Difference and Linkage in Climate Change Modes from East and Central Asia at Multi-Time Scales

Previous studies argued that climate change modes from East and Central Asia (EA and CA) are out of phase at multi-time scales. However, in recent years, dry/wet changes in CA which contradict traditional views have provoked further discussion. The synchronization of rain and heat periods is a common climate phenomenon in most regions of East and Central Asia. In this paper, we selected EA and CA to carry out a comprehensive study of modern observations, paleoclimate records, and model simulations at multi-time scales. EOF analysis results of modern grid precipitation and self-calibrating Palmer Drought Severity Index (scPDSI) demonstrate the synchronization of rain and heat periods in EA and the east of CA at the short-term timescale. Meanwhile, paleoclimate records indicate parallel dry/wet changes in EA and the east of CA since the Last Glacial Maximum (LGM), also reecting the synchronization of rain and heat periods at long-term timescales triggered by the insolation. The climate mechanism of difference and linkage in climate change modes from EA and CA, under the framework of the synchronization of rain and heat periods, is analyzed by PMIP3 simulations between the LGM and Mid-Holocene (MH). Overall, we suggest that, in addition to the regional differences caused by different circulation systems (the westerlies and Asian summer monsoon), climate change modes in EA and CA universally have inter-regional connections affected by the synchronization of rain and heat periods at multi-time scales. means northward wind; zg geopotential Height; hus near-surface relative humidity; psl means sea surface pressure; pr means precipitation; tas means near-surface temperature

The synchronization of rain and heat periods is an important phenomenon at the multi-time scale climate change in EA and CA, behaving as that the summer half-year at the short-term timescale and warm period at the long-term timescale has more precipitation than the winter half-year and cold period respectively. In this study, utilizing modern observations, paleoclimate proxies, and model simulations, we conducted a comprehensive analysis for the differences and connections of climate change modes at multi-time scales between EA and CA based on the synchronous phenomenon of rain and heat periods and reasonably explain the "pattern differences" and "spatial connections" between the two regions. Empirical orthogonal function (EOF) is a powerful method for dimensionality reduction and pattern extraction. EOF can decompose the multidimensional climate data from different locations into spatial (EOF modes) and temporal functions (principal components). Therefore, to investigate the spatiotemporal variations of precipitation at the interannual timescale over EA and CA, the EOF analysis was applied to the gridded precipitation data and scPDSI. We focused on the rst two leading modes that objectively account for the majority of dry/wet changes in EA and CA (Lorenz, 1956).

Regional and global paleoclimatic proxy data
Here we compiled various LGM paleoclimate records to reconstruct long-term climate variability and primarily paid close attention to paleo-precipitation and moisture changes since the LGM. The δ 13 C, δ 18 O, TOC, C/N was mainly selected to indicate the climate changes. As all, in this paper we have selected records based on three criteria: (1) the selected records must have a reliable chronology. (2) The record length should cover most of the LGM without documented depositional hiatuses. (3) Proxies should have clear meanings showing dry/wet status. Following the above criteria, we collect typical proxy data from various paleoclimate records in EA and CA (Table 1 and Fig. 1a) since the LGM.

