The Ultimate Showdown of Weather Websites for Photographers!

Landscape Photographers are always looking for perfect weather conditions, but how reliable are the weather forecasters?

Hi, my name is Rich Dyson from Edinburgh Photography Workshop, and this is Coffee Break Photography.

Landscape photographers can have the best camera body, the most expensive lens, and the highest-quality filters. However, the weather can make or break a photo, and we can do little to influence it. That’s why I try to find the most reliable weather forecaster to plan for conditions that might help my picture. When I’m running my photography workshops, particularly my Switch to Manual sessions for beginners, I don’t want us to have a day of pouring rain.

I don’t know about you guys, but it feels like the last 18 months have seen a deterioration in the reliability of the forecasts. There have been occasions in that period where I have cancelled a workshop based on the predicted weather, but the actual weather could have seen it go ahead. A couple of times, the forecasts said it was going to be dry, and instead, we had a horrid day of pouring rain.

Every few years, I like to do a semi-scientific survey to see which forecasters are getting their predictions more right than others. I last checked the forecast quality in April 2023, so it feels like now is the time to update it.

Before I talk about the methodology I will use, let’s look at the forecasters I will pit against each other. 

First up is BBC Weather, part of the UK’s public service broadcaster, the British Broadcasting Corporation. They get their weather data from DTN, a US-based forecaster that bought the Meteo Group in 2018.

Next, we have Clear Outside, the UK-based company called First Light Optics, which sells astronomical tools such as telescopes, astronomy cameras, and observatories. Its sixteen astronomers use the same data as the BBC from DTN but their own model to predict the conditions specifically for astronomers.

Metcheck is a US-based forecaster, but it has thousands of weather stations around the world. They use a variety of models to predict the weather. They state that their reliability is checked through a methodology called Data Assimilation Comparison. Every day, their servers randomly pick locations and then monitor them for a minimum of 7 days for temperature, wind speed, wind direction and sea level pressure. The server then marks each location every single hour based on what they forecast and what was observed. Any predictions which fall below 90% accuracy are flagged to their model builders to find out why they got it wrong and they then adjust their models.

The Met Office is the UK’s National weather and climate service. Established in 1854, it gathers information from weather satellites in space and observations on Earth and processes the data through its in-house Unified Model.

Weather Underground is a commercial company based in the US. It started as an offshoot of the University of Michigan weather database. It provides forecasting services to the Associated Press and Google. Weather Underground pulls data from the US Government’s National Weather Service and 250,000 personal weather stations worldwide.

The penultimate service is Windy.com. It was started in 2014 by Ivo, a kiter, helicopter and jet pilot. As its name suggests, it was established to predict the wind conditions that are essential for pilots. It now has a large number of overlays, such as rain and thunder, clouds or waves, that can be used by anyone interested in weather. One of the unusual aspects of windy.com is that they provide multiple weather forecasts using different data models. I’ll be using the following models. The GFS model from the National Oceanic and Atmospheric Administration. The ECMWF, or European Centre for Medium-Range Weather Forecasts model. Meteoblue’s model which is recognised for its wind and temperature accuracy. The ICON-D2 model from Germany’s weather service, DWD. Finally, the Met Office provided a downscaled version of the Unified Model.

Our final prediction comes from XC Weather. It was recommended to me a few years ago by a friend and uses the same GFS data as windy.com

While the data sources overlap, each forecaster applies their oversight to provide what can often be widely different forecasts.

Now that we know who I will evaluate let’s examine my evaluation process. I will concentrate on four elements of weather that will impact photo-taking: rain, wind, temperature, and cloud cover. For each of these four elements, I will record the predictions at about the same time each day for the five days preceding 21 January 2025 at sunset, which is around 4 p.m. here in Edinburgh.

For rain, I’ll use a fairly simplistic methodology. If a predictor suggested there would be more than a 50% chance of rain, and it does rain, they’ll get 2 points. They’ll also get 2 points when they predicted a less than 50% chance of rain, and it doesn’t rain. If they get it wrong then it’s nil points.

When it comes to wind, I’ll take a wind speed reading, and if they are within three mph, I’ll give them three points. They'll get one point if they are within six miles per hour and no points if they are outside six miles per hour.

The temperature methodology will be three points for getting within two degrees of the actual temperature, one point if they are within five degrees and nothing for being more than 5 degrees incorrect.

The conditions assess how accurately they predicted the amount of cloud cover. I have tried to create a grid of the various symbols used by the sites to develop an equivalent setting for each site. If they get it spot on, they’ll get three points. They'll get one point if they are one step away from the prediction. When they are totally wrong, I am afraid it’s nil points again. 

To complete the scoring system, I want to give a higher weighting the further they are from the reference date. So, a prediction five days before the actual date will get a multiplier of 1.5. Four days from the predicted date is 1.4, and so on, all the way to the prediction on the day, which will get a multiplier of 1.1.

OK, so let’s start recording the weather forecasts.

Now that we have the source data to compare, we need to visit one of Edinburgh’s favourite photo spots.

Hi there. So here I am on Calton Hill. It's one of the most spectacular locations in Edinburgh. You get a great view from the hill over to the castle It’s a view I'm sure you've seen on many, many photographs when people come to Edinburgh.

