Rating Forecast Performance And Looking At The Next Potential Storm

Another snowstorm is now over and in the record books. One of the things I do after each storm is go back and evaluate how the forecast worked out. There’s always a range of where the best part and the, “needs improvement part” of the forecast were located.

There are all sorts of ways to look at performance. Forecasting is of course based on understanding the models and then taking one’s own knowledge of local patterns and applying it to each unique situation. It would take me 2 or 3 years to be a good forecaster in another part of the country because I wouldn’t understand or have experienced the nuances of that location.

Sometimes the forecast is good for the majority of the land mass, but poor for a large number people. If I forecast a foot of snow for the entire area and everyone got a foot of snow, except Suffolk County, I’d have been wrong for a lot of people. The converse is also true. I take these things into account when making a forecast because it’s impossible to get it right for everyone. I try to capture what the majority of people will experience and not focus on the outliers too much except to acknowledge the high or low end of an event. As a side note, I think this is the problem with mass media. They tend to focus on the most extreme possibility, hype the situation and then the public thinks it’s all overblown.


Below is the snow forecast I made before the storm began. I updated this map and updated my totals on Twitter @growingwisdom late in the morning. That update included 18” amounts into Boston, but for valid evaluating purposes let’s look at what was forecast before that update and then what occurred. Again as a side note, updating during a storm is called nowcasting and also an important piece of this business.


Overall, the forecast was about 2 to 4 inches too low. The best accumulation forecast was for Metro West and Cape Cod where most places fell in the range. Boston was too low by 4 inches and Worcester County was also underdone. The North Shore snow was under forecast by 3 to 4 inches on average, but several places did receive a foot, the upper range of the prediction.

As I think about it, there are two reasons for the higher snow totals. First, the warm air stayed about 40 miles further south than expected. This kept it colder for about 2-4 hours longer and therefore snow ratios were higher. The amount of moisture was close to what the models predicted. My map below from yesterday shows where I thought the mix line would end up on its northward progression.

fluffy snow2215.bmp

To give you an idea how much temperature matters, on average 10 inches of snow when melted will yield and inch of water. Yesterday, many areas saw snowfall at twice the ratios of average. Boston’s 16 inches of snow when melted yielded about .80 inches of water, a very high ratio.


The second reason was the snow bands that set up after 3PM. These bands dropped 2 to 4 inches of snow and put the numbers above the forecast. Snow band signals do show up on the models, that’s why I was so concerned for the heavy rate of snow during the AM commute and through noon. The type of banding we saw at the end of the storm was a surprise. Part of the reason I believe was when the warmer air failed to move further north, it shifted the location of temperature contrast south and brought in more arctic air from the north. Subsequently,this acted as a focal point of the heavy snow bands. There was also a bit of lift (air rising) at around 10,000 feet and this was further enhanced by the arctic air.

Forecast for snow was overall too low and was heavier in the afternoon than forecast. Morning was well forecast. Cape Cod had a good forecast. Impact to evening commute higher than expected.

Into The Future
Meteorological models are really amazing. It still is incredible we can know how the atmosphere will behave in a general sense days or even weeks in advance. Predicting the location of snow bands, coastal fronts, sea breezes and other smaller scale features is getting better, but still far from perfect. Better computers, more initial data and increased modeling of small scale features will all continue to help improve the forecast further. If the leaps we have made since the days of the first weather models are any indication, the improvements ahead will be breathtaking.


Next Storm (s)
There will be a few flurries tomorrow, mostly in the morning and then the next chance of snow is Thursday. Most models are keeping this system a weak one and moving it out to sea. Right now, the European continues to bring another plowable snow event to the region with 2-6 inches of snow. I’m not ready to commit to that until at least one other model comes on board. After that there is the potential for a storm Sunday or Monday. More on all of this later today.

Jump To Comments


This discussion has ended. Please join elsewhere on