Front page    Table of  Contents   Abstract Chapters  1   2   3   4   5   6   7   8   9   Bibliography

1. Overview

The Great Lakes, bordering on eight US states (Minnesota, Wisconsin, Illinois, Indiana, Michigan, Ohio, Pennsylvania, and New York) and the Canadian province of Ontario, comprise the world’s second-largest freshwater system (after the polar ice caps), containing about 18% of the world’s supply. The Great Lakes basin, some 500,000 km² (200,000 sq. mi.) in area, is home to over 37 million people, over 10% of the US population and a quarter of Canada’s population (EPA-GLNPO, 1995; IJC, 1995).

Despite environmental improvements since the 1960s, the Great Lakes ecosystem remains a matter of concern due to toxic air and water pollutants, nutrient loading from shoreline development, and introductions of non-native ("nuisance," or "exotic") aquatic species (IJC, 1995), all exerting pressure on a largely closed system: for instance, Lake Michigan’s retention time is 99 years. These concerns have led to several inter-state and bi-national agreements, and have created "a legacy of cooperation" among the region’s state, province, and federal agencies since the 1950s (Ralston, 1995).

The Great Lakes Commission is one outcome of these regional agreements: an interstate compact of the eight Great Lakes states, it was formed in 1955 with a mandate to promote "orderly, integrated, and comprehensive" management of Great Lakes water resources. In mid-1993, the Great Lakes Commission set out to link the region’s various stakeholders via networked communications and information interchange. This effort, known as the Great Lakes Information Network (GLIN), quickly enlisted the participation of dozens of U.S. and Canadian Federal agencies, state and provincial agencies, and other public and private organizations with an interest in the ecology and economy of the Great Lakes. The resulting network has gained quite high visibility as a "one-stop shopping" resource for information on Great Lakes-related organizations and activities. As such, GLIN provides an interesting look at the challenges of growing a highly networked, mostly unplanned information infrastructure within a complex institutional context.

The following sections first sketch the history of the GLIN infrastructure, its institutional and technological context, and the choices made in building it. Subsequent sections then evaluate GLIN by the quality criteria outlined in the previous chapter, and by its impacts on organizations and policy. The last section outlines the challenge of drawing on the information resources of participating organizations, and of interacting fruitfully with other, overlapping information-infrastructure efforts.

2. Brief history

"The little pilot has become a messy beast!"
- GLIN Board minutes, July 1994

The pilot project allowed them to leverage additional grants from the U.S. Environmental Protection Agency (EPA) and the National Telecommunications Infrastructure Administration (NTIA); Ameritech renewed and increased its funding in 1996. In April 1994, an interagency Advisory Board was formed to represent about 20 different organizations in guiding GLIN’s growth. As expressed at this first Advisory Board, GLIN’s intent was to help organizations all over the Great Lakes put up information servers and to link them to each other: "You will keep the information on the computer in your office—and the Internet, a common tool, will connect us all." (GLIN Board minutes, April 1994).

During the two-year pilot phase, the network attained high visibility, with significant media coverage (WSJ, 1994) and data links from Canada’s Environmental "Green Lane," the U.S. EPA, and others. Its use grew more rapidly than anyone expected, and in mid-1995, as the two-year pilot stage drew to a close, the Advisory Board was faced with several challenges for the network’s future: building financial and technical self-sufficiency, making effective use of GLIN’s growing web of information, encouraging holders of ecological and economic data to put it online, and defining GLIN’s scope in relation to other, overlapping data sharing efforts in the region. The next several sections explain how they arrived at these challenges, and suggest some possible solutions.

