Surface Temperature Reconstruction Over North Hemisphere During Last Millennium:Merging Method At Multi-Scales

Studies on surface temperature over the North Hemisphere during the last millennium are greatly important for understanding the mechanisms of climate change over the past hundreds of years,for evaluating global warming in the 20^th century,and for projecting future climate.This study aims to reconstruct surface temperature based on proxy records?most of which are tree ring chronologies?and merge these records with climate system model outputs by applying multi-scale methodologies and data assimilation techniques.The dissertation includes the following aspects.1)A new climate field reconstruction method was proposed.The ensemble empirical mode decomposition?EEMD?method combined with power spectral analysis was applied as core process to extract the signals at different timescales.Pseudo proxy tests?PPEs?have proved that the reconstruction method is a significant improvement over the traditional method.2)The merging method based on optimal interpolation?OI?algorithm was proposed to merge the CCSM4 outputs and tree ring chronologies.The results of verification tests over North America indicated that the standard deviation of errors gets lower than tree-ring reconstructions and model simulation by 0.3°C and about 0.5°C,respectively.In addition,the simulated decadal warm–cool periods were readjusted in the merged results to make the decadal variability more reliable?e.g.,the 1940–60s cooling?.3)The merging method was further developed by applying the typical mode decomposition.Two datasets were reconstructed based on CMIP5/PMIP3 model ensemble and Last Millennium Ensemble?LME?by CESM1.1 at resolutions of 5?by 5?and 2?by 2?,respectively.The temperature errors at annual and decadal timescales were reduced by 0.17 and 0.11? respectively.4)Disturbing tests showed that more than 60%of error variance was shared by annual mode that is much higher than other components.5)Both datasets indicate that the global warming during the 20^th century was unprecedented,and the anomaly is about 0.36? with respect to average temperature during the little ice age?1451-1850 AD?.The response to the volcanic eruption during the last millennium in the merged datasets become medium between pure model simulation and proxy records based reconstruction.The verification tests have shown that the greatest improvement lies in the component at inter-annual timescale,and the improvement gets less significant as timescale increases.6)By applying the time averaged data assimilation method and adopting the forward model as observation operator,the surface temperature field over Tibetan Plateau was reconstructed at monthly scale,and PPEs turned out to show clear improvements over the pure model simulation,and the errors of monthly mean could be reduced 0.2⁰.6?.To sum up,the reconstruction and merging methods proposed in this study was an attempt to combine proxy records and climate model and results of a number of PPEs have proved their feasibility and reasonability.Through the multi-scale methodology,the contributions from components at different timescales have been made clear.However,the problem concerning the low frequency variability still remains,and more explorations on the relationship between the model and proxy data are need in future.