Beta Weather Site now available

Hi all,

A beta of our new weather site is now available at beta.wx.photoephemeris.com.

Highlights

  • Supports multiple weather models (Arome France HD, ARPEGE Europe, ECMWF, GFS, HRRR, Met Office UKV, RAP)
  • Frequent updates (e.g. hourly for HRRR and RAP)
  • Multiple forecast runs available for comparison purposes
  • Initial selection of forecast layers

Why use this?

There are a gazillion weather sites already out there that work very well. Why are we bothering to do this?

The main driver arose from some frustrations I had with other weather services:

  • Non-interactive maps: many sites present static maps which are often frustrating to use and donā€™t let you zoom in on areas of interest
  • No animation over time: few sites animate smoothly over time with interpolation, and but the interpolation can help your ability interpret weather trends
  • Hard to ā€˜readā€™ cloud forecasts: often you canā€™t see multiple layers, or if you can, differentiating between low/med/high cloud is tricky. Photographers need a reliable ā€˜readā€™ on cloud layering
  • Lack of Sun and Moon context: we often care about the weather at the times of sun/moon rise or set
  • Time zone uncertainty: if youā€™re not busy trying to reverse engineer UTC into your local time zone, you may still be left wondering whether the displayed time is in your computerā€™s time zone or the time zone at a location shown on the map.
  • Convenience: wouldnā€™t it be nice to be able move seamlessly from your shot planning in the web app to the same map view for weather data with the same time and date selected?

This initial beta hopefully begins to address some of these points and lays the foundation for adding much more.

How can I use the site?

Open beta.wx.photoephemeris.com in your desktop browser (mobile support is coming soon).

If youā€™re not already signed into app.photoephemeris.com, youā€™ll be prompted to do so: sign in and youā€™ll be redirected back to the weather page.

From there, you can choose your weather model, forecast run and forecast layer using the controls on the right of the page (click the ā€˜side panelā€™ button at the lower right to open this if required). Then you can scroll through time using the interactive Sun/Moon chart at the below the map.

Non-PRO users can select the ECMWF and GFS models. PRO subscribers can choose from any available model. All weather functionality will be included in the PRO subscription at no additional charge.

Hereā€™s an over of the basic controls:

Controls

  • Blending: when enabled, the map shows an interpolated blend of two forecasts, e.g. if you choose 12:45pm with blending enabled, youā€™ll see 25% of the 12pm forecast and 75% of the 1pm forecast. If disabled, youā€™d see only the 12pm forecast. This works very well for layers such as temperature, but you may see ā€˜smearingā€™ for cloud and smoke layers (anything that is advective) - however, the smearing animation can help you visually intuit how things are evolving during the hour.
  • Preloading: when enabled, the app will pre-request an additional hour before and after the displayed times so that when you adjust the selected time, the tiles are ready and preloaded. This consumes additional bandwidth, but provides a smoother animation experience. We suggest leaving this disabled on slower connections or older computers.
  • Day/night terminator: this shows the approximate line of sunrise/sunset across the globe - very useful when evaluating weather at these times
  • Lunar day/night terminator: the same, but for the Moon. Note: the dark are corresponds to when the Moon is set - it can be easy to mix this up. Refer to the Sun/Moon altitude chart if youā€™re unsure what youā€™re looking at!

You can adjust the selected time of day by clicking and dragging left or right on the time chart below the map. The span of the time chart spans all available models and forecast runs. The hours covered by the selected model and run are highlighted in yellow:

Make sure you select times that are within the coverage of the model and run you choose!

Itā€™s a BETA!

Remember - this is a beta, so there will undoubtedly be issues. If you spot a problem, please let me know!

Also, please post your comments, feedback, suggestions, and questions too. Iā€™ll be updating the beta site regularly, so expect things to evolve as we go.

If the forecasts are slow to load, it could be because youā€™re the first person to view the data and the ā€œfirst pullā€ is always a little slower. The more people who view the data, the faster it will get for everyone, via the wonders of layer caching content delivery networks!

Iā€™ll post more highlighting forecast models and layers that I think will be of particular interest to photographers.

This is a very nice presentationā€“easy to use/understand. And this will be very useful.

