Skip to Content

Is the Forecast Really as Accurate as the Flip of a Coin?

Clichés, Jokes, and Guesses

If you tell a weather joke to a meteorologist, there’s a 70% chance they have heard it a thousand times before. (see what I did there?)

If I had a dollar for everytime someone said to me something along the lines of “I wish I could be wrong half the time and still get paid”, I could retire, buy a yacht and move to my own personal island.

How meteorologists feel when somebody says that

 

In all seriousness, I understand why people think that. Weather is really important. It is the main reason people watch local news. If a forecast doesn’t go exactly as one might think it should go or want it to go, said person might discard the entire forecast as being inaccurate. After all, this is 2019, why can’t we nail a forecast down hour by hour to the exact wind, temperatures, rainfall, dew points, etc.?

We can get pretty close in our weather app

 

On the other hand, viewers will occasionally call up the weather center and ask how the weather will be for their daughter’s outdoor wedding in two months. The answer is always: “well to be honest I don’t know, but temperatures are typically x and precipitation is typically y at that time.” That far out, we can only make a guess based on climatology while maybe factoring in some of the outlooks that the Climate Prediction Center provides. We have to strike a balance between a long range forecast that lets viewers plan out their week while feeling confident that that forecast will be close to reality.

How Far Can We Forecast?

An NCAR (National Center for Atmospheric Research) scientist, Falko Judt, and author of a recent paper says that we lose all predictability of the weather after 17 days. The randomness and chaos of so many different variables involved in the atmosphere make it impossible to predict beyond that. But wait, hold on, why aren’t we doing 17 day forecasts instead of 7 days at the end of 6 PM news?

https://news.ucar.edu/132673/confronting-unpredictable

 

Even though we may have some semblance of an idea of what might happen up to 17 days out, if we showed a 17 day forecast on the news the saying would instead be: “I wish I could be wrong 9 times out of 10 and still keep my job”. Notice in the graphic above how the line above starts to drop dramatically after about a week? In order for a forecast to be usable by a viewer, we need to keep that line as high as possible, meaning that forecasts are only good up to about a week out, ideally 6 days out. You’ll see many stations these days showing 9, 10, or even 14 day forecasts and honestly most of the time, the data beyond 7 days out can be worthless.

https://news.ucar.edu/132673/confronting-unpredictable

Let’s Get Personal

I have been a meteorologist at KWWL for less than 3 months. If you’ve read my bio you know that I came from a station in northern Iowa and that I have experience in forecasting the weather in Iowa. Moving stations is always difficult because distance, topography, viewing area shape and quality/number of weather stations (just to name a few) all play a roll in forecasting. For me, this is still true even if it is only about 60 miles as the crow flies from station to station.

I take great pride in what I do and want to give you, the viewer, the best forecast possible. I am also naturally a curious person so in an attempt to become a better meteorologist, I gave myself a checkup to answer the question: how is my forecasting? I crunched some numbers and am going to be transparent with anybody reading this and show you the answer to that question.

Disclaimer

Before I show you the numbers, I have to note what went into producing this. I typically work Saturday through Wednesday with my weekend shift covering the morning newscasts and my Monday and Tuesday shift covering the noon show. Depending on staffing, I will also forecast for the noon show on Wednesday. All of those forecasts involve forecasting for the day of; “today” or “High 1”. “Low 1”, or “tonight” would be the night following “High 1”, or “today”. For the noon shows, my forecasts are just a tweak of the morning meteorologist’s forecast. For my weekend shows AND days that I filled in on the morning show, the forecast is produced around 3 AM for the day of and maybe tweaked later on. The latest forecast version is used. If I filled in for the evening shift, the forecast would start at “Low 1” or “tonight”.

This checkup only takes into account temperature forecasting, since that is the clearest benchmark of a forecast and temperatures are generally impacted by other weather conditions. This only includes about 48 of my forecasts since I have come on board. Also you’ll see that there are not even data points for each forecast daypart. That is because my record keeping was not very good early on. I tried to reach back and recover as many forecasts as I could. Forecasts and data are valid from June 17, 2019 to August 28, 2019.

