It's Sunday, 4/23/2023 and I'm sitting here in Titusville listening to the wind howl. Windy predicted 10 to 13 kt NE winds but I'm seeing 20 to 25 kts sustained winds at my dock. I'm sure it's more than that in the wide-open spaces of Indian River Lagoon heading north to New Smyrna. Did any weather forecast predict the high winds I'm seeing today?
Let's take a pictorial walk with the photos below. It's interesting to compare the predictions between the major weather models used by the weather apps to see which one predicted what I'm seeing right now at my dock in Titusville at 6:00 am. Based on what you will see in the comparisons, I now know which ones I will depend on in the future.
I used three weather apps showing nine predictions based on seven weather models:
- PredictWind can show the predictions by seven weather models that you can switch between with a tap.
- Aqua Map can show the predictions from NOAA's HRRR model, NOAA's coastal forecast, and the local NOAA weather forecast with a tap on a chart location.
- Windy can show several weather models but I used it for the ECMWF model (the European model) just to be sure that PredictWind and Windy agree when using that model.
The weather model used is more important than the app used to display the results. I checked the display of the ECMWF model prediction by PredictWind and Windy just to be sure and as expected, both showed the same prediction.
Now let's get to the visual comparisons. First up is Windy. It is a clear miss when using the ECMWF model, it does not show the 20 to 25 sustained winds I saw for several hours that morning.
The PredictWind display of the ECMWF model shows 10 kts. I didn't tap the exact same spot on the chart so it's off by 2 kts but it's pretty close. It certainly doesn't predict the 20 to 25 kts sustained winds. It's still a miss.
How about the GFS model? Here's the prediction from PredictWind using that model for 11 kts. Another clear miss.
Now we've shown the two most popular models: the European ECMWF and the US GFS model but both failed to show the 20 to 25 kt winds I saw that morning. PredictWind has a proprietary model based on ECMWF called PWE in their app. Let's take a look. Well that's more like it. It predicted 22 kts at 6:00 am. I would call that a hit!
PredictWind has another proprietary model based on the US GFS model that they call PWG. Let's take a look. It predicted 19 kts, not bad.
Another model to look at is the HRRR model which is NOAA's latest and greatest high resolution weather model for the US that's updated once an hour. How did it fair? Look at the results. It predicted 20 kts that morning.
Next up the Aqua Map forecast using the same model, they ought to agree - and they do but it's presented as a grid of arrows with velocity numbers, easy to understand.
Let's check out the two remaining models in PredictWind. First up is NAM. The North American Mesoscale Forecast System (NAM) is one of the National Centers For Environmental Predictions (NCEP). It missed the high winds.
The one remaining model is UKMO, the United Kingdom Meteorological Office model? It's another miss.
But there are other sources. How about the NOAA coastal forecast? I wouldn't say they got it right either.
Okay, one last source. The local forecast by NOAA that's displayed in Aqua Map with a tap on the chart for that location. I would say that's pretty good, it showed 20 kts of wind!
So, what have we learned? The models may work well most of the time but when a ridge or front comes through, there can be big differences. PWE, PWG, and NOAA's HRRR models fared pretty well predicting the high winds I saw on 4/23 at 6:00 am. I'll be paying more attention to them in the future. The NOAA local forecast gave good results too.
You might want to add the models that did well to your toolbox for predicting weather. It's no fun waking up to unexpected winds, and even less fun being hit with unexpected high winds while underway. There are also weather services you can use that provide forecasts by a meteorologist for your specific needs, at a price, but using the weather models that performed best in this exercise is a good first step.