 2013b).In contrast, Ts-min and Ta-min were less sensitive to variations in the precipitation.AsFigs.6 shows, during 1960-2003, the trend in the precipitation over the 1977 stations had obvious spatial heterogeneities.China's precipitation during this period showed a slight increasing trend with an increasing rate of 0.112 mm 10yr −1 .Precipitation in Northwestern China and Southeastern China experienced an increasing trend, whereas precipitation in the North China Plain, the Sichuan Basin, and parts of Northeastern China experienced a decreasing trend.However, the trend of precipitation was insignificant in most regions (see Figs. S6).Variation in precipitation had significant seasonal differences (see Figs. 6b and Figs.6c).The seasonal and spatial characteristics of these precipitation variations are consistent with those identified in Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2016-1022,2016 Manuscript under review for journal Atmos.Chem.Phys.Published: December 2016 c Author(s) 2016.CC-BY 3.0 License.
Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2016-1022,2016   Manuscript under review for journal Atmos.Chem.Phys.Published: December 2016 c Author(s) 2016.CC-BY 3.0 License.and quantified the impact of SSR and precipitation on the temperature.The main results are as follows.The national mean warming rates of Ts-max, Ts-min, Ta-max, and Ta-min were 0.227, 0.315 °C 10yr −1 , 0.167 °C 10yr −1 , and 0.356 °C 10yr −1 , respectively, from 1960 to 2003.The warming rates of Ts-min and Ta-min were significantly greater than those of Ts-max and Ta-max (seeFigs.2 and Figs.3).Warming rates of Ts-max and Ta-max in South China and on the North China Plain were significantly lower than other regions (seeFigs.4).The spatial heterogeneity in the warm seasons was greater than in the cold seasons.During the study period, SSR decreased by −1.502 W m −2 10yr -1 in China, with higher dimming rates in South China and the North China Plain.Using partial regression analysis, we found that SSR was the primary cause of the spatial pattern in the warming rates of Ts-max and Ta-max.After adjusting for the impact of SSR and precipitation, the warming trend of Ts-max increased by 0.146 °C 10yr −1 and that of Tamax increased by 0.055 °C 10yr −1 .After adjustments, the trends of Ts-max, Ts-min, Ta-max, and Ta-min became 0.373 °C 10yr −1 , 0.315 °C 10yr −1 , 0.222 °C 10yr −1 , and 0.356 °C 10yr −1 .The reduction of SSR resulted in the warming rates of Ts-max and Ta-max decreasing by 0.139 °C 10yr −1 and 0.053 °C 10yr −1 , accounting for 95.0% and 95.8%, respectively, of the total impact of SSR and precipitation.Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2016-1022,2016 Manuscript under review for journal Atmos.Chem.Phys.Published: December 2016 c Author(s) 2016.CC-BY 3.0 License.