3. Context of the infrastructure

a. Institutional context

In devising GLIN, the Great Lakes Commission was building on a long tradition of cooperation among the states and with federal agencies and others in the region. Past agreements (rules, treaties, and the like) on the Great Lakes environment include (i) the Boundary Waters Treaty of 1909 (water use and allocation between the US and Canada, overseen by the International Joint Commission); (ii) the Great Lakes Water Quality Initiatives of 1972 (phosphorus runoff), 1978 (toxic pollutants), and 1987 (airborne pollutants); and (iii) the Great Lakes Binational Toxics Strategy of 1997 (joint US/Canadian pollution targets). The two federal agencies responsible for protection of the Great Lakes are the US Environmental Protection Agency (EPA) and its counterpart Environment Canada. Also at the federal level, the US Army Corps of Engineers, Detroit District, is responsible for maintaining shorelines and shipping channels in the Great Lakes. The governments of the eight Great Lakes states and the province of Ontario form another important part of GLIN’s institutional context. Finally, several regional and professional associations (Great Lakes Protection Fund, Council of Great Lakes Governors, and others) have a long-standing interest in the Great Lakes, along with non-profit groups (National Wildlife Federation, Nature Conservancy), and academic and research facilities (the Great Lakes Environmental Research Laboratory of the National Oceanographic and Atmospheric Administration (NOAA)). These few examples suggest the organizational complexity within which GLIN was formed.

b. Technological context

GLIN came into being amidst a significant amount of ongoing information technology effort in the Great Lakes region. Several groups had begun to build sizable information resources on the Great Lakes; and a few had embarked on interagency information sharing infrastructures of their own. The following are salient examples:

First, the Environmental Protection Agency (EPA) had lengthy experience with environmental databases, dating back to the 1960s, with STORET, "the world’s oldest and largest environmental data system," running on an IBM mainframe (EPA, 1997). An important EPA emphasis in the Great Lakes was a data warehouse named Envirofacts, containing information on air and water quality, regulated facilities, permits, and Superfund sites. In 1994, EPA staff began building several Internet gateways to Envirofacts (EPA, 1997b; Adhikari, 1997), and integrating it with custom mapping tools ("SiteInfo") on the Web (EPA, 1997c). Furthermore, EPA's regional Water Division had a large set of digital geographic data on the Great Lakes, with plans to make it publicly available via the Internet (EPA, 1996).

The EPA also began a multi-year contract in 1993 with the Consortium for International Earth Science Information Network (CIESIN—pronounced "season"), a non-profit research organization largely funded by the global climate change research program of the National Aeronautic and Space Administration (NASA). Under this contract, CIESIN was cataloging the environmental data holdings of the EPA and other agencies in the Great Lakes region, and building a "Great Lakes Regional Environmental Information System" that would allow users to tap into the EPA’s information services via the Internet (i.e. via the Web as well as a special-purpose "CIESIN Gateway" client). (CIESIN, 1996b).

The U.S. Army Corps of Engineers, Detroit District, had a large GIS facility (as of March, 1995, about 7 Gb of vector data and 320 Gb of raster data, including a complete set of digitized shoreline airphotos for the Great Lakes) which they used for shipping lane and shoreline operations and for analyzing the impacts of Great Lakes water levels. The Corps had begun putting real-time hydrological information online (at http://sparky.nce.usace.army.mil), and was working with both GLIN and CIESIN staff to catalog its data holdings and provide networked access to them. In mid-1995 the Corps’ headquarters put a server online (http://corps_geo1.usace.army.mil) providing Web and WAIS access to metadata on their geographic data holdings.

The Corps also worked closely with the State of Michigan's Department of Natural Resources, which had about 30 Gb of GIS data files, and distributed them via a statewide consortium known as IMAGIN ("Improving Michigan's Access to Geographic Information Networks"), as well as through sales and cooperative agreements. IMAGIN was a carefully institutionalized consortium of public and non-profit organizations in Michigan, funded by the W. K. Kellogg foundation, with formal agreements for exchanging and using geographic information held in a central data archive at the Library of Michigan (Beaulac et al., 1994).

Lastly, Environment Canada and the Canada Center for Inland Waters had begun building the Great Lakes Information Management Resource (GLIMR), a set of environmental data accessible through the World Wide Web and indexed through a multi-server keyword system. Officially launched in March 1995, GLIMR (at http://www.cciw.ca/glimr) was the first of several such systems within Environment Canada's "Green Lane" networking initiative.

As a result, GLIN’s efforts in information sharing had many precedents among the region’s government agencies. This meant that plenty of expertise was available to them; but on the other hand they didn’t have the luxury of starting from a clean slate: several competing styles and affiliations for information sharing already existed. (The three information sharing initiatives above (CIESIN, IMAGIN, and GLIMR) formed an unusually important part of the GLIN story: Section 7 below explains how they made GLIN’s context both fruitful and chaotic, and suggests the benefits that GLIN might have reaped from a more intentional collaboration with these information sharing efforts.) One final, important element of GLIN’s technological context: many state agencies in the region had little or no experience with Internet access, and were eager to get online through GLIN.

