The standard MM5 modeling system contains a Newtonian nudging FDDA method (NNM). Nudging to both gridded analyses and/or conventional observations is examined in a companion paper. The NNM method has the disadvantage, from a forecast systems point of view, that the nudging is done before the observation time and it is difficult to use the nudging coefficients afterwards for forecasting purposes. Therefore other FDDA approaches need to be investigated for such purposes.
The so-called 'intermittent data assimilation' (IDA) method is one such alternate approach. Schemes following Cressman (1959) and Bratseth (1986) are two possible ways of obtaining the analyses used in IDA. The Bratseth (1986) scheme avoids errors with the Cressman (1959) or other successive correction schemes associated with the fact that in such schemes the analysis always converges to the data, which should not be the case when errors exist in both the observations and the background. The solution of the Bratseth scheme converges toward a solution obtained by Optimal Interpolation (OI). This method alleviates the shortcomings of the Cressman scheme and of other successive correction schemes and also requires much less computational costs than performing a full OI procedure.
In this study we apply the IDA method using the Bratseth scheme to the summer Alaskan heavy rain event examined in the companion paper with the NNM method. The 6-hour model forecast is reanalyzed by ingesting new observations at the forecast time. Then the analysis is initialized for an MM5 forecast run using the Bratseth scheme. Analysis and forecast results will be presented, as well as an evaluation ofthe usefulness of intermittent data assimilation with the Bratseth scheme at high latitudes.