A Case for Operational Utility in Non-Traditional NWP Output Fields

Numerical Weather Prediction (NWP) model output provides forecasters with valuable guidance that often forms the basis of public forecasts. Post-processing of NWP output also adds value, and recently there has been a rapid advance in the application of artificial intelligence to operational forecasting. Despite these advances, I contend that some highly valuable model fields are rarely output, let alone plotted or analyzed in a forecast process (automated or otherwise). For example, NWP models compute tendency fields of temperature, moisture, and momentum; these fields may provide valuable information about model error growth and uncertainty. Model microphysics schemes compute detailed tendency fields for different hydrometeor classes, along with their size distributions and terminal fall velocities. Radiation parameterizations compute the radiative fluxes and their divergence. Frequently, these quantities are not output, and thus are not made available to human forecasters or to machine learning algorithms.

In order to illustrate the value of some of these non-traditional model fields, I present a case study of a heavy snow event from the Great Plains from early February, 2016. Model output fields include dendritic snow growth, vertical snow flux, and terminal hydrometeor fall velocity. These additional fields provide a more complete picture of physical processes during this snow event. I show that areas of dendritic snow growth are located upwind of some areas that experienced heavy snow accumulation, and that horizontal snow advection exerts a strong influence on surface snowfall distribution. Visualization of the dendritic snow growth in three dimensions is useful in the analysis of the event evolution. Interpretation of point forecast soundings produces incomplete results for this case, owing to the spatial displacement between where snow forms and grows aloft, and where it reaches the surface.

In general, I suggest that a more complete set of model output variables could aid in forecaster interpretation of NWP model output, and that these data can potentially enhance the accuracy of forecasts derived from artificial intelligence applications.