Here’s a new video giving a complete walkthrough of Photo WX. It has full chapter markings, so you can skip to parts of most interest directly.
After the capacity challenges we had during the live webinar last weekend, it seemed to best to have a do-over to avoid all the distractions.
Since the webinar, we’ve added more capacity and made a number of changes to how the map tile servers are deployed, managed and synchronised. While we’ve seen a huge decline in the error rate and generally much improved performance, it’s still not where I’d like it to be, so we’ll be making more changes in the coming weeks.
I’ll be holding off a while on adding more weather models and layers until we’re fully confident everything is performing as intended.
Thanks for your patience while we continue to work through this - there are some lessons you only learn when a site starts to receive production traffic!
You can send them a direct link to the video - I’ll try to paste it here, but it might convert to a video player when I post it - if it does, then just click the preview to load the player, and then it’s the share button third from top on the right hand side: Introduction to Photo WX
Just watched the Photo WX video on YouTube and also shared it with my photographer friends and groups. This is a great addition to Photo Empheris and will really come in handy when doing my Full Moon moonrise photo shoots over Cleveland. Will the weather feature be added to the iPhone and iPad apps?
Thanks Ron - great to hear! Re iOS apps - yes :slight_smile. Here’s a preview - we have still quite a bit to do on this major new release, but progress is being made.
I just watched this as I missed the livestream - all looks great!
A quick question regarding the models, and in particular the UKMO model. The UK Met Office API site says that the UKMO 2km data is updated every hour, but I understand that data needs to be processed into a model and then uploaded, and it looks like that is normally at intervals of 3 hours.
However, when I checked at around 8am UK time this morning it was showing the most recent model as -8h and now (1pm UK) it’s showing -13h. Is there some kind of initial delay each day before models start getting built & pushed?
This looks fantastic, and I will definitely upgrade to a paid tier. However, I am unsure whether I need the web supporter or pro tier to get uk met office 2km and fog? Please let me know. Alternatively, is there a free trial so I can have a try-out and decide which level suits me best? Thank you!
Hi Nick - you’re correct, the model is published every hour. I was pulling the forecast every 3 hours, primarily while we were scaling the platform up, but I think it should be OK to update to this to an hourly update schedule at this point (it increases the processing load, but not the number of forecast layers we’re serving up).
I’ve made this change just now in our staging environment and will see how it performs today. If all looks good, I’ll get this in to production too.
Regarding delay: the Met Office only publishes their open data feed with an automatic 4 hour delay, which is something we could only address by adding a gazillion new customers and paying them lots of money. Maybe one day!
I just checked on Amazon S3 (where they publish the data) and the 11Z run showed up just 15 minutes ago, i.e. at 17:00 UTC, so I think the 4 hour delay starts at the end of the model run at best:
Hi Stephen, many thanks for all the info, much appreciated
Understood re the Met Office’s 4 hour delay, but I’m currently seeing a last model time of -12h (see screenshot) - is that correct, or am I misunderstanding what the -12h means?
Nick - fair question. I double checked and we had a slightly overly-pessimistic data availability delay configured for this model (9 hours). I’ve just dropped that to 6.5 hours in staging, so that should account for some of the delay. Looking at recent forecast runs, it seems like they arrive fairly consistently just over 6 hours after the model run start time (see my example in my last reply). I’ve allowed a 30 minute margin of error.
In theory, with these two changes, we should, at best, see forecasts available around 7-8 hours after the model run time.
BTW, if you switch the Time zone option to UTC, the model run badges show the raw model run time (e.g. 11Z, where 11 is 11:00 hours UTC).
Hi @nickjroper - quick update on the UK Met Office model. So it seems that while the model is updated every hour, it’s only every third hour that the forecast run goes out 54 hours. The intervening hourly updates only go out 12 hours.
I think we can include those, but I need to complete a bit more testing of the forecast run retention logic so that we don’t accidentally discard a 54 hour run for later 12 hour runs.
That said, the reduced data delay should help carve a couple of hours our of the schedule. That’s been working fine in staging and I’ve just made the change in production now.
I’ll post more here if/when we make further updates.
Thank you Stephen for all you do. It is all quite incredible. Just a thought, we have a major total eclipse coming up in a bit over two years in Northern Africa. Is there any kind of weather mapping available in that area that you might be able to include when we start getting close to that window?
I’m planning on adding eclipse paths to the weather site too - I think the cloud forecasts will be something folks will be keen to study in the days leading up to solar eclipses.
And for TSE2026, the cloud obstruction markers may be very helpful, given how low the Sun will be in the sky for that eclipse (see Cloud obstructions).
Thank you Stephen. I have also noticed in the WX view, if I am in full screen mode, the sun/moon cycle displayed at the bottom of the screen is correct. But if I switch to any smaller screen, it moves and becomes inaccurate for the selected time. In the regular panel, it just stretches/compresses the time line rather than moving it. Just an FYI.