Synchronization of rain and heat periods at the shortterm timescale
We calculated the precipitation difference between the summer (April, May, June, July, August, and September) and winter (January, February, March, October, November, and December) half year over 1965-2014, and then de ned the region greater than 0 mm as the synchronous region of rain and heat periods (Fig. 1, gray slash). The East Asian summer monsoon boundary was depicted by the seasonality in precipitation according to the de nition of the summer monsoon index by Wang et al. (2012) (Fig. 1, red line). Based on this, we de ned the monsoon region as EA, and the non-monsoon region as CA, in particular, the asynchronous region of rain and heat periods is the core region of CA (Fig. 1). It can be seen that the synchronous region of rain and heat periods covers most of the non-monsoon region (i.e., the east of CA). To obtain the spatial distribution characteristics of the precipitation and scPDSI anomalies in EA and CA under the context of synchronization of rain and heat periods, we conducted an EOF analysis on the precipitation and scPDSI standardized anomaly eld over 1965-2014. The center of positive values is in the CA core mainly belonged to the asynchronous region of rain and heat periods, while the center of negative values is in the south of EA located in the synchronous region of rain and heat periods (Fig. 2a). The opposite signals indicate that the mean annual precipitation changes in the asynchronous and synchronous regions show a see-saw pattern. The rst mode exhibits the interdecadal and interannual changes according to the PC1 (Fig. 2b). The variance contribution rate of the second mode is 9.08%, indicating zonal dipole distribution characteristics (Fig. 2c). The center of positive values is in the north of CA, and the center of negative values is in the east of CA and northeast of EA, displaying the spatial diversity of climate change in the synchronous region of rain and heat periods (Fig. 2c). The PC2 shows that the second mode has signi cant interdecadal characteristics (Fig. 2d). Alternatively, we analyzed the spatial distribution and time series of EOF decomposition of precipitation difference between summer and winter half-year ( Fig. 1e-h). The variance contribution rate of the rst mode is 10.47%, which is a meridional bipolar distribution (Fig. 2e). The center of the negative value is located in the north of EA and CA, according with the synchronization of rain and heat periods in the non-monsoon region. The variance contribution rate of the second mode is 8.35%, displaying a dipole mode (Fig. 2e). The positive load in EA illustrates the effect of the synchronization of rain and heat periods (Fig. 2e). Simultaneously, the PC1 and PC2 exhibit that in contrast to mean annual precipitation, the amplitude of dry/wet changes and the characteristics of interannual changes in EA and CA are more prominent (Fig. 2f and h).
The variance contribution rate of the rst mode of mean annual scPDSI is 13.58%, showing an obvious dipole mode (Fig. 3a). The centers of positive values are mainly distributed in the core and northern part of CA, while the negative values are mainly distributed in EA and the east of CA (Fig. 3a). This spatial distribution also shows that EA and the east of CA, both of which belong to the synchronous region of rain and heat periods, have the same dry/wet status, in contrast to the core and northern part of CA. The rst mode exhibits the interdecadal changes according to the PC1 (Fig. 3b). The variance contribution rate of the second mode is 11.58% (Fig. 2c), with the center of positive values in EA and the east of CA. The PC2 shows that the second mode has signi cant interdecadal characteristics opposite to the mean annual precipitation (Fig. 3d and 2d). The variance contribution rate of the rst mode of scPDSI difference between summer and winter half-year is 8.63%, which is a bipolar distribution (Fig. 2e). The center of the positive value is located in the north of EA and northeast of CA, indicating the generally consistent dry/wet change in the synchronous region of rain and heat periods, which is different from the core region of CA. The variance contribution rate of the second mode is 8.35%, displaying a zonal dipole mode between the synchronous region of rain and heat periods and the core region of CA (Fig. 2e). The PC1 and PC2 of scPDSI difference between summer and winter half-year the intense amplitude of dry/wet changes and interannual changes ( Fig. 3f and h).
In all, the spatial distribution and time series of EOF decomposition of mean annual scPDSI are signi cantly similar to that of precipitation, collectively indicated that EA and the east of CA, which belong to the synchronous region of rain and heat periods, have similar dry/wet changes. Therefore, we suggest that synchronization of rain and heat periods is an important factor linking climate change mode in EA and CA at the short-term timescales.

