Temperature and precipitation are two of the most important variables describing climate ( Isaac and Stuart, 1992). Observed changes in regional temperature and precipitation can often be physically related to one another (e.g., Huang and Van den Dool, 1993; Trenberth and Shea, 2005; Adler et al., 2008). The relationships between temperature and precipitation have been investigated in various regions (e.g., Madden and Williams, 1978; Isaac and Stuart, 1992; Stuart and Isaac, 1994; Zhao and Khalil, 1993; Trenberth and Shea, 2005; Jones et al., 2007; Wu et al., 2013). The results showed that their relationships strongly depend on the region and season. For example, over tropical land, part of the subtropics and mid-latitudes, the correlations between annual mean temperature and precipitation appear to be negative, while in the higher latitudes the correlations are positive. In addition, a strong seasonal variation seems to appear in the Northern Hemisphere higher latitudes ( Déry and Wood, 2005).

Also, daily maximum air temperature ( T_max) and minimum air temperature ( T_min) have been documented to physically relate to precipitation at a variety of space and time scales (e.g., Dai et al., 1997, 1999; Power et al., 1998; Nicholls, 2004; Zhou et al., 2009). The daytime T_max is strongly affected by surface solar heating and the partitioning of sensible and latent heat fluxes, while the nighttime T_min is largely controlled by net longwave radiation ( Dai et al., 1999). Precipitation and associated changes in clouds, soil moisture, and other land surface conditions can modify the surface energy balance, thus affecting the daytime T_max and nighttime T_min (e.g., Dai et al., 1999; Zhou et al., 2007, 2009; Zhang et al., 2008a, 2009; Zhang and Wang, 2008; Wu and Zhang, 2013). Changes in T_max and T_min play important roles in influencing precipitation by modifying interactions among land surface, atmospheric boundary layer, and clouds ( e.g., Betts., 2004).

The 1976-1977 climate regime shift has been widely recognized (e.g., Francis and Hare, 1994; Miller et al., 1994; Hare and Mantua, 2000; Marcus et al., 2011; Sabeerali et al., 2012). Precipitation over East Asia has been reported to undergo significant changes after the climate shift in the late 1970s (e.g., Huang et al., 1999; Wang, 2001; Li et al., 2002; Wu and Wang, 2002; Lu, 2003; Dai et al., 2003; Hu et al., 2003; Zhao and Zhou, 2006; Ye, 2013). In connection with a weaker Asian summer monsoon, the summer precipitation increased over central China, the southern part of northeastern China, and the Korean peninsula, while decreased over North China after the climate shift (e.g., Huang et al., 1999; Wang, 2001; Gong and Ho, 2002; Ho et al., 2003; Ding et al., 2008). In winter, precipitation significantly decreased over the northern part of eastern China but increased in the South (e.g., Zhai et al., 2005; Zhang et al., 2008b). Although the observed changes in precipitation over East Asia have been widely investigated, changes in the covariability between surface air temperature and precipitation associated with the late 1970s climate shift are not yet well understood. The observed relationships between surface air temperature and precipitation are useful for further testing climate simulations of the relationships between surface air temperature and precipitation (e.g., Isaac and Stuart, 1992; Stuart and Isaac, 1994; Trenberth and Shea, 2005; Wu et al., 2013). Here, we explore this issue using observed gridded temperature and precipitation data. Besides mean surface air temperature ( T_mean)-precipitation covariability, we are also motivated to examine if the covariability of precipitation with T_max versus T_min respond differently to the climate shift because surface fluxes associated with large variation of solar radiation are quite different between day and night.

We used a gridded surface air temperature and precipitation dataset provided by the Climate Research Unit Time Series (CRU TS 2.1), University of East Anglia, Norwich, United Kingdom ( Mitchell and Jones, 2005). The dataset contains information on observed surface air temperatures and precipitation over land at a resolution of 0.5° grid and includes T_mean, T_max, T_min, and precipitation on a monthly basis from 1900 to 2002.

