Chapter 10

Analysis of the Specificities of Software Development in Geomatics Research

This chapter offers a synthetic analysis of the specificities of the work that has been presented: AROM-ST, an object-oriented knowledge representation language devoted to the representation of spatiotemporal phenomena; GENGHIS, an application to design spatiotemporal information systems in the fields of risk which rely on this language; GENEXP-LANDSITES, an application to generate space tessellations which present specific characteristics; GEOLIS, an application for geographical data exploration-querying; GEOXYGENE, a platform to develop object-orientated processes on geographical data; MDWEB, a module to catalog data and geographical services; ORBISGIS, a GIS focused on the city; and WEBGEN, a generalization Web service platform.

We would like to organize this analysis along the headings of the description canvas of these projects which have proven, after analysis, the most relevant to identify the project classes: origin and motivations, major functionalities, and reusability (architecture and field).

10.1. Origin and motivations

10.1.1. Targeted users and uses

The projects presented in Chapters 2 to 9 can be organized into two major categories, depending on whether specific users were targeted at the beginning of the project, and on whether there are developers participating in the project or not.

Thus, GENEXP-LANDSITES, GEOLIS, MDWEB, and ORBISGIS aim for precise thematic uses for users who are not direct contributors to the tool itself. More specifically, GENEXP-LANDSITES is meant for agronomy and agroecology researchers; GEOLIS for geographers; MDWEB for thematic mapper technicians using and producing geographical data on a single territory; and ORBISGIS for city thematic mapper technicians. GENGHIS is also a part of this category, since it is designed for information system designed in the field of risk, and not for GENGHIS contributors.

AROM-ST, GEOXYGENE, and WEBGEN are mainly meant for participants to the development of the tools or for researchers and engineers in the same field. AROM and AROM-ST, for instance, are usable by the community of researchers working on knowledge representation languages. The two projects GEOXYGENE and WEBGEN have for their part evolved to also be open to outside users: GEOXYGENE's matching module is used outside of a matching research context, and WEBGEN is used by the industrial sector.

10.1.2. Motivations and foundations

As for strategic motivations, we have identified two main categories depending on the dominating motivation: projects (1) which aim to drive computer science forward and for which geomatics is rather considered as a field of application which presents new challenges, and projects; (2) which are more thematic and aim to adapt computer science techniques to solve geomatics issues.

In the first category, we have the AROM-ST and GEOLIS projects. AROM-ST's goal is to drive the field of object-oriented representation languages forward, by giving them elements devoted to the expression and the manipulation of spatial characters. The GEOLIS developers want to progress in the field of data exploration. Let us remind the reader that these projects are the projects that show the highest level of declarative representation of knowledge, in compliance with the symbolic artificial intelligence principles at work here. That said, in the specific case of spatial analysis, algorithmic geometry is a field in which declarative representation of knowledge is still hard to apply.

The second category is made up of projects which aim less to drive computer science forward and more to solve a geomatics issue. Let us remind the reader that these projects often opt for an innovative computer science technology: the Web services for WEBGEN, the Web services and cataloguing metadata for MDWEB, the object orientation and the agent-based modeling for GEOXYGENE, and the AROM-ST language for GENGHIS. Among these projects, GEOXYGENE aims for a computer science modeling with a certain level of expressiveness and MDWEB also integrates ontologies. In this latter category, we can distinguish a certain number of projects which we will qualify for development integration, since they aim first to build on developments within a community and then to share them. GEOXYGENE, thus, aims first to integrate and build on proposals designed in the COGIT laboratory. Its mission to transfer the knowledge outside and extend the community with scientific collaboration came after, as a natural consequence. WEBGEN aims to integrate generalization tools, designed within a clearly identified researcher community, and make them interoperable and comparable. As for ORBISGIS, it was designed to build and capitalize on the proposals for urban information systems carried out at the IRSTV.

Let us also remind the reader that more and more developments are carried out or changed under the double impetus of queries from thematic mapper technicians and breakthroughs in terms of model design and computer science development. This is particularly the case for GENGHIS that aims both to promote AROM-ST and to deal with an important geomatics issue. This is also the case for WEBGEN and MDWEB. The GENEXP-LANDSITES project also aims to answer a specific thematic question and leads to joint progress in thematic, statistic, and computer science research.

10.2. Major functionalities, fields, and reusability

10.2.1. Functionalities

As for functionalities, the projects presented are generally found in one of the two following categories. We have not classed AROM-ST which is a language and whose main functionality is thus to help implement software.

The “GIS platform” type projects aim to integrate most of the functionalities of a GIS into various modules (which can be imported), by developing them more or less depending on the platform (ORBISGIS and GEOXYGENE).

Other projects focus on a classical functionality of GIS but decline it innovatively GEOLIS studies the classical geographical data exploration and query functions, but offers to implement them in an innovative way through a logical expression system. GENGHIS focuses on the spatiotemporal data visualization and query functions but implements them innovatively in many ways (Web, etc.). WEBGEN offers analysis functions (generalization) but allows the user (of the GIS client) to use them remotely. GEOXYGENE aims to offer functions which integrated the IGN expertise in vector data management into automatisms.