One initial suggestion. Would it be possible to have an opacity control? I was looking at the multi cloud layer and trying to scroll the map to the point I am interested in. But, the clouds are so thick that I couldnā€™t easily see the underlying map.

2 Likes

Hi David - I agree, that would be a useful thing to add. Iā€™ll try to get it in the next couple of days.

1 Like

Hi @david - the opacity controls are now added: Low/Med/High buttons at the bottom of the control tray on the right.

1 Like

Wow, that was quick!

Thank you! This makes a world of difference.

1 Like

This is so very cool!!! I just donā€™t understand the details of what Iā€™m seeingā€¦ is it displaying the weather 1, 2, Xā€¦ hours from now?

What do the Forecast Run buttons mean/show?

E.g. using HRRR model, when you hover over the cursor over

[:clock3: -03] button, it says Oct 9 10:00 - 07:00
[:clock3: -04] button, it says Oct 9 09:00 - 07:00
[:clock3: -05] button, it says Oct 9 08:00 - 07:00
[:clock3: -06] button, it says Oct 9 07:00 - 07:00

What do those annotations/notes mean? Are those values supposed to be tied to the time youā€™ve chosen at the bottom? It doesnā€™t seem to beā€¦

Which Forecast Run button yields a display closest to current conditions?

Please add a hover annotation to the Cloud (low, med, high) buttons that shows the height range/definitions for each buttonā€¦ Like the info for the Met Model in the Using the Multi-Level Cloud Forecast Layer topic/post.

Is there a model for the US that shows Fog forecasts? Iā€™d would really find it helpful to know when fog is or will be on the ground at different elevations around Portland, Oregon. E.g. fog at Council Crest elevation ~1100ft equals low clouds when you are down by the river. Currently I use ODOT, local news station webcams, and some private webcams to see if the fog is at various locations. Fog here can be very different over very short distances or changes in elevation.

1 Like

Hi Victor,

Each weather model performs a forecast run periodically - some are hourly, some every 3 hours, some every 6. The Forecast Run buttons show which model runs are available.

A ā€œrunā€ means the numerical model was initialized based on up to date actual weather conditions - inputs typically include satellite measurements and ground measurements (e.g. METAR reports and weather balloon soundings). The model algorithm then forecasts forward from that time generating predicted conditions at a certain time interval, e.g. run time + 1 hour, run time + 2 hours, etc.

The buttons indicate how long ago the model run was started. The tool tips shows the date, time and time zone of the absolute time of the run.

Those times are not really related to the time selected using the time chart, other than youā€™re not going to find any data for times preceding the selected model run.

Generally, you would refer to the most recent model run (i.e. :clock3: -03 in your example above) to get the most recent forecast predictions. However, you can refer to earlier model runs to see how the forecast is evolving - this can be useful to see if things are appearing consistent between runs or not. More consistency between runs generally implies the forecast can be expected to be more reliable.

Iā€™ll be enhancing the UI bit by bit over the coming weeks adding things such as additional layer descriptions.

I havenā€™t found a dedicated fog layer as yet. The only model so far in the UI that outputs a fog layer is the UK Met Office 2km, but thatā€™s not going to help so much for Oregon.

Fog forecasting is a tricky thing it seems. Iā€™m looking into seeing if we could a derived ā€œfog potentialā€ layer using a combination of other layers, such as (dew point ~= temperature AND wind speed < x mph) or similar. However, thatā€™s not likely to capture all potential fog forming conditions or other factors that might inhibit fog formation.

Thanks so much for this very helpful detailed explanation. The only thing I still donā€™t understand is when i click on a Forecast Run button, e.g. if at 9AM my time, I press on the :clock3: -03 button for a particular model, how far in the future is the displayed forecast data Iā€™m looking at? I.e. if the button names indicate how long ago the model was run, how do I know how far forward the displayed forecast data is?

Hi Victor - check out the original post above - itā€™s explained there. Look for the yellow area in the timeline - that shows the time period available for the selected forecast run.

Regarding Fog. I was surprised that there was a Fog button for any model.

Fog here in Portland can be extremely localized, there can be thick ground fog along the Willamette River, but clear at my home (on the other side of the west hills) just 2mi away. Or clear at my home at 400ā€™ elev, but thick ground fog at Council Crest Park only a mile away at 1100ā€™ elev .

