DESIGN AND IMPLEMENTATION OF THE VALID TIME FOR

SPATIO-TEMPORAL DATABASES

Jugurta Lisboa Filho, Gustavo Breder Sampaio, Evaldo de Oliveira da Silva and Alexandre Gazola

Universidade Federal de Viçosa, Departamento de Informática, 36570-000, Viçosa, MG, Brazil

Keywords: Geographic database, GIS, Spatio-Temporal Data Modeling.

Abstract: Three different types of time are identified in the literature on Temporal Database Management Systems:

valid time, transaction time and existence time. This paper describes the design of the valid time for Spatio-

Temporal Databases in Geographic Information Systems, based on the UML-GeoFrame conceptual data

model. It is also presented two translation rules of the valid time from the conceptual to logical level, im-

plemented for the TerraLib Spatial Components Library.

1 INTRODUCTION

A number of government and business organizations

need to handle data that vary simultaneously with

time and space, which is most of the time carried out

using Geographical Information Systems (GIS). In

some government agencies as well as the depart-

ments linked to mapping and monitoring of envi-

ronmental risks, urban planning, residential and in-

dustrial zoning, of epidemic control, among others,

the time factor is a fundamental element.

These agencies use GIS to execute operations of

spatial analysis and make decisions based on results

of these analyses, however, a great number of times

the temporal aspect of the used information is not

appropriately considered. This comes from the great

difficulty GIS designers and users find when design-

ing a spatio-temporal database and developing ap-

plications that allow the representation of phenom-

ena that vary with time and space.

There are several conceptual data models for

geographical database design, some with support for

modeling temporal aspects. MADS (Parent. et al.,

1999), Geo-ER (Tryfona. et al., 2003) and Percep-

tory (Bédard, 1999) are examples of these models

and tools. These conceptual data models seek to

consider all the possible variations of spatio-

temporal data, making them highly complex models

and of little practical use for most geographical da-

tabase designers.

Three different types of time are identified in the

literature on Temporal Database Management Sys-

tems: valid time

- instant or interval of time in which

a phenomenon is valid; transaction time

- time in

which the datum is inserted into or eliminated from a

given database; and existence time

- time in which

the object really exists in the reality.

The conceptual data model UML-GeoFrame ex-

tends the UML class diagram to geographical data-

base design (Lisboa Filho and Iochpe, 1999).

Rocha et al. (2001) proposed an extension to the

UML-GeoFrame model for modeling temporal as-

pects that comprises these three types of time. How-

ever, as for the other models previously mentioned,

the enormous possibility of combinations among the

geographic phenomenon properties regarding tem-

poral features, become the understanding, modeling

and implementation of all these types very complex.

This article presents a simplified approach for

modeling and implementation of spatio-temporal

phenomena based on the valid time, using the UML-

GeoFrame conceptual data model. It also presents

the incorporation of these aspects into a CASE tool

called ArgoCASEGEO. This tool is capable of trans-

forming conceptual spatio-temporal data schemas

into logical data schemas for the main data models

of commercial GIS (eg.: Shape File Format, Oracle

Spatial) and also for the data model of TerraLib spa-

tial library (Vinhas and Ferreira, 2005).

569

Lisboa Filho J., Breder Sampaio G., de Oliveira da Silva E. and Gazola A. (2007).

DESIGN AND IMPLEMENTATION OF THE VALID TIME FOR SPATIO-TEMPORAL DATABASES.

In Proceedings of the Ninth International Conference on Enterprise Information Systems - ISAS, pages 569-573

DOI: 10.5220/0002386405690573

 SciTePress

2 THE COMCEPTUAL DATA

MODEL UML-GEOFRAME

The conceptual modeling of geographical database

based on the Unified Modeling Language (UML)

and on the GeoFrame framework (Lisboa Filho and

Iochpe, 1999) generates an easy-to-understand data-

base schema, improving communication among de-

signers and/or users.

GeoFrame is a conceptual framework that pro-

vides a basic class diagram to assist the designer on

the first steps of the conceptual data modeling of a

new GIS application. The mutual use of UML class

diagram and GeoFrame allows the solution of most

requirements of GIS application modeling. A con-

ceptual schema of geographical data, based on the

UML-GeoFrame model includes, for instance, the

modeling of spatial aspects of the geographical in-

formation, in other words, the type of spatial repre-

sentation of each phenomenon, and the difference

between conventional objects and geographical ob-

jects/fields. The specification of these elements is

based on the stereotype concept, which is an UML

extension mechanism (Booch,, Jacobson and Rum-

baugh, 1998).

As example, Figure 1 illustrates an UML-

GeoFrame data schema, where the UML package

represents a theme relative to the real world that is

being modeled and all of them concern the same

specific geographical area (eg.: a street). The theme

Street Mesch comprises four classes of an urban

cadastre application. In the theme some classes are

geographicals objects, wich are modeled with stereo-

type of object view (<3>).

The spatial representation of classes of geo-

graphical phenomena perceived in object view can

be of a point <>, line <>, polygon <> or com-

plex <>, in this case when it is formed by more

than one spatial object. The geographical fields can,

in turn, be represented by one of the six types of

representation of attributes varying in space, de-

scribed by Goodchild (1992), which are: Irregularly

sampled points <8>, Adjacent polygons <1>, Con-

tour lines <;>, Triangular irregular network, or

TIN, <:>, Cell grids <<> and/or Regularly sam-

pled points <9>. The existence of multiple repre-

sentations is modeled through the combination of

two or more stereotypes in a same class. For exam-

ple, a class Municipality can have two abstraction

forms of its spatial component (a point or an area),

which is specified by the pair of stereotypes <>.

Figure 1: Example of UML-GeoFrame schema.

3 THE GEOFRAME-T PROPOSAL

GeoFrame-T is a GeoFrame temporal extension for

modeling conventional, spatial, temporal and spatio-

temporal objects (Rocha et. al., 2001) based on the

TUML concept (Svinterikou, 1997). One of the main

GeoFrame-T characteristics is to include the aspect

time also for attributes and relationships.

The representation of temporal aspects in

GeoFrame-T is carried out by the TemporalObject

class, where there is an association of aggregation

with the class TimeObject, which can have or not

instances of the class TemporalMetadata associated

with it. The TemporalMetadata class has attributes

that describe the concepts of objects existing in the

TimeObject class. These attributes qualify informa-

tion on the reference systems based on time values

such as the Greenwich meridian and local coordinate

references. Other TemporalMetadata class attributes

are granularity and calendar, which describe the time

(month, year, day, second, etc.), and the type of cal-

endar considered for the time. Rocha et al. (2001),

defined the TimeObject class as a generalization of

Static and TemporalType classes, which is special-

ized in ValidTime, TransactionTime and Bitemporal.

The Static class represents objects with no tem-

poral variation associated. Through the class Trans-

actionTime, it is possible conceptually to model ob-

jects that should have their historical information

retained, whose time variation is linear, considering

the time in which the information was entered into

the database. The class ValidTime enables the tem-

poral modeling of a spatial object, whose time is

considered in accordance with the modeled reality,

being likely to occur linear or ramified variation.

Finally, the Bitemporal class can model spatial ob-

jects, considering both the recording and the valid

time, inheriting thus characteristics of Transaction-

Time and ValidTime classes.

ICEIS 2007 - International Conference on Enterprise Information Systems

570

Following the same fundamentals of

GeoFrame modeling, GeoFrame-T uses stereotypes

to facilitate the modeling of objects and geographi-

cal phenomena. In the GeoFrame-T model, the Tem-

poralObject class is considered as a time structure

possible of being represented by the modeled

classes. GeoFrame-T enables us to model conceptu-

ally the relationships between geographical phenom-

ena and their temporal characteristics. These rela-

tionships are represented by the Relationship class.

All the static and non-spatial relationships existent

among the modeled classes are considered instances

of the Relationship class.

4 MODELING THE VALID TIME

The descriptive, temporal and spatial dimensions are

orthogonal. Temporal properties can be defined by

the three categories of geo-spatial data: conventional

objects, geographical phenomena perceived in field

view and geographical phenomena perceived in ob-

ject view. A model that considers the three types of

time (valid, transaction and existence) for both

classes and attributes and relationships provides the

designer with great power of expression. However,

the great number of possible combinations also

makes the model difficult to learn and understand,

eventually compromising the implementation of

applications needed for handling spatio-temporal

data.

Experience has showed that in most GIS applica-

tions only the valid time of the geographical phe-

nomena has been considered, since this is the most

significant piece of information in most queries in-

volving temporal data. This article shows, in this

way, how design aspects spatio-temporal based on

the valid time, using UML-GeoFrame model, in

which only the valid time is considered.

Valid time is the time instant or time interval

when an object of the real-world is considered valid.

For example, the strike against the World Trade

Center took place on September 11, 2001, in turn,

the Gulf War occurred in the period between August

1990 and February 1991. Hence, another important

factor is the granularity of temporal information.

The UML-GeoFrame model considers three types of

time granularity: Date, Time and Timestamp. Speci-

fying the granularity of a temporal attribute is the

same as defining the domain of a descriptive attrib-

ute value (e.g. Char or Boolean). Finally, two types

of temporal classes are discriminated: Instant Class

and Interval Class.

If a class is modeled as being of Instant type

(stereotype <

>), it means that its objects are only

valid at a particular point in time. In this case, the

object does not evolve, since its validity is con-

densed into an instant. It is the case of a road acci-

dent representation, in which is essential to associate

an instant of time with the object.

