Call for Feedback

[jdemaeyer]

Bright Sky has entered a second round of public funding and will be in much more active development over this summer!

I have some core development targets planned for this time:

• DWD Alerts / Unwetterwarnungen (#106) and other area-based data like UV index (#119)
• New weather parameters, in particular precipitation probability for forecasts (#95)
• Area-based querying (#97)
• Radar data (#144)
• Daily aggregates (#117)
• Improved reusability of our parsing components (#89)
• Improved documentation
• Low-hanging fruit issues (#126, #135)

But there is plenty of wiggle room for other development! If you have ideas and preferences for missing features and improvements, please create new issues, vote :+1: on existing issues, or reply in this discussion issue! :)

(/cc'ying other feedback issues active in the past year, pardon the spam)

111 #124 #125 #129 #130 #131 #132 #133 #134 #137

[poetaster]

Alerts are an area that I would love to implement as is a Rainradar directly in app. In my basic wrapper of Bright Sky, I added a rainviewer.com WebView to add a radar. Thanks for your efforts and congrats on another round of funding!

[poetaster]

I'm not sure if it's of value for you to know, but I have about 800 users of https://github.com/poetaster/harbour-dwd which is a SailfishOS native (QT/QML) mobile app.

[kossaFounderblocks]

Congrats for the funding. Would love to see the precipitation probability and radar data for the rain radar. That would basically complete my use case

[jdemaeyer]

Thanks @kossaFounderblocks! The precipitation probabilities are already available as of last week as an experimental / undocumented feature. Would love for you to try them out and give feedback over in #95 if you find the time! Radar and alerts are on the roadmap for the coming weeks :)

[jdemaeyer]

Weather radar is now live :) #144

[poetaster]

It does mean I'll have to rebuild the current version :) Thanks!

[nils-werner]

• Maybe I missed it in the docs, but I would love to see JSONP support, as an alternative for CORS.
• And also maybe sunrise/sunset times?
• Precipitation type, not just amount
• Wind speed seems to be sampled every 2 or 3 hours, not hourly. It would be great if that could be changed.
• Completely out there and probably not even provided by the DWD: lightning strike coordinates, like lightningmap.org

[jdemaeyer]

@nils-werner, thanks a lot for the feedback! Could you elaborate a little on the wind speed sampling issue? Here's a graph with hourly wind speeds from the past (blue) and forecasted (orange):

While the forecasted values often change hour-to-hour, they do look a little more stair-like, I guess that's what you're referring to? My guess is that that's a just a result of the impossibility of predicting wind in detail, unfortunately. Even a hyperspecialized site like windfinder.com defaults to showing wind forecasts in three-hour intervals...

[nils-werner]

yes, that's what I mean... it looks more irregular than the other plots so I was wondering if there was something weird going on.

[Bragor95]

Hey, this is such a cool project. Thank you so much for your work! :) I'm planning to use the API for my Masters' thesis on forecast based Peak Shaving.

My understanding from the docs is that on the weather endpoint if I enter a date that lies in the future I will obtain forecasted data whereas when I enter a past date I will receive actual realized values. Is it somehow possible to obtain the forecasted weather from past dates as well? I'm asking because my ML models will in practice only be used on forecasted data, thus I would like to train the models on this data as well. If this ain't possible I should still be fine.

[jdemaeyer]

Hi @Bragor95, thanks for reaching out and for the kind words :)

My understanding from the docs is that on the weather endpoint if I enter a date that lies in the future I will obtain forecasted data whereas when I enter a past date I will receive actual realized values.

That is correct! Timestamps that are in the very recent (less than ~90 minutes) past will sometimes still contain (obsolete) forecasts, as the DWD takes some time to publish its real-world measurements. You can find out whether a record comes from a forecast or a measurement by checking the observation_type of the associated source.

Is it somehow possible to obtain the forecasted weather from past dates as well?

Unfortunately not through Bright Sky, as that'd be far too much data for our little database (roughly 5 GB of new data per day). You can recover some outdated forecasts by directly parsing the raw data from the DWD open data server (MOSMIX S, MOSMIX L), e.g. using dwdparse, Bright Sky's parsing backend. Even the DWD deletes their forecasts after 48 hours, though.

Good luck with your thesis! Feel free to reach out to me at jakob@brightsky.dev any time if you have any questions regarding Bright Sky :)

[demetz]

Hi @jdemaeyer

thanks for your amazing work.

Maybe as an addition to your list: integration of heating degree days and cooling degree hours