In late August 2018 we launched our new Continuous Learning (CL) auto-calibration system for Humidity. The system performs a quality control inspection and ensures that your humidity values are both unique to your location, but also correctly governed by the mesoscale conditions in your immediate area on an ongoing basis. Here’s how it works:
The Humidity CL system begins by comparing raw humidity data from your AIR to numerous data sources in your immediate area. The initial comparison data set is passed through a custom-tuned clustering algorithm that rejects outlying data points (ie. bad reference data, a device located inside when it is marked as outside, etc). Once qualified, hundreds of humidity data points from many days across a range of conditions are plotted and then evaluated with a linear regression analysis. If the coefficient of determination (r-squared value) of the comparison data set in your immediate vicinity is greater than 0.75 (in sufficient consensus agreement) then and only then the CL system will approve a daily calibration to be applied. The calibration includes the mathematically determined slope and intercept for the linear calibration curve.
The humidity data you see displayed from your station comes directly from the sensors in the AIR and will reflect all the micro-climate nuances in your yard. The CL system is always collecting and evaluating comparison data (that’s why it’s called “Continuous Learning”) and approved calibration tweaks are applied once per day.
What you may or may not realize is that humidity sensors in all weather stations tend to drift out of calibration over time. The CL system not only prevents this drift, but also results in a homogenous data set across the network allowing for consist analytical comparison. Redefining smart.