To estimate the differences in the covariability between surface air temperature and precipitation before and after the climate shift, we split 1953-2000 data into the pre- (1953-1975) and post-shift period (1978-2000) groups. Previous studies (e.g., Francis and Hare, 1994; Miller et al., 1994; Hare and Mantua, 2000; Sabeerali et al., 2012) have presented that 1976 and 1977 are transition years in the climate shift. To clearly identify the differences in the covariability between surface air temperature and precipitation before and after the climate shift, the two years were excluded in the study. We computed monthly anomalies on a 0.5° grid for the pre- and post-shift periods, and combine the months from December, January, and February as winter season months, and June, July, and August as summer season months, so there are 23 years and a total of 69 values in each correlation computed for pre- and post-shift periods, respectively. The value of 95% confidence level is 0.24, and 0.31 for the 99% confidence level (Two-tailed). Negative correlation coefficients indicate warm and dry or cool and wet conditions, whereas positive correlation coefficients represent cool and dry or warm and wet conditions.

Firstly, we analyzed spatial patterns of the 1978-2000 minus 1953-1975 differences of precipitation (Fig. 1). In winter, after the climate shift precipitation largely decreased over Mongolia, much of northern China, Korea, and Japan, while increased over South China and many areas of Northeast Asia. In summer, the precipitation differences exhibited a negative-positive-negative pattern over China, with increased precipitation in the Yangtze River valley and decreased precipitation over northern and South China. Over Korea and part of Northeast Asia precipitation increased, while precipitation decreased over much of Japan. Previous studies ( Gong and Ho, 2002; Lu, 2003; Ding et al., 2008) used monthly precipitation data from the China Meteorological Administration to analyze variations of precipitation in summer over China for last fifty years ( since 1951). They found that the North China precipitation showed a remarkable decrease and there was excessive precipitation over the Yangtz River valley after the 1970s climate shift. In general, our results in the present study are consistent with previous studies. Also, some differences are seen over some areas ( Zhang et al., 2008b).

Figure 1

	Figure Option ViewDownloadNew Window
	Figure 1 Precipitation difference in percentage between 1978-2000 and 1953-1975 with respect to the 1953-1975 mean in (a) winter and (b) summer.

We then looked at how the covariability between temperature and precipitation responds to the climate shift. Figure 2 presents the correlations between T_mean and precipitation during the pre- and post-shift periods and the differences between the two periods. In winter, significantly negative correlations mainly occurred over North China and Mongolia, while significantly positive correlations appeared over Korea, much of Japan and some isolated areas in southern China during the pre-shift period. After the climate shift, negative correlations occupied more areas over Mongolia and China. In contrast, in summer significantly negative correlations almost covered the entire East Asia during the pre-shift period. After the climate shift, the negative correlations were mostly weakened, while negative correlations were only enhanced in some isolated areas. These results indicate that the covariability between T_meanand precipitation experienced remarkable changes over certain areas of East Asia after the climate shift with evident seasonal dependencies.

Figure 2

	Figure Option ViewDownloadNew Window
	Figure 2 Correlations between mean surface air temperature and precipitation for groups of winter (upper panels) and summer (lower panels) months: (a), (d) 1953-1975; (b), (e) 1978-2000; (c), (f) differences between 1978-2000 and 1953-1975. The value of ±0.24 represents that the correlations are above 95% confidence level.

The correlations between T_max and precipitation during the pre- and post-shift periods and their differences are shown in Fig. 3. In winter, the correlations between T_max and precipitation north of 40°N showed similar features to those of T_mean-precipitation during both the pre- and post- shift periods. South of 40°N, negative correlations for T_max-precipitation over China were significantly stronger and more expansive than those of T_mean-precipitation. In summer, negative correlations between T_max and precipitation also covered almost entire East Asia, and were much more significantly than those of T_mean-precipitation. The spatial patterns of the changes in the covariability between T_max and precipitation showed similar features to those of T_mean-precipitation in both winter and summer.

Figure 3

	Figure Option ViewDownloadNew Window
	Figure 3 Correlations between daily maximum surface air temperature and precipitation for groups of winter (upper panels) and summer (lower panels) months: (a), (d) 1953-1975; (b), (e) 1978-2000; (c), (f) differences between 1978-2000 and 1953-1975. The value of ±0.24 represents that the correlations are above 95% confidence level.

