Advances in numerical modeling, computer speed, and high-resolution observing systems are driving a trend toward high resolution numerical modeling using grid resolutions of 30 km and less. Recently, for example, the Center for Analysis and Prediction of Storms (CAPS) has undertaken real- time forecast efforts at scales of 27, 9 and 3 km. There is a need for data assimilation techniques that can be used at these scales and take maximum advantage of data such as the Oklahoma Mesonet and the operational Doppler radar network. It is also necessary that the assimilation techniques be efficient, so that lead time can be maximized for 0-12 hour forecasts. Two assimilation methods are presented to address these needs: the ARPS Data Assimilation System (ADAS), and a method of directly correcting phase errors in a forecast, as part of a data assimilation strategy.. ADAS uses a modified Bratseth successive correction scheme to analyze surface and upper air data. Unique terms allow for the proper handling of raw Doppler radial velocity data. Because the Bratseth scheme converges to the solution found by optimal interpolation, the relative accuracy and density of data from the diverse sources is taken into account.

One of the unique attributes of the sensitive operational Doppler radars is that they can observe atmospheric boundaries (such as fronts and thunderstorm outflow boundaries) well. It is common for numerical forecasts to have position errors in such features as well as in thunderstorm cells and convective complexes. It is proposed that the problem of correcting a numerical forecast field can be simplified if such phase errors are directly addressed. An objective method of determining and correcting phase errors in forecasts is described and tested under various scenarios, including simple dynamic models, an observing system simulation experiment (OSSE) involving the forecasting of ongoing convection, and a real data case involving predictions of the development of severe thunderstorms.

It is shown that the phase correction is effective in producing an analysis field that agrees with the data and preconditions the fields for the application of objective analysis. In the OSSE, the scheme is quite successful in achieving a long-term positive effect on the thunderstorm forecast. In the real-data case, the phase correction had a positive influence on subsequent forecasts for both a front and a complex severe thunderstorm situation. The improvement lasted between 2 and 3 hours for a 3-km forecast, which is at or beyond the expected limit of predictability for thunderstorms.