25 January snowstorm

Draft V_2.0

Richard H. Grumm

And

John LaCorte

Introduction

On Tuesday, 25 January 2000 a storm moved up the East Coast producing heavy snow from South Carolina into New England. A mixture of snow, sleet, freezing rain, and occurred along the immediate coastal plain with mainly snow inland. Initially, this storm was poorly forecast, due to the poor quantitative precipitation and surface cyclone track forecasts. No NCEP model was able to produce the precipitation, which was observed to the west of the surface storm track. Despite the errors in these initial forecasts, the models did contain some useable trends in successive forecasts.

Individual model trends can be a useful forecast tool, however these trends may be biased by the models physics and initialization schemes. Additional, and often very useful, information can be obtained by using different models and producing a model consensus forecast (Fritsch et al, 2000). A suite of models, using various initial conditions and physics, would produce a likely range of solutions. Ideally, one of the solutions would close to the observed outcome. A deterministic forecast could be derived from a consensus of all the model solutions. Typically, the consensus forecast will be more skillful than any of the individual members (see Fritsch et al, 2000 Figure 3). NCEP is currently working on providing forecasters a mesoscale ensemble suite from which these types of forecasts could be derived. However, using the operationally, available model data, a consensus forecasts can be derived from the NCEP model suite.

This page focus is on the model trends produced by the operationally avialable NCEP Eta and AVN models. Additionally, forecast consensus is derived from these data. A forecast method, using a reverse logic approach is presented. The concept provides forecasters the tools to identify model trends. Once the trends are idenitified, the forecaster then must determine the potential reasons for these trends. It is beyond the scope of this page to determine what errors cause the trends presented.

 

Also included in this page are 14-km Eta runs using the operational Eta from 1800 UTC to see how the higher resolution forecasts would handle this event. Preliminary results suggest this mesoscale run offered no additional useful information with which to forecast heavy snow over the populated cities along the eastern seaboard. We plan to run the 0000 UTC 25 January version of the model in the near future.

Method:

Model trends were obtained by using NCEP model output displayed using Grads. Trends are shown for each individual model. Model consensus forecasts were derived by regridding each model to the same grid and producing a consensus forecast in Grads.

NCEP Eta files were saved in real-time for analysis. Additionally, Eta data on the 212 grib were archived in near real time to allow running of the workstation Eta and MM5. We used 3-hour boundary conditions for this model run. In addition to these data, snow depth and sea surface data were also archived. The workstation Eta (WS-ETA) data are available in 3-hour increments verses only 6-hourly data for the operational Eta (ETA).

Some data were displayed using GEMPAK.

Model Trends

a. 6-hr forecasts of Eta Model trends

Using model trends to improve upon a forecast.

There are too many fields to examine, so we will stick to few fields.

 

When dProg/dt shows a trend,

Examples from 1200 24 Jan through 0600 UTC 25 Jan 2000

Surface and QPF

700 hpa height and relative humidity

500 hpa height and vorticity

250 hpa height and vorticity

b. 12-hour forecasts of Eta dProg/dt

c. 12-hour forecasts of AVN dProg/dt

The forecaster must address:

Workstation Eta results:

QPF forecasts from the 14-km Eta initialized at 1800 UTC 24 January are shown. The first figure, valid at 0300 UTC 25 January 2000, showed the 3-hourly accumulated precipitation. The figure shows the surface low off the North Carolina coast and locally heavy precipitation along the coastal plain.

By 1200 UTC, the forecast showed the surface low off of Cape Hatteras, NC with heavy precipitation forecast along the coastal plain from northern North Carolina to the Delmarva. These 3-hourly amounts show up to 0.75 inches of precipitation northwest of the surface low. Note only light precipitation is forecast over Washington, DC and Baltimore, MD where moderate to heavy snow was observed.

 

By 1800 UTC, the surface low was forecast to be off the Delmarva Peninsula with heavy precipitation forecast along the coastal plain. The satellite imagery, valid at 1800 UTC showed the surface cyclone tucked in closer to the coast. This allowed precipitation bands to develop farther west as seen on radar. More imagery can be found at this website.

The storm total precipitation is shown in the final figure. Contours are 0.25,0.50, .75, 1 and 2 inches. Values exceeding 1 inch are shaded. The 14-km Eta forecast the heavy precipitation along the immediate coastal plain, just west of the track of the surface low. The observed snowfall is shown in the snowfall figure.

 

Conclusions:

The higher resolution WS-ETA provided more detailed QPF’s then the Eta. However, this version of the model, using 1800 UTC 24 January data was insufficient to forecast heavy snow over areas west of the storm. We plan to examine dProg/dt to see if the models showed some kind of useful trend.