By contrast, T_min demonstrated a different spatial correlation pattern with precipitation from that of T_mean (Fig. 4). During the pre-shift period, in winter positive correlations dominated the areas south of 40°N, while negative correlations were mostly observed over the north, with significant correlations appearing over Mongolia. After the climate shift, the positive correlations in China shrank to smaller areas with large decreasing over southern China. On the other hand, the negative correlations expanded to larger areas, and the southern edge of significant correlations shifted northward near 43°N. Seasonal variations of the covariability between T_min and precipitation over East Asia were also evident during the pre- and post-shift periods. The relationships in summer were mostly reversed compared with those in winter. During the pre-shift period, significantly negative correlations appeared over the areas south of about 35°N in China, Korea, and much of Japan, whereas positive correlations dominated the areas north of about 35°N. After the climate shift, positive changes were observed in the Yangtze-Huai River valley and in Korea, while negative changes were found over South China and Japan and northern part of East Asia. As a result, the changes in T_min exhibited a negative-positive-negative pattern over East Asia.

Figure 4

	Figure Option ViewDownloadNew Window
	Figure 4 Correlations between daily minimum surface air temperature and precipitation for groups of winter (upper panels) and summer (lower panels) months: (a), (d) 1953-1975; (b), (e) 1978-2000; (c), (f) differences between 1978-2000 and 1953-1975. The value of ±0.24 represents that the correlations are above 95% confidence level.

Previous studies have demonstrated that temperature and precipitation are physically correlated. It has been also found that precipitation over East Asia has undergone noticeable changes since the climate shift occurred in the late 1970s. Considering the abrupt changes in precipitation, one might expect that the covariability of surface air temperature and precipitation should change after the climate shift. This study investigated the changes in the covariability between surface air temperature ( T_mean, T_max, and T_min) and precipitation associated with the climate shift by using 1953-2000 observed gridded data.

Our results show that the covariability between T_mean and precipitation experienced remarkable changes over certain areas of East Asia after the climate shift with evident seasonal dependencies. Before the climate shift, in winter significantly negative correlations between T_mean and precipitation mainly occurred over North China and Mongolia, while significantly positive correlations existed over Korea, Japan, and some isolated areas in southern China. After the climate shift, significantly negative correlations occupied more areas over Mongolia and China. In summer, significantly negative correlations appeared over almost the entire East Asia region before the climate shift. After the climate shift, the negative correlations were mostly weakened, while only over some small isolated areas negative correlations became stronger.

For both the pre- and post-shift periods, the relationships of precipitation with T_max in winter showed similar features to those with T_mean over the areas north of 40°N, while significantly negative correlations with T_max covered larger areas south of 40°N. In summer, significantly negative correlations with T_max, which covered nearly all regions of East Asia, were generally stronger than those with T_mean during both pre- and post-shift periods. The changes in the relationships between T_max and precipitation appeared to be similar in both magnitude and spatial pattern to those of T_mean precipitation.

In contrast, the relationships between T_min and precipitation differed significantly from those of T_mean-precipitation. Before the climate shift, in winter positive correlations dominated the areas south of 40°N, while negative correlations mainly appeared over Mongolia and northern China. After the climate shift, the positive correlations over southern China were markedly weakened, whereas significantly negative correlations occupied more areas over Mongolia. In summer, a very different pattern was seen. Before the climate shift, negative correlations existed over areas south of 35°N in China, Korea, and much of Japan, while positive correlation dominated the north. After the climate shift, changes in the relationships between T_minand precipitation showed a negative-positive- negative pattern over East Asia, with positive changes in the Yangtze-Huai River valley and Korea, and negative changes over South China and Japan, and northern part of East Asia.

Here, we discuss the possible mechanisms for changes in the covariability between surface air temperature and precipitation after the late 1970s climate shift. Our results show that negative temperature-precipitation correlations dominated over East Asia in both winter and summer before and after the climate shift. More precipitation accompanies more clouds, leading to reduced downward shortwave radiation, subsequently decreases surface air temperature ( Trenberth and Shea, 2005; Wu et al., 2013). In addition, more precipitation leads to more soil moisture, providing more latent heat, and further reducing surface air temperature ( Zhang and Dong, 2010; Zhang et al., 2011). The two mechanisms are possibly responsible for negative temperature-precipitation correlations. The two mechanisms have much stronger effects on surface air temperature during the daytime than those during nighttime, thus resulting in larger T_max-precipitation correlations than those of T_min-precipitation ( Wu and Zhang, 2013). Positive temperature-precipitation correlations also existed over some areas, which may have been caused by positive soil moisture feedbacks and other physical processes ( Trenberth and Shea, 2005; Zhang and Dong, 2010). Changes in these mechanisms after the late 1970s may explain the changes in the covariability of temperature and precipitation found in the present study. The proposed mechanisms need to be further tested by model simulations.