Wind gauge accuracy

the tempest samples the windspeed every 3 seconds. gust is just calculated as the maximum value it measured. The graph is very confusing as the period over which it calculates the max, varies with the zoom level. In my opinion it should be something like the max wind speed in the last 10 minutes and the graph should just average those values. same with lull. So zoom in and compare again.

1 Like

That is very enlightening because until you pointed this out, I had assumed the zoom buttons just expanded the timeline. So, when I wanted to compare the two stations, I had to hunt down matching 5 minute intervals because the Ecowitt graphs (and tables) plot data in 5 minute periods.

One thing I did try a few weeks ago was sitting under them with 2 phones, each phone displayed real time data from each station. The problem is the Ecowitt seems to just sample every minute, so I had to factor that into my observations. Again, I noticed that there was a large difference in the gust speed with the Tempest about half of the Ecowitt reports, and because the Tempest wind direction was every 3 seconds it was all over the place and hard to judge. However, the average for direction between them seemed to be pretty close in the end. I also had a piece of flagging tape on the poles for each, and I got the impression that the flagging tape was more often at odds with the Tempest than the WS90 unit. So I started to play with the wind direction offset in the Tempest and it seemed to help, but in the long run made the average values even more inaccurate so I reset it back to defaults.

In terms of windspeed, I got the seat of the pants impression that the actual speed of the gusts was higher than the Tempest was reporting, but of course because that is a subjective observation with me sitting about 7 meters underneath them I cannot say by how much.

With a Kestrel WeatherMet on the way, I am evaluating these units (when combined with a Starlink installation) for suitability to be left on a ridge in winter where the only important factor is wind speed and direction for guides/avalanche forecasters in backcountry lodges to select the days ski runs and avoid lee slopes where cornices, sastrugi, and snow deposits have potentially built up due to aeolian factors and thus represent increased existential risks and avalanche potential.

Thanks for your valuable observation on that graph!

1 Like

My Tempest wind speed reports are very inaccurate. We live in a very windy area, and during some events we have gust over 50 mph, some much higher. Neighboring weather stations (neighbors have stations visible on WeatherUnderground) are reporting speeds that align with those of the National Weather Service stations at the airport (just 5 miles from me). Yet, my Tempest is saying the highest speeds are about 8 mph. When my palm trees are nearly blown over, I think it would take more than 8 mph wind to do that. My Tempest is mounted on a pole above the roof line of the house with nothing impeding it.


I would love to see your comparisons with the Kestrel Met.

They told me last July they were sending me a KestrelMet station to test, a few weeks ago, they said that a station ‘was on its way’, but alas no tracking and no station up to this point. Once again, they are not responding to emails either.

In the meantime I have been doing some seat of the pants calculations. They seem to indicate that the Tempest is under reporting wind gusts by a factor of about 28%.

Oh wow! I would reboot the Tempest sensor. If that doesn’t work, you might need to contact support directly.

Thanks for the suggestions. Have rebooted a few times during the various testing regimes, but it would not hurt to try again as I can’t recall if it had a reboot at the current location. As I will eventually do a review of these consumer grade stations,I have been collaborating with one of the Weatherflow meteorologists, so I sent him some questions yesterday. One of the questions was with respect to calibrations, as it would seem simplest if the end user could make these calibrations. Another question is regarding alerts. I understand there is a retail subscription based app that also has a free version that will push wind alerts, but it seems natural that most end users would expect to be able to configure some basic alerts from within the native interface.

1 Like

Good deal! I do know that a properly working Tempest ultrasonic should match up with an accurate reference very closely.

I did a reboot about an hour ago. It is rather gusty here right now so I should have a clearer picture in a few hours. Initially however it does seem like the gusts are under estimating by about 40%.

Has the Tempest had wind from all four quadrant? I understand that it needs this to calibrate. Maybe reboot it again and then rotate it slowly before mounting it.

That’s interesting. I just moved it to a lower position that has less potential to sway during the gusts. I have an LED Shop light that is adjacent the solar panels to try an experiment to see if it will charge the battery tonight when it gets dark.

While I was in the process of setting this up, I got this alert:

|### Ecowitt Weather [
|to jmkellas

Wed, 08 Nov 3:13 PM Wind: Wind Gust at Ws90 is 46.4 km/h.

