Introduction
Ephemeral Cities is a historical digital atlas project, currently under construction, using a geographic information system (GIS) as a map interface for access to digital library content and as a tool for understanding local and state history. Based on historic fire insurance maps created by the Sanborn^® National Insurance Diagram Bureau, the atlas integrates documents, ephemera, maps, museum objects, and photographs to populate a rich, geo-temporal mosaic. Funded in part by the Institute for Museum and Library Services, this prototype project focuses on three key Florida cities: Gainesville in northern Florida, the site of the state’s largest university; Tampa in central Florida, the west coast hub of commerce and finance; and Key West in southern Florida, an island rich with Bahamian and Cuban influences. The time span is 1884–1903.

Partnerships
Ephemeral Cities was born of existing and new partnerships. Lead and primary partners, the University of Florida in Gainesville, the University of South Florida in Tampa, and Florida International University in Miami, are all members of the Publication of Archival, Library & Museum Materials (PALMM) digital library collaborative. Each has contributed to Florida Heritage and Florida Environments and other Florida-related collections for more than five years. Each also maintains relationships with local libraries, historical and genealogical societies, museums, and other cultural institutions. Several of these local institutions, in turn, have partnered to contribute additional content. Some, including the Alachua County Library District, the Alachua County Clerk of the Court, and the Monroe County Public Library, have created and maintain their own digital libraries.

Partnerships were vital to the project; no single cultural institution held a sufficiently complete historical record for any given place. Museum collections, in particular, lent a tactile flair to an appreciation of the past. And, the regionalized hub-and-spokes approach to construction of the most complete collection ensured that those new to digitization had immediate access to an experienced partner and gave less experienced partners access to digital technologies and services. Typical of the PALMM experience, digitization and object description were carried out by the holding institution, using common guidelines, while more complex and more expensive tasks, e.g., text conversion (optical character recognition and mark-up) and programming, were centralized. Figures 1-3 illustrate partner relationships. Programming, for the description and contribution of digital objects, helped to ensure consistency among the partners.

The partners include five libraries, four museums, and two public records offices: (in Gainesville) Alachua County Clerk of the Court, Alachua County Historic Trust/Matheson Museum, Alachua County Library District, and the University of Florida; (in Tampa) Henry B. Plant Museum, Tampa Bay History Center, and the University of South Florida Libraries; and (for Key West) the City of Key West, Florida International University, Key West Art & Historical Society, and the Monroe County Public Library. The Florida Center for Library Automation (FCLA) provides digital library technologies and digital archiving services. The Florida Center for Instructional Technology at the University of South Florida creates the educational modules. The Digital Library Center (DLC) at the University of Florida provides text services and project coordination. The University of Florida’s Government Documents and Systems Departments, in collaboration with the DLC, provides GIS and associated programming and database services. As with any project of this size and complexity, communication is both a constant necessity and problem.

Additionally, because many resources are held privately, the project partners sought the assistance of local collectors. Although many allowed access to their collections in-situ, many more collectors brought their resources to “My Town” events, where objects were imaged and descriptions collected. “My Town” events, while allowing collectors to share their sense of history, suffered one limitation: many of those participating were elderly. Their participation represented tremendous physical effort on their part and many potential participants, undoubtedly, were dissuaded by the effort. The project would have benefited from additional planning and out-reach to this community of “living history.”

Collective Content
Ephemeral Cities references previously and newly digitized resources. Content ranges from published materials—the strength of existing digital collections—to diaries, letters, oral histories, and photographs. Museum and “My Town” partners contribute artifacts, a gamut of every-day objects: kitchen utensils, photographic equipment, writing implements, clothing, etc. But, the project’s work could not have been completed without maps and name-rich resources.

Sanborn^® Fire Insurance Maps of Florida, historically published every seven years on average, establish the base layers of the GIS. Sanborn^® maps are geographically precise and historically accurate. They provide construction information on each mapped structure and additional use information for many of the same. Each sheet of all map sets published for each of the three cities was commercially geo-rectified. Rectification allowed programmers to associate known places with geographic (and temporal) coordinates and, subsequently, to layer one map atop another. Layering revealed change, often pervasive, over time across the city landscapes.

New uses changed the character of Tampa between 1884 and 1903 (Figure 4). This block near the waterfront changed from shopping district to an after-hours entertainment center, as transportation took more goods inland and as more goods took more people to handle and deliver them. In 1884, Tampa’s economy and population was centered on Tampa Bay. The southern corner of Franklin was occupied by a barber, cabinet maker, and a seller of fancy oils. By 1903, the railroads had finally reached Tampa. The southern corner of Franklin supported a café, two saloons, and a liquor store.

Without name-rich resources the reasons for change would remain mute. Those resources allow causative factors, such as people and events, to be examined. They answer the question “Why?” Ephemeral Cities postulates that giving voice to history can lead to a sea-change in our understanding of history. More importantly, it suggests that, by bringing together whole communities’ resources we can see change; and, if we can see it, we can understand it. Name-rich resources come from three sources: (1) newspapers, which fix people and events in specific time at specific places; (2) government records, particularly transactional, legal, and census records that capture people in civil actions (e.g., birth, marriage, and death); and (3) city directories, telephone books and the like, which fix people to known places (e.g., homes, workplaces, churches, and clubs). Many other written and printed sources provide names but most of these are low-yield sources. The parsing of name-rich resources is discussed below.

Assumptions
Ephemeral Cities makes several assumptions, most of which have thus far proven true, some, however, with considerable effort. First, it is assumed that citizens of Florida and tourists have general knowledge of the state’s geography. At present, information about the use of the PALMM Aerial Photography: Florida collection, another GIS project, supports the validity of the assumption. This largely anecdotal information also suggests that the GIS interface requires simplification. Second, it is assumed that everything can be fixed in time and place. This has proven to be true generally, at least among name-rich resources. The project continues to explore means of collaborative information building in the web environment. Similar to the (analog) experience of the “My Town” events, such methods would serve to fix date and place to other resources and to collection of personal histories that would otherwise be lost. Methods of establishing and attaching levels of trust and independent verification remain outstanding work items. Third, it is assumed that semantic processes can be applied to metadata and searchable text to facilitate tagging and discovery. Although software enabled processes (e.g., Gate and Annie; cf, http://gate.ac.uk/) facilitate tagging, they work well only against structured authorities. Augmentation of existing authority systems is discussed below; the task suggests the need for gazetteers and enhancement of name authority records.

Technologies
Ephemeral Cities makes use of several technologies for GIS, for distribution of digital objects, and for tagging. The GIS interface is built upon the ESRI Map Server and the Tomcat servlet container for ESRI’s ArcIMS. Digital object distribution is supported by an FCLA-administered XPAT engine and SQL (Structured Query Language) applications of some of the independent partners.

Because research-discovery occurs within the text of documents rather than against search-discovery based on descriptive metadata, all textual resources have been converted to searchable text. Prime Recognition, configured with six optical character recognition (OCR) engines, ensures optimal text conversion quality.

Selected name-rich resources were parsed into enhanced authority records in an SQL database. These records are based on MARC name authority record structure and on a variety of common geographic schemes (e.g., Alexandria Digital Library [ADL] Feature Type metadata, Geographic Names Information System [GNIS], etc.). Authority records associate persons, named things, and events with places at specific times. Augmented with information from city directories, land ownership files, and other information sources, these records become the base of a gazetteer. Geographic coordinates in the gazetteer allowed linkage to the Sanborn^® maps in the GIS.

Gate and Annie scripts subsequently use these records to identify and tag names in texts, thereby enriching associations. This both augments the authority records and links archival, library, and museum resources to the maps. Taken together the process of parsing name-rich resources and using parsed information to uncover related information in other texts and metadata is akin to sampling DNA to establish a family tree.

Access
A query engine is currently under construction and a prototype is working in a test environment. The user enters the “collection” through a map interface, potentially zooming in on an area of interest. Guided queries help the user to find specific information. The GIS helps them to explore relationships. An illustrated example follows.

Towards the Future
Ephemeral Cities suggests the creation of a historic city atlas encompassing the nation. Sanborn^® and similarly precise historical map sets were created for U.S. cities with populations over 2,500. Updated periodically, they represent snapshots of the growth of cities. Upon public launch planned for mid-2005, fully elaborated procedures will be released. The project’s planners anticipate extending the model to other Florida cities and hope that others will do so elsewhere.

Programs not yet able to support GIS might also look to the future. Research resources valued by genealogists are an excellent place to begin. City directories and similar content, surprisingly, have been widely overlooked. These resources begin to tell us who lived where and when. Development of services supporting conversion and text-searching of newspapers is also key. Partnership with local government records offices unlocks a wealth of temporal and geographic information for future discovery.¹

As we free the content of our digital collections for temporal and geographic discover, Ephemeral Cities also makes possible a more fantastic vision. The geographically aware cellular telephones and digital cameras of the future will enable a form of education and tourism that turns historic sites into museums and libraries that serve-up information as one passes through a location (cf, www.uflib.ufl.edu/digital/collections/EphemeralCities/EPCfuture.avi).

Issues
Ephemeral Cities is an attempt to address research functions rather than to provide additional research objects. Its most apparent means is the GIS-enabled map interface. It joins a number of similarly enabled digital libraries.² Each is predicated upon the notion that regardless of the type of research or information-use conceived, time and place are constants of the process. Nothing exists outside of time or place. Ephemeral Cities distinguishes itself both in its depth and attempts to mine data from digital texts. No other GIS-enabled project has attempted to establish layers of information at the building level that can be peeled back or piled on in sheets of time.

Like previous geo-temporal digital library projects, Ephemeral Cities identifies particular difficulties as development needs. Digital gazetteers, both more and more-historically detailed gazetteers, are needed. Enhancement of authority record structures (e.g., Library of Congress Name Authority and Geographic Names Information System) is also needed, as are methods for trusted, distributed augmentation of those records. Authority systems currently lack sufficient historical record of name changes and of uses. And there is a need for a distributed system of contribution and record augmentation, if only because of the sheer volume of names and name information that can be compiled through machine processes. Finally, though not discussed here, Ephemeral Cities identifies a need to move from association of named places with points, a single set of coordinates roughly identifying the center of an object, toward association with bounding boxes or polygons, roughly identifying the boundaries within which an object rests. A point is an existential designation that says nothing of the shape that an object takes either at any given time or as it moves through time.³ As cities grow and counties shrink, a point is a virtual needle in a hay stack.

2 Reports of other notable projects examining issues of space and time include:

• Gregory Crane. “Designing Documents to Enhance the Performance of Digital Libraries: Time, Space, People and a Digital Library of London,” D-Lib Magazine, v.6, no. 7/8, July/August 2000, http://www.dlib.org/dlib/july00/crane/07crane.html;
• Scott R. McEathron et al. “Naming the Landscape: Building the Connecticut Digital Gazetteer,” INSPEL 36(2002)1, pp. 83-93, http://www.fh-potsdam.de/~IFLA/INSPEL/02-1mcsc.pdf;
• Michael Buckland and Lewis Lancaster. “Combining Place, Time, and Topic: The Electronic Cultural Atlas Initiative,” D-Lib Magazine, v.10, no. 5, May 2004, http://www.dlib.org/dlib/may04/buckland/05buckland.html; and
• Patrick McGlamery. “A Decade of Spatial Metadata Content Standards,” Readex Digital Institute (2^nd: 2004 October 7-9). Unpublished.

3 For illustration of the problem, see Alachua County (FL) Ancient Records’ Census Maps http://www.clerk-alachua-fl.org/Archive/AncientJ/1830map.html.