Lex Berman Data Model for Historical GIS 1 A Data Model for Historical GIS: The CHGIS Time Series Merrick Lex Berman China Historical GIS Harvard Yenching Institute October 2003 Abstract: The CHGIS project was established to provide a fully documented database of historical administrative units in China, from the formation of the first Chinese Empire (222 BCE) to the fall of the last Dynasty (1911 CE). Records in the relational database are also linked to geographic objects in GIS. In other words, for each record there are associated spatial objects in GIS layers which can be used to represent the historical places on a digital map. To accomplish this task, we had to clearly define what constitutes an historical administrative unit, and we had to devise a method of querying the resulting database to answer several fundamental questions about historical places, namely: can we find all the units that were in existence at a particular date? can we search for placenames and find the correct record among ambiguous results? can we filter the results by feature type? can we determine the place in the administrative hierarchy for a particular place? can we show how a particular place changes over time? The following description of the CHGIS time series database examines the ways in which the current data model satisfies these basic requirements, with illustrated examples of the basic problems encountered when trying to keep track of spatial objects as they--and their attributes!--change over time. 1.1 Rules Regarding Creation of Records in the Database According to the basic requirements for the CHGIS project, there were three primary types of change to be recorded in the database: change in placename change in administrative type change in location, area, or boundary This proves to be a sound basis for tracking named places, as it has also emerged as the basis for tracking places in large digital gazetteer projects, such as the Alexandria Digital Gazetteer Content Standard. [ADL-Gaz] The ADL standard defines the core elements needed for any digital gazetteer entry as: placename, category, footprint. These are directly comparable with our placename, feature type, and spatial object. For historical records, we need only add a valid date range to complete the minimum required attributes. Let us briefly clarify our definition of the digital gazetteer core elements with some examples. Our placenames must allow us to know that unique instances of places existed sometime and somewhere called Constantinople and Istanbul. Similarly, unique records must exist so that we can distinguish between multiple instances of the identical placename, like Paris #1, the capital of France, and Paris #2, the small town in Texas. Our feature types must distinguish for us that Paris #1 was of the type = national capital, while Paris #2 was of the type = town. Finally, the spatial objects have to be defined as unique records in various GIS layers, which are separated thematically and also, for technical reasons, into data types of point, lines, and polygons. Our record for Paris #1 may be represented by a single point of latitude and longitude coordinates,Lex Berman Data Model for Historical GIS 2 corresponding to the administrative office location. Paris the point would be placed at the City Hall building, for example. Or, Paris #1 could be represented BOTH by the point just mentioned AND by a polygon, which would be the area of jurisdiction of Paris, ie. the municipal boundary of the city of Paris. The CHGIS project records the three core elements for each historical place in the main table, to which a new row is added whenever one of the three core elements change. For our purposes each row is called an historical instance and is defined as a record with a valid period of time during which the placename, feature type, and spatial object all remained unchanged, lasting from a specific begin date to a specific end date. Should any of these three attributes change, a new record is added to the database. In Figure 1 we show actual records from the CHGIS database. Each of these records constitute a single historical instance and have their own unique ID number. Notice that Jianning Fu is listed three times. The first instance of Jianning Fu was valid from 1368 to 1454, and the second instance of Jianning was valid from 1455 to 1733. Figure 1: rows in the main table Since the placename and feature type are identical for all three instances of Jianning Fu, we need to provide some more information about what constituted the changes. To do this the records in the main table include a begin change type and an end change type for each row in the database. In the Figure 2 the relevant columns are shown: Figure 2: definition of change types From these columns we can now see that the first record for Jianning Fu appears in 1368 because its name was changed to Jianning Fu. In 1455, the jurisdiction area increased, which means that the spatial object or boundary used to represent Jianning Fu with ID # 90004, was no longer valid. As a consequence of this change, the new instance with ID # 90005 is created to reflect the new, larger area, which remains valid until the year 1733, when the area is reduced. The final record with ID # 90006 remains valid up to the last year of the database coverage, 1911. By recording the change types we can now distinguish why Jianning Fu, as a prefecture, has three separate instances, even though its name and administrative unit type did not change. 1.1 Areas of Jurisdiction Contrasted with Administrative Seats In most cases, the geography of administrative units in China are defined in two ways: by the area over which they have jurisdiction, and by the location of the administrative seat. Using GIS technology it is possible to define both a polygon (to represent the area of jurisdiction) and a point (to represent the administrative seat) for a single record. However, in the production of CHGIS datasets, we found that boundary changes and changes of administrative seat locations veryLex Berman Data Model for Historical GIS 3 often do not occur at the same time. In addition, there are various special cases, such as an office serving as administrative seat for a particular jurisdictional area but which is not actually located within that area. These and other factors forced us to keep track of the area of