Monday, January 5, 2015

A Major Breakthrough in Numerical Weather Prediction for the United States

Monday, January 5, 2015
1:10 pm

There is currently major flooding on the Tolt and Snoqualmie Rivers near Carnation, and I had just Google'd 'NWS' to get to the Seattle NWS homepage ( is too much work) when I stumbled upon a rather important headline.

Retrieved from the original press release from NOAA.

No hyperbole can describe how euphoric I was when I saw this. The United States has lagged behind other countries (most notably a coalition of European countries that have formed the 'ECMWF' model) in numerical weather prediction for as long as I can remember, and the reason is largely because we have a lack of computer power compared to them. This means that we cannot run our models at as high of a resolution and that our initializations are not as accurate as they could be, as a good model initialization (i.e. a good representation of the initial conditions from which the forecast is based off) is necessary for a good forecast and improves as the number of observations (and the computer power to process them) increases.

Additionally, the additional computer service has obtained will greatly assist in developing ensemble forecasting and helping it become a more viable tool to potentially overtake traditional deterministic forecasting. Deterministic forecasting is the type we are all familiar with; you take your initial conditions, plug them into the models, and voila! You get a forecast. With an ensemble forecast, you put a slightly different set of initial conditions into each of the ensemble "members," and as a result, you get a variety of results. Ensemble prediction is the prediction of the future as it offers many advantages that simple deterministic modeling does not, such as accounting for uncertainty in a forecast.

The computers we have now are not mediocre by any means; we have two of them, and each one can perform 213 trillion operations per second. However, by the end of the month, they are expected to triple their operating capacity, and by October, each should be able to perform 2,500 trillion operations per second. That's a lot. Imagine playing a ridiculously hi-resolution version of Call of Duty on one of those things. Unfortunately, that's not what these computers were made for.

This upgrade will cost 44 million bucks, which sounds like a lot of money. However, I did some calculations, and seeing as the U.S.' GDP in 2013 was 16.8 trillion (thanks Google), the amount of money the U.S. pulled in was over a half million bucks per second. Setting aside a minute-and-a-half of the year to fund something that, by providing accurate forecasts, could save billions of dollars in the long run, sounds like a good plan to me. It should have been done far earlier, but I'm glad it has been done now.

I'm not as "in-the-know" about this stuff as one of my professors at the UW, Cliff Mass. He has a blog, and I noticed that he wrote a post on it this morning as well. Read his blog here for more information on this monumental achievement by the NWS.


1 comment: