Climatology and variability of tropical rainfall in the 20th Century Reanalysis

We propose to validate the climatology and variability of tropical rainfall in the 20th Century reanalysis, version 2 (20CRv2, Compo et al., 2010). To do so we will compare the reanalyzed precipitation to global satellite-based records for the most recent past, to several land-only datasets that go back to the beginning of the century, and to statistical reconstructions of marine precipitation (Smith et al., 2008, 2010).

Our first task will be to document rainfall biases both in terms of their monthly and seasonal patterns and in terms of the characteristics of daily precipitation. We will try to better understand the source of these biases by replicating the work of Biasutti et al. (2006), which diagnosed the strength of dynamical (i.e. convergence) and thermodynamical (i.e stability) control on rainfall in several GCMs and reanalyses.

Second, we will contrast the time-series of regional rainfall in the reanalysis and in observations. Some preliminary analysis for the African Sahel suggests that the 20CRv2 precipitation monthly anomalies depend strongly on the density of the observational records of surface pressure that are assimilated: when the synoptic spread in surface pressure is relatively low, the ensemble-mean monthly rainfall anomalies track observations very well, but when the spread is large, the correlation with observations drops below what we would expect from AMIP simulations. 

Given the limited skill of the 20CRv2 reanalysis in reproducing historical variations of precipitation in data-sparse regions and periods, it is desirable to understand whether a MOS correction could be applied to the reanalyzed fields to create a trustworthy, complete record of rainfall. The idea behind a MOS correction is that models that do a poor job of simulating rainfall might still have skill at simulating the large-scale circulation associated with an observed rainfall anomalies.

If the reanalyzed time series of the tropical circulation is found to be accurate, and the relationship between circulation and rainfall is found to be stationary, then it would be appropriate to create a MOScorrected record of rainfall. Although the short duration of this project prevents us from producing such a record, we propose to test whether it would be possible. To do so, we will both validate the historical variations of key circulation indices (for example, the Southern Oscillation Index and the monsoon indices), and test the fidelity and stationarity of the relationship between regional rainfall anomalies and circulation anomalies.