[M-GI26] Data assimilation: A fundamental approach in geosciences

[MGI26-04] Ultra-High Spatiotemporal Resolution Reconstruction of the Climate around Japan with Large Ensemble of Climate Simulations and Old Documents

*Atsushi Okazaki¹, Satoshi Watanabe², Koji Ito^1,2, Yoshinori Tajiri² (1.Chiba University, 2.Kyushu University)

Keywords:Data assimilation, Old documents, d4PDF, Particle filter

Many of the old documents in Japan, whether it be private or public, contain the records of weather types (e.g., sunny, cloudy, rainy) on a daily basis. Previous studies have used such records to reconstruct monthly to seasonal mean climate at the point where the document was written. Only a few studies reconstructed point-wise climate on a daily scale due to the difficulty of using such qualitative data. Climate field reconstructions with old documents are also limited.

This study proposes a data assimilation-based approach to use such data effectively for daily weather field reconstruction. We use the particle filter and a pre-existing large ensemble climate simulation known as d4PDF as a model prior ensemble. The d4PDF provides simulations that are longer than 3000 years with horizontal resolution of 20 km and enables ultra-high resolution climate reconstruction in the field of paleoclimate.

In the presentation, we show the feasibility of the approach for the present climate using AMeDAS data. We found that the prosed method reconstructs daily precipitation fields successfully. The reconstruction skill is robust to the number of observations to some extent, suggesting the feasibility of the weather reconstruction with actual weather daily data.