Finally, the other projects offer functions which currently (in 2012) do not exist in the definition of a GIS and will probably become core functionalities in future GIS. The GENEXP-LANDSITES project thus offers to generate data with the desired characteristics, allowing us to carry out sensitivity studies (test data) and long-term planning. The MDWEB project offers to research geographical data or geographical services online.

10.2.2. Fields

The notion of “field” in geomatics can refer to application fields (hydrography, sustainable city, etc.) or methodological fields (artificial intelligence, remote detection, etc.). A single field is sometimes considered like an application field – for instance, cartography is an application field of generalization – and a technical field – for instance, cartography is a technical field for many GIS projects.

Some projects are clearly positioned on thematic fields and others more on methodological fields. Thus, ORBISGIS is essentially positioned on the city field, and makes, in a lesser way, the methodological choice of raster modeling. GEOLIS is clearly positioned on a methodological field, the field of logical information systems. WEBGEN and MDWEB are positioned strongly in technical (generalization, knowledge sharing) and methodological (Web services and Web service cataloguing) fields. GENEXP-LANDSITES is also strongly thematic and technical (generating field patterns through spatial tessellation methods which are in the field of algorithmic geometry and spatial statistics). GEOXYGENE is mainly positioned on the methodological side (object-oriented programming and geographical data vector modeling).

A project's field is important for pooling. Two projects covering the same thematic field can look for function complementarity (build a new process using the functions of one and the other), method comparison or network pooling. Two projects with the same technical field can look for complementarity in terms of method comparison and application comparison: improving our knowledge of a technique by learning how it is applied to other issues, assessing the applicability of our own methods to other thematic fields. For instance, there is a specific literature focusing on the application of agent-based techniques in the spatial field where researchers can confront their proposals.

The restrictions that a field brings to a piece of software's reusability seem to be threefold:

• – The software interface (programming or graphical) might require expertise inherent to the technical field to be used. For example, the COLORBREWER library in GeoTools, devoted to the offering of ranges of color to write thematic maps, requires the user to understand the notion of sequential, differential, and ordered relation. In the same way, to correctly set the online accessible processes of the WEBGEN platform, we must understand the meaning of the parameters to assess them accurately.
• – The data model can be very specific to the field. For example, an itinerary calculation tool that takes in input all the transportation networks and interconnections as well as the user's preferences (on foot, etc.) seems to be extremely specific to that field. However, a tool to calculate the shortest path is still very generic. On the contrary, in GENEXP-LANDSITES, an extremely narrow project, the methods used are spatial statistics and tessellation methods which are generic and can be applied to objects different from agricultural fields.
• – The work distribution network. There are cases in which a piece of software using a certain technique to deal with an issue in a certain thematic field will be completely unknown to researchers working in a different field and using different techniques, when it could be of interest. Let us imagine, for instance, that a researcher wishes to transmit hydrographic data over a mobile network. He/she can make relevant use of road network generalization methods, extrapolate them into his/her water system network to simplify it and transmit it in a more simple manner so that it takes up less of the client's memory.

10.2.3. Reusability

Reusing a development may differ depending on the project. Beyond the reusability limits linked to the specificity of a field, which was mentioned previously, the reusability depends on the choice of models and architecture, language, open source or proprietary aspect, or even the coding quality.

AROM-ST initially had a very low level of interoperability since it was a new language, but its authors aim for its conversion into OWL, which has become, in the last few years, the new standard in knowledge representation on the Web. As for GENGHIS, it is interoperable within a small measure, on the level of the result, since it uses standard Web technologies and the generated interface can thus be retrieved and modified.

GENEXP-LANDSITES is meant to be coupled with software which loads space tessellations to simulate agroecological processes and focuses the issue of interoperability of these client modules and their input format.

GEOLIS can read GML and be used remotely (independent of the client platform). That said, to use GEOLIS's capacities to their fullest, we must use its language to describe its field.

GEOXYGENE and ORBISGIS were both designed as interoperable platforms inasmuch as the data models used are ISO norms. This interoperability is however limited at the model design level since during the development there was no norm implementation. The data coupling can happen due to XML implements but the API coupling is more complex. That said, ORBISGIS does not implement certain complex elements of the ISO models and is thus more interoperable. Both offer extension mechanisms to their interface.

MDWEB is the project which can be most reused among all the projects proposed in this book, to us. It has been designed as a component for an open and extensible modular architecture. It focuses on specific functions and deliberately relies on standard model implements, whether in terms of manipulated data formats or of interface, i.e. of proposed functionalities.

WEBGEN belongs to a context of sharing and reusing processes and Web services. It was built around an ad hoc communication model which then evolved to take up OGC proposals in terms of geographical data processing service description.