The reference period will be used to represent the typical climate. Therefore, the data you choose should be consistent over the period, without unnatural changes due to how the data was constructed that do not reflect the actual climate.

You should find this out from your data source before using it for your background reference.

For example, CM SAF's long term Climate Data Records are carefully quality controlled and calibrated across all the sensors involved in the time period to prevent artificial changes over time due to factors such as changes to instrument or improvements to retrieval algorithms. Therefore, they are well suited for calculating reference periods because they should be homogeneous.

Whereas their Interim Climate Data Records, which provides more up to date data, could be used for a review time point, but it's consistency and continuity has not been thoroughly checked and therefore is not suitable for a reference period.

Another example is the CRU gridded station interpolated rainfall and temperature data, which is noted as not being specifically homogeneous at all points for various reasons, including potential inhomogeneity in the original station data, and different number of stations contributing to the final data at different points in time.

This highlights the importance of reading technical documentation and publications on the data source, which often provide helpful information about the strengths and limitations of a data product. If in doubt, contact the data provide help or support desk directly.

As you have now seen, there are no fixed rules that you can apply when choosing your reference period. But there is advice you can use as guidance to make your decisions. It is important is to clearly document the decisions you make in defining your reference period.

The Bureau of Meteorology, Australia has a good example of this in their webpage on Periods used in climate maps and statistics where they state statistical (extreme values and variability) and practical (limited historical satellite data available) considerations involved in producing maps and statistics on their website.

By clearly communicating the choices you make, it means that others can assess the usefulness of your results for themselves, and place them in the right context, even if they disagree with your choices.