Fix constraints on relationship between gauges and measurements/predictions measure_date is the time of measurement, not date_issue. There can be only one measurement at a time for a gauge location, independently of gauge versions with differing validity. The 'measure_date' of a prediction can well be beyond the validity of the gauge, since a new gauge version might appear within the time range of the forecast. There can be only one prediction at a time per date of issue, independently of gauge versions with differing validity.