Reconstruction of Daily Weather in the 1810s by Diary Data Assimilation

Large-scale volcanic eruptions have a profound impact on climate. The Tambora eruption in April 1815 was among the largest in recent history. This eruption led to a global temperature reduction in the following year, commonly known as "the year without summer". Due to the limited availability of meteorological measurements before 1900, the atmospheric condition during the pre- and post-eruption years has been rarely investigated. In Japan, the Historical Weather Database (HWDB) provides descriptive daily weather information recorded in diaries at many locations since the 1660s. This study reconstructs daily weather during 1813-1818 by assimilating the descriptive weather information for the first time. We use the Global Spectral Model (GSM) as a forecasting model, and a local ensemble transform Kalman filter (LETKF) as an assimilating scheme. We first categorize the descriptive records into "sunny", "cloudy", and "rainy". These categorized data are then transformed into a Gaussian distribution to satisfy the error distribution assumption in LETKF. We use documentary evidence and the 20th Century Reanalysis (20CR) dataset to validate the reconstruction results. The reconstructed cloud cover by assimilating weather categories corresponds well with the diary observations in Japan, with an averaged correlation coefficient of 0.82 in summer 1813-1818. The surface air temperature anomaly indicates significant cooling in 1816, demonstrating the climate response to the Tambora eruption. This study shows the potential of diary data assimilation to reconstruct daily historical weather, providing the basis to understand the cause of short-term variability in the past climate.