Overview
Historic Fires Near Me visualises digitised newspaper reporting on bushfires across Australia between 1850 and 1900. Each mapped point corresponds to at least one newspaper article that mentions a fire event associated with that location.
Data sources
Newspaper reporting was sourced from digitised newspapers available through Trove. Relevant articles were identified using a keyword search for the term bushfire. Placenames surrounding this search term were extracted and geolocated using Named Entity Recognition (NER) and a custom geocoding algorithm built using infrastructure provided by the Time Layered Cultural Map of Australia.
How the map works
The map displays and clusters locations dynamically. At lower zoom levels, clusters indicate broader areas of reporting about historic fires. As users zoom in, clusters separate into individual points. The Results panel lists the articles corresponding to the currently visible points and applied filters.
- Select clusters to filter the Results panel to relevant articles.
- Use the timeline to restrict the period displayed on the map.
- Use the search box to locate towns or suburbs mentioned in reports.
Further information about the project, including background and methodology, is available on the project website: fiannualamorgan.com/historical-fires-near-me.
Interpreting results
This map shows patterns in historical newspaper reporting about bushfires between 1850 and 1900 and provides a valuable resource for understanding the history of fire in Australia. It does not constitute a complete or definitive record of all bushfires that occurred during this period.
Newspaper reporting changed substantially over the nineteenth century as populations grew, new technologies such as the steam train and the telegraph transformed how news was gathered and disseminated, and editorial practices evolved. As a result, some regions and time periods are more visible in the data than others.
Interpreting mapped locations and dates
- Locations shown on the map are placenames that appear near the word bushfire in newspaper articles, not the precise sites where fires occurred.
- In many cases, these placenames refer to nearby towns or communities that reported on, were affected by, or responded to fire events.
- Dates associated with mapped points reflect when fires were reported in newspapers, not necessarily when the events themselves took place.
Despite these constraints, the dataset offers meaningful insight into the spatial and temporal dynamics of historical bushfire reporting and provides an important evidentiary base for fire history, environmental history, and media history research. The workflow used to construct the dataset is outlined below.
The following steps outline the workflow used to identify and isolate bushfire-related content from the broader corpus of digitised newspapers available through Trove.
Build the corpus
The Trove API was used to collect digitised newspaper articles about bushfires between 1850 and 1900. This accommodates variations in spelling of the term bushfire.
Apply initial named entity recognition
The Stanford NER three-class model was applied to extract PERSON, ORGANISATION, and LOCATION entities from the corpus, producing an initial set of prospective placenames associated with reported bushfires.
Establish a context window
A context window was defined around occurrences of the term bushfire to capture placenames appearing in close descriptive proximity. A span of approximately 600 characters was used to accommodate variation in article structure, genre, and nineteenth-century journalistic conventions.
Apply relevance rules
Placenames that were too geographically broad to indicate a specific bushfire location, such as Australia, were excluded to improve geolocation accuracy.
A custom-developed geocoding algorithm was used to assign approximate coordinates to historical placenames while making uncertainty explicit. The heuristic draws on the Gazetteer of Historical Australian Placenames provided through the Time Layered Cultural Map of Australia. This approach avoids opaque black-box geocoding services and explicitly encodes uncertainty within the resulting dataset.
Retrieve all gazetteer entries
All historical gazetteer entries corresponding to each identified placename were retrieved, including associated colony or state, geographic class, and coordinates. Minor spelling variation was retained to accommodate OCR errors.
Identify the most likely colony
Colonies were evaluated using an above-chance threshold to determine whether a single geographic context could be assigned. Placenames failing this test were classified as ambiguous and excluded from automated geolocation.
Aggregate coordinates within a colony
When a single colony could be confidently assigned, coordinates from all matching gazetteer entries were averaged to produce a single approximate location representing the placename.
Detect duplication within colonies
Median distance thresholds were used to identify cases where multiple unrelated locations shared the same name within a single colony. Such cases were flagged as ambiguous.
Handle OCR and spelling variation
Fuzzy matching was applied incrementally to align misspelt placenames with gazetteer entries, after which the same disambiguation steps were applied.
Label ambiguous cases
Placenames that could not be confidently resolved were flagged as ambiguous and excluded from automated spatial analysis unless manually reviewed.
Selected outputs associated with Historic Fires Near Me, grouped by type for ease of scanning.
Public scholarship and media
- Dwyer, Graham and Fiannuala Morgan. “What 174 years of bushfire records teach us about emergency management.”Pursuit, June 2025.
- Episode 2: Bushfires: To Be Continued.A Lost Literature Podcast, 2023.
- Reading Ecological Decline in Nineteenth-Century Bushfire Serials and Reporting.Overland, 28 June 2023.
- Mythologised, Memorialised Then Forgotten: A History of Australia’s Bushfire Reporting.The Conversation, 18 January 2022.
Monographs
- Black Thursday and Other Lost Australian Bushfire Stories.Canberra: Orbiter Publishing, 2021.
Peer-reviewed scholarship
- