A FRAMEWORK FOR TEMPORAL GEOGRAPHIC INFORMATION

GAIL LANGRAN AND NICHOLAS R CHRISMAN

University of Washington I Washington USA

ABSTRACT This paper defines the critical components of cartographic time and compares tempo
ral and spatial topologies. Because time is topologically similar to space, spatial data structuring
principles can be adapted to temporal data. We present three conceptualizations of temporal geo
graphic information and select one as the most promising basis for a temporal geographic information
system. This conceptualization creates a spatial composite of all geometric information (at all times),
where each object has an attribute history distinct from that of its neighbors.

INTRODUCTION

Parkes and Thrift (1980) distinguish geography from geometry because in

geography, space is indivisibly coupled with time. While perhaps an understate
ment, this observation highlights the importance of time to geographic analysis.
Despite this fact, cartographers have not achieved notable success in representing
the temporal information so vital to understanding geographic processes.
A limited set of static and dynamic mapping methods are available to repre
sent geographic change or motion graphically (for a discussion, see Muehrcke
1978, p. 138141; and Moellering 1980). T o analyze spatiotemporal information
digitally, however, current cartographic theory and methods are inadequate. One
reason is the strong allegiance of digital maps to their analog roots. For example,
Sinton (1978) describes geographic data as having three components - theme,
location, and time. T o measure one component requires that a second component
be controlled and the third component be fixed (Table 1). Not surprisingly, on
maps it is the temporal component that is usually fixed. Muehrcke's (1978)
explanation is that cartographers maintain their composure in the face of a
continually changing world by making static maps of relatively static phenomena,
thereby shifting the burden of dealing with temporal phenomena to the map user.
Because this practice has been transferred from the analog to the digital world,
the data structures in use today are designed for, and limited to, the representa
tion of static phenomena.
How to express change and motion graphically is a problem we leave to
future discussions. This work's focus is on digital methods of storing and manipu
lating sequent states of geographic information. O u r intent is to select a concep
tual model of geographic temporality that can serve as a basis for effective digital
representation. T h e first sections of this paper define the components of carto
graphic time and compare their relationships with those occurring in space. T h e
paper then presents three conceptualizations of time and offers reasons why one
is more promising for future developments.
BACKGROUND

Time, a phenomenon that can be perceived only by its effects, has intrigued and
preoccupied philosophers through the ages. Time is commonly viewed as a line
without endpoints that stretches infinitely into the past and the future, although
NICHOLAS R CHRISMAN is an Assistant Professor in the Department of Geography, University of
Washington. GAIL LANGRAN is a PhD candidate in the Department of Geography, University of
Washington. T h e thoughts presented here were influenced by the perceptive comments of Ric Vrana,
Morgan Thomas and Tim Nyerges. Partial funding was provided by the University of Washington
Graduate School Research Fund. MS submitted May 1988; accepted August 1988
CARTOGRAPHICA Vol 25 No 3 1988 pp 1-14

GAIL LANGRAN AND NICOLAS R CHRISMAN

Table 1. THE REPRESENTATION OF GEOGRAPHIC

FORMATS (EXTENDED FROM SINTON I 9 7 8 ) .

Soils data
Topographic map
U.S. Census data
Raster data
Weather reports
Flood tables
Tide tables
Airline schedules

DATA

IN

VARIOUS

Fixed

Controlled

Measured

Time
Time
Time
Time
Location
Location
Theme
Location

Theme
Theme
Location
Location
Time
Time
Location
Theme

Location
Location
Theme
Theme
Theme
Theme
Time
Time

coherent arguments exist for such alternate topologies as multiple parallel lines,
tree structures, circularity, discreteness, and nonexistence (Rescher and Urquhart
1971 and Newton-Smith 1980 provide excellent discussions of this topic). Cartog
raphers, however, can sidestep debates on what time is, and instead focus on how
best to represent its effects.
Temporal Data Processing in Cartography
The cartographic literature includes one previous attempt to structure spatiotem
poral data. In 1978, Basoglu and Morrison designed a data structure to retrieve
the county boundaries of a given state for any date since that state achieved
statehood. T h e structure is overtly hierarchical, as shown in Figure 1. States own
counties, and counties own boundary records. Each boundary record holds a line
segment, and the time interval during which that segment formed part of that
county's boundary. This hierarchical structure cannot recognize that one line
segment might no longer bound a particular county, but remain in use as the
boundary of a different county through historical subdivision. In addition, the
structure cannot produce an answer to the query "What boundaries changed
between two given dates?" T h e structure's fundamental weakness, however, is
that it cannot ensure that all mapped space belongs to some county at all possible
times. This paper provides some solutions to these problems.
Aspatial Temporal Data Processing
The database systems literature has witnessed an explosion of publications con
cerning temporal and historical databases. With few exceptions, however, re
search has focused on aspatial applications. Nonetheless, a review of this literature
is instructive.
The goal of a temporal database is to make the time dimension accessible to
users. Clifford and Warren (1983) describe such a database as "a model of the
dynamically changing world," in which data are never 'forgotten.' T o design such
a database, one must develop procedures by which data are superseded but never
deleted, thereby avoiding what Copeland ( 1982) has termed "the agony of delete."
Virtually all temporal database research builds on the relational database
model. T h e relational model stores information in matrices called tables. Each
entity is represented by a matrix row, called a tuple; matrix columns represent
entity attributes. When temporal information is modeled in this way, the design-

A FRAMEWORK FOR TEMPORAL GEOGRAPHIC INFORMATION

States

FIGURE 1. Morrison and Basoglu's (1978) hierarchical representation of U.S. county boundary changes.

GAIL LANGRAN AND NICOLAS R CHRISMAN

er's dilemma is whether to propagate new database versions at the table, tuple, or
attribute level. T h e earliest work generated a new version of a table each time one
of its cells changed; however, researchers quickly progressed to the more spaceefficient approaches of tuple- and attribute-level versioning.
The relative merits of tuple- and attribute-level versioning are best summa
rized as a tradeoff between processing speed and storage. T o generate a complete
new tuple version each time an attribute changes means that redundant informa
tion is stored. Conversely, to store multiple attribute versions within a tuple
violates the relational model's stricture against variable-length fields. Presently,
several different tuple- and attribute-level versioning methods have been pro
posed but only one method has been implemented experimentally (Ahn and
Snodgrass 1986); none is used operationally.
There is little doubt that aspatial temporal database methods will play a role in
spatiotemporal GIS; however, the treatment of spatial data is a problem that
remains to be addressed by cartographers. A fundamental problem is how to
conceptualize the geographic sequences we will represent.
COMPONENTS OF CARTOGRAPHIC TIME

Cartographic time is punctuated by 'events,' or changes, which are recorded along

two axes. One axis marks when changes occur (or are discovered) in the real
world; we will refer to this type of temporality as 'world time.' T h e second axis
traces when changes are recorded in the database; we have named this type of
temporality 'database time.' These essentially orthogonal times are widely noted
in the temporal database literature, and are variously referred to as 'valid' and
'transactional' (Snodgrass and Ahn 1985), 'logical' and 'physical' (Lunn et al.
1984), 'extrinsic' and 'recording history' (Ariav 1986), and 'object level' and
'system level' (Bolour and Dekeyser 1983).
In practice, world time would begin on the date of the earliest known in
formation and end with the most recent information, although predictive models
might be employed to extend world time beyond the present. Database time
would begin when the first data entered the database and end with the last data
entry. To clarify the distinction between world time and database time, consider
two queries:
- t o the best of our current knowledge, how did this region actually appear on a
given date?
-what information on this region existed in the database on a given date?
Maintaining information on both world and database time is critical. While
the intention of a temporal information system might focus on the historical
sequence of events, there is no guarantee that this information will arrive in the
database in the same order. In addition, new information might shed light or cast
doubt on existing database information, necessitating retroactive amendments.
The most practical approach to all eventualities is to adopt the 'supersede-butnever-delete' principles of Accounting, which document both the history of an

A FRAMEWORK FOR TEMPORAL GEOGRAPHIC INFORMATION

FIGURE 2. The relationship of map states and object versions.

enterprise's finances and the sequence of its recordkeeping. An erroneous entry is

not corrected by erasure, but by posting an adjustment at the later date.
THE NATURE OF TEMPORAL OBJECTS

Time, like space, must be subdivided, ranked, or measured before it can be

analyzed. Drawing analogies between spatial and temporal entities is one useful
device for understanding their commonalities.
The temporal parallel of'map' is 'state.' We borrow this term for aggregate
conditions from systems theory, which considers the history of a system to be a
series of states punctuated by 'events' that transform one state into the next (see
Ferg 1985, and Bolour and Dekeyser 1983). A cartographic state consists of a
spatial configuration of objects, each of which can change somewhat independently of the others. Just as a map is transformed from state to state by events, an object
is transformed from one 'version' to the next by 'mutations.' Thus, each map state
freezes geographic evolution into a configuration of object versions.
Of course, each object mutation is an event that causes a new map state. This
interrelationship results in a world-time topology comprised of many parallel
lines (Figure 2) a view of time that coincides with that of other time specialists.
For example:
Given that temporal information is expressed in space, and that intervals between events
are subject to variability in the recording...we can begin to envisage a topology of temporal
information in which we have rubber strings or nets or sheets on which events are recorded
as knots or other singularities, expressing certain necessary relations between events,
relations which hold good even though the rhythms of all the clocks concerned are variable
(Meredith 1972, p. 260).

GAIL LANGRAN A N D NICOLAS R C H R I S M A N

1982

1984

1987

FIGURE 3. Spatial and temporal neighbors.

Aside from the obvious interrelationships between object versions and map
states, the succession of each object's versions and mutations (or one map's states
and events) has an internal structure. A version or state can be seen as a line
segment that represents the duration of a condition, while a mutation or event is a
point that terminates that condition and begins the next, forming a zerodimensional boundary between two one-dimensional 'regions.' Thus, temporal
units that share a boundary can be considered contiguous neighbors in time
(Figure 3). Finally, convention allows us to draw spatial boundaries firmly, regard
less of whether they represent gradual transition zones or formal lines of de
marcation. While temporal boundaries are, in reality, no more discrete than are
spatial boundaries, sharp lines also prove useful in representing temporal bound
aries.
Table 2 summarizes some analogies between space and time, building on
those offered by Parkes and Thrift (1980).
Table 2. PARALLELS IN SPATIAL AND TEMPORAL STRUCTURE

Overall configuration
Configurations separated by ...
Regular sampling units
Meaningful units
Subdivisions separated by ...
Size measured by ...
Position described by ...
Contiguous neighbors are ...
Maximum number of contiguous neighbors

Cartographic
Space

Cartographic
Time

map
sheet lines
cells
objects
boundaries
length, area
coordinates
adjacent objects
infinite

state
events
hours, days, decades, etc.
versions
mutations
duration
date
previous and next manifestations
two

A FRAMEWORK FOR TEMPORAL GEOGRAPHIC INFORMATION

SPATIAL VS. TEMPORAL STRUCTURE

Because many geographic questions concern spatial structure, GIS algorithms

draw on basic search procedures that identify for a given object:
-what are its neighbors?
-what are its boundaries?
what encloses it?
what does it enclose?
When geographic data structures treated objects in isolation, searches for
neighbors were exceedingly long and inefficient (Chrisman 1974). As geographic
information processing applications became more demanding and geographers
grew more skilled in working with spatial data, new structures were developed to
encode the topological as well as the geometric character of the mapped area. The
effectiveness of the early implementations of topological constructs demonstrated
in the U.S. Census Bureau's DIME files (Cooke and Maxfield 1967; see also Corbett
1979) led to their widespread adoption. Now most geographic databases pro
duced by government institutions for analytical purposes employ some variant of
the topological structure, including the U.S. Geological Survey's Digital Line
Graphs, the Defense Mapping Agency's Standard Linear Format, and the Census
Bureau's TIGER system. Additionally, most modern commercial software systems
use topological data structures.
Topological structures can improve spatial information processing capabili
ties in three ways. First, spatial structure is evident, which means that many
algorithms are simpler and speedier because spatial searching is reduced (as
demonstrated in White 1978). Second, data errors can be trapped, which protects
data quality. To trap errors, software checks whether the data's spatial structure
adheres to the model's rules for topological integrity, thus identifying the gaps,
slivers, and overshoots that plague less descriptive models (Chrisman 1983; White
1983). Finally, minimal data redundancy reduces storage requirements.
Similarly, a temporal information processing scheme should exploit temporal
structure to facilitate quality control and temporal analysis. By considering con
tiguous temporal neighbors to be topologically connected, a temporal topological
data structure can be devised to link one to the next, thereby avoiding exhaustive
searches through layers of time and space. Thus, just as the spatial topological
data structure provides a means of navigating from an object to its neighbors in
space, the corresponding temporal data structure would provide a means of
navigating from a state or a version to its neighbors in time. This capability would
facilitate such queries as:
-what was the previous state or version?
-what has changed (during a period, or at a place)?
-what is the periodicity of change?
what trends are evident?
In sum, a higher level of information would be encoded in the data, since the

GAIL LANGRAN AND NICOLAS R CHRISMAN

FIGURE 4. Time-dice snapshots representing urban expansion into a rural area. Intervals between time slices are not
necessarily equal.

connectivity of time is as important to temporal analysis as is the connectivity of

space to spatial analysis.
The next three sections describe three methods of conceptualizing
geographic temporality. Each is judged by how well it represents temporal
topology, since this criterion is indicative of data compactness, completeness, and
descriptiveness.
IMAGES OF CARTOGRAPHIC TIME: TIME-SLICE SNAPSHOTS

A common spatiotemporal model is a sequent set of data-state snapshots (Figure

4). Wood and Fels (1986) supply a useful description of such snapshots in their
description of maps as time-exposed photographs that record only fixed
phenomena because moving phenomena become transparent blurs on the film.
Attempts to create dynamic maps often turn into the creation of a sequence of
snapshots to be displayed using motion picture technology or video display.
Snapshot sequences are an intuitive spatiotemporal model but they have their
limitations. While finding 'what exists at T- is easy, finding 'how has T, changed
from 7V,' or 'what is the frequency of change' is not, because the pertinent
snapshots must be compared exhaustively to detect differences.
The root of the problem is that snapshots represent states, but do not
represent the events that change one state to the next. Even worse, snapshots
provide no explicit representation of object versions or their mutations. While it
may be possible to detect different versions of an object visually as snapshots flash
by, the model provides no means of recording their temporal structure or
assistance in locating temporal neighbors (i.e., an object's previous and next
manifestations). In many respects, time-slice snapshots are the temporal
equivalent of the formless spaghetti data structure; in both cases, the database
objects reflect the graphic, not its underlying meaning. Interestingly enough,
time-slice snapshots share the three major shortcomings of spaghetti data.

A FRAMEWORK FOR T E M P O R A L G E O G R A P H I C INFORMATION

Hidden structure. We have defined temporal neighbors to be an object's

previous and next versions, which are bounded by the mutations that caused the
metamorphosis. Because snapshots capture states, not versions, the boundaries
between versions are difficult to locate.
No error trapping. With no understanding of temporal structure, rules to
enforce logical integrity are difficult to devise. Change for each particular map
must be interpreted through the rules of the theme portrayed. For instance, a
land use change from pasture to residential may be more likely than the reverse.
Redundant storage. Regardless of the extent of change, a complete snapshot is
produced at each time slice, which duplicates all the unchanged data.
IMAGES OF C A R T O G R A P H I C TIME'. B A S E M A P W I T H O V E R L A Y S

Standing back from the time line and examining time-slice snapshots is not the
only way to view geographic temporality. Alternately, we can stand directly upon
the time line and look up or down it into the past or future. From this perspective,
a more versatile means of representing the time slices is possible. We place an
opaque base map at To to define the data's original state. At appropriate (and not
necessarily even) intervals, we interject a clear overlay and record changes that
have occurred since the previous update. When used this way, the overlays are
events that bound the map's states, while the marks on the overlays are mutations
that bound object versions.
Now from a point on the time line looking backward into the past, we see one
or more transparent overlays that amend the base map, which stands at the far end
of the time line. By backing away from this overlaid time series, as we did to view
the snapshot time series, we would see profiles of the base map at one end, a final
overlay at the other end, and intermediate overlays interspersed along the line
(Figure 5). Eventually, spatial change is comprehensive and the mutations
recorded on the overlays supersede the entire base map.
This image of time is not unlike a Minkowski diagram in its spatialization of
time. It is also reminiscent of Rucker's ( 1984) vision of reality as a time-space cube,
and Hagerstrand's (1974) time-space 'aquariums,' through which people and
objects trace paths of a given life span. Rucker argues that the passage of time is an
illusion; all of space/time is present at any given moment. This conceptualization is
a reasonable model for a temporal geographic information system, in which all
states of the study area are equally accessible to the analyst. Recently, Szeg (1987)
has developed Hagerstrand's time geography diagrams into a model of
spatiotemporality. While philosophically interesting, this model offers little to the
practical matter of digital representation.
To use the base map/overlay construct, a new overlay would be created for
each database update session to represent transactions in database time. World
time would be expressed using dates or color codes on the overlay's change
notations. Thus, neighbors in database time are located on adjacent overlays while
neighbors in world time are not, unless both times are effectively synchronized.
The base map/overlay construct can answer queries on both states and
versions:

10

GAIL LANGRAN AND NICOLAS R CHRISMAN

FIGURE 5. A base map/overlay representation of urban encroachment into a rural area.

- To find "what was the data state at Tit" merge the base map at To with the
overlays at T, to Tt.
To find "what has changed between T\ and 7}," create a composite overlay T;... 7}
by merging the overlays from Tt+1 to 2}.

A FRAMEWORK F O R T E M P O R A L G E O G R A P H I C I N F O R M A T I O N

11

- To find "what versions has this object had and when did it mutate?" check each
overlay for amendments in that object's location.
- To compute the frequency of change, accumulate the number of mutations at
each location.
Because it represents the boundaries of both states and versions, the base
map/overlap construct is superior to snapshots. Just as the astructural style of
snapshots shares the spaghetti structure's problems, the temporal structure of the
overlays mimics the topological structure's assets.
Temporal structure is evident. Temporal neighbors (i.e., a version's previous or
next forms) are located by finding the mutation that separates them.
Errors can be trapped because improbable events can be prohibited.
Minimal redundancy. Use of storage is spartan; each object version is stored
only once.
IMAGES OF CARTOGRAPHIC TIME: SPACE-TIME COMPOSITE

The final image of time is a variation on the base map/overlay theme. Rather than
retaining change notations on separate overlays, the base map becomes a
temporal composite built from accumulated geometric changes. Each change
causes the changed portion of the coverage to break from its parent object to
become a discrete object with its own distinct history. In other words, the
representation decomposes over time into increasingly smaller fragments the
coverage's greatest common spatiotemporal units each of which references a
distinct temporal attribute set (Figure 6). This method of temporal decomposition
was originally suggested by Chrisman (1983). T h e spatiotemporal units created by
this intersection should be called greatest common units to correct the misleading
terminology promulgated by Peucker and Chrisman (1975) for a related case.
The mechanics of space-time compositing begin with a base map that repre
sents a region's geometry and spatial topology at some starting time. An overlay is
generated during each database update session, as described previously for the
base map/overlay model. Once accepted for permanent inclusion, the overlay is
incorporated into the system using the same intersection procedure currently
used for polygon overlay (Dougenik 1980). New nodes and chains are added to
the historical accumulation, forming new polygons that have attribute histories
distinct from those of their neighbors. Each unit's attribute history is represented
by an ordered list of records. A record contains an attribute set and the database
and world-time intervals during which that attribute set was valid. While this
temporal attribute concept seems logical, it is difficult to implement in current
relational database software, because the temporal ranges create a host of variable
length anomalies from the 'normal form'.
Accessing temporal information stored in the space-time composite is con
ceptually straightforward. T o compile a single time slice from the composite, one
has only to 'walk' the attribute history list of each polygon to locate the attribute
that was current at the desired time slice. If polygon neighbors in the time slice
share a single attribute, the chain that separates them is dropped.

