Reconstruction of Historical Weather by Assimilating Old Weather Diary Data
Abstract
Climate can control not only human life style but also other living beings. It is important to investigate historical climate to understand the current and future climates. Information about daily weather can give a better understanding of past life on earth. Long-term weather influences crop calendar as well as the development of civilizations. Unfortunately, existing reconstructed daily weather data are limited to 1850s due to the availability of instrumental data. The climate data prior to that are derived from proxy materials (e.g., tree-ring width, ice core isotopes, etc.) which are either in annual or decadal scale. However, there are many historical documents which contain information about weather such as personal diaries. In Japan, around 20 diaries in average during the 16th - 19th centuries have been collected and converted into a digitized form. As such, diary data exist in many other countries. This study aims to reconstruct historical daily weather during the 18th and 19th centuries using personal daily diaries which have analogue weather descriptions such as `cloudy' or `sunny'. A recent study has shown the possibility of assimilating coarse weather data using idealized experiments. We further extend this study by assimilating modern weather descriptions similar to diary data in recent periods. The Global Spectral model (GSM) of National Centers for Environmental Prediction (NCEP) is used to reconstruct weather with the Local Ensemble Kalman filter (LETKF). Descriptive data are first converted to model variables such as total cloud cover (TCC), solar radiation and precipitation using empirical relationships. Those variables are then assimilated on a daily basis after adding random errors to consider the uncertainty of actual diary data. The assimilation of downward short wave solar radiation using weather descriptions improves RMSE from 64.3 w/m2 to 33.0 w/m2 and correlation coefficient (R) from 0.5 to 0.8 compared with the case without any assimilation. Non-assimilated fields are improved as well (e.g., RMSE from 40% to 29% and R 0.1 to 0.5 improved in TCC). Similarly, the assimilation of other variables shows improvements in atmospheric fields. These findings indicate the potential to reconstruct historical weather for the last five centuries by assimilating available weather descriptions.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2017
- Bibcode:
- 2017AGUFMPP31A1264N
- Keywords:
-
- 3305 Climate change and variability;
- ATMOSPHERIC PROCESSES;
- 3344 Paleoclimatology;
- ATMOSPHERIC PROCESSES;
- 1620 Climate dynamics;
- GLOBAL CHANGE;
- 4928 Global climate models;
- PALEOCEANOGRAPHY