Reconstruction of winter temperature of southwest China over the past 300 years based on a Bayesian approach

Abstract. A Bayesian approach was applied to reconstruct winter temperatures in southwestern China from 1700 to 1949, using 1950–1999 as the reference period. Within this methodological framework, documentary data provided weather and climate information to generate the Cold Weather Index (CWI; 1 = warm, 6 = extremely cold), two paleoclimate simulation ensembles served as priors, and uncertainties in the documentary records together with the dependence of observations on climate contributed to the likelihood estimates. The reconstructed CWI identified 20 extremely cold, 34 cold, and 15 warm winters, with the 1890s emerging as the coldest decade of the past three centuries due to a succession of severe winters. The posterior reveals pronounced interannual variability with an amplitude of ~ 2.77 °C, slightly smaller than existing reconstructions from an individual station, where the coldest winter was ~ 2 °C colder than the reference period. On longer timescales, the reconstruction captures a cold phase in the latter part of the nineteenth century and the warming in the twentieth century. An alternative reconstruction using time-independent priors demonstrate the capacity of the approach to disentangle the contributions of simulations and documentary evidence. This study provides a new regional climate reconstruction for southwestern China and highlights the potential of Bayesian approach for obtaining climate reconstructions from documentary climate data.