If you already read the about the OGC WFS standard, it should be pretty obvious that adhering to the standard alone is not enough for making services and clients using different data models interoperable. The data models also need to be agreed on at least within user communities handling the same kind of datasets.
Both the data providers and the users of data provided by different providers benefit from standardisation of the data models: For data providers it makes publishing easier, because fewer different data models need to be supported for the delivery. The data users are able to retrieve similar datasets from different data providers with the same software components or applications. However, if too much many different data types are fitted in one model, the model probably becomes too generic or difficult to understand and use.
The use of Observations & Measurements (O&M) data model for providing the atmospheric and oceanographic datasets is required by the INSPIRE technical guidelines is a good compromise between a general "object" model and a data-specific models. FMI's Open Data WFS service returns all data using the O&M features.
The central entity of the O&M is an Observation event. Don't be fooled by the name, the O&M Observation is designed to capture events of making estimates of some properties of a real-world entity. It does not matter if the estimation is trying to evaluate the values of those properties now, in the past or predict them in the future. The real-world entity is called feature of interest in the O&M model, and the evaluated property is called observed property. Information on how the estimation was made is recorded in another entity called the procedure or "process". The estimated value of the properties is recorded as the result entity. The Observation entity is linked to all these entities, and itself provides the information about the time the estimation event happened, and possibly some other parameters used for the estimation process for the specific observation event.
The O&M model also provides an intermediate Sampling feature entity, which is used to model the geometry and properties of a "tool" used for sampling the values of the actual target feature. For a temperature observation of atmosphere near the ground in the center of Helsinki (the feature of interest) the sampling feature could be the Kaisaniemi weather observation station where the thermometer is located.
The INSPIRE guidelines define seven specialization classes of the basic Observation class to be used for delivering O&M data. These classed restrict the use of some of the properties, like the phenomenonTime (single instance in time or a time period) and sampling feature geometry (point, area or volume in space) and the type of the Result (simple timeseries or a GML coverage). The following feature types are allowed for atmospheric data:
Of these type only the GridSeriesObservation and the PointTimeSeriesObservation types are currently used by the FMI Open Data WFS. The PointTimeSeriesObservation is used for timeseries data for a single location, like weather observations from single observations stations between a given start and end times. The GridSeriesObservation is used for delivering radar reflection and precipitation data and numerical model data, like the the weather forecast model values or values from various numerical models related to oceanography.