A Collaborative GIS Decision Support Model on Internet

Jeff Wang
Lockheed-Martin Service Group
Visualization Center of National Environmental Supercomputing Center
U.S. Environmental Protection Agency

Biography

Jeff Wang, Ph.D. is a technical leader in the Visualization Group of National Environmental Supercomputing Center of USEPA. Dr. Wang is an advocate for the use of Internet technologies in the environmental decision support community. He organized the first-time-ever MBONE broadcast of an EPA workshop to the Internet community and developed several World Wide Web (WWW) information service sites to promote decision support in different industries. Dr. Wang also have extensive experience in GIS applications. His research interest is collaborative GIS model on the Internet.

1. Introduction

The international Geographical Information System (GIS) community has experienced enormous growth in the past decade because of the advancement of computer, remote sensing, and space technologies. The typical size of a GIS dataset has grown in magnitude from megabytes a decade ago to gigabytes now. The typical use of GIS has evolved from simple cartography to multimedia decision making support in the same period. Typical users have expanded from the geographical survey community to virtually all communities. It is no longer possible to put all GIS data into a centralized location and use one computer to handle the data processing, let alone intensive data analysis for decision support. However, the society demands more GIS data and better utilization of these data.

The most promising solution to the GIS growth problem is the Internet. The Internet, which is defined as interconnection of computers around the world, has grown from tens of computer hosts two decades ago to tens of millions of computer hosts today. It is one of the fastest spreading technology/communication tools in the United States and around the world. Meanwhile, the Internet community has been agonizing over how to develop applicable programs to use the full capacity of Internet in the future.

Collaborative Spatial Decision Making (CSDM), or Collaborative GIS, will advance to be one of the most popular applications on the Internet. Future GIS data can be managed and served in a distributed system at the county/metropolitan or township level. In such a system, the data can be updated constantly by the people who are most familiar with the environment. A huge computer network will connect all of these computers. A user of the GIS data will run a client anywhere and be able to discover and filter the GIS data, retrieve wanted data, conduct analyses, and archive results. The analytical results are then used for decision making. An intelligent agent will be used for optimizing the solutions and audio-visual tools will be used to conduct conferences and utilize human expertise.

Fortunately, the GIS community does not have to reinvent the wheel in order to build collaborative GIS. The Internet community has developed many of the needed technologies to accommodate the future growth of collaborative GIS. These include research and development of the WWW, gophers, the Wide Area Information Service (WAIS), Computer-Supported Collaborative Work (CSCW) environment, object-oriented programming (OOP), client/server paradigm, and information agents.

This paper will propose a CSDM model and discuss the related major issues.

2. Important Issues in a CSDM

In this section, we will discuss some of the important issues in development of a CSDM model. We will only provide some brief discussions on each issue and will provide in-depth discussion of each issues in separate studies.

2.1 GIS Data Model

Most of the current GIS data models are based on relational database technology in a centralized location. This can no longer continue because the "modern" (compared to 10 years ago) GIS applications demand more efficient use of memory and a shorter application life cycle. Therefore, the future data model will be object-oriented and reusable. Furthermore, the GIS data model will take "open system design" approach, so the data model is interoperable, reconfigurable, and can be grouped with other data models to form a more powerful meta-model.

One of the current practices in telecommunication applications, which will be continued through the next few decades, is bandwidth on demand. Users can choose their bandwidth configurations based on cost. Therefore, GIS applications will face different users groups in different geographical locations with different bandwidth capacity. One of the solutions to this problem is to implement sharable objects in the data model, so we can build a collaborative GIS data model.

2.2 Distributed vs. Centralized

Although many GIS applications are still using a centralized approach for data storage and management, few will dispute that we are unable to continue using a centralized approach. The most popular technology paradigm nowadays is the enterprise client/server environment, in which the more powerful data processing machines are used as data servers, and the less powerful desktop PCs are used as clients. Obviously, the old, centralized GIS applications will be obsolete, and the new, client/server GIS environment will be established.

