Features properties vs Observations
About the distinction between Features properties and observations.
A video introduction by Kathi Schleidt of the observation concept can be found here: https://www.youtube.com/watch?v=bYDSgs2fKLk
It notably proposes an explanation of what is an observation vs a property of a feature. The picture below makes it clear.
Feature property: one object called the feature (here the smily face) is identified and is associated to a property (here Expression = Happy)
Observation: a dedicated object called the observation is set. It is more exactly an act which itself have some properties that enable to define the context and also the result of the observation act (here that the Smily Face looks to be of a yellow color according to Kathy's experience).
Feature properties are a direct way to associate an information to an object. The association is "absolute" in the sense that it is always true, no matter when, who/what is involved in describing the object. For example, the name of a person, its date of birth.
On the contrary, the observation concept offer the capacity to contextualize an information that is provided. For example, the time, the location, the method, the observer: that is to say conditions in which the information was obtained and may have an important impact on the validity of the result.
For those reasons one can see that:
-
feature properties perfectly fit domain where the knowledge of the objects are well mastered. This is typically the case in industry that propose to build from hand or machine made well known components.
-
observation is very relevant in domain where the knowledge is incomplete, or when only a small part (either in terms of space or time) of an object is studied. This is typically the case in science where each information must be kept in its context before (possibly) deducting a more general law or rule.
Information obtained through a process of observation, measurement, interpretation or even calculus fit the concept of observation and shall be exposed that way.
A basic need for semantized objects is to be able to associate them with information.
Approach 1: (Direct) Feature properties, as in GeoSciML Lite
In this simplified version of GeoSciML, a direct association between the "ObservedProperty" and the "Result" is provided.
Approach 2: Association of one ObservedProperty to a Measurement, as in GroundWaterML2
Here the gwUnitThickness of the GL_GeologicUnit is intended to be provided as a Measurement, following ISO19156.
To realize this association, the "ObservedProperty" shall be associated to the identifier of a Measurement Object. A clean way to do that is to have the measurement exposed through an [OGC API](OGC APIs) and use the link to it.
The picture below illustrated it with a link to a WFS service.
Extracted from https://www.youtube.com/watch?v=L1L2u3u35cY
Approach 3 : Association to an OGC API
A drawback of the second approach is that you need to have all the ObservedProperties described in the feature and to associate the results one by one.
A solution to avoid this constraint is to directly associate the feature to an "Observation Oriented API" that will list the available observations associated to the feature.
The following request to an OGC SensorThingsAPI enable to list the "DataStreams", that is to say collection of observations associated to the Object of interest (here a borehole or BhTrajectoryThing):
This is the prefered approach for this IE.