D. Digital map database development
5. Digital map integration
(a)Introduction
2.316. A census mapping project should take advantage of all suitable cartographic data sources.
These are likely to be stored in different formats, using varying map scales and cartographic projections.
Integrating these heterogeneous data sources requires considerable knowledge of GIS data integration methods if the goal is to produce a complete and seamless digital census map database. The following sections discuss the most important methods that facilitate digital map data integration (for more details see Hohl 1998).
(b) Georeferencing
2.317. The coordinates captured with a digitizer or scanner are relative coordinates measured in the x and y direction usually in centimetres or inches from the data input device’s origin—usually the lower left corner. If several adjacent map sheets are digitized, they will clearly not fit when their digitized map sheets are later pasted together in the database. In fact, they will be drawn on top of one on other since they are all referenced in the same segment of the digitizer’s coordinate system. Similarly, existing georeferenced GIS layers for the same area or coordinates collected using a global positioning system will not be compatible with the digitized maps since they are referenced in a
real-world coordinate system. For this reason, the digitized point and line coordinates need to be converted from digitizing units to real-world map coordinates are measured in metres or feet (see, also, annex II). As pointed out earlier, this step can be done in most systems either at the start of digitizing or after spatial data automation has been completed.
2.318. Nearly all GIS packages provide the functions necessary for georeferencing. The user needs to specify a number of control points for which the real-world coordinates are known. Based on the input coordinate data in digitizing units and the real-world output coordinates, the system computes a set of transformation parameters that perform the following transformations (see Figure II.32):
• Translation. The geographic feature is moved to a new position simply by adding (or subtracting) constant values to the x and y coordinates. The offset will usually be different for x and y;
• Scaling. The feature is enlarged or reduced by multiplying the x and y coordinates by a factor for the x and y coordinates respectively. The scaling is usually done relative to the origin of the coordinate system;
• Rotation. The geographic feature is rotated about the coordinate system's origin by a given angle.
Rotation will make sure that the resulting digital map has the proper orientation even if the paper map has not been correctly aligned on the digitizing board.
•
Figure II.32: Translation, scaling, rotation
Rotation about fixed point (1,1) * x' = x cos(320) - y sin(320) y' = x sin(320) + y cos(320) Scaling
x' = x * 2 y' = y * 2 Translation
x' = x + 5 y' = y + 4 Input feature position
* Requires translation before and after rotation about the origin.
Rotation is positive anticlockwise.
2.319. Note that the shape of the digitized features does not change in this transformation as it would in a
projection change. Only the relative size and orientation of the objects is modified. After the correct translation, scaling and rotation parameters have been computed,
the system applies these parameters to all point and line coordinates in the database. The output is a map that looks very similar, but is now registered in the proper coordinate system that was used in the production of the original base map (see Figure II.33). It is important to ensure that the error in this operation is minimized. The
system usually provides information on the error in the estimation of transformation parameters for each point, which is helpful to detect errors in specifying the control points’ real-world coordinates. More technical details are given in an in annex II.
Figure II.33. Map in digitizing units; map in real-world coordinates
0
0 x x
y y
50 25,000 35,000
40 240,000
232,000
Input map in digitizing units (Centimetres)
Output map in real-world coordinates (Metres)
2.320. A serious problem occurs when the map projection and coordinate system of the source paper map is unknown. Unfortunately, this problem is encountered quite frequently since many paper maps, especially thematic maps, do not contain this information. Two options available in this case are to try a large number of possible map projections (the standard projection used in the country’s mapping programs is a good candidate), or to use so-called rubber sheeting.
2.321. Rubber sheeting requires a large number of control points that are well distributed across the map.
Sometimes, a digital map of country and administrative boundaries, or any other clearly defined points that are also present in the digitized map, can be used to find links between corresponding points. The system then uses the coordinates of the input and output coordinates to compute higher-order polynomial transformations.
Typically, the error introduced in rubber sheeting is quite large, and this operation should therefore be avoided if at all possible. However, in some instances, where the input maps clearly do not conform to any well-defined projection, rubber sheeting is a viable
option to make use of available geographic information.
A good example in the context of census mapping is the georeferencing of hand-drawn sketch maps. Section F, Annex II, provides a practical example of georeferencing that illustrates the process of converting, for instance, a digitized map into a properly referenced digital database.
(c) Projection and datum change
2.322. Related to the transformation process that converts the coordinates of digital map features without changing their shape is projection change. When converting from one projection to another, the shape and distortion of map features do change, although the changes may be all but imperceptible at large cartographic scales.
2.323. Projection change is necessary when maps that were digitized from different map sheets need to be assembled into a seamless database. Often, maps issued at different map scales use different projections. In other instances, a mapping agency may have changed the standard projection used for mapping in the country, so that older map sheets may use a different projection
from those map sheets that were revised more recently.
Similarly, the mapping agency may have modified the geographic datum, which establishes the reference framework for cartographic work in the country, so that older topographic maps, for example, use a slightly different coordinate system than do newer maps.
2.324. Projections and geographic datums are discussed in more detail in annex II. It will be useful for the census mapping agency to have a trained cartographer on staff or to have access to experts from the national mapping agency who can advise on the most appropriate strategy for reconciling projections and related issues to produce a consistent national census map base. The actual technical steps of projection change will require relatively little effort, since all commercial GIS provide the required projection change functions.
(d) Coding
2.325. After the previous steps have been completed, the map database consists of a structured set of points, lines and polygons. Each geographic feature—that is, each point line, or area—has a unique identifier, which is used by the system internally. This internal identifier is not usually accessible by the user and should not be modified externally. What is needed is a more meaningful identifier that can be used to link the geographic features to the attributes recorded for them. For the enumeration areas and administrative units, this link is the unique EA or administrative identifier that is listed in the master file of all geographic areas relevant in the census.
2.326. How this identifier is entered is again software-specific. It can be added during the digitizing process by entering the identifier before the feature is digitized. Or it can be added at a later stage by selecting the feature interactively and adding the identifier through a menu interface. For polygon features, some systems require the user to add a label point that is contained in each area unit. While conceptually simple, coding may require considerable time and resources.
(e) Integration of separate map segments
2.327. The purpose of a digital mapping project is to produce a seamless database for a large region or an entire country. At medium or large cartographic scales (e.g., 1:250,000 or larger), base map information will be contained on separate topographic map sheets. These are digitized separately and the resulting digital map sheets are joined in GIS (see Figure II.34).
Figure II.34. Joining adjacent digital map sheets
A B
C
A B
C C
B
2.328. Usually this is straightforward. But the match between map sheets may not always be perfect. Features that span both sheets—for example,, roads or boundaries—might be displaced at the map boundaries (see Figure II.35). Errors could have been introduced during digitizing, or the errors may actually be present on the source map sheets. For instance, adjacent map sheets may have been produced at different times, so that newer features such as new roads do not continue across map sheet boundaries or they are represented by different symbols.
2.329. The problem is particularly serious if there is no complete coverage for the entire country at the desired map scale, so that map sheets of different scales with different feature densities, need to be integrated.
This problem is often encountered when integrating map sheets at the urban/rural interface, where large-scale city maps need to be matched to smaller-large-scale rural maps. Owing to the variations in cartographic generalization, features may or may not be present on the smaller-scale maps, or their symbology may be different in the two map series. Integration of such maps requires considerable judgement and experience.
2.330. The process of fixing these errors is called edge-matching. It is usually performed manually, involving a considerable amount of editing. If the displacement is not too large and the features are compatible across map sheets, features can be connected using automatic edge-matching functions provided by some GIS packages.