If the class is of the Interval type (stereotype

<>), it means that its objects are valid in a period

of time, i.e., between an initial and a final temporal

value. These valid intervals do not necessarily have

the same size. In addition, the object evolution is

maintained, because its attributes can vary in the

period corresponding to its valid interval.

If a class is temporal of Interval type, it indicates

that every change in any object attribute will gener-

ate a new version of the object, and the old one will

not be lost.

Each temporal class must have its granularity

specified. The option Date is the default option and

indicates that a value of the date type has to be

stored. The option Time indicates the need for stor-

age of an complete hour value. The option Time-

stamp indicates that a pair (date, hour) must be

stored into the database. Each object (or object ver-

sion) is associated with a piece of temporal informa-

tion, characterized by a stereotype, along with the

chosen granularity.

Besides allowing the modeling of temporal

classes, the UML-GeoFrame model enables the rep-

resentation of temporal associations, which is identi-

fied by the stereotype <<time>>. The validity of an

association can be defined as the intersection of ob-

ject’s valid periods of classes involved in this asso-

ciation. This is because it is impossible for a rela-

tionship to exist in an instant of time in which a re-

lated object is not valid. Thus, considering t as this

period of intersection, it follows that:

Temporal Association - its validity must be con-

tained into the interval t, i.e., the association must be

valid at the most for the period of time in which both

objects coexist in time;

Non-temporal Association - its validity is same

as interval t. In this case, the interpretation is the

same given to conventional relationships, where the

association is valid while the objects coexist in time.

Some association examples involving classes of spa-

tio-temporal phenomena can be presented, when two

classes of geographical phenomena are perceived in

object view, where both were modeled as temporal

classes of the Interval type.

For example, an association between the classes

Country and Epidemic, where the class Country

could have information of name, population and

DESIGN AND IMPLEMENTATION OF THE VALID TIME FOR SPATIO-TEMPORAL DATABASES

571

GDP (Gross Domestic Product). These attributes

could vary in time, generating different versions of

an object from the Country class. The Epidemic

class describes epidemics that could occur in these

countries, through the temporality. This temporality

is important because of the need for observing its

evolution in time, as well as possible periods of out-

breaking and eradication. This relationship is de-

fined as temporal (<<time>>), indicating that a same

type of epidemics could occur in more than a period

in the same country.

The association in Figure 2 illustrates the Fire

and Building classes, the former being of non-geo-

graphical object type, but temporal of Instant type.

The Building class describes the existent buildings

and their use, such as museums, hospitals or schools.

The Building class is temporal of Interval type,

which means that a building has an existence period

characterized by the time interval relative to each

type of use. The association (non-temporal) between

these two temporal classes will make each Fire oc-

currence be associated with the most recent version

of a Building object. Therefore, in case one wants to

be possible to associate it with the correct version of

the Building object.

Figure 2: Non-temporal association with temporal classes.

5 TEMPORAL MODELING IN

THE ARGOCASEGEO TOOL

ArgoCASEGEO is a CASE tool built to support

modeling and design of geographical databases with

the UML-GeoFrame model (Lisboa Filho et. al.,

2004). To support the modeling of temporal classes,

the modules Graphic Design, Data Dictionary and

Automatic Generation were changed.

The ArgoCASEGEO Graphic Module was ex-

tended to support the temporal builders modeling, in

other words, to enable the specification of the stereo-

types Instant <> and Interval <>, as well as al-

lowing the designer to specify a type of granularity

for each temporal class.

A conceptual data schema drawn up using the

ArgoCASEGEO tool is stored in a XMI file (XML

Metadata Interchange), which is automatically cre-

ated in the Data Dictionary Module. New tags were

added to the XMI schema to store the definitions of

temporal characteristics of a class.

The Automatic Generation Module (AGM) is re-

sponsible for the application of rules that transform a

conceptual data schema into a logical-spatial data

schema. As there is still no standard model adopted

by the available GIS software, the ArgoCASEGEO

tool has an AGM for some commercial systems and

also an AGM for the Open Source Library - Terra-

Lib (Vinhas, 2005).

Aiming at validating the temporal extension pro-

posed to the UML-GeoFrame model, in this work

the AGM-TerraLib was extended to enable the

specification of logical data schema. Thus, starting

from the XMI file, the AGM-TerraLib transforms an

UML-GeoFrame conceptual schema into a TerraLib

logical schema. In the following section the trans-

formation rules implemented are described.

6 RULES FOR TEMPORAL

ASPECTS TRANSFORMATION

The transformation rules are defined using the rela-

tional model as basis, following the TerraLib

schema. Therefore, a primary key must be specified

in each class in order to carry out the transformation

correctly.

6.1 Rule 1: Transformation of

Temporal Classes

The attributes defined for a class generate attributes

in a corresponding relation. A temporal attribute

with the domain defined in accordance with the

specified granularity is added to temporal classes of

Instant type.