12

GAIL LANGRAN AND NICOLAS R CHRISMAN

FIGURE 6. A space-time composite of urban encroachment. Each polygon has an attribute history distinct from that
of its neighbors.

DISCUSSION

Both the base map/overlay and space-time intersection are reasonable models
upon which to base a spatiotemporal data-structuring methodology. However,
because the base map/overlay model is not space filling, it is more prone to error
unless an alternate way of checking for internal consistency can be devised.
Additionally, no simple means of structuring base map/overlay data is evident
(which is not to say that a means does not exist). The arguments for retaining
separate overlays vs. storing a composite follow those of retaining categorical
coverages on separate overlays vs. merging them into one coverage integrated by
an overlay processor. A composite incurs preprocessing costs and offers direct
response to any query, while separate overlays must invoke an overlay processor
to respond to any query. In addition, a composite minimizes polygon-processing
error, while separate overlays permit greater flexibility in modeling the error
inherent in each overlay.
For both the base map/overlay and the space-time composite models, performance efficiency might be improved by maintaining a current data state
separately from previous states, especially when queries on the current state are
more frequent than are forays into the past. Conversely, a temporal GIS used by
historians or archeologists might maintain the oldest data separately on the
assumption that it will be accessed and altered most often. In either case, the
temporal partitioning would prevent the longitudinal data from adversely affect-

A FRAMEWORK FOR TEMPORAL GEOGRAPHIC INFORMATION

13

ing the most frequent operations. Aspatial methods of temporal partitioning are
discussed by Ahn and Snodgrass ( 1986), Ferg ( 1985), Ariav ( 1986), and Lum et al.
(1984). Further performance improvements will be possible through indexing
schemes. Inspiration for such schemes can be found in persistent data structures
(Sarnak and Tarjan 1986) and nondestructive data structures designed for writeonce-read-many-times (WORM) optical storage (e.g., Rathmann 1984).
In operation, both database time and world time will be important for
different purposes. T h e composite approach can accommodate two time axes if its
attribute database is structured appropriately. Even with skilled design, however,
database records with a temporal range strain the capability of most current
database systems, and two such axes will add further complications.
Y

CONCLUSION

Cartographic time includes the separable components of world and database

time. Treatment of either temporal axis can borrow from the treatment of spatial
data. An important concept is that of a temporal boundary (called a mutation)
which separates object versions and provides an effective means of describing
temporal structure. Temporal topology should simplify queries and should per
mit enforcement of data integrity.
Two reasonable methods of representing geographic change were pre
sented. T h e more promising method allows a study area to decompose over time
into polygons that represent 'greatest common spatiotemporal units,' each with a
distinct attribute history. Such an approach is amenable to enforcement of logical
consistency because of its space-filling and time-filling nature.

REFERENCES
AHN, ILSOO and SNODGRASS, RICHARD 1986. Performance evaluation of a temporal database manage
ment system. Proceedings of the SIGMOD '86 Conference: 96107.
ARIAV, GAD 1986. A temporally oriented data model. ACM Transactions on Database Systems 11/4:
499-527.
BOLOUR, A. and DEKEYSER, L.J. 1983. Abstractions in temporal information. Information Systems 8, 1,
4149BASOGLU, UMIT and MORRISON, JOEL 1978. T h e efficient hierarchical data Structure for the U.S.