4. Infrastructure choices

Partly in response to its institutional and technological context, GLIN embodied several choices of organizational structure and technology.

a. Institutional arrangements

GLIN embraced its complex institutional context from the day it was launched. According to an attendee list on the GLIN gopher server (gopher.great-lakes.net), the Great Lakes Commission invited over 100 people to a "GLIN kickoff meeting" in June 1993, representing 60 different organizations. The Advisory Board, formed a year later, was quite broadly defined and represents GLIN’s institutional context quite well (Table 4-2).

b. Technological design

In mid-1993, GLIN began as a gopher server, the Internet technology du jour at the time. In mid 1994, as the World Wide Web gained prominence on the Internet, GLIN began creating a few Web pages, but didn’t make the Web its primary mode of information sharing until late 1994. (In early 1995 the gopher pages still provided more complete information, and any comprehensive search had to include both information spaces.) The Great Lakes Commission intended this Web/Gopher server as a single access point to information on the Great Lakes, but not as an information "hub" in the usual sense: although the Great Lakes Commission’s server hosted the pages of several GLIN partners, decentralization was to increase as partners put their own servers online and move their files onto their own machines. When GLIN partner agencies did have a server online, the GLIN front page at www.great-lakes.net maintained hypertext links to it.

Much of GLIN’s early efforts were focused on helping and training less-experienced agencies to set up their own servers. To get basic communications started quickly, however, many people (about 400, well in excess of the 30 initially expected) were given accounts on the Great Lakes Commission’s server, which they reached via dial-up. (This resulted in a greater-than-expected training and technical support burden on GLIN staff: in early 1995, GLIN’s technical director expressed concern about having "soft-pedaled" the actual cost by providing as much technical support as they did to these agencies, instead of helping them sooner to rely on local Internet services and support.)

Less visible, but equally important forms of information sharing on GLIN included file transfer protocol (ftp) servers and several electronic mailing lists: for instance, one public list, glin-announce, kept a broad community abreast of regional events; another, closed, list, air-toxics, allowed a few dozen specialists to collaborate on air-pollution monitoring, as part of the RAPIDS project described below.

5. Information sharing characteristics

The information infrastructure that resulted from these choices was quite informally structured, with mostly fixed documents and images, with a few more detailed data sets emerging in early 1995. The following paragraphs relate the characteristics of the infrastructure, including its size and quality.

a. Forms of information sharing

As of early 1995, GLIN's information sharing was of four kinds. Most prominent were summaries (reports, fact sheets, brochures, etc.) that presented quick facts on the Great Lakes and its stakeholder organizations. According to GLIN’s director at the Great Lakes Commission, the audience for this information was the "intelligent layperson," including staff in other public agencies, educators, and the general public. Electronic mailing lists and sharing of data and software through ftp were another form of information sharing on GLIN. The result was akin to an atlas or reference library, with a lot of facts within easy reach, and coming from many reliable sources.

Yet many respondents to the pre-GLIN user survey in 1992 had expressed the need for "technical data" and "research findings" on a frequent basis; and in early 1995, a few detailed, dynamic data holdings began to come online: the Great Lakes Environmental Research Laboratory had daily satellite images from the "CoastWatch" program of the National Oceanographic and Atmospheric Administration (NOAA); and the US Army Corps of Engineers had begun putting reports and forecasts of Great Lakes water levels online. (Interestingly, not everyone expected their detailed data to be of much interest to anyone else. A member of GLIN’s Advisory Board recalled that federal agencies invited to a GLIN meeting in September 1993 conceded that they could make data available to GLIN with some effort, given that it "belonged to the scientists"—but that "no one would likely be interested.") The Great Lakes Commission had always had an interest in providing detailed geographic data from the start; but as of early 1995, GLIN’s forays into distributing GIS data were still limited to one person’s experiments with Arcview software. (However, Commission staff continued their experimentation; in mid-1996, as off-the-shelf toolkits became available for integrating GIS into Web services, they evaluated MapInfo Corp.’s ProServer and ESRI’s ArcView Internet Map Server. As a result, in May 1997 the Commission was invited by the EPA’s Great Lakes office to submit a proposal for a four-year project to tie GLIN to networked GIS services.)