For temporal classes of Interval type, having a

primary key (PK) and a set of attributes {att}, a rela-

tion with PK is created. The foreign keys, if existing,

are also added to this relation. In addition, a second

relation is created to store the versions of its objects.

To guarantee the integrity, a relationship 1..N be-

tween the first relation and the version relation is

also created.

The primary key of the version relation is PK,

plus a temporal attribute. Thus, the version relation

will have the following schema:

RELATION_VERSION = (PK,begin,end,{att})

ICEIS 2007 - International Conference on Enterprise Information Systems

572

6.2 Rule 2: Transformation of

Temporal Associations

Given a temporal association, regardless of its mul-

tiplicity, a new relation is created with the primary

keys of the involved classes. This new relation is

required for storing the validity period of the asso-

ciation. With this objective, two temporal attributes

are created to indicate the beginning and the end of

the association validity.

One wants to create an association between them

in order that it is possible to store data on river sam-

ple collection for each collection point, with sam-

pling being done at different instants.

The first alternative is to model a non-temporal

association representing that each collection point

has several samples.

Another alternative is to model a temporal asso-

ciation, indicating that each collection point has only

one sample at a same instant in time. In this case the

relations schemas shown in Figure 3 will be gener-

ated.

Figure 3: Schemas generated from a temporal association.

The designer can take the second option to express

better the association. However, there is a question

to be discussed; the attributes begin and end of the

relation PointCollection_CollectionSample are re-

dundant, since the only value that they can take is

the valid instant of the sample collection. This re-

striction must be guaranteed by the application. It

also must not allow that a point has more than one

sample at a same instant, because this is incompati-

ble with what was specified.

7 CONCLUSIONS

In this article we presented a proposal to simplify the

conceptual modeling of temporal aspects in spatio-

temporal database. The concepts were presented as

an extension to the UML-GeoFrame model, however

they can be adapted to any conceptual data model.

Modeling the temporal aspect of an application

has not been a trivial task, probably due to the com-

plexity of the existent temporal models in the litera-

ture. The completeness of the temporal aspects is

actually necessary in a very restricted set of prob-

lems, for the great majority of temporal applications,

modeling valid time of real-world objects is enough.

Some examples of modeling of classes and tem-

poral relationships were presented and the Argo-

CASEGEO tool support for modeling of spatio-

temporal aspects was described.

At the present, only the Automatic Generation

Module for TerraLib library is dealing with temporal

aspects. An AGM for the Oracle-Spatial system is

being implemented and the temporal extension will

be drawn up for AGM-ShapeFiles.

ACKNOWLEDGEMENTS

This work was partially financed by the Foundation

for Research Support of the Minas Gerais - Fapemig.

REFERENCES

Bédard, Y., 1999. Visual modelling of spatial databases

towards spatial extensions and UML. Geomatica,

v.53, n.2.

Booch, G., Jacobson, I. and Rumbaugh, J., 1998. The Uni-

fied Modeling Language User Guide. Addison-

Wesley.

Goodchild, M. F., 1992. Geographical data modeling.

Computers & Geosciences. 18(4).

Lisboa Filho, J. and Iochpe, C., 1999. Specifying analysis

patterns for geographic databases on the basis of a

conceptual framework. In Proc. of 7

ACM Sympo-

sium on Advances in GIS, Kansas City.

Lisboa Filho, J. Sodré, V. F., Daltio, J. Fidelis, M. R. and

Vilela, V., 2004. A CASE tool for geographic data-

base design supporting analysis patterns. In Workshop

on conceptual modeling for GIS, Shanghai. Springer-

Verlag LNCS 3289.

Parent, C, Spaccapietra, S. and Zimányi, E., 1999. Spatio-

temporal conceptual models: data structures + space +

time. In Proc.7th ACM GIS, Kansas City.

Rocha, L. V., Edelweiss, N. and Iochpe, C., 2001.

GeoFrame-T: A temporal conceptual framework for

data modeling. In Proc.9

ACM GIS, Atlanta.

Svinterikou, M. and Theodoulidis, B. 1997. The temporal

Unified Modeling Language (TUML). Dep. of Compu-

tation, UMIST, UK, (TR97-1).

Tryfona, N., Price, R. and Jensen, C. S., 2003. Spatiotem-

poral conceptual modeling. In Spatiotemporal Data-

bases: the Chorochronos approach, Springer-Verlag.

Vinhas, L. and Ferreira, K. R., 2005. Descrição da

TerraLib. In Casanova, M. A. et al. Bancos de Dados

Geográficos, Curitiba: MundoGeo.

DESIGN AND IMPLEMENTATION OF THE VALID TIME FOR SPATIO-TEMPORAL DATABASES

573