Move to a timeseries concept
We are currently based on single timestamps. It would now make sense to move to a timeseries approach: a single object would hold all the timestamps for all parameters for a given station (this requires issue 690 for a cleaner implementation). It would also hold metadata: the location, the sensor height for each sensor, a list of which timeseries exist, etc If the location changes, a flag could be set and the object would fill a vector with the time-dependent location information.
For cases like NetCDF, this would make it possible to export metadata such as if a given timeserie has been measured or comes from a model, etc