A new last two millennium reanalysis based on hybrid gain analog offline EnKF and an expanded proxy database

Abstract

This paper systemically assesses the performances of two assimilation methods, i.e., the Offline Ensemble Kalman Filter (OEnKF) and the Hybrid Gain Analog Offline EnKF (HGAOEnKF) with three proxy databases, on reconstructing the temperature and precipitation during the last two millennia. The results show that, among three databases, increasing the number of proxy records significantly improves the reconstruction skill for both assimilation methods, with a larger improvement in HGAOEnKF. In the instrumental era, six reconstructions have comparable skill (similar correlation coefficients, CEs, and RMSEs) when validated against out-of-sample proxy records and instrumental reanalyses. During the pre-industrial era, HGAOEnKF shows better assimilation performance by improving the background field of assimilation when the number of proxy records is limited. Compared with the temperature reconstruction, the skill of precipitation reconstruction is relatively lower. The proxy records from the ocean contribute more to the temperature reconstruction skill with both assimilation methods. Finally, a new reanalysis product (NNU-2ka Reanalysis) is generated through the HGAOEnKF with the expanded proxy database.