Tempest is massively under reporting rainfall. We’re having a rare early rainstorm here in Southern California, and I have several friends in the area with Tempest stations. Some of them also have CoCoRaHS gauges and some have Davis and Accurite stations as well. ALL of the non-Tempest rain totals are at least double what the Tempest reports. Two stations are triple what the Tempest says. In addition, 3 people reported very high winds, which didn’t occur.
I wrote the Teapot app for iOS that many people use, so I watch these things very closely. Weatherflow is no longer reporting raw (non-Nearcast) rain observations so all we have is Nearcast. I have sent emails to them asking for clarity. They know who I am, and I expected them to answer. No one has ever replied.
I can tolerate a small discrepancy in rainfall observations But not 100%, 200% or even more. At this point, Tempest rainfall and wind are completely unreliable.
I have been running 3 different brands of weather station for about 3 years, and have 2 Tempest stations (the second was sent to replace the first I was having the same issues with). The wind a rain are so consistently unreliable I have given up on both of them.
I knew that Nearcast is a manipulated value, but I now believe that the wind is too. There is no explanation for what I saw. 4 of us saw insane wind speeds of 40 mph or more when there was no significant wind. The next 2 days we all saw normal, credible wind speeds. All 4 of us are in different parts of town. Weatherflow was foolish to think they could crowdsource these things. Their sensors are simply not capable. It hasn’t improved over the years, it has gotten worse. Garbage in, garbage out.
The Tempest uses a Decawave DW1000 chip and have to convert values from it using an algorithm because this chip is not a standalone wind sensor, but a low-level timing engine enabling ultrasonic wind measurement via an algorithm that translates the data provided by the chip. So far the complexities of calibrating and writing an appropriate algorithm seem to be beyond the capabilities of Weatherflow engineers. This is dichotomous to other products such as those from Ecowitt who do not use a generic off-the-shelf commercial sensor in units such as the WS-80 or WS-90, but a custom-engineered unit tailored for their weather stations.
There is no manipulation to the wind data. Although, nearcast wind is something we’re developing for those who can’t install at a professional standard height but want to see a 10m wind value.
Erroneous wind gusts and direction readings are typically caused by something blocking the signal path of the ultrasonic anemometer. I suspect standing water or debris was causing a bad sample. That’s interesting all of you saw an erroneous gust spike. Would you mind sharing the station pages?
Anything that obstructs the ultrasonic signals can cause issues with a wind sample, e.g. water droplets or other debris like pollen, dirt, detritus, spider webs, etc. Calm wind conditions may allow water droplets to remain present in the signal path. There are filters in place that prevent incorrect readings from being accepted as valid data; while rare, it’s possible some afflicted samples can sneak through the filters. If and when erroneous measurements occur, it should only be temporary. The sensor should automatically correct itself and readings will return to normal after some time.
For your rain readings, it sounds like the Tempest’s haptic rain sensor may have missed some period of light rain. Post-install calibration can be very beneficial. If your raw rain readings are way off, please reach out to our support team and request for calibration. See: Rain Accumulation Calibrations
Regarding rain totals, I’ve also noticed that the Weatherflow API is unfortunately now reporting nearcast rain totals, even when that option is deselected in the app.
This is easy to verify if you send your data to any other service like WeatherUnderground. The rainfall totals on my WU station page do not match my station rain totals, but they DO match the nearcast values when I enable that option. This is disappointing, as my station-measured rainfall values have actually gotten quite accurate after a period of calibration. But it seems there is no ability to send the real values to other services… they always get the nearcast values.
Voted. But I’ll note that it does not matter whether you use 3rd party solutions for uploading data to WU and other places or not (I do, as the built-in Tempest utility to do this only updates 4 or 5 times an hour). Whatever utility you use, it’s going to pull the precip value from the Weatherflow API, and it’s always going to be Nearcast rain.
This is a completely bizarre lack of user control over their data, right? Isn’t the point of having a PWS and sharing its data that you are contributing your own data to a network? If we’re just sharing Nearcast rain, we aren’t sharing anything about our own data at all - we’re just sharing an arbitrary value calculated by the Weatherflow servers. What’s the point of us even sharing that?
If you run a third-party app that listens to the LAN UDP broadcasts, such as weewx using @vreihen’s weatherflowUDP driver, that app will publish what the hub emitted, which is ‘unaltered’ and does not include anything coming from the WF servers such as nearcast rain.
I’m pretty sure that there is at least one other third-party WF-T app that uses the local UDP broadcast data and uploads to various services including CWOP…
Well, I was going to post that something had improved with Nearcast but I’m glad I waited. In my original post I commented about Nearcast’s incredible inaccuracy as well as a wildly unreliable wind sensor. There were 4 of us in my area all reporting bogus readings. All of us also have Davis systems as well. We’re all ham radio guys and we routinely activate during heavy rain storms.
We determined that one of the 4 has a bad Tempest unit, It reports intermittently.
Yesterday we had a small storm move through, Davis reported .71 inches at my house, and it was similar at the other guts locations, Last evening the Davis and Tempests were either exactly the same or .01 inches different. Almost perfect agreement. All of us were encouraged thinking maybe Weatherfow had magically fixed the problem.
Nope! No such luck. Sometime overnight Weatherflow applied their Nearcast algorithm and my almost perfect .71 inches dropped to .44 inches. The other guys report similar drops.
SO… Tempest was reporting accurately and Weatherflow screwed it all up. As an app developer of an app for Tempest owners, this kind of crap is intolerable. I’ve come to the conclusion that their programmers are either incompetent or they are using some Dollar Store budget AI to ruin their data.
WEATHERFLOW - PUT BACK THE NON-NEARCAST PRECIP REPORTING IN YOUR API. MY CALIBRATION IS BETTER THAN YOURS.
UDP data is not a reliable solution. It will only work on your local network, and if your device hangs your data is gone. You have no way to fill in any data you may have missed. It is a very narrow approach to provide Tempest observation display. Using the API is more difficult, but it allows you to use a lot more types of data,
no chance WF will alter my sensor readings in any way
I have extremely stable power, so no risk of missed data and no backfill needs
the consumer grade lightning detector is not really any good anyway, so not having that reading unavailable (since it is server-only magic-math calculated) is fine with me
It’s nice we have that option. Davis removed that option in their newest consoles. WF gets a thumbs-up for giving us options.