Synchronization of rain and heat periods at the longterm timescales
In the last decade, several paleoclimate records with a relatively high resolution, reliable chronology, and unambiguous proxies have been published to discuss the long-term timescale climate evolution in EA and CA. According to the model simulation, Yu et al. (2000) concluded that the low temperature in the cold period causes the decreasing evaporation, with the enhanced westerlies driven by expanding land ice sheets, forming the high lake level in western China and the low lake level in eastern China during the LGM. The reconstructed precipitation covering the past 22,600 years from Achit Nuur suggests the wet periods from 22,600 to 13,200 cal BP   (Fig. 4h). In the Holocene, pollen record of the Caspian Sea displays that the terrestrial vegetation around the Caspian Sea changed from desert and desert steppe during the last glacial to dry shrubland and forest from the early-Holocene to late-Holocene, revealing the continuous wetting process since the early Holocene and the wettest late-Holocene (Leroy et al., 2013) (Fig. 4j). Meanwhile, results of climatically-sensitive magnetic properties from the Xinjiang loess demonstrate that the relatively moist conditions are generally formed after ~6,000 cal BP, with the wettest climate occurring during the late Holocene, and that a dry climate prevailed during the early Holocene (Chen et al., 2016) (Fig. 4i). To sum up, most views believed that the climate change mode in CA (non-monsoon region) affected by the westerlies, characterized by wet climate conditions during the LGM and mid-and late-Holocene (Fig. h-j), is opposite to that in monsoon-dominated EA (Fig. 4b).
However, there are still partially contradictory long-term timescales dry/wet changes in CA, which are different from the climate change mode in CA but similar to that in EA. Herzschuh. (2006) comprehensively analyzed 75 paleoclimatic records in CA and revealed that wet conditions occurred during early-and mid-Holocene, while the LGM was characterized by dry climate conditions in the region (Fig. 4c), indicating the similarity with monsoon climate represented by the speleothem δ 18 O records from Dongge Cave and Hulu Cave (Fig. 4b). High precipitation observed between early-and mid-Holocene, indicated by δ 18 O records of ostracod shells from Qinghai Lake, shows that the climate in Qinghai Lake since the late Glacial re ects the monsoon change (Liu et al, 2007) (Fig. 4e). The climate in Ulaan Nuur was most humid during the early Holocene, humid during the mid-Holocene and dry in the late-Holocene, embodying a typical characteristic of the East Asian monsoon (Lee et al., 2013) (Fig. 4f). Based on a sediment core from Lake Karakul, the early-to mid-Holocene is characterized by moister conditions in the region, and the lake level remained low during the LGM (Heinecke et al., 2017) (Fig. 4d). Furthermore, the regional climate in western China, inferred from speleothem oxygen-carbon isotope in Kesang Cave, suggests a close coupling with the Asian summer monsoon (Cheng et al., 2016) (Fig. 4g). The lake level and climate reconstructed results also suggested that the cold-dry LGM climate triggered a substantial lowering of lake level in most of arid western China, challenging the traditional view of cold-wet climate and high lake levels in arid western China during the LGM (Zhao et al., 2015).
Considering the spatial distribution of the above paleoclimate records, we found that the Central Asian records similar to the monsoon evolution are located in the modern synchronous region of rain and heat periods (Fig. 1a). At the same time, the dry climate condition during the LGM and the wet climate change mode during the early-and mid-Holocene also re ect another meaning of the synchronization of rain and heat periods at long-term timescales triggered by the insolation (Fig. 4a), namely the dry-cold period and wet-warm period.
3.3 Climate mechanism of difference and linkage in climate change modes from EA and CA at the longterm timescale The results of paleoclimate simulations explain the asynchrony of the long-term climate change mode in EA and CA and the climate linkages under the framework of the synchronization of rain and heat periods at the long-term timescale. During the LGM, lower summer insolation increases the meridional difference of temperature and sea level pressure in the summer half-year (rain period) large ( Fig. 4a; 5a-b), leading to the strengthening of the westerly jet stream and further increasing the precipitation in the core region of CA ( Fig. 5c and h). Given the weakening of the LGM summer monsoon and the complex control factors (Fig. 5g), however, the precipitation in the east of CA is weaker than that of MH (Fig. 5c), which is consistent with the climate change mode in EA and re ects climate linkage between CA and EA caused by synchronization of rain and heat periods. Although the westerly jet stream weakens in the LGM winter half-year (Fig. 5i), the higher winter insolation contributes to the general warming in CA and EA ( Fig. 4a; 5d), resulting in lower relative humidity (Fig. 5e). According to climatological theory (Barry and Richard, 2009), a decrease in relative humidity means an increase in saturated water vapor pressure, which ultimately leads to an increase in precipitation (Fig. 5f). Therefore, this elaborates the asynchrony of the long-term climate change mode in EA and CA under the background of synchronization of rain and heat periods.

Conclusion
Using EOF method, this study analyzes the spatiotemporal variations of precipitation and short-term scPDSI in EA and CA. Results reveal that the synchronization of rain and heat periods is an important factor linking climate change modes in EA and CA at short-term timescales. Concurrently, paleoclimate records re ect the synchronization of rain and heat periods at long-term timescales triggered by the insolation. The model simulations of multiple climatic elements explain the climate mechanism of differences and linkages in climate change modes from EA and CA. In the context of asynchronous dry/wet changes between monsoon-and the westerlies-dominated regions, we believe that regional linkages also exist in EA and CA. Figure 1 The difference between summer and winter precipitation over 1965-2014 (shade), and the major trajectories of atmospheric circulation. The gray slash represents the synchronous region of the rain and heat periods, the red line represents the locations of the climatological summer monsoon boundary over