Physical processes of cooling and megadrought in 4.2 ka BP event:

16 It is widely believed that multidecadal to centennial cooling and drought occurred from 17 4500 BP to 3900 BP, known as the 4.2 ka BP event that triggered the collapse of several 18 cultures. However, whether this event was a global event or a regional event and what 19 caused this event remain unclear. In this study, we investigated the spatiotemporal 20 characteristics, the possible causes and the related physical processes of the event using 21 a set of long-term climate simulations, including one all-forcing experiment and four 22 single-forcing experiments. The results derived from the all-forcing experiment show 23 that this event occurs over most parts of the Northern Hemisphere (NH), indicating that 24 this event could have been a hemispheric event. The cooler NH and warmer Southern 25 Hemisphere (SH) illustrate that this event could be related to the slowdown of the 26 Atlantic Meridional Overturning Circulation (AMOC). The comparison between the 27 all-forcing experiment and the single-forcing experiments indicates that this event 28 might be caused by internal variability, while external forcings such as orbital and 29 greenhouse gases might have modulation effects. A positive North Atlantic Oscillation 30 (NAO)-like pattern in the atmosphere (low troposphere) triggered a negative Atlantic 31 Multidecadal Oscillation (AMO)-like pattern in the ocean, which then triggered a 32 Circumglobal Teleconnection (CGT)-like wave train pattern in the atmosphere (high 33 troposphere). The positive NAO-like pattern and the CGT-like pattern are the direct 34 physical processes that lead to the NH cooling and megadrought. The AMO-like pattern 35 plays a “bridge” role in maintaining this barotropic structure in the atmosphere at a 36 multidecadal-centennial time scale. Our work provides a global image and dynamic 37 background to help better understand the 4.2 ka BP event. 38


Introduction
Understanding the causes and mechanisms of the 4.2 ka BP event can provide      The double peak centennial cooling and drought are still obvious when the 31-year 165 running mean is applied to the time series (not shown), which indicates that the 166 simulated climate events potentially comparable to the 4.2 ka event. Moreover, the 167 centennial warming periods right before and after the cooling event indicate that this 168 event might be included in a quasi-millennium variation. Therefore, the 4.2 ka BP event 169 could be a multiscale event, i.e. from multi-decadal to millennium.

170
The seasonal mean NH surface temperature changes show that the annual mean year running mean), which is significant above the 99% confidence level, much higher 175 than the correlation coefficient between the annual mean and the boreal summer (June-  To identify the characteristics of the 4.2 ka BP event, two centennial cool periods 181 and two centennial warm periods that exceeded ±0.5 standard deviations are selected.

182
The two centennial cool periods span from 4320 BP to 4220 BP and from 4150 BP to   warming over the SH could be related to the orbital change (Fig. S3), which induces 199 insolation increasing over the SH but decreasing over the NH.

200
The spatial distribution of annual mean precipitation differences between the cold 201 periods and the warm periods is shown in Fig. 3b. During the cold periods, significant 202 drought is mainly located over many land regions of the NH, especially over Europe,   The correlation coefficients between the annual mean NHT derived from the 285 TraCE-ALL run and the NHT derived from each single-forcing run are listed in Table   286 2. A two-sided Students t-test is used for the statistical significant test, assuming 20 where Ndof is the effective degree of freedom regarding to the two correlation samples,

310
On the other hand, the annual mean NHT difference between the TraCE-ALL run 311 and the sum of the 4 single-forcing sensitivity experiments shows variation similar to 312 the NHT derived from the TraCE-ALL run from 5000 BP to 3000 BP (Fig. S5). The 313 correlation coefficient between these two time-series is 0.66, which is significant above the 95% confidence level (assuming 20 degrees of freedom). We assume the difference 315 between the TraCE-ALL run and the sum of the 4 single forcing runs to be the internal  There is a reduction of NH temperature and precipitation at around 4600 BP in the 4000 BP shows cooler NH high latitudes and a warmer SH (Fig. S7), especially the 375 cooling over the northern North Atlantic Ocean, Europe, East Asia and North America.

376
The geopotential height at 200 hPa regressed against the SST over the two North 377 Atlantic outstanding regions (Fig. 4) shows a CGT-like pattern after application of a 378 31-year running mean (Fig. 9), which is similar to the conclusion from Lin et al.

394
We notice the centennial-millennial variation of the AMOC after the mid-

395
Holocene in the all forcing run (Fig. S4a). There also exits a double peak variation 396 during the period of 4.4-4.0 ka BP, accompanied by the similar spatial patterns of 397 temperature and precipitation anomalies as the simulated 4.2 ka BP event (Fig. S4b, c).