A COMPARISON BETWEEN
SNAPSHOT AND COMPOSITE CHANGE DATA

by

 

MARK G. SCOTT

A research paper submitted in partial fulfillment of the requirements for the degree of
MASTER OF SCIENCE in GEOGRAPHY Portland State University 2001


CHAPTER 5

CONCLUSIONS

This study draws 4 important conclusions based on examples presented in Chapters 3 and 4.

- Database models are adapted to the behavior of phenomenon.
- Database models reflect availability of data sources.
- Snapshot databases can be enhanced by temporal joins.
- Composite databases will become more useful with temporal GIS operators.

The snapshot and composite models both are useful in determining the fact of change. Database models are intended to reflect the behavior of the phenomena. One is static through time and the other is defined by events in time. There is no need to have a temporal database built for source data where there is no means of representing events as database objects. The surface water source data contained no evidence of the events that created or modified the mapped phenomena. There is no evidence within the source data that indicates a specific event caused an object to change from one state to the next. For example, how did the dredging and filling turn Swan Island into a peninsula? The source data did not indicate any particular dredging that made the feature a peninsula. These source data used alone lacked the pertinent information required for the construction of a composite database. Therefore the snapshot model works best in instances where the source data contains no representation of the events that cause change. To compensate for the lack of temporality in the snapshot model, a temporal join can be used to link event information.

The snapshot model can work well with surface water in Portland, Oregon, but what about other themes that exhibit different spatial and temporal characteristics? When historic source data contain event driven changes, the model must integrate how these events destroy one state and replace it with another. The snapshot database is not be effective in this case because it does not have the ability to model individual object events. The robustness of the database would increase by placing such data into a more complex data model.

The composite database is a more robust representation than the snapshot database because spatial entities and time data are used in a different way. Because of this capacity, the composite model is better suited for source data that contain evidence of discrete, independent, or repetitive events that are discrete and cause change to geographic features between map versions. The database models various land cover changes in direct relationship to the events that caused change. For example, clear-cutting events are directly responsible for the removal of vegetation and subsequently cause change to the object state that follows in the database. Composite data include time dependent attributes including age. The database updates the value of time-dependent attributes each time an event changes an object state.

Mapped surface water used in Chapter 3 does not readily contain any attributes, it just simply exists, or it does not. Air-photo data for the Gray's River watershed provided evidence of event data making it a more complex source of information than the surface water data in historic maps for Portland, Oregon. Because the air-photo data source contained the required event evidence and spatial topology, it was better suited to the composite data model. The composite model is the only method providing a way to represent the chronology of events independently. The composite data model has the flexibility to use classification schemes, model age, measure size, shape, and location relative to transaction and valid time. The added robustness comes with a cost - much larger processing overhead and increasing spatial complexity over time.

 

5.1 Recommendations for Future Use

"The maps only provide a picture of the situation, so changes between the dates have to be inferred from other sources" (Hooke and Redmond 1989, 80). Improvements to the database could be made in two ways. First, in order to determine more accurately when an event that triggered change occurred, valid time could be incorporated into the database. Second, event data could be added to the composite to expand the range of these data beyond 1996.

Adding valid time attributes to the present temporal database increases the level of information to what Armstrong (1988) describes as the historic database. This is the most complex database design for temporal information because it relies on valid time. This added functionality comes at the expense of researching, obtaining and integrating valid information into the database. These data would most benefit from information obtained from such sources as:

- The Washington Department of Natural Resources timber permits, which contain information including the location, extent of activity, and date of application.
- The Washington Department of Ecology timber records, which include pertinent site plans and harvest schedule.
- Timber processing documents, including the amount of harvested material.
- Tax assessment records for yearly harvest activity.

 

Further, the composite database as it presently exists does not contain truly dynamic data because queries are not updating the attribute information. Static values make inference to events as recorded in the database not in the real world. Ultimately we need to simulate temporal operators by experimenting with algorithms that continually update the values contained in the database by the last event. If this is done, then programming the database can be done with SQL.

Lastly, we have spatial operators that perform functions on the database, but no temporal operators. A temporal operator asks the question to the database. The answer must be given and modified to the present. If this operation stores the value in the attribute table, all objects have the same attributes. The example queries suggested earlier would benefit if a temporal operator were able to first return the present time and then the value of forest age after subtracting the clear-cut data. Instead the database requires the 2001' in the query expression.


<-- Chapter 4
Return to Menu
Works Cited -->