Sunday, January 30, 2011

Why Forecasts Vary

BUMPED TO TOP:
There are so many questions about the pending winter storm in the 
central and eastern United States, including "forecaster X says 
one thing and forecaster Y says another" that I thought I would bump 
this posting to the top.


I have received a TON of emails and Facebook comments/complaints today pertaining to the great disparity of forecasts for the developing storm in the East. Apparently, there are cities where the current TV forecasts vary from “rain only” to “12 inches of snow.” I’d like to take a moment to explain why this is sometimes the case.

There are four techniques commonly used to forecast the weather:

  • Computer models
  • Scientifically valid derivative techniques
  • Experience
  • Instinct
The computer models can have wide disparities. Take a look at these three models all current and valid at the same time:
5km model. For D.C., 8 - 11 inches. Parts of Delaware, 15-19 inches. 
ARW WRF model. D.C. 8 - 10 inches. The same part of Delaware, 10-14". 
NAM model. D.C. 5-7 inches. Delaware 5-7 inches. 
Scale. Click to enlarge scale or any of the maps. 
While the range in model forecasts above for the District of Columbia could be reasonably stated as 5 to 11 inches (although some will complain that range is too wide), in Delaware the range in these three models is from 5 to 19 inches! And, there are many more models than are shown here.

So, how does a meteorologist make a determination as the final forecast?

One way is to use derivative techniques. Without getting technical, there are a number of techniques to try to assist with improving on the models or choosing between the models such as relating the path of heaviest snow to the path of the  500mb vorticity maxima, 200mb temperature advection, etc. Your local meteorologist might or might not be using these.

Experience. What are the biases in the model? For example, in early spring, experienced Kansas meteorologists know the Gulf moisture usually replenishes farther east than the models indicate, thus the thunderstorms form farther east than a simple model technique indicates. The human can improve the forecast by using experience.

Finally, instinct. I knew, every bone in my body knew, that if thunderstorms were able to develop the night of the Greensburg tornado, they would be very severe with large tornadoes possible. I can’t tell you why I knew, but I did. Really good forecasters develop a knack for forecasting. 

There are so many ways meteorologists can weigh this information that it is surprising the forecasts don't vary more. 

How can you make sense of all of this?

I suggest that people informally keep track of the accuracy of the various weather sources for a month or so and then select one or, at most, two. For example, if you are in one of the cities where this disparity is occurring jot down the various forecasts and compare them to what actually occurs. Do this the next several times the forecasts vary. If there is a clear winner, you have your answer. 

You will be better served by choosing the more accurate forecast and sticking with him/her/it (obviously, we hope you choose AccuWeather). If you try to follow the weather hour by hour and sort all this out yourself, you'll just drive yourself crazy!

Finally, if you have found this interesting, I guarantee you'll enjoy Warnings: The True Story of How Science Tamed the Weather

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.