Ah, all based on another model. A model for which none of will see the code, and know how it handles the data that is fed into it.
All sensors use models. There are models for the sensor and models for the surface being sensed. All sensors and sensor models have error. The error factor is always there and unknown, because if the error was known it would be compensated for. The error ‘probability’ is larger for a larger size of the area being looked at. What are the odds that all of this so-called ‘change’ is within the margins of error?