Historical Boundary File. Harvard Papers for GIS, volume 4. G. Dutton, ed. Reading, Massachusetts:
Addison-Wesley.
CHRISMAN, NICHOLAS 1974. T h e impact of data structure of geographic information processing.
Proceedings, Auto-Carlo T. 165 176.
1975. Topological information systems for geographical representation. Proceedings, Auto-Carto
2:346351.
1983. T h e Role of Quality Information in the long-term functioning of a Geographic Informa
tion System. Cartographica 21/2&3 7987.
CLIFFORD, JAMES and WARREN, D.S. 1983. Formal semantics for time in databases. ACM Transactions on
Database Systems 8/2, J u n e : 214-254.
COOKE, D.F. and MAXFIELD, W.F. 1967. T h e development of a geographic base file and its uses for
mapping. Urban and Regional Information Systems for Social Programs, Papersfrom the Fifth Annual
Conference of the Urban and Regional Information System Association: 207218.
COPELAND, G. 1982. What if mass storage were free} IEEE Computer 15/7: 2 7 - 3 5 .
CORBETT, j . p . 1979. Topological principles in cartography. Technical Paper 48, U.S. Bureau of the
Census, Washington DC.

14

GAIL LANGRAN AND NICOLAS R CHRISMAN

DOUGENIK, JAMES 1980. WHIRLPOOL: A geometric processor for polygon coverage data. Proceedings
Auto-Carlo IV 2: 3 0 4 - 3 1 1 .
FERG, STEPHEN 1985. Modelling the time dimension in an entity-relationship diagram. Proceedings of the
Fourth International Conference on Entity-Relationship Approach: 280-286.
HAGERSTRAND, T. 1974. On socio-technical ecology and the study of innovations. Elhnologia Europa 7:
2537.
LUM, v. et al. 1984. Designing database support for the temporal dimension. Proceedings of the ACM
SIGMOD '84 Conference: 115126.
MEREDITH, PATRICK 1972. T h e psychophysical structure of temporal information in The Study of Time I:
Proceedings of the First Conference of the International Society for the Study of Time. New York:
Springer-Verlag: 259273.
MOELLERING, HAROLD 1980. Strategies of real-time cartography. Cartographic Journal 17/1: 12-15.
MUEHRCKE, PHILLIP C. 1978. Map use. Madison, Wisconsin: J P Publications.
NEWTON-SMITH, W.H. 1980. The structure of time. Boston: Routledge and Kegan Paul.
PARKES, DON and THRIFT, NIGEL 1980. Times, spaces, and places. New York: John Wiley and Sons.
PEUCKER, THOMAS K. and CHRISMAN, NICHOLAS 1975. Cartographic data structures. The American

Cartographer 1/2: 5569.

RATHMANN, PETER 1984. Dynamic data structures on optical disks. Proceedings of the International
Conference on Data Engineering, IEEE : 175 180.
RESCHER, NICHOLAS and URQUHART, ALASDAIR 1971. Temporal logic. New York: Springer-Verlag.

RUCKER, RUDY 1984. The fourth dimension: Toward a geometry of higher reality. New York: Houghton
Mifflin.
SARNAK, NEIL and TARJAN, ROBERT 1986. Planar point location using persistent search trees. Communica
tions of the ACM 29/7: 669-679.
SINTON, DAVID 1978. T h e inherent structure of information as a constraint to analysis: Mapped
thematic data as a case study. Harvard Papers on GIS, vol. 7, G. Dutton, ed. Reading, Massachusetts:
Addison-Wesley.
SNODGRASS, RICHARD and AHN, ILSOO 1985. A taxonomy of time in databases. Proceedings of the SIGMOD
'85 Conference: 236-245.
SZEG, JANOS 1987. Human cartography: Mapping the world of man. Stockholm: Swedish Council for
Building Research.
WHITE, MARVIN 1978. T h e cost of topological file access. Harvard Papers on GIS, vol. 6, G. Dutton, ed.
Reading, Massachusetts: Addison-Wesley.
1983. Tribulations of automated cartography and how mathematics helps. Proceedings, AutoCarto 6, vol. 1 : 408418.
WOOD, DENIS and FELS, JOHN 1986. Designs on signs: Myth and meaning in maps. Cartographica 25/3:
54-103.