So it's five to four. It's the time that we've been measuring the temperatures, the wind speed readings, the conditions and the precipitation over the last five days. And now's the time to evaluate which one can come up with the best forecaster.

I'm going to take a couple of readings. First of all, I'm going to take the temperature reading. You can see from my anemometer that the temperature is, hopefully you can read that, just between 11 and 12 degrees, So I'm going to take 11 degrees Celsius as being the temperature that we're going to measure here. And now for a few seconds, I’m just going to hold the wind speed reading up. And that's going to give us a wind speed reading that we can then add in officially. I use this device when I'm flying my drone. So it takes 20 seconds to 30 seconds just to give me a good average. So the average at the moment is looking to be about 6.3, 6.5 miles an hour. So six and a half miles an hour is going to be the wind speed reading that we're going to use for here.

In terms of precipitation, well, it's dry. So that's definitely anybody who was predicted less than 50%, they're going to score a point for these conditions. And then finally, what I'm going to do, I'm going to do a quick 360 degrees spin around here. So that way is north, that's towards Fife. Behind me you can see Arthur's Seat and the Salisbury Crags, that's south. Up towards the castle, so you probably see, you won't actually see the castle from this particular location, that's going to be west and east where the night is going to start coming in any second soon. So i'm going to spin around. If i was looking right now i'm probably going to say this is a four on the rating system that i created to try and measure. There’ quite a bit of blue sky above us but we do have some partial cloud all the way around so i'm going to call this a four for that rating, so we've got everything i think we need, and i'll just prove it to you by doing a quick spin around.

The next step is to head down to the office. I'm going to plug all these readings into the spreadsheet and let's see who is the best weather predictor in Edinburgh on the 21st of January.

Well, here I am, back in the office, and I have plugged in the readings I took up on Calton Hill, and it’s time to see what the data tells us. Before I do the big reveal, I admit this is only a semi-scientific study. It’s only based on one day and in one location, so it’s by no means a definitive measure of the accuracy of these services. I will leave a link to the spreadsheet I used to record and evaluate the forecasts, so if you’d like to have a go yourself, all you will need to do is to record the forecasts for five days and then plug in the actual conditions. If you’d like to email me your results and the location you were evaluating, I’ll pull together a consolidated league table and post it sometime in the future.

OK, before I reveal the overall winner, let’s see which services did best on each of the five days.

The best longest range  forecaster was shared between BBC Weather and the MeteoBlue model from Windy.com; the MetOffice and Clear Outside were snapping at their heals.

Now for the four-day result. Again, BBC Weather wins, this time sharing with the ECMWF model from Windy.com.

We have a new winner with the three-day result, and this time, it’s Weather Underground on its own in the first place, closely followed by a shared second place of the MetOffice, XC Weather and again, the ECMWF model from Windy.com. This is usually the number of days before one of my workshops when I take the call to go ahead or postpone, so this is quite interesting to me.

The two day result sees two of the Windy.com models leading the way. The MeteoBlue and ECMWF win this day and are closely followed by the GFS model from Windy, too. So, if you want to know who to believe when looking at the next day, it seems that Windy.com is the way to go.

The readings were taken just four hours before we assessed the actual conditions. Four forecasters get it more right on the day. BBC Weather, Weather Underground, ECMWF model from Windy and the Windy.com MetOffice model.

So, that brings us to the overall winner and officially the best weather forecaster for Edinburgh on 21 January 2025, in true TV talent show format, I will count down from third to first.

In third place, the performance was only let down by a poor prediction four days before 21 January. We have the MeteoBlue model from Windy.com. With 49.3 points, it is a strong contender for the trusted weather forecaster for photographers.

And now, the runner-up. With strong consistency in the last three readings, the ECMWF model from Windy.com scored an overall total of 49.9 points. It seems that Windy.com could be a good choice for photographers looking for accurate weather predictions.

So, we now come to the overall winner for the 2025 Weather Forecast challenge. On the 21st of January in the City of Edinburgh, particularly at 4pm, the most accurate forecaster scored a spectacular 51 points, aided by a strong performance on days 5, 4 and 1. In first place, and officially the most reliable forecaster in Edinburgh, at 4pm on 21 January 2025…….

Is……..

The BBC Weather App. Well done, BBC. You will now be the first forecaster I rely on for the next couple of years to determine whether I run (or not) my workshops. I hope that my confidence in you isn’t misplaced. I’ll also review Windy.com to give the BBC a second opinion.

While this is a bit of fun and not a wide-ranging scientific study, I have created a simple approach to determining which forecasters are more reliable. It would be great if more people tried this approach using the spreadsheet I linked under this video. The results would give a broader overview of the forecasters.

Also, if you recommend another forecasting site that provides the four data points I have used, please add a comment below. I will consider including them next time I do this review.

If you’ve enjoyed this video, it would be great if you could like it and share it by clicking on the thumbs-up button below. That way, a few more people will get to see it. You can also subscribe to the channel by clicking on this button here. I send out a monthly newsletter to my subscribers with news about photography, as well as exclusive offers. Scan this QR code to sign up. My name is Rich Dyson, from Edinburgh Photography Workshop, and this has been Coffee Break Photography. See you next time.