An acceptable range for the forecast is +/- 3°. If I forecast a high of 75°, I am ecstatic if the actual high turns out to be 75° but I am also pleased if the actual high turns out to be 72° to 78°. A range of +/- 5 isn’t bad but could be better. If the actual high/low is more than +/- 5° than the forecast is a bust.

If I forecast within the  +/- range 75+% of the time that is GOOD and is marked in green. 50% to 74% is OK and marked in yellow. Anything below 50% is BAD and marked in red. These are just marks that I have set, of course you can draw your own conclusions.

How Accurate are those Forecasts?

We will start with the 7 day forecast, which are the same numbers you will see when you open up the extended forecast in your kwwl storm track 7 weather app. Typically this is the forecast for Waterloo, a.k.a. homebase.

Click on the picture to enlarge

As expected, the average forecast error increases the further out we get. Today, tonight, and tomorrow forecasts are really accurate with a +/- 3° prediction over 80% of the time. A forecast within that range becomes harder to predict by the second night, but if we take a look at the +/- 4° margin, we hit green all the way out through the 4th night, or 80+% of the time. The +/- 5° numbers are extremely reliable out to day 6, as I hit it 85+% of the time. That means if you look at the forecast on a Monday, the number you see all the way out to Saturday will end up being within 5° of that at least 85% of the time. That’s pretty darn good.

Forecasts degrade with time because it is like playing Plinko

 

Similar to that study I was talking about earlier, the forecast is usually very reliable out through Day 6 and really good within a 5° margin through the Day 7 high (note that low 7 only has 4 data points). Some other interesting notes: I tend to forecast about a degree cold for today’s and tomorrow’s high but warm for the lows day 3 through 7. Overall, the average error for all highs and lows are under 3°. That means your KWWL forecast is usually very accurate.

We also produce the city by city forecasts for Dubuque, Cedar Rapids and Iowa City. Each town that you see on the today/tonight/tomorrow graphics gets their own individualized forecast. At most, these individualized forecasts will go through Day 4. We will take a look at Dubuque and Cedar Rapids below.

Dubuque

The +/- 3° range hits from High 1 (today) through tomorrow night 75+% of the time. The only drop into yellow is on Day 3 when I am biased a little warm.We are in green for each forecast period through High 3. The average high and low error is below 3° so Dubuque, our short term forecasts for you are highly accurate! Note that there are low data points for Low 2 and High 4.

Cedar Rapids

Using the same methods from Dubuque we can look at the individualized forecast for Cedar Rapids. I appear to have a warm bias for Cedar Rapids likely due to some forecasting bias when I think “hey, Cedar Rapids is further south, therefore must be a little warmer”. Sensor location, topography, and overall pattern may play a role in why that isn’t always the case.

Still, the forecast is very accurate for all time periods and ranges except for the +/- 3° range for Low 1. I must not be off by much because we are back to the green for the +/- 4° and 5° ranges. This is still an area that I need to improve on. The average high and low error is still below 3° (below 2° for high).

Other Areas

The Waterloo, Dubuque, and Cedar Rapids sensors all are very reliable and have reasonable access to data through the National Weather Service. The data from other sensors across the area, including Iowa City, are not always as reliable or may have sketchy data records through the NWS and thus makes it more difficult to draw conclusions from the data.

Conclusion

Sometimes us meteorologists deserve more credit than we get. Sure there are a few misses here or there but have you seen other professions that predict the future? Sports analysts, financial analysts, they are wrong sometimes and it happens.

Coming from another market in Iowa was definitely beneficial from a forecasting standpoint but there is always room for improvement. I will continue to track my forecasts to see which areas I am improving in and where I need to make changes. My hope is to earn your trust and give you the most solid forecast I can give you! Of course, I also want you to draw your own conclusions and if you think I am doing a poor job, feel free to let me know about it on twitter or facebook, or shoot me an email at blibby@kwwl.com. I am always up for constructive criticism.

Author Profile Photo

Brandon Libby

Meteorologist

Skip to content