I think this is where current thinking diverges the most. But I also don't really know how to describe the division. I'll try anyway.
Some people see the objective as creating a "classification system" as a series of class mappings. Whether and how that is "instantiated" is a separate concern.
Some people (including myself) do not see a strong distinction between a classification system, and geospatial data artefacts where a classification system is put into practice.
I think my persepctive is borne out of experience in implementing other's classification systems (such as ALUM) as spatial data. That is, what people want and need is a geospatial dataset, a map, of land use—and the land use has to adhere to some hierarchical system of classes using particular labels, and the decision of how to do this is made by interpreting the class definitions and coming up with rules as to how data is best combined to fulfill it.
For example, I don't see how it can be possible to come up with a land use classificaiton system without also deciding the appropriate geographcic scale or object of classification. If you say that the object of classification is "property parcels", well property parcels are a concrete dataset. It is therefore inherently a spatial dataset. Therefore decisions about what format is suitable, whether it is raster or vector, etc. are perhaps not inseparable from, but at least related to, the particular system of hierarchical classification.
I think this is why my current draft "procedures" pre-suppose some kind of spatial data artefact. This is why I talk about "validation" implying that this exists. Because that is the ultimate end. It's not what we are producing per se; in fact I can imagine someone implementing a classification system by simply driving around in a car and looking at things to decide what class is most suitable to describe the nature of the thing. However I can't imagine the utility of a hypothetical classification system that can't be realised ultimately as spatial information.
I think this is where current thinking diverges the most. But I also don't really know how to describe the division. I'll try anyway.
Some people see the objective as creating a "classification system" as a series of class mappings. Whether and how that is "instantiated" is a separate concern.
Some people (including myself) do not see a strong distinction between a classification system, and geospatial data artefacts where a classification system is put into practice.
I think my persepctive is borne out of experience in implementing other's classification systems (such as ALUM) as spatial data. That is, what people want and need is a geospatial dataset, a map, of land use—and the land use has to adhere to some hierarchical system of classes using particular labels, and the decision of how to do this is made by interpreting the class definitions and coming up with rules as to how data is best combined to fulfill it.
For example, I don't see how it can be possible to come up with a land use classificaiton system without also deciding the appropriate geographcic scale or object of classification. If you say that the object of classification is "property parcels", well property parcels are a concrete dataset. It is therefore inherently a spatial dataset. Therefore decisions about what format is suitable, whether it is raster or vector, etc. are perhaps not inseparable from, but at least related to, the particular system of hierarchical classification.
I think this is why my current draft "procedures" pre-suppose some kind of spatial data artefact. This is why I talk about "validation" implying that this exists. Because that is the ultimate end. It's not what we are producing per se; in fact I can imagine someone implementing a classification system by simply driving around in a car and looking at things to decide what class is most suitable to describe the nature of the thing. However I can't imagine the utility of a hypothetical classification system that can't be realised ultimately as spatial information.