If alerts are driven by model data they will generally not be as good as those issued by the forecasters at weather services using local knowledge. This needs to be considered in the design as a matter of safety and for general user education. It could potentially have a legal liability aspect in some jurisdictions that should be considered.

It would be good to add some kind of advice on a risk acceptance page when creating an alert. Any generated alerts could include a link to the local weather service's alerts web page.

I'd love to be able to set (and clear) the default view in Android. For me this is Rain. I have the particle animation and isobars on my rain view which gives most of the quick/basic forecast information I want. This would save me a couple of presses just about every time I load the app. I spend a fair bit of time outside and around here wind is not the key weather variable for assessing the day. This would be different for other users and locations.

The placing of isobar labels seems a bit too random. Maybe this could be looked at.

AROME would be the most interest here, plenty of mountains and coastline here in Tasmania

I've been thinking about the temperature display. It has a fundamental problem that to display a geographical range, and therefore a wide range of temperatures, you end up with very little useful detail. As an alternative, a temperature anomaly display would be better, since you would be showing a deviation from a standard local temperature. I'd like to see (eg) a color change when a front passes if the temperature scale goes from the polar to tropical temperatures the effect of a front is barely visible. If you are using a localised scale rather than a global scale it becomes possible to show more gradations.

There are multiple ways to do this, for example, deviation from the same time yesterday (or last few days), or from the climatic average for the place/month/time of day, or from the climatic average for this place/time of year. Different approaches have advantages. Including the diurnal change in the base will show the deviation better, but won't show the things heating up and cooling during the day which might be a bit weird. Another option would be to set the temperature scale shown relative to a local value.

I'd love to see the sinusoidal sea level curve. This gives a quick impression on the times and amplitude of the tide. Also, in keeping with Windy's visual style :)

@robbrooks33

To add to @idefix37 fronts are sometimes unmistakeable phenomena, but sometimes a bit vague and sometimes very vague. There may be a single front or there may be several lesser fronts succession, or just a weather change over a period of an hour or two to cooler temperature and a different wind vector.

There'a judgement and even a psychological component to what constitutes a front. A human forecaster will take the human impact into account in choosing whether to draw a front, how far it extends horizontally and when it has degraded sufficiently to drop it. It's a tough job for a NWP model - more of a neural problem - but expect it one day.

It's probably better to look at rain, temperature and wind change then decide if that's what you call a front.

I'd love to be able to login with Google (Facebook, for others I guess). It's simple, quick and doesn't require a password.

Every additional password I need a management problem and a security risk. I think twice about register on sites without a Google login. From your side it add a little bit of user data quality, and makes you less of a hack target for password harvesting.

Fantastic site. Cheers.