2.3 GIS Information Archive, Retrieval, Discover, and Management on Internet

In a client/server computing environment with reasonable bandwidth, we can do much more on GIS information archive, retrieval, discover, and management. For example, WWW (also called "the Web") uses universal resource locators (URLs), hypertext transmission protocol (HTTP), and search engines to facilitate information archive, retrieval, discover, and management. The efficient use of computers and bandwidth contributes a great deal to the popularity of the Web; and of course the user-friendly graphical user interface (GUI) contributes a lot too.

Without a major development effort, we should be able to achieve satisfactory results by using the existing software environment. In fact, a prototype of the GIS information discovery and management model can be implemented using WWW and its associated software programs. In WWW, a data server is a Web server that serves text, HTML documents, and images; a client is a Web browser such as NCSA Mosaic or NetScape. We can run a mail robot around the Web to pull out information we want at any time. If we implement URLs in a certain format on all the GIS data servers, the mail robot program can identify these special URLs and discover what we want in a reasonable time frame. And we can archive our data by File Transfer Protocol (ftp) into the data server. These can all be done by clicking on the homepage.

Later in this paper, we will propose a CSDM model with 3-tier client/server environment. This model will make use of the WWW as well as the local client/server environment on the Local Area Network (LAN) or Metropolitan Area Network (MAN).

2.4 Audio-visual Conference and Other Means of Communications

The necessity of having an audio-visual conference capability depends on the intended application. In some applications, such as Intelligent Vehicle Highway System (IVHS) and emergency road service, a conference system is necessary because the scene is too chaotic to be described clearly by other means. But in most of the cases, especially those involving scientific research and analysis, audio conference with very low bandwidth video conference should be sufficient.

Use of the Differential Global Position System (DGPS) will be very helpful in gathering and updating the GIS data, as well as the utilization of GIS data to do analysis.

Even the traditional mail delivery can make a difference: if we take the sharable object approach, the users can order the data, presumably on a CD-ROM, by mail. When the decision makers call on a conference, each party can insert the CD-ROM, and only use the network to pass control commands, so all parties can share applications and conserve the use of bandwidth.

2.5 Visual Interface and GUI

A consistent GUI and visual interface across all platforms is one of the secrets of success in this client/server environment. In fact, the success of NCSA Mosaic is not purely luck. It is the only program that can run on IBM-compatible PCs, Macintosh computers, and UNIX Workstations while retaining a remarkably consistent interface. And it is free of charge!

In a 3-tier client/server environment, the Web server on the Internet can be hidden from the user clients, so we can still give the users the impression of a consistent GUI. The GUI should also hide the underlying file management process from the users, and yet give the users freedom to troubleshoot the results.

2.6 Intelligent Agent

Anyone who has ever run a search engine or mail robot program on the Web has been overwhelmed by the huge amount of information available for his/her search subject. But all these information may or may not be related to the subject, depending on how people find them. For the GIS datasets, things can be more complicated because of the size of the data, and it may take much longer to search for a piece of data. Therefore, an intelligent search engine is needed to filter out some unwanted information before they reach the users. This intelligent search engine is called "intelligent agent," or "information agent."

The building of the intelligent agent is a combination of computer network, universal resource location, data organization and indexing, and artificial intelligent. The GIS community should take the lead in applying these advanced technologies to solve the distributed GIS problem.

2.7 Automation of Decision Support Process

Automation of the decision support processes is crucial to the success of CSDM. A decision making process usually involves a tremendous amount of data analysis and "what-if" simulations. Actually, the major problem is not the analysis, but the file management. A decision support system should hide all of these low level works from the users, but at the same time still give users the flexibility to change parameters, conduct quality assurance of the process, and extract comprehensive information through the decision making process.

Furthermore, an optimization program should be available to the decision makers so they can run "what-if" scenarios in batch mode and obtain the "best-guess" solutions. This is itself a big topic in the decision support theory, and it is beyond the CSDM research domain. Yet CSDM should utilize this knowledge for its own benefit.

