Abstract

The Twentieth Century Reanalysis (20CR) project provides global, subdaily weather estimates stretching back over 150 years, by assimilating only surface pressure observations into the Global Forecast System constrained by SSTs and sea ice concentration. In addition, the data assimilation method provides quantification of uncertainty around these estimates via the ensemble spread. The forthcoming Version 3 system will incorporate several improvements over the data set that is currently publicly available, Version 2c. For example, we show that the uncertainty in the current (Version 2c) system tends to be too low in data-rich regions, and too high in data-sparse regions. Conversely, many of these effects are diminished in the future Version 3 system, which uses an improved algorithm for estimating this uncertainty. I will also discuss ongoing work to improve 20CR, particularly regarding the effects of biased marine observations in the 19th Century. These biased observations have been shown to significantly impact the global sea level pressure in this time period. I will present results from Version 2c, in which many of these uncorrected observations were assimilated, and discuss possible solutions for Version 3.