b. Size of the infrastructure

There are many ways to measure GLIN. Some participants saw size or usage data as relatively unimportant; but others had carefully assessed the number of Web pages on their server, the number of users, the size of their data holdings, etc. In early 1995, the www.great-lakes.net server had 250 user accounts, 500 email list participants, and some 1,400 files online. In February 1995, the gopher server (gopher.great-lakes.net) reached a peak of 12,600 file retrievals a month, before declining gradually as Web access became more widespread. In April 1997, the Webserver at www.great-lakes.net received 67,000 "hits" (requests), transferring a total of 442 Mbytes. These are clearly only rough measures: not all hits represent Great Lakes users; there are always more transactions than actual pages retrieved; and hits tend to lump together large and small files. GLIN’s director at the Great Lakes Commission typically paid more attention to another measure of growth, the rate at which state agencies in the region were putting information servers online. This showed much slower growth, so she concentrated her efforts there.

c. Quality of the shared information

As outlined in the previous chapter, the quality of an information infrastructure depends on how well the shared information matches the needs of its users, in particular its accuracy, timeliness, and encapsulation.

i. Precision and accuracy

ii. Concurrency and timeliness

iii. Usability and encapsulation

d. Quality of the information infrastructure

The above quality measures are from the perspective of the data recipient. Several additional "quality" aspects of information sharing pertain not to the data, but to the infrastructure that provides it.

i. Reciprocity: what it takes to publish data

ii. Flexibility: how people join the network

iii. Scalability and sustainability

iv. Non-intrusiveness and transparency: how much change is required

6. Infrastructure impacts

Finally, perhaps the most important measure of an information sharing infrastructure is its impact on people and organizations. GLIN's potential was always clear. To one EPA data specialist, it offered a chance to see other people's information for free and in a "one-stop shopping" mode. For a researcher at the Great Lakes Environmental Research Laboratory, it afforded opportunities for publicizing current laboratory experiments and finding other researchers to work with. To the EPA and many others, it provided a vital public relations function. GLIN's success stories often revolved around its high-visibility users (high-ranking officials), its inclusion of many diverse participants, or its rapid rise in usage, beyond its founders’ expectations.

a. Interdependent, "ecosystem" thinking

One important impact of GLIN on people and organizations was the interdisciplinary "ecosystem perspective" it presented on the Great Lakes. GLIN staff sought out information providers from a wide variety of sectors: pollution prevention (EPA), industrial development (the Council of Great Lakes Industries), business and economics (Federal Reserve Bank of Chicago), and nature conservation (National Wildlife Federation), to name a few. Putting all these kinds of information near each other, linked by hypertext, may have highlighted the complexity, diversity, and interdependency of all these parts of the Great Lakes ecosystem. (However, as one researcher at Environment Canada observed, this "ecosystem thinking" effect depended on the user: some might simply "raid" GLIN to get what they wanted regardless of GLIN’s educational intent.)

A good illustration of how GLIN could communicate Great Lakes issues in a holistic way is the interactive map of "Areas of Concern" on the Web (Figure 4-3), a page jointly maintained by Environment Canada and the US EPA. On this map, color-coded dots representing harbors or shipping channels identified as needing remedial action by the 1987 Great Lakes Water Quality Agreement. Each of these dots is linked to a Web page on either EPA or Environment Canada servers, depending on which side of the border it’s on. This simple artifact communicated powerfully and creatively the interdependence of the two country’s actions, and emphasized the cooperation between counterpart environmental agencies (insofar as making joint maps indicates cooperation).

This begs the question: has GLIN actually changed any tasks or processes of environmental management? Tangible evidence is still too scarce to tell. GLIN may still be too new a facility for people to depend on it for their work, and to change their work processes accordingly. Another view is that GLIN's loosely organized growth to date has made it difficult to find needed information across its many data sources. GLIN staff point to one clear impact, the Rapid Air Pollutant Inventory Development System (RAPIDS).

b. RAPIDS: a clear impact?