Hi @jmkellas,
I am very interested in wind gauge accuracy and so I mounted the Tempest above the front of my car to test it and I considered that it was accurrate to my requirments. But testing gusts depends on your time span required. If one instrument requires an instantaneous value to measure the maximum value and a different instrument requires the gust to be susstained for 3 seconds or 10 seconds or any other time period then the same gust will record different values for different time periods. Because the same gust or a maximum speed can be measured in many different ways.
I believe the Tempest uses hundreds of ultrasonic measurements during each 3 second period to determine the average strength and direction for that 3 second reading. If a different instrument uses a shorter time period it would have a higher maximum speed because it is not including the slower speeds during the 3 seconds.
Being able to watch the 3 second readings from the Tempest can be valuable if you are trying to analyse the behavior of the air.
I do wish to analyse the air for my paragliding decisions so I regularly look at my one minute resolution graph and can tell what the thermals are doing. The angle and strength some times makes a wave pattern on the graph which tells me where the thermals are building and decaying. Thermal behaviour also occurs on icy mountains. If I analysed the 3 sec values into my graph I could improve my understanding further.
If I was trying to determine the behaviour of snow on a mountain I would be more interested in how detailed I could ‘see’ the movement.
But I guess it is more practical to keep it simple.
Now another important item of data is to know what direction your gust was from. Usually the gust is from a different direction than your average wind during the time period. It might be valuable to know that the gusts come up a different side of the mountain than the gentle wind for the rest of that time period.
Much of historical analysis bases its requirements on the data that was available when the person creating the formula came up with a method to measure it.
Now that the Tempest can deliver 3 second data of both speed and direction there is oppertunity to understand what the air movements are doing more than most people consider.
Back to keeping it simple…what time period do you use when measuring a gust to then compare accuracy?
Cheers Ian :slight_smile:

Thanks for the thoughts Ian. For purposes of choosing safe terrain for skiing, there are many variables such as slope profiles, temp, and specific gravity of the snow. However, fundamental to all of this is the initial consideration, were the aeolian factors have to be present. EG, if a strong wind blew from the west all night during a snowfall event, then lots of snow was deposited on a lee slope, If that slope was convex for example, then skiing over the brow of that slope could have undesirable consequences. Wind also fragments snowflakes and can cause high density slabs that can fracture to depth and slide on an unstable layer, etc., etc. So where a minute may influence your decision making, it’s more like hours and the average of the prevailing wind that is important to an avalanche forecaster or guide.

My interest in this discussion is more moot in that if I see a gust that is strong enough to knock the neighbors fir tree onto my house or car, I’d like my Tempest to send me an alert so I can move my car, or consider spending the night in the back bedroom (we have fatalities from falling trees in this area almost every year, but most visitors worry about being mauled by a bear, which is maybe a once in 20 year event). Whether that gust was sustained for 50 milliseconds or 3 seconds is probably irrelevant to the tree if at any time it reached 100km/hr and the regolith was not strong enough to hold the tree’s root ball in place.

Sample frequency and post/report frequency/rate are two completely different things!

Sampling every 3 seconds with an ultrasonic sensor is extremely limiting. The sample rate on professional instruments is much higher, typically 10Hz (ten samples every second). This is then averaged and posted every so many seconds to minutes. This way you have numbers such as max, gust, wind speed and average over period. I find this way during violent storms the peaks are captured and are much more realistic to what someone would experience if they were next to the instrument.

The Tempest is sampling many times during the 3 seconds and then reports every 3 seconds.

Somehow the reported gust is the same as the average during that 3 seconds. if it did 30 measurements during that period, it could have reported a better gust value.

What are the specifics though?
Is that shared?

I remember a thread that discussed how the ultrasonic anemometer takes many samples to determine the 3 second value which David confirmed.
I have not found any specific detail on how many samples are taken each 3 seconds.
A discussion regarding the Sky and how it determines the maximum and minimum wind speeds to display can be found here

And differences to spinning cups here

And if you want to know how wind direction is calculated Gary discussed his findings here

I have not found where David confirmed that many samples are taken to calculate a 3 second value. If you search the forum far enough I am certain that he confirmed it. But he never said exactly how many samples contribute to a 3 second value, that I have seen.
Cheers Ian :slight_smile:

1 Like

I have found the same thing with under reporting gusts even compared to Davis stations. This seems to be more true after 20mph. WF needs to figure out how they can improve on this sensor on the unit regarding the gusts.
Statements member(s) have made from another weather forum.
“The tempest also underestimates gusts vs the Davis, sometimes by a significant margin.
Totally agree especially regarding the wind gusts.”
Other forum member reply back “Had a wind gusts of 51 mph during a thunderstorm on my Davis Vantage Vue and the Tempest peaked at 32 mph. NOT ACCEPTABLE! Sold it the following day.”