2.8 Joint Application with Other Discipline

Even a very powerful CSDM cannot solve all of the current problems. If CSDM works with other decision support systems in other disciplines, the system can be very successful. Therefore, CSDM should be open, object-oriented, reusable, and able to accommodate major data format in the market (e.g., HDF, NETCDF, etc.)

A good example of CSDM cross-discipline application is the ecosystem management system in the environmental field. On one hand, an ecosystem management system demands a lot of GIS data, on the other hand, the ecosystem management system needs to use environmental modeling and simulation to support the decision making. Hence, by using the existing environmental decision support systems and the CSDM, one can easily extend the decision making to the earth's ecosystem. However, at this point, the two communities have been unable to build something together that is truly integrated and interoperable.

3. A CSDM Model

3.1 Data Model

The data model of this CSDM system will be object-oriented and closely disseminate real world geographical objects.

3.2 System Architecture

The system will be open, modular, and object-oriented. It will be expandable so it can work with other decision support systems to form a meta decision support system for large scale, complex problems.

3.3 Distributed Desktop Environment

The system will use 3-tier client/server structure. The first layer is the client layer, the local clients will be connected with a local GIS server, which usually be the file server on a LAN. The collection of these GIS servers is the second layer. The third layer is the Internet layer, which connects all the GIS servers together. The command and control centers of the GIS server network are strategically placed in different locations all over the country (or world). These centers are called "GIS meta-centers."

Under this distributed environment, the clients can be run by a wide range of UNIX workstations, IBM PCs, and Macintosh computers. The local GIS servers can be run on typical server machines, which have more memory and storage capacity. In the meta-center, supercomputers and high density storage are used to handle the management tasks.

3.4 Collaborative Environment

For co-authoring, document sharing, and collaborative work, the model will use the shared object approach. The users may obtain/purchase a copy of GIS data via CD-ROM or via Internet applications like ftp and WWW before the collaboration, then load the software and data to their computers when the conference begins. This approach is a collaborative process in asynchronous mode.

For the video and audio conferencing, the model will use the Internet (e.g. MBONE) or ISDN to conduct video-audio conference. So in the collaboratory process, users can start the video-audio conference, run the CSDM model, and apply remote control to manipulate the shared objects. This approach is a collaborative process in distributed synchronous mode.

3.5 GUI and Visual Interface

X/Motif-like GUI is the de facto standard in current GUI development industry, therefore, the model will use X/Motif to construct the GUI and Visual Interface of the GIS model.

3.6 Information Archive, Retrieval, Discovery, and filtering

The model will use the newly-built or existing Internet tools to archive, retrieve, and discover GIS information on the Internet.

4. Concluding Remarks

In this paper, we presented a CSDM model and discussed issues we are facing currently. We proposed a comprehensive CSDM model that will work on the Internet and will provide different levels of opportunity of participation. We used the Internet client/server environmental as our backbone and decision support as our goal, and we conclude that this model, if realized, will benefit the GIS and environmental science community at large.

Today, Internet technology is utilized at an ever increasing rate. USEPA has already used satellite TV and MBONE to transfer environmental technology to the user community. Further, more and more divisions inside EPA and other government agencies are providing their staffs with WWW access and establishing their own WWW home pages in order to provide increased access to public information as well as convey information to the public and other government institutions. At the same time, the GIS community has been spearheading the utilization of Internet to make GIS data available to the general public. It is inevitable that CSDM model will use Internet as a major testbed for applications.

In the future, the U.S.EPA Scientific Visualization Center (SVC) pursue the support of scientific visualization of ecosystem management and combine GIS into environmental scientific visualization. SVC's goals are to support the decision making processes used by environmental decision makers, and to convey the scientific research results more effectively to the general public. We will be actively supporting and participating in the development and deployment of the CSDM models.

Jeff Wang, Ph.D., System Manager/HPCC, U.S. EPA Scientific Visualization Center
jfwang@vislab.epa.gov                  P.O. Box 14365
Voice: 919-541-7662                    86 Alexander Drive, MS24/ERC1
Fax:   919-541-0056                    Research Triangle Park, NC 27709