RAPIDS was a complex, science-intensive piece of software developed by air-quality specialists and software engineers around the Great Lakes region. Contributors communicated with each other via the air-toxics mailing list housed on www.great-lakes.net. They used this list to discuss various aspects of the analysis, and to provide feedback on successive versions of the product, which they shared via the ftp server at ftp.great-lakes.net (electronic exchange of software enabled new versions to be distributed to the whole group in minutes). Another link with GLIN was that RAPIDS was coordinated by GLIN’s director at the Great Lakes Commission. Many involved felt that the RAPIDS product was of exceptional quality and that it was produced exceptionally quickly; and they attributed both of these aspects to the networked collaboration that went into developing it, and it became known as a clear impact of GLIN. However, RAPIDS was a separate effort from GLIN, with its own budget, personnel, and goals; its only links to GLIN were the machines it used as mailhub and software locker, and its coordination by GLIN’s director. So, while RAPIDS does show the impacts of collaboration over the Internet, its role as a GLIN success story is not nearly as clear. It does, however, suggest the kind of joint projects that could occur through the use of GLIN’s diverse information and expertise.

7. Challenges and lessons

By any standard, GLIN was an ambitious project, and it surfaced several dilemmas related to supporting its growth, populating the servers with detailed environmental data, and defining the network's purpose amidst a complex network-of-networks.

a. Supporting GLIN's growth

As the GLIN community grew, so did the burden on its staff. In particular, the growth of user accounts on great-lakes.net had a drastic impact on GLIN’s technical director. As the lone technical contact for all current and potential users, he expressed concern that GLIN staff were becoming, in his words, "a bottleneck to the network." For instance, supporting such a large user community had kept him from launching longer-term GLIN projects such as indexes, interfaces, improved search strategies, or anonymous electronic discussions. He felt that GLIN staff should have been more strict about (merely) training and equipping GLIN partners, and about pushing them to become real Internet nodes with local support, instead of providing so many with Web and Gopher files and accounts on great-lakes.net.

b. Populating GLIN with "real" data

One of GLIN's primary goals was to bring together data from many different sources in the region. However, holders of detailed information were slow to publish it on the Internet. Some were wary of Internet security breaches; some saw little outside demand for their data; some were concerned about losing control over their data; still others were concerned about the staff burden of putting data online and providing support to users.

i. Security concerns

ii. No perceived need

iii. Risk of misinterpretation or misuse

iv. Staff burdens

c. Adjusting to a complex network-of-networks

GLIN was initially conceived as a single autonomous network, but its content and purpose overlapped with several other ongoing efforts to share environmental data in the Great Lakes region. One of its more difficult challenges was how to define its role in light of these other, significant information-sharing projects, some of which had quite different styles and views of information sharing: Environment Canada's Great Lakes Information Management Resource (GLIMR), EPA and CIESIN’s Great Lakes Regional Environmental Information System (GLREIS), and the State of Michigan’s IMAGIN ("Improving Michigan's Access to Geographic Information Networks") consortium.

i. Distinct approaches to data sharing

First, Environment Canada’s Great Lakes Information Management Resource (GLIMR) was essentially GLIN’s Canadian counterpart, in that it sought to interchange information on the Canadian side of Great Lakes with anyone on the Internet at no cost. It targeted Environment Canada’s offices as sources of information, but it was also creating partnerships with other agencies to link to their information. Rather than overlap with GLIN, it complemented GLIN’s focus on the eight Great Lakes states by covering the remainder of the basin; furthermore, GLIMR’s chief technical architect was developing multi-server search schemes and shared his technical expertise freely with GLIN staff.

Second, the IMAGIN consortium, led by Michigan’s Department of Natural Resources, had a carefully institutionalized policy for sharing geographic data produced by that Department, and for ensuring its appropriate use by (only) Michigan government agencies (local, county, state, federal), public universities, and some non-profit groups. Yearly dues ranged up to $450; members would agree not to "misuse" the information (i.e. use it in a way that would misrepresent its accuracy, or use it for personal or private gain). For those outside the consortium, there was a detailed price list: $120-$900 for geographic datasets, depending on size, and $50 per county for tabular data. A primary concern was to protect the ownership of information and to ensure returns on investment for Michigan’s Department of Natural Resources, the creator of the data. Yet IMAGIN’s manager preferred not to charge money, but to make agreements for enhancing the information in return for its use. For example, he experimented with free distribution to some engineering firms engaged in public-sector contract work. Nonetheless, some in the region saw these rules as restrictive or expensive: for instance, a GIS specialist suggested that obtaining IMAGIN data for use on GLIN would be difficult.

