Limitations in the Eta-12 Forecast from 00 UTC 24 December 2001

 
Discussion

Horizontal resolution

Shown below is a graphic of the native grid resolution for the Eta-12 in the Buffalo, NY area, a radar picture from the Buffalo radar site from 15:01 UTC on 24 December to its right, and the 00Z 24 December 2001 Eta-12 forecast verifying at 15:00 UTC 24 December (15-hr forecast).

Comparing the native grid resolution on the upper left with the clear-air radar reflectivity on the right, we see that the banded structures in the radar are at most about 4 grid points across. There are multiple bands both in and north of Lake Erie and over Lake Ontario, some of which are marked with white ovals. The band over Lake Ontario is downwind from a concavity in the lake shore shape, and thus in a favored area for snow band development according to research studies.

Comparison of the radar reflectivity with the model forecast valid at the same time shows that the model forecasts a single band oriented from southwest to northeast across the northern portion of Lake Erie and the adjacent shoreline, extending through the northern Buffalo suburbs and Niagara Falls into Lake Ontario. Additionally, the width of the model-forced band is a few grid points wider than the radar indicates. Finally, the narrower bands highlighted in the radar do not appear, and small-scale disturbances (which appear as waves in the band) are not able to be captured. The forcing for these features is too small in scale to be emulated by the Eta-12.

Both the horizontal spreading out of the band and the loss of small scale features which may produce locally heavier precipitation rates, will result in problems with obtaining a good maximum for the QPF. The wider band might account for about 10-20% of the difference between the model forecast and the observations. Precipitation generated by disturbances smaller than the model can resolve and predict might account for another 10-20%.

Convective scheme

A more serious problem than the horizontal resolution perhaps is the convective scheme used in the Eta-12. From the same 15-hr forecast, we can see that the Eta-12 convective scheme has been triggered in the single lake effect snow band, consistent with the positive values of surface CAPE indicated in the forecast field at that time. The convective precipitation rates at this time are about 0.02-0.04" per 3-hrs, or about half of the total 3-hr model forecast precipitation.

The Eta-12 uses the Betts-Miller-Janjic (BMJ) convective scheme, which when triggered by sufficient CAPE and a deep enough unstable layer (200 hPa), adjusts the convectively unstable grid column toward a pre-determined reference profile over an assumed convective time scale, without drawing any water vapor from the planetary boundary layer. The reference profile is representative of post-convective environments in the warm season extratropics and in all seasons in the tropics, and shows a moist-adiabatic lapse rate with dewpoint depressions of 4-6°C.

The problems with the BMJ scheme in lake effect snow situations are:

• Moisture is not drawn directly from below the convective cloud base.
  • If that cloud base is relatively high, then much of the moisture source will be left out of the convective scheme.
  • In lake effect situations, vertical moisture fluxes below the cloud base are essential to development and maintenance of the snow bands.
• Potential impact of convective scheme on grid-scale (explicit precipitation with ice processes) scheme:
  • Convective scheme includes no microphysical effects, which are especially important in lake effect snow events
  • No condensate is created and passed from the convective scheme to the grid scale scheme.
  • The profile to which the convective scheme relaxes may not be representative of lake effect snow situations, and in fact may be too dry
• Potential impact of convective scheme on boundary layer scheme:
  • No vertical momentum mixing by convection
  • Winds will be mixed down to the surface only by boundary layer processes
  • Near-surface winds may be too weak

The last two major points' impact will be discussed in the next two sections.

Grid-scale scheme

The grid-scale scheme explicitly predicts water and ice condensate, and includes improved parameterization of ice and snow microphysics over the former scheme. Precipitation is explicitly tracked in the model grid column, which allows for the advection of precipitation between grid columns. Fall velocities of ice/snow and liquid are also differentiated. Such advances should allow for improved precipitation forecasts for lake effect snow.

In lake effect snow events, there is combined production of precipitation from both convective and grid-scale (i.e. stratiform) processes. In the Eta-12, however, there is currently no direct linkage between the convective and grid-scale schemes, since the convective scheme does not explicitly predict hydrometeors, and only makes adjustments to the vertical profile of temperature and moisture to mimic the redistribution effects of convection. The assumed post-convective vertical profile may be too dry to allow the grid-scale scheme to produce ice clouds and snow. If the latent heat flux is sufficient from the lake, the post-convective profile will not be reached, however, and grid-scale precipitation will indeed be produced. Since there is grid-scale precipitation and cloud condensate predicted by the model, there is no obvious impact from too much drying. At times when the lake effect has reached a quasi-steady state (e.g. when a single lake effect snow band remains over a number of hours), the shortcomings in the convective parameterization may not have as much of an impact, if

• The convective time scale is not too short and
• The moisture/heat fluxes from below are at least qualitatively correct.

However, there is reason to believe the moisture/heat fluxes may be lower than observed, as discussed below.

Boundary layer scheme

The Eta-12 surface fluxes are forecast based on the lowest model layer wind and gradients of moisture and temperature, for the latent and sensible heat fluxes, respectively. Errors in either will result in errors in the surface latent and sensible heat flux. The graphic below shows the surface layer wind compared to the observations at 15 UTC 24 December 2001. Pay particular attention to observed versus model winds off the lake (circled).

From the graphic, it can be seen that the Eta-12 surface winds are consistently underforecast, by up to 20 kts for winds with a long fetch off the lake. This will result in latent and sensible heat fluxes that are too low, perhaps by as much as a factor of 3. The steady state single lake effect snow band will thus get too little moisture and heat from below, which will result in too little instability, too little convection, and too little precipitation. Additionally, as previously stated, the convective scheme does not draw moisture out of the boundary layer, which may leave it too moist and warm. This reduces the vertical gradient of moisture and temperature, and thus even further reduces already low latent and sensible heat flux from the lake.