Rain seem off (auto-calibration system not yet deployed)



For as much we know the details, yes. The learning process is multiple and will never cease to adapt to your situation locally, but it needs several rain episodes to start. Several of us are seeing the gap slowly closing between the Sky and other calibrated instruments.

Also WF coders are still making the AI better … they are now using data from more and more stations as they become available but you can imagine the amount of data to handle is equally growing :slight_smile:


how many rain events for it to work the right way ??? got 0.81 from sky — other vantage pros got 0.11 –


There is no fixed number for events required but you will notice an improvement over time. Thanks for your patience!


I have noticed improvement in my readings over time. Very pleased with what I’m seeing so far.


Has been so dry up here in Ontario that the rain sensor hasn’t had much action. Hoping for rain for the lawn and trees, and for Sky testing here.

split this topic #70

5 posts were merged into an existing topic: Rain sensor and vibrations


I also am getting abnormally high rain readings, had alot of rain today, local stations indicate around .86 while the Sky is saying 3.34 in past 24hrs.


Same here. It’s not accurate at all


Same here. Still getting very high readings. For instance, yesterday Sky reported 3.111" of rain. Standard rain gauge next to it reported 0.65" . There has been no noticeable improvement in the accuracy of the rainfall readings since it was installed. Start and end times appear to be accurate and there does appear to be a somewhat linear relationship of the readings with the rain gauge suggesting a simple calibration may all that is required in my case to at least get it close to reality. (Note: Installation is standard with Sky mounted on about a 8 foot high metal pole cemented into the ground.) I probably should mention that very heavy rain rates are more inaccurate that lighter rain rates. It is little hard to pin down since most rainfall of late has been convective in nature. Sky often rates the rain intensity as “extreme” when I would call it no more than a normal heavy down pour. (I understand that WeatherFlow continues to work on correcting this previously identified problem which they cite as each individual station haptic sensor setup causes the rainfall amount to vary depending on a number of factors related to how it was installed.)


We had a day of constant slow rain. All of the near by stations reported approx 16-18mm mines report 48mm , not even close ![:frowning:]


My rain and wind readings are off. It is showing way more rainfall than what is taking place. Last major rain recorded 88mm but we received way less than that maybe 20mm if lucky. And the wind readings are always less than Environment Canada states.


Hello mister insurance.

Can you tell us more about your setup, your station’s page, to what you compare etc.
A weather station is a measuring of a very local situation and many factors can influence it easily.

It is known that rain can be off and WF is working on calibration to make it better over time. Wind calibration kicks rather quickly in once it gets several windy moments coming from all sides.

Again share some more details so we can try to help you.


Morning - just adding to the observations. Last week our Sky reported almost 200% over the rainfall. This week its notably under reported (comparing to a Vue - side by side - so should be similar enough to compare).

Vue - 16mm
Sky - 6.4mm

The 16mm figure is much closer to the actual rainfall…

As a side note - i have two cloned sites to monitor - one with a

Vue - http://www.finchamweather.co.uk/vue/

Sky -http://www.finchamweather.co.uk/



Hi Eric


The unit is mounted in my backyard on a pole about 6 feet above a shed. It regularly shows more rain than we get and way less wind than we normally get. This

is in comparison to Environment Canada’s local weather monitoring station.




Regarding rain, I won’t say much as indeed it is still not ok. In heavy rain conditions it goes way of for now and WF is as quickly as they can analyse the data make it better.

Wind is normally ok for most. I have no clue how much distance there is between your Sky and th reference you give but know I have 2 sky units 70 cm apart and even that is enough to give different readings. Every single obstacle,hieght of the device can really change a lot the result.

Other testers have several different stations (like VP2 etc mounted on the same pole etc) and will tell you that instant readings differ though slightly … but there is no way to read exactly the same. It is more on average that 2 units set very close will be equal.

here is a page from WF explaining in depth wind


This is my first post. I am glad to see there is some discussion on the rain ‘sensor.’ The numbers I have been getting from the sensor don’t even come close to what my official CoCoRaHs rain gauge measures. I cringe at the thought of a smartweather user providing their weather stations rain observation to the NWS. I’ve seen it report both extremes… 0.06" when I had 0.25" in the gauge to yesterday when 3.43" was actually 1.32" in the gauge. This is a HUGE problem. The data is useless and wrong, and new users should be warned of this.

Weatherflow "Continuous Learning" Algorithm

My rain sensor reports about 6X what my CocoRahs rain gauge is reading.

Also, I suspect my anemometer is reading low. We had a recent dust storm event where my highest gust reported was 22 mph even though there were multiple trees down within a few blocks, and this was a very wide scale event - a monsoonal haboob in Phoenix, AZ. The event was August 30th about 7:15PM MST ( UTC-7). Here is the link for that: https://www.wunderground.com/personal-weather-station/dashboard?ID=KAZSCOTT344&cm_ven=localwx_pwsdash#history/s20180730/e20180730/mdaily

What do I do about these things?


Hi. Thanks for your feedback and welcome to the forum! Please see my previous post, above:


I bought a NOAA rain gauge to compare my results after noticing how wrong Weatherflow was. They don’t need AI to solve the problem, although it could work. They need proper benchmarking and a good set of observations to make a linear regression fit. That should have been done before our units shipped.

They needed to calibrate against old, well known instruments. From all my emails to support, it is not clear they did that sort of benchmarking and adjustments. To make it worse, it’s not clear they understand that AI wont fix the problem without the right strategy.

For your reference, my rain gauge is between 1/3rd to 1/4th of the WF measurement consistently. Happy to provide photographic evidence as needed.

Current WF measurement 1” today. Current rain gauge .31 inches. Rain type- thunderstorm


Last night my sensor recorded 5.414" of rain, actual CocoRahs 1.05". And what is it with that 22 mph? I estimated gale force winds associated with a supercell thunderstorm but the peak gust was 22 mph with peak actual wind only 14 mph. In fact, since June peak wind gusts generally strangely average around 22 mph per hour. I had assumed that the sensor simply isn’t high enough off the ground although well exposed to an large open field but now suspect the readings are indeed too low. Unfortunately, don’t have a wind device to compare to be sure.