Third, CIESIN’s Great Lakes Regional Environmental Information System (GLREIS) was a Web-based system that provided access to metadata, data, and data utilities of relevance to the Great Lakes region." Part of building this system involved cataloging the region's data holdings (primarily EPA’s, but those of other data providers as well, such as the Army Corps of Engineers Detroit District and Michigan’s Department of Natural Resources). Within this project, CIESIN also provided networked access to EPA’s Oracle databases (ENVIROFACTS and others), modeling and decision support capabilities on the Web, interactive map servers using ESRI’s Arc/Info software, and demographic data viewer for Census data. As part of their cataloging effort, CIESIN created simple Web pages describing several agencies in the region (e.g. Michigan DNR and the International Joint Commission). Initially, some in the region mistook CIESIN’s work as a shortcut to getting online; but CIESIN’s contract with EPA left little room for technology transfer to the region’s agencies, and the contract’s short time horizon meant that CIESIN couldn’t afford to wait for agencies to put their own servers online before cataloging and describing their information. So most of the datasets and guides delivered by CIESIN were on a single server, (epawww.ciesin.org), and it had little impact on the network capabilities of agencies in the region. This led some GLIN participants to think CIESIN was promising easy Internet solutions instead of helping clients to build their own. However, CIESIN’s agenda was to provide networked access to EPA data and EPA-related data on the Great Lakes, independently of the agencies themselves. Many GLIN participants had long recognized GLIN’s need for additional "hard data" of this sort; however, some of them saw CIESIN not as a data-intensive complement to GLIN, but as an overlapping, possibly competing regional information infrastructure, and thus a matter of some concern. According to minutes from the GLIN Board’s April 1994 meeting, GLIN’s representative from EPA’s Great Lakes office "appealed to the Advisory Board to help EPA manage this effort, to make sure that CIESIN's effort is something worthwhile and not redundant with GLIN." In an effort to ensure that the two efforts would be coordinated and mutually beneficial, not long after that meeting the GLIN Board circulated a proposed memorandum of understanding between GLIN and CIESIN, that called for a "common public access mechanism (...) to GLIN and the Great Lakes REIS under the GLIN banner." No such mechanism was ever put in place, and the two programs existed alongside each other, with a few hypertext links to each other but a fair amount of overlap, for instance between GLIN’s agency pages and CIESIN’s organization guides, or between GLIN’s work with geographic data services vs. those that CIESIN had built over a year earlier.

ii. Reconciling the different approaches

Members of GLIN’s Advisory Board had suggested making GLIN's focus more specific, cautioning that "we can't be all things to all people." Others were more reluctant to narrow GLIN's scope, suggesting that "GLIN may not define Great Lakes information, but merely deliver it." There was some debate as to whether GLIN's focus was congruent with the GLC's charter, or whether it "may leave the Great Lakes Commission and become more region-wide." (This was resolved for the moment in mid-1995 with a formal Great Lakes Commission Policy Statement endorsing GLIN.) Whatever form GLIN takes in the future, adapting to an increasingly networked environment may require a careful evaluation of GLIN's mission as part of a larger system of Great Lakes infrastructures: what it does best and how that complements, builds on, or reinforces these other data sharing efforts.

8. Conclusions

As a pioneer in networked information sharing among governmental agencies, GLIN made impressive strides in a short time. It was committed to modern networking technology and to linking loosely coupled organizations. Its regional focus on the Great Lakes watershed provided a powerful shared concern among a wide array of participants spanning all levels of government in two separate countries. Yet it had more trouble than expected drawing on the information resources of data providers in the region, and it unexpectedly faced the challenge of other, competing views of information sharing. Only after several years of existence did it begin tackling the question of networked geographic data. As such, GLIN will continue for some time to be a fruitful example of the challenges and tradeoffs of building infrastructures for sharing geographic information.

Previous chapter | Top | Next chapter

Front page    Table of  Contents   Abstract Chapters  1   2   3   4   5   6   7   8   9   Bibliography