Weather Prognostication

[Michael Beard]

I didn't want to clutter up the Keystone thread anymore than what we already have, so I'll move the discussion here. I think it's important, as in recent years, I have frequently seen racers talking about making decisions about an event based on "forecasts" 10-15 days in advance on an event. There is no such thing as an accurate 10-15 day forecast. About the closest half-decent guess for a weekend forecast is going to be made late Tuesday the week of an event -- and even those are subject to major changes. Just in the last few weeks, we have seen them miss a forecast for *the next day* by 50-70%. It's WEATHER, folks. It changes. The number of variables involved are astronomical. (no pun intended)

To recap the discussion from the Keystone thread:

Quote:
Originally Posted by david ring View Post
the problem is the forecast for next weekend is not promising either

(Note: This was posted on a Wednesday, so it is referencing forecasts for 8-11 days out)

Posted by Superfan1
From someone who studied meteorology in college, any forecast for more than 3 days into the future has a diminished accuracy probability, and that probability diminishes even more as the forecast range increases. At this point, a forecast for next weekend shouldn't be taken literally.

Posted by Gearhead55
I hear you Bill,, but in the interest of fairness,, I'm assuming college was a few years back and forecasting has come a long ways in the last couple decades. I hope theres no problems for those racers already parked, but the potential for some rigs getting blown around does exist. We've all seen campgrounds, racetracks, etc, get hit by storms and have seen awnings ripped off or worse! I'll keep my fingers crossed for all from out here on the "left coast". Take care,, my friend.
Danny Durham

Posted by Michael Beard
Exactly right. It's Chaos Theory. There are infinitely too many variables involved, and even the absolute tiniest of error in the initial assumptions multiple over time exponentially.

Posted by Superfan1
Danny you are right, when I was in college, meteorology was in its infancy. The current computer guidance and associated technology didn't exist when I was studying. However, I stay current and up to date on all of the new technology because I have always been fascinated by the weather. Believe me a forecast for 7-10 days into the future has a very low degree of accuracy. Take care my friend, I hope that you are doing well.

Posted by Superfan1
Well said Michael, very well said.

Posted by jwsamuel
A forecast for later the same day should not be taken literally.

Posted by david ring
Bill,

I understand that, but decision makers have to make decisions based upon the information they have now. Suppose you have this information.

1. forecast for this weekend-50 percent chance of rain; forecast for next week-0 percent chance of rain;
2. forecast for this weekend-50 percent chance of rain; forecast for next week-100 percent chance of rain;

If you have case 1, you postpone until next weekend; if you have case 2, you race this weekend and take your chances. Sure, the weather next weekend is likely to be different from the forecast, but you can't wait until next weekend to make the decision about this weekend. Given the forecast that next weekend's weather was not likely to be great, I think NHRA made the right call to try to get the race in. Lee Zane is there with my comp car ready to race.

Posted by Superfan1
David, I agree with everything that you said; my point was that anyone planning to attend the Dutch, at this point, shouldn't take next weekend's forecast literally.

Posted by Michael Beard
False premise. There is no such thing as a forecast for next week, particularly one that would ever say 0% or 100%. They have a hard time denoting a 0% or 100% chance within hours, let alone 10 days away. Again, it is scientifically impossible to generate a 10-day forecast with any accuracy better than "pin the tail on the donkey". It is a guess, period. I call it weather entertainment. They publish it just to have something to talk about. I've tracked it for numerous events for several years, and they have never bared any resemblance to actual conditions.

NHRA's procedure historically is to make every possible effort to complete an event on the original scheduled weekend due to scheduling, staffing, logistics, advertising, commitments of the racers, fans, sponsors, etc. It's rare to lose 3-4 complete days straight, so they choose to utilize whatever windows that present themselves. It's a lose-lose situation for most involved, but it is not possible for them to operate as nimble as a local bracket track. While they are forced to plow ahead and make the best of situations, they're generally pretty consistent about it.

We postponed our March race by a week this year, not because we were able to tell what the following weekend was going to offer, but because the original date offered no chance whatsoever. When the postponed date rolls around, you have to once again simply deal with the hand you are dealt, and adjust accordingly (which we had to do).

Visual demonstration of chaos theory:
My professor had a similar apparatus, with two sets of double bars. When dropped, they would start falling in the same direction at the same pace, but the error in the initial constraints multiply exponentially over time, quickly resulting in wildly different patterns. He explained things much better than this guy, but you get the idea.

Posted by Ron Ortiz
Percent of chance of rain is a misleading statement.

If it rains one drop, then no matter what the percent chance is or was, it has met the percent chance.

50% chance does not mean that it will rain 50% of the day. Only that there is a chance it will rain, so if it rains one drop, then the percent chance has been met.

You racers chose to chance it out by going. Hope all goes well.

Several years ago we got hit with a hurricane on FL. west coast. In my little town of Englewood the storm went right through us but there was a pocket of clear area that stayed over us as the storm went through, no rain, the storm just cycled around us.

After that experience, I look at weather predictions a lot different. And that is a 50 % guess, which is not bad odds.

Ron Ortiz
U/SA chances of winning in the rain with my car, 100%

Posted by david ring
First of all, the reason for choosing 0 and 100 percent were to use the extremes. Second of all, the point of my example was to indicate that any decision today should use all information, including information about the future. Third, I know a little bit of math, maybe even more than a little bit and I understand about non-linearities and for example the stability of a system (for example a system could be locally stable and not globally stable-it is much easier to prove locally stability because you can use a linear approximation via the first few terms of a Taylor expansion). And while national events are rarely postponed in advance, points races have been due to forecasts.

1) Those 'extreme' examples are a false premise. They do not exist in a 10-day forecast. They are extraordinarily rare even in a 24-48-hour period (outside of extreme climates like deserts and rainforests). Again, just in recent weeks, we have experienced a swing of 50-70% change in a next-day situation alone!

2) Yes, decisions should use information about the future, IF such data has an expectation of reliability. Accurate 10-15 day forecasts are scientifically impossible. While you can frequently get decent data out of a 3-day forecast, and might get a good guess out of a 5-day forecast, long-range forecasts are demonstrably garbage. Garbage in, garbage out. If you make decisions based on bad data, you make a bad decision.

IHRA pulled the plug on their Div. 3 Bracket Finals in Pittsburgh on Tuesday due a forecast that showed rain from Thursday straight through til Monday. The forecast now says rain from 10PM Friday through 2PM Saturday. They could have gotten the race in. The made the best decision possible based on the information they had, but the data turned out to be bad, on a forecast just 3 days out.

In April, racers panicked on a Sunday night due to a forecast for 80% Fri-Sun for the Piedmont Pro-Am, and wanted to know when it was going to be rescheduled. It turned out that we had an absolutely beautiful, perfect weekend.

Additionally, as Ron pointed out, in terms of event planning, percentages tell you nothing, until the hourly forecast comes into view. "80%" can mean rain overnight or in the morning and have little effect on an event.

Without looking at historical records, can you predict the actual weather for March 13-15 based on these long-range forecasts?