Summary and concluding remarks

Caveat Emptor

These results apply to this case only, which was a strongly forced event with relatively large scale features compared to the NWP model being used. Other more localized lake effect events, or lake effect events involving features of smaller scale, such as multiband lake effect events (bands of 1-20 km in scale) will not be predictable in detail. The environment favorable for the occurrence of these smaller-scale events, however, may still be well-predicted, especially at short-range forecast time ranges.

Summary

The first 36 hours of the 0:00 UTC Eta-12 forecast from 24 December 2001 mark a significant improvement in operational forecasting of lake effect events. Some of the characteristics of the forecast during this period are listed with possible causal factors (and despite the limitations) found in the bullet list below:

• The Eta-12 captured the timing and location of the development of a single lake effect snow band forced by cold air moving over warm LSTs over Lake Erie at between six and nine hours into the forecast.
• The Eta-12 accurately forecast the lake effect snow band's initial shift southward after 15 hours.
• During the first 36 hours, forecast precipitation amounts were generally half as much as the point verifications, consistent with other mesoscale model simulations of single band lake effect events at similar resolution (e.g. Ballentine et al., 1998). The NWS Buffalo observations of 25.2" of snow and 1.99" water equivalent were localized effects resulting from interaction of the lake effect snow band with local topography (Mclaughlin 2002, personal communication). The general precipitation deficiencies seen may result from:
  • Latent and sensible heat fluxes from Lake Erie that were lower than occurred because surface winds used to calculate the fluxes were too light, and
  • Surface model layer that stayed too warm and moist because convective scheme does not draw latent or sensible heat below the convective cloud base, so there is no direct connection between generation and release of convective energy.
  • No microphysics in the convective scheme.
  • No microphysical connection between the convective and grid-scale precipitation scheme
• In spite of the physics limitations outlined above, the model captures the essence of the mesoscale circulation generated by the passage of cold air over Lake Erie, including:
  • The proper vertical convergence/divergence and vertical motion structure
  • The expected surface pressure field perturbations, with strongest horizontal sea level pressure gradient located to the south of the snow band
  • Stretched out warm core over and downwind from Lake Erie
  • Convergence near the edges of the lake effect snow band

The model forecast for the position and strength of the single lake effect snow band in this case is sensitive to a number of mesoscale and synoptic-scale features. It has also been shown in experiments that all the Great Lakes (of which there are 3 upwind of Lake Erie) influence the synoptic and mesoscale flow in cold air outbreaks, in ways that will not be reproducible in detail by NWP models, especially in the longer time ranges. Because of these factors, we would not expect the Eta-12 to be able to accurately predict the position or strength (or even the existence) of a single lake effect snow band beyond a day or so. So, there is no surprise that as the forecast moves forward in time beyond 36 hours:

• Model errors in synoptic flow and in mesoscale forcing result in problems with forecast strength and location of single band relative to verification.
• Model single band is maintained early in the extended period when verification was for weakening. Movement of the band to the north was reasonably predicted.
• During the period when the lake effect snow band was redeveloping on 26 December, the Eta-12 has a relatively broad and weak (though still significant by model standards) snow band in the correct general location.
• The southward shift of the lake effect snow band back to Buffalo during the evening of the 26th, and subsequent settling to the south of the city is not completely captured. In the model, the snowband remains north of the reported verification throughout.

We can conclude, however, that the Eta-12 got the synoptic-scale and gross features of the mesoscale environment correct, since it did predicted a single snow band.