How do we think about effects?

A metrological approach to uncertainty analysis for an FCDR requires us to consider what different effects are and to evaluate the uncertainty associated with each effect. This information may not be readily available, particularly when dealing with historical sensors, and may need to be estimated using relevant information.

For example, we may be able to estimate noise relatively directly, but to do so we need repeat measurements that are often not available from the ground. We can, however, look at measurements of on board calibration targets to try and estimate noise in the time series. In contrast, the best way to evaluate some effects may be through modelling.

This process of critically evaluating sources of uncertainty and effects will not only lead to a more robust uncertainty analysis, but often leads to more insight into the workings of the sensor, which can lead to an improved instrument calibration.

Of course, there are times when information is not available. In these cases, we need to make an educated guess based on what we do know. Some effects may have negligible effects on the overall uncertainty, but, even so, the decision to exclude an effect should be made based on suitable evidence.