I've been working with Kalman filter for a while. In this particular case I'm trying to estimate the velocity or rising air at a specific location (a square km) on a hot summers' day. I expect the velocity to be more or less constant during the day.
Multiple measurements are performed during the day (by aircraft flying over the square km). I have some idea of the accuracy of the measurements performed by the aircraft, but there are many factors that can influence the situation that are hard for me to quantify. I figured I might be able to derive something from the spread of the samples around the model (in my case a fixed value) that I use for prediction in the kalman filter?
I wonder if there are ways for a Kalman filter to take into account how well the data seems to fit the model? Are there ways to take into account the variance between the samples and the kalman estimates? If so, how is this called? And how is this used in practice?