I donā€™t know what a reasonable/meaningful forecast interval for fog might be, even an hour advance notice could be helpful. Iā€™ve seen fog on traffic cams, only to have it disappear in the 20 minutes it took me to get thereā€¦

One proxy for a fog layer is the ā€˜visibilityā€™ layer included with HRRR (and also in NAM, which I hope to add soon).

Hereā€™s how it appears for the Oregon Coast:

You have to always remember that is an ā€œall causesā€ visibility layer primarily used for transportation and aviation safety purposes. ā€œAll causesā€ can include: rain or snow, near ground smoke, low cloud, but alsoā€¦ fog.

In coastal areas, particularly if youā€™re familiar with local weather patterns, you can often infer that the likely cause of limited visibility is fog.

I think I will be able to add an ā€œinferred fogā€ layer for HRRR based on a heuristic along the lines of ā€œif visibility is low, and thereā€™s no precipitation, and no smoke and no low cloud, then itā€™s likely to be fogā€. All those other variables can be checked independently in other forecast layers.

DOH! Canā€™t believe I missed that the Yellow section of the timeline shows the time period covered by the selected forecast run. It all makes sense and is so simple and straight forward now that I understand that! :wink: Thanks again!!!

1 Like

re: "One proxy for a fog layerā€¦ " wow!! that is a really cool way to use/interpret the visibility layer! Next time I see fog on a web cam or outside my window Iā€™ll check the visibility layer.

For us on the West Coast the Goes West GOES-West - Sector view: Pacific Northwest - GeoColor - NOAA / NESDIS / STAR and https://atmos.uw.edu/current-weather/satellite/ NW Satellite views can be helpful for disambiguating the current cause; fog often shows up really well on those images, particularly at night.

1 Like

Dew Point, Temperature, and Astrophotography.

The subtle change in colors of the temp and dew point layer donā€™t seem very useful for local purposes. What is of importance to astrophotography is when might my lens be affected by dew. Perhaps there is a way to combine the two factors plus nocturnal cooling into a dew warning index. I usually figure that if nighttime temps are going to fall within about 5 degrees of DP I need to take my lens warmer. Here in the SE US they almost always do - but I think zipping a timeline to see when it happens would be useful.

1 Like

Hi @takingsley - funny you should ask!

In our testing environment this week, we added Dew Point Depression (ā€œTDDā€) which is defined as temperature minus dew point temperature. At values near zero, the air is moist and condensation is likely. These are shown in blue on the map. Where the dew point temperature is significantly below the air temperature, i.e. more positive TDD values, the air is dry and condensation will not occur.

This video shows the progress of the dew point depression forecast through the hours of tonight - you can see the damp conditions spread as we get later into the hours of darkness:

For your use case of 5 degrees (F or C?) that would be anything in blue/green/yellow would be of concern for dew:

I can adjust all these color maps as needed, so let me know if you think this data can be better represented by other color choices.

Iā€™ll also be adding conversion to Fahrenheit in due course.

The Dew Point Depression layer will hopefully make it to the beta site next week.

Yes, very exciting!. I am considering degrees F but given that you are correlating the change to a gradation of colors it will be easy to see how quickly the change is occurring as well as when critical occurs. The colors make sense. I like it.

1 Like

Great! Iā€™ll post again here as soon as itā€™s available in the beta site.

PS. I discovered the correct acronym for this - ā€œTDDā€ - via this handy NWS glossary of weather terms.

I just pushed a small update to the beta weather site. Whatā€™s new:

Dewpoint Depression, HRRR

This is now available - see the discussion above. Simply put: red means dry air, blue means moist air with higher probability of condensation.

Preload option removed

The preload option has been removed. It only really made any difference to performance if you were the first person to view a newly generated forecast, but then had the unfortunate side effect of overloading the origin server with too many parallel requests.

NAM now available

The North America Mesoscale (NAM) model is now available - just a few layers to get started, but weā€™ll add more. Itā€™s a 5km resolution model and right now we have 48 hours available - I hope to extend that further out soon.

Weā€™ve been doing quite a lot of work on the back end processing capacity and have parallelised some of forecast map generation, which sets things up to add more models and data sources - what this space for further updates.