GREAT-ER 2

The vision

GREAT-ER 1.0 demonstrated successfully the applicability of a geography-referenced modelling approach. The final version published in March 1999 goes beyond what was expected when the project started in 1996.

GREAT-ER is now prepared for the next goal to provide a modular system of database (storing geography-referenced data as well as non-geography-referenced, e.g. substance data), a user interface with general GIS capabilities and a model server providing operations on the data specified through the interface and provided by the database. The general approach of GREAT-ER has to be multi-platform capable with network features to allow each part to run on the optimal environment.

The model server covers a modular set of environmental faith models to establish GREAT-ER as a tool in the official process of risk assessment for chemicals and substances. GREAT-ER could also be used as a system for use by competent authorities and others as a screening tool to aid the implementation of the Water Framework Directive or as a decision support system providing comfortable access to large geography-referenced data collections. Use as a tool supporting the design of monitoring studies or survey campaigns are also possible. And there are many other applications.

The user interface has to deal with this range of applications and associated user profiles. The requirements vary from simple visualisation and options to interrogate the data up to GIS features like geography-referenced data editing, data analysis and customized map creation. The different user groups have to be served with homogeneous interface/dialogs (at least for the available functionality) but with different technologies in the background.

GREAT-ER 1.0 demonstrated that the model system has to be implemented following a modular approach. The more open and modular technique presents the opportunity to interconnect models flexibly. It offers developers of other models the option to integrate their own tools. The multi-platform approach enables the model server to run models on their most adequate platform. This can range from a single computer to a network with distributed objects or even parallel computing, to provide the reasonable computing power for each application. The ultimate goal is to have GREAT-ER available on the Internet.

The database also has to deal with different user requirements. Since the risk assessment process also considers confidential data, security is an important feature. Ultimately, this could mean running the DBMS on a stand-alone computer. But other, more scientifically oriented user groups need options for data exchange which can be realized by central hosts with trusted competence in the evaluation of the served data. Since this scalability does not affect the data model itself the DBMS has to provide an interface which can serve local as well as network access.

Realising the identified parts of GREAT-ER with the described flexibility with easy options to extend the results of GREAT-ER 2.0 will attract other groups to incorporate their work into GREAT-ER. Hence it will benefit the whole system.

How to achieve this vision

As mentioned above, GREAT-ER consists of three major parts: database, user interface and model (application) server. The first two are tools in the GREAT-ER system whereas the last one can be seen as the connection to interested groups. The goal of GREAT-ER 2.0 is to design the interfaces between the three parts and to develop and implement the parts with sufficient flexibility to be extended in the future.

This extension procedure means that other projects connect their tools to the application server. To meet the requirements of the new tool the database and the user interfaces then have to be adopted.

As proposed, each part of GREAT-ER should be running on the most adequate platform. Hence the interfaces have to be designed considering multi-platform features. The communication technologie has to be chosen with focus on efficient and reliable data transfer.

General requirements for the three parts to gain acceptance of GREAT-ER and to facilitate future development of the system is to reduce related costs to minimum. I.e. third party software has to be selected carefully by criteria like availability/licensing and spread of usage. All software and concepts developed within GREAT-ER 2.0 must be well documented and published, including the sources. This enables and moreover attracts other projects to include their tools into GREAT-ER.

To substantiate these general requirements for the major parts of GREAT-ER:

• Database
  GREAT-ER 1.0 demonstrated the need for a powerful DBMS, the required features like access permission management or remote maintenance of data via Intra-/Internet are obvious. The database should be set on top of an existing DBMS which should fulfil either broad use or free availability. The database has to be easy to enhance in the future, even by third parties/projects. This requires documentation of the data model and the implementation.
• User Interface
  GREAT-ER 1.0 has identified different user profiles: One group expects full GIS functionality, they usually already have the tools available. The second group works more substance related and is therefore satisfied with a simplier interface to visualize geographic data. Both groups want options to access the data (i.e. interrogate or edit). Hence both options have to be developed: starting from the existing GREAT-ER 1.0 a Desktop GIS-based version with connections to the database and the model server allows power users full GIS-featured access to the data and all other Arcview capabilities. In addition a browser based version has to be developed with similar interfaces to database and application server. Both user interfaces must have common dialogs. These dialogs have to be realized using a web-able scripting language.
• Model Server
  The simulation module developed within GREAT-ER 1.0 demonstrated successfully that a geography-referenced simulation is possible and furthermore that it is usable. It was designed straight-forward, with speed optimization for running on a single computer. However, the applied optimizations limited the extendability of the system. Therefore a new system has to be designed following a modular approach. The seperate models have to be implemented as modules. They all are invoked by a model server instance driven by the users data selection. An efficient method has to be designed or selected to meet all requirements specified.

The key issue for the success of GREAT-ER beyond the period of CEFIC LRI funded development is the documentation and publication of all sources. GREAT-ER is a scientific toolkit and science depends on the share of knowledge. On the other hand, GREAT-ER and the incorporated techniques have to be protected against commercial abuse. An appropriate licence will serve the funders and third party contributors interests.

Development strategies

The development strategy needed for GREAT-ER 2.0 differs from the one applied for GREAT-ER 1.0, which can be best described as evolutionary. Since it was a scientific project it covered permanent communication between the developers of the seperate modules and frequent releases of development snapshots for revision by a scientific steering commitee.

The development of GREAT-ER 2.0 consists of three different strategies:

• The user interface is already defined by GREAT-ER 1.0. Slight adaptions are straight-foreward and hence dialogs can be implemented according to the specification.
• Since two groups are working on the interfaces between the three major parts of GREAT-ER, it is important to ensure concordantly development as specified. Therefore multiple integration tests have to be applied at pre-defined milestones.
• Much development goes into the model server, which only can be realized following a evolutionary strategy. This ensures to keep on track with the requirements of related CEFIC LRI projects and other interested groups.

For several other software projects Intevation applies the Open-Source development model (which is also known as the Free Software development model) with success. It has proven superior to several conventionally development models, especially for projects with many interested parties involved (Eric S. Raymond, The Cathedral and the Bazaar, O'Reilly and Associates Inc., 1999). For the model server we propose to involve related parties early in the project progress.