As anyone who has shoveled knows, all snow is not created equal. Some winter storms produce the light, powdery kind that results in more inches on the ground. Then there's the snow blower-defying type that's wet and heavy.
The difference, says Paul Roebber, a professor of mathematical science in the atmospheric science group at the University of Wisconsin-Milwaukee, is the snow ratio -- the amount of water that's contained in each inch of snow. But forecasters, who first predict how much liquid a storm will produce and then convert that to snow amounts, have long struggled with the second half of the process.
Roebber and one of his students, Sara Bruening, recently published a method of predicting the snow ratio that shows a significant improvement over current methods. It is so much better, in fact, The National Weather Service is now using the Roebber/Bruening method in their forecasts all over the country.
Roebber says the ratio of snow to liquid water (what you would get if you melted a certain depth of snow) can vary tremendously, anywhere from 3:1 (3 inches of snow per inch of liquid water) to 100:1. That's one reason why meteorologists give a range of snow amounts in their forecasts.
Forecasters can already predict how much water a storm is likely to produce. But when predicting snow, forecasters often default to an "ordinary" snow ratio of 10:1, a standard set in the 1800s that is generally regarded as arbitrary. For a storm that results in a half-inch of rain, for example, the forecast would be for 5 inches of snow at a ratio of 10:1, but 10 inches for a ratio of 20:1 (lighter, fluffier snow).
A web technician at the North Dakota bureau of the National Weather Service helped the pair build an online formula that would calculate results if certain information was plugged in.
The web site calculates the probability of the snow being heavy, average or light for each of the models of weather conditions that the Weather Service already uses. The forecaster simply has to input the predicted rain amount and the wind speed.
Roebber and Bruening applied a statistical process involving artificial neural networks (ANN) in designing their forecasting tool. ANNs can create a computer simulation of a "brainlike" system of interconnected processing units. Though ANNs have not been used in meteorology before, Roebber thought the system could successfully take into account the many variables involved in forecasting snow. He is now collecting cases that will verify strengths and weaknesses of the project.
The American Meteorological Society contributed to the information in the TV part of this report.