Operational guidance on the ratio of snowfall depth to liquid water is severely limited. The familiar 10:1 rule is unreliable in many circumstances and guidelines based on surface temperature alone are flawed. In a recent study of snow ratio (Roebber, Bruening, Schultz and Cortinas 2003; Weather and Forecasting), a method for producing superior forecasts of depth of snow was developed, based on artificial neural networks. For your convenience, this technique has been adapted on this website such that given a model forecast sounding (obtained from the WRF/NAM and GFS), you can determine the likelihood that the snow will fall into one of three density classes (heavy, with ratios up to 9:1; average, with ratios from 9:1 up to 15:1; light, with ratios exceeding 15:1).
This realtime system was made possible by an informal collaboration between UWM and the National Weather Service (NWS) office at Grand Forks, North Dakota (KFGF). Programming was accomplished by Richard Hozak of KFGF.
The UWM realtime snow-ratio forecast page is designed as a quasi-operational system. The intent of the system is to provide a test bed for ongoing research into the predictability of snow ratio. In addition, the output from the system is provided on this web page as a service to the operational forecast community and other interested persons. However, usage of these products remains entirely at the discretion of the user and the responsibility for decisions made (good or bad) based upon the forecasts rests entirely with the user.
How do I use this page?
On this page, you simply click on the map to view all potential forecast sites, based on the output from the NOAA operational models. Then click on the blue triangle to obtain a forecast for the site. Alternatively, simply enter the site id in the box to go directly to the forecast. In either case, select the desired forecast model from the pull-down menu and click continue. Then, choose the sounding valid time, enter the forecast liquid equivalent QPF and surface wind speed. Click continue again. The probability of snow being in the heavy, average or light categories, given that snow occurs, is displayed.
1) The forecasts are based upon average vertical motions (more info here). Should the vertical motion be focused in a particular layer, some adjustment of the forecast may be required. For example, if the vertical motion is particularly strong in a layer that contains supercooled liquid water, then riming may be increased, and snow densities will be higher.
2) The sounding profiles are based on the selected model forecast. Two different model forecasts for the same valid time may well produce different results. Obviously, you should choose the model forecast that you believe best represents the observed conditions at the forecast valid time.