1. Introduction

Associated with global warming, the occurrence and durations of drought show an increasing tendency. Its impact is becoming more and more severe in the region of eastern China and its occurrences are of great concern ( Huang et al., 2015 ; Dai, 2012 ; Zou et al., 2005 ;Jiang et al., 2012 ). Because observational datasets are too short to assess the temporal variability of hydroclimate under the “natural ”and“anthropogenic ”condition and to address the decadal—centennial variability and mechanisms for predicting drought change in the future, understanding the decadal—centennial hydroclimate variability and mechanisms during the last millennium has attracted increasing attention.

Estimates for the climate of the last millennium are usually either based on proxy records obtained from natural archives and historical documents or on climate model simulations with a variety of climate forcing factors. Based on historical documents, tree rings, and speleothems, a number of high-resolution hydroclimate reconstructions over eastern China during the last millennium have been carried out in recent years ( Zhang, 2004 ; CNMA, 1981 ; Zheng et al., 2006 ; Cook et al.,2010 ; Yang and Tan, 2009 ; Man, 2012 ). Work to reconstruct the drought status over eastern China has revealed that severe large-scale droughts on decadal—centennial timescales have occurred many times during the last millennium ( Zhang, 2004 ; Zheng et al., 2006 ; Cook et al., 2010 ). Although paleoclimate proxy data provide a critical reference history for hydroclimate over eastern China, interpretation is challenged by considerable errors: different statistical methods used in the reconstructions lead to different results, non-climatic factors influence the proxies, and the poor spatial coverage of the climate proxies lead to errors in large-scale means ( Jansen et al., 2007 ; Jones and Mann, 2003 ;Matsikaris et al., 2015 ). Moreover, a lack of dynamic variables and complete physical constraints does not allow for diagnostic studies of past hydroclimate variability ( Steiger et al., 2018 ). However, global comprehensive datasets that are consistent with atmospheric dynamics can be provided by climate model simulations.

Based on modeling results, considerable efforts have been made to examine precipitation changes over eastern China and possible explanatory mechanisms during the last millennium ( Liu et al., 2011 ; Man et al.,2012 , 2014 ; Ning et al., 2017 , 2019 ; Shi et al., 2016 ; Peng et al., 2010 ,2014 , 2015 ; Peng, 2018 ; Yan et al., 2015 ; Zhou et al., 2020 ). These results show that changes in solar radiation, volcanic activity, the El Nino—Southern Oscillation (ENSO), Interdecadal Pacific Oscillation, North Atlantic Oscillation, and Atlantic Multidecadal Oscillation (AMO) have played important roles in precipitation over eastern China throughout the last millennium. Because of model bias ( Zhou et al., 2017 ), errors in the forcings and the uncertainties caused by the models of chaotic components of internal variability ( Peng et al., 2014 ), models cannot reproduce the exact variability to replicate the information obtained from proxy data, and the interpretation of such simulations is challenged. For example, the modeled linkage between ENSO and severe drought over eastern China during the last millennium remains to be revealed. Our previous work ( Peng et al., 2014 ) showed that a temporally consistent relationship between the drought and sea surface temperature (SST) pattern in the Pacific Ocean could not be found. Meanwhile, another modeling analysis, the ensemble mean of the CESM (Community Earth System Model)—LME (Last Millennium Ensemble) simulations, shows that ENSO plays an important role in precipitation variability over eastern China( Peng et al., 2019 ; Ning et al., 2019 ). It suggests a temporally consistent relationship between drought and negative phases of ENSO can be found in model data ( Peng et al., 2019 ; Ning et al., 2019 ). Thus, it is unclear if the underlying climate dynamics of such events are accurately produced in climate model simulations. This issue is due in part to a lack of constraints from the known proxy record ( Steiger et al., 2018 ).

There is growing interest in applying data assimilation to estimate pre-instrumental climate. Data assimilation provides a mathematical framework that combines the empirical information from proxy data with the representation of the dynamical processes that govern the climate system given by climate models ( Matsikaris et al., 2015 ). Therefore, data assimilation is both consistent with the model physics and with the empirical knowledge. It provides clear hypotheses that can be tested with additional model simulation and independent proxy data and checks the physical consistency of the proxy-based this study, the Paleo Hydrodynamics Data Assimilation (PHYDA)product is used to elucidate more clearly the mechanisms of the decadal—centennial hydroclimate variability over eastern China during the last millennium.