Abstract
A method was developed, based on Bayesian model averaging (BMA), to reconstruct regional mean temperature. Different from the arithmetic mean, which gives equal weight to each chronology, BMA weights the chronologies according to their contributions to the actual temperature variances. Thus, BMA holds advantages in integrating chronologies to reconstruct the regional mean temperature. The regional mean temperature for the Yunnan–Guizhou Plateau was reconstructed for the past four centuries (1628–2005) using BMA, which performed better than the simple arithmetic mean. The reconstruction explained 41.33% of total observed temperature variances during the period of 1961–2005. The warmest decade was found to be 1840–50 and the coldest 1810–20 prior to the instrumental period. The reconstructed temperature showed a high correlation (r > 0.7, p < 0.001) with gridded observed temperatures in most grid cells of the Yunnan–Guizhou Plateau, suggesting that the regional temperature changes were captured well by the reconstruction.