Frozen Sky Devices?

I am located in Southern MN and we are in the middle of the deep freeze. Right now I have -19F and strong winds. The sky is reporting light winds of less than 2mph. Out of curiosity, I located several other Weatherflow stations in the region and found they are all reporting similar light winds. Real wind observations are 20 - 30 mph with higher gusts. Is it likely ice crystals are forming on the sensors or do they just not work in temps this cold? Seems more than coincidence that all area stations are having similar issues.

We have gone below -20F on several occasions the past couple weeks, but this is the first with wind.
My station ID is 2360

LRM_EXPORT_95775278929389_20190129_184322139 LRM_EXPORT_95743645797942_20190129_184250506 LRM_EXPORT_95716445367484_20190129_184223306 LRM_EXPORT_95682909781976_20190129_184149770 LRM_EXPORT_95826081020567_20190129_184412941

1 Like

My Sky is covered with snow. On top and sitting inside the sonic sensors. That throws off all readings.

Do a search for snow.

Here are a couple screenshots of the wind and temp charts. Temps started dropping around 2:30. Winds kicked up just before that. After 3pm, despite even stronger winds, the graph shows winds dropping as the temps dropped.

Screenshot_20190129-185956_Smart%20Weather Screenshot_20190129-185928_Smart%20Weather

I am aware of snow. Not climbing on the roof to check, but from ground level, no snow on top and best I can tell none inside the opening. Wind readings were working before the drop in temps and the last light fluffy snow was Sunday night.

Send an email to and ask that they look at your station.

Hi Trent, our support team will get in touch with you - thanks for opening a ticket.
I wanted to share some food for thought here…
Remember wind chill is calculated when the temp goes below a certain threshold and when the wind goes above a certain threshold. When average wind speed is <5mph there is no wind chill calculation. The wind chill values you are seeing on the news, from other sources, etc. are taken from meteorological equipment nearby which is probably mounted at a standard height of 10m (33ft) above ground level with clean, open wind fetch. Winds are much quicker at this altitude relative to winds at the ground surface level and faster winds correlate to a lower wind chill temperature value. So, you’re not going to see same wind chill reported by the news or from other airport weather stations, etc. As long as SKY and AIR are reading correctly, you’re getting accurate weather data from your location, not somewhere within ~5-10 miles from you. Also, one thing to keep in mind is that wind chill is not a measurable value; there are dozens of ways to interpret how humans feel cold but we use a standard calculation method also used by NOAA.

Regardless of whether or not your SKY is reading low wind speeds, the average wind speeds it was reporting were mostly below the 5mph threshold until you started seeing some of the stronger winds and gusts. Only then did your wind chill calculation start kicking in. Keep in mind we calculate using the average wind speed, not the gusts.

That being said, the electronics in AIR and SKY are rated to -40 degrees. Everything should be working fine but we’ll ask our technical team to take a look at the data.


The issue has nothing to do with wind chill accuracy. It has to do with the wind sensors not working in these cold temps. When the Sky started reporting low wind speeds, the actual observed winds were increasing. Until this cold snap, Sky has been working wonderfully…even when it’s snowing. Here is a chart from a different brand PWS located about 1 mile from me. Normally, our stations track fairly close. The winds reported at this station were not as high as the NWS, but our town is in a valley, so that is normal. Less than 2 mph on the Weatherflow last night was not accurate.



just a little side note. The speed of sound varies with temperature. I assume the algorithm weatherflow uses takes care of that. If not, it might explain a drop of a few percent in windspeed (definitely less then 10%)
Speed of sound at 21 degree Cesius is 343 m/s and at -20 degree celsius it is 319 m/s.
It depends a little bit an humidity, but at sealevel that change is less then 0.3% between 0 and 100% humidity. But let me repeat myself, I think weatherflow does take care of it as it is such an obvious thing to do.

Thanks for the report. We’re going to dig into your data and see if we can figure out what’s going on. This cold weather is a real boon to test data. One question for you: can you tell if there is any ice, frost or frozen snow accumulation within the wind sampling area (the “gap” at the top) of your SKY? This could be affecting the wind readings.

Indeed! The principle behind the wind speed measurement in the SKY, while it does measure the time of flight between transducers, it does not depend on the local speed of sound!


No snow that I can see. I can’t get on the roof to confirm 100%, but no precipitation when this started yesterday and the snow on the ground is so dry, it doesn’t stick to anything.

I thought it emits a sound pulse on one transducer and measures the time for it to arrive on the other receivers. If wind blows through, the sound travels with the wind and arrives earlier. If it is that simple, shouldn’t it depend on the temperature, because at higher temperatures the sound pulse also arrives earlier (due to increased speed of sound).
O wait, you probably measure the same thing in the opposite direction.
so if Time_forward and time_backward are measured, and T_zero is the time it takes when the windspeed is 0 (Time_forward would be equal to time_backward). Time_forward = T-zero - windspeed, Time_backward = T_zero + windspeed. So Time_backward - Time_forward = 2 * windspeed. (but only the component of the windspeed in that particular direction and asuming that you measure at exactly the same time, or windspeed is constant). T_zero is not part of the equation, and the result is independent of temperature.

Funny that you could actually calculate the temperature from Time-forward (ignoring the influence of humidity on speed of sound), Unless you do the subtraction in hardware before you measure anything. Or if you have accurate temperature and speed you could even calculate humidity.
Speed_of_sound = 331.4 + 0.6*Temperature + 0.0124*Relative_Humidity


Exactly, this is the way WF does it. In detail for each pair of transducers, hence they use transducers instead of senders and recievers.

Yes, quite funny to use the data to calculate the other values. Especially, as those values are not directly measured in Sky, but in Air.
Maybe it is or could be used for crosscheck the air values within the CL-System, @dsj ?

The wind readings started to come back to life this morning once the temps reached around -8 F. A quick check of other WF stations in the region showed wind readings returning there as well.

Considering -8 is well below the freezing mark and the fact it was snowing at the time the wind readings came back to life, it is probably safe to rule out snow obstructing the sensors. Hopefully as others are suggesting, it’s just a matter of calibration for the cold temps.

1 Like

My wind data has also been performing badly with the cold weather. Very high winds speeds and wrong direction. Lots of blowing snow so that could also be an issue.

there might be snowflakes in between, that could disturb it.

interesting question just came to mind. Apparently windspeed is effected by rain/snow/ice in the sensor area. But how does that work? They measure speed of sound from sensor A to B and from B to A, and from the difference the windspeed is calculated. So if there some rain in between and the distance between A and B appears to be smaller, so would the distance between B and A, and the difference would be the same. There must be something else involved to make rain effect the windspeed. But what??

Thanks for the follow-up. It’s certainly possible this issue could be caused by a component of the anemometer failing at such a low temp. We’ll dig in and see if and how that can be addressed.


Following up on this… we have indeed identified a defect present in at least some of the ultrasonic sensors out there that limits their operation to roughly -22 C despite being rated to -40 C. We are working with the manufacturer to rectify the defect.


Weather is difficult. Someday I may actually understand a wee bit of this.


9 posts were merged into an existing topic: Sky battery life