Dark Sky Developers Help

By Adam Grossman on May 28, 2013.

Today the third major version of Dark Sky hits the App Store.

Along with a bunch of performance improvements and bug fixes, the app is now powered by the Forecast API, which means a few things: better accuracy, intelligent text summaries that cover more than just precipitation, and support for the UK and Ireland.

In addition, we’ve added a cool new feature: the Personal Weather Reporter, which lets our users submit their own hyperlocal weather observations to help make our forecasts even better.

Intelligent Text Summaries

When we launched Forecast last month, a big focus was on making our text summaries for hourly and daily forecasts as intelligent as possible. And now we’re bringing those improvements to Dark Sky.

We’ve collected historical weather observations from tens of thousands of ground stations going back almost a century, and have used that data to figure out what “normal” conditions are for any time of day and any day of the year. When we identify that the current conditions aren’t normal, we’re able to roll that into our text summaries. For example, most people don’t really care about things like wind speed and humidity unless it’s exceptional for their location. Using the historical data, we know what “exceptional” is for any time and place. Instead of simply saying “Clear” when it’s not raining, for example, we might say “Partly Cloudy and Humid”.

In that way, we can deliver the most relevant information without cluttering up Dark Sky’s simple interface.

The British Isles

Over the past year, we’ve been inundated with requests to bring Dark Sky to the UK, and now we (finally) have.

The UK Met Office has given us access to their radar data, which covers England, Scotland, Wales, Ireland, and the surrounding areas. Dark Sky can now be downloaded from the UK and Irish App Stores, effective immediately.

Personal Weather Reporter

Dark Sky’s forecasts are generated using data primarily from doppler radar stations, informed and sanity-checked by a suite of global computer models. The problem has always been how to verify the truth of the forecasts on the ground. Sometimes light rain fails to show up on radar, other times the radar indicates precipitation but it fails to reach the ground. Occasionally our radar cleaning algorithms act up and either remove too much noise, or not enough.

All of these contribute to a certain amount of error in our forecasts.

Now we have a way of verifying forecasts by allowing users to submit their own weather reports (I’m resisting the urge to call it “crowdsourced cloudsourcing”). You can record the intensity and type of precipitation, the amount of cloud cover, and even submit a photo of the sky or weather conditions using the built-in camera. Users are encouraged to submit reports as often as they can, even when we nail the forecast.

In time, we think the accumulated database of user observations will have a huge impact on our forecasting abilities, both in Dark Sky and in Forecast. To bring up the Personal Weather Reporter, just launch Dark Sky and shake your phone!

Available Right Now

We’re super excited about the great new stuff in Dark Sky Version 3, so go ahead and update or download it now, and let us know what you think!

“Storming over Two Trees” image by …-Wink-…