Effective false lightning filtering

There is some filtering going on in the lightning sensor, but occasionally it still reports false lightning. Now one strike might just not be best limit to send out a warning or even let it count as a strike. Wouldn’t it be easy and better to just ignore or withhold the first strike, until there is another one in the next 10 or 15 minutes?

Remember that most probably the sensor already filtered out about a few hundreds before it let pass one as being a strike.

The alerts are on 3 distance levels and you should not get a new alert for another strike in a distance circle unless it is coming in the closer circle.

You ask to filter 1 out … but some will be asking to be alerted as soon as possible for al kind of reasons. Problem is, what is ok for you, won’t be for someone else … and from there you go to user defined options with no limits. The app is designed simple and will remain as simple as possible for the lambda user. If you want more it is possible via third party options of by coding more your side. All data is open and can be accessed localy or from WF servers

Maybe disable the alert on the app, use the IFTTT option and make something there suiting your needs ?


I understand all that, but the sensor filters only on a very, very short time scale. Currently it doesn’t come across as being very reliable.

The detector is meant as an early warning, nothing more. For more detailed lightning stuff you need to use way more sophisticated material like those used by Blitzortnung.

The Air works alone and not in a network … i does a best guess (and does a pretty good job if not faulty)


I wonder if it would be possible to add a “gain” setting, so each user could set the sensitivity to fit their own circumstances? I get a number of lightning detections during perfectly clear, sunny days; and I’ve tried moving the AIR sensor around and at various orientations. Nothing seems to reduce the number of false positives. So for now, I just turn off lightning alerts unless there’s a storm moving in.

I’m not sure a “gain” setting will work as some of the false lightning detections are closer then 35km.

There are several discussions here about the lightning sensor. They make informative reading.

I already spend many hours on reading all kind of useful stuff from this community. Very informative indeed.

False lightning is one of the issues that people haven trouble with. Because of the false lightning alarms, people are even disabling it, which kind of defeats the purpose of having a sensor. Other people try to rotate it, or move it around in order to minimize the false alarms. But I figured that the sensor is very sensitive and for some situations it is very hard or impossible to find a spot without false lightning detections.

My suggestion to only raise an alarm and suppress or withhold the first lightning from the data unless there is a second detection within the next 10 or 15 minutes, seems simple and effective enough and wouldn’t effect any real thunderstorm alarms. (except that the alarm might come a bit later for storms with very infrequent lightning)

I think its discussed or implemented here Lightning Alerts logic

1 Like

That discussion is about the logic that is currently implemented to prevent people from getting too many alarms when a real storm is present. The suggestion mentioned above is to prevent alarms when no storm is present.

So you know WeatherFlow is actively working on the firmware to correct false alerts.


It’s actually more sophisticated than firmware tuning. Our applied machine learning / continuous learning systems can inspect data and should (eventually) be able to disqualify and/or correct a range of suspect data. This is a double edge sword as instantaneous lightning alerts are really cool…but to apply a layer of data inspection to filter data takes time, thus delaying “instantaneous” alerts. Hummmm.


indeed, real time alerts are super cool, but for people that are unable to find a spot or orientation that suppresses the false alerts this feature might be really appreciated. Note it only delays the first alert and only by as much as the time between the first and the second strike. If the second strike isn’t detected, the first alert will not be shown at all (suppressing the false detection), if the second strike does arrive you can still report the first strike, but indeed a bit delayed.

Sounds like a good conditional filter to build with IFTTT.

1 Like

perhaps, but given the number of people with false detections, it is worth considering to be part of the supplied app.

It’s a small number and initially on setup. Once the Air is properly positioned false readings drop to zero.

I have not had a false reading all this year.

1 Like

Think I wrote this numerous times here and on other fora. If you have a little am radio (old school type) tune it to some free frequency and listen to ploks (strokes or distant disturbance)

And now walk around your house and appliances and enjoy all the noise (little transfo, telly, induction plate, computer, printer, your screen … … ) you will be amazed by all the things that produce what you call false positives.

Now that you know this, go back to where you air is and do the same around it (all direction and don’t forget up and down) and try to find the sweat spot where there is least ‘noise’ …

And don’t think you’ll find it within minutes … you’ll be busy for several weeks because what seems ok at a moment can be a complete mess 10 minutes later when a neighbour starts his plasma tv … a small shop uses some machine that sparks … don’t even hope for a sweat spot when you have a welder nearby …

You can filter all you want but that will never work in the end. Blitzortung members know all to well, after trying all kinda tricks, you finally end by reducing noise candidates to a minimum in your house (change transfo, light bulbs (especially those economic ones, most leds are ok btw …) and then you go walk around the house and try to negotiate with neighbours if nearby … and finally you start again to find the sweat spot …

Blitzortung devices can detect 5000 km and more in some cases … so imagine how to filter which such sensitive devices ??? And the little device in your air is pretty sensitive to …


Don’t know about sensitivity of WF gear. Still waiting for mine but I run the Blitzortung Blue receiver in Galway.
They are very sensitive to stuff like electric fences, switch gear and arcs. Even a cigarette lighter with a piezo sparker held close to the antenna will cause a false.
A lot depends on antenna positioning. A lot of noise can be subdued by rotating the antennae or moving it. It can also be shielded.

I know. But i don’t think it comes from my house. It happens when I’m not even at home. The only thing that turns on during that time, is the fridge. Now surely that will induce a small emp, but the air is outside and at least 10 meter away from the fridge. If it were the fridge, it most likely would generate more false detections. But even one or two false detections each day makes it almost useless.

Delaying only the very first alarm doesn’t do a lot of harm I would say. A storm with only one strike might be interesting, but is not worth giving a warning for anyway.

Funny story - I have had a few false lightning reports this weekend. Turns out that one of the fittings on the air compressor in my garage had a tiny leak, causing the compressor motor to cycle on/off about once per day and generate the spark that was picked up by the Air on the outside garage wall…