Feedback mechanism to confirm/dispute readings

Has there been any thought of adding a feedback mechanism to some Notifications and readings? For example, if there’s a report of rain, give me an option to confirm/reject the report. And following a rain event, let me enter the reading I get from my other rain gauge. Of course, then use this feedback as part of the WeatherFlow adaptation process. I only mentioned rain, but something similar could be done for other readings, for folks with alternative measurements for those items.

1 Like

Good idea. In fact, this is the basis for our applied machine learning / continuous learning / auto-calibration system. We compare the raw data from your sensors against a range of trusted data in your immediate area. This analysis is done dozens of times daily. The result is a non-linear calibration which is applied to your sensor if R-squared values meet a certain threshold.

We unleashed the first version of the system to all station last week for temp and humidity. It’s a really cool system and initial results have been very positive. It will get better and better over time. We hope to deploy the system for the other parameters (except wind, which doesn’t have reliable reference sources as wind varies widely due to siting) by the end of this month. Stay tuned.

8 Likes

We had a rain event yesterday and the readings from my station were much better. It recorded 1.23 inches, and my other gauge (and other stations in the area) recorded 1.5 inches. Gettin’ better!

I noticed that the rainfall rate indications have changed from quantitative (X in/hr) to qualitative (light, medium, heavy). A recent update, I assume. Is this permanent, or are there plans to try to go back to exact readings? I imagine it’s difficult to convert haptic input to instantaneous rainfall rates.

That would be a great thing to have configurable - it’s one of the things I like best about my VP2 measurements. Fascinating seeing X.XX inches/hour and comparing that with my timelapse weathercam and seeing the dark approach for that interval.