Satellites provide uniform data coverage globally. Thus, their data have the potential to reduce analysis errors in data sparse areas significantly, thereby improving numerical weather prediction (NWP) model forecasts. We describe a methodology to generate NWP model initial conditions (ICs) from satellite total column ozone data based on three principal steps: 1) convert a chemical total column ozone field to a dynamical mean potential vorticity (MPV) field via linear regression, 2) convert the 2D MPV field to a 3D potential vorticity (PV) field via vertical mapping onto average PV profiles, 3) invert the 3D PV field to obtain model ICs. Each step of this methodology has been refined to increase the accuracy of the process. Refinements include: the experimental determination of step 1)'s regression scheme parameters, including the MPV bounding levels (400, 50 hPa), regression time period (14 d), and latitudinal bands used (all latitudes together); the synthesizing in step 1) of ozone-derived MPV troughs and analysis MPV ridges, in recognition of the fact that diabatic processes are a source/sink for the latter field alone; the use of level-varying mapping coefficients in step 2), in recognition of the fact that an MPV trough represents a phenomenon located primarily below 150 hPa; and, finally, the employment of inversion subdomains in step 3) to prevent the generation of height field dipoles.

The validation of the ozone-influenced model ICs involves simulating the record-breaking 24-25 January 2000 east coast snowstorm using the Mesoscale Compressible Community (MC2) Model. Ozone-influenced initializing fields consist of inverted heights, temperatures and winds from 600-30 hPa. Moisture fields at all levels and all fields at and below 700 hPa are provided by the ERA-40, GEM and Eta (re)analyses. We find that ozone-influenced upper-level initializing fields improve the quantitative precipitation forecast for two of the three (re)analyses. Furthermore, the best forecast of all utilizes ozone-influenced upper-level initializing fields.

We believe that the methodology presented, which generates NWP model ICs from total column ozone data, is of particular use for the forecasting of weather systems originating in data sparse areas.