A winter temperature reconstruction based on the Bayesian approach in southwest China during the 18th-19th century

Some classic approaches to climate reconstruction, such as employing transfer functions, have stringent requirements for the quality and continuity of historical records. As a result of limited data sources and complex topography, it’s hard to estimate the winter temperature in southwest China. However, the Bayesian approach allows integrating probabilities into the temperature indices and assimilating documentary information with different uncertainties with climate modeling data. Some gap years with less or even absent information in narrative sources could also be evaluated.

Based on Bayes’ theory, a large-scale simulation ensemble containing 20 members called ModE-Sim serves for the estimate of prior atmospheric states. The documentary data, including abnormal phenomena records in the local history and official reports of precipitation, contain various information on snow, rainfall, flower phenomena, and personal feelings. They are used to generate Indices from extremely cold to warm as well as the associated likelihood of each winter and contribute to the reconstruction of the posterior probability. Finally, a series of winter temperature with uncertainty in southwest China during the 18^th-19^th century is generated from the prior and posterior probability.

This new 200-yr reconstruction in this study fits well with an independent dataset called ModE-RA, which is a global monthly reanalysis also employing ModE-Sim as the background state of the atmosphere. The cold winters of 1700/1701, 1783/1784, 1809/1810, and 1892/1893 and the cold late 19^th century are expressed in this reconstruction. This study presents a new application of the Bayesian approach in the historical climatology field and has the potential to contribute to the analysis of large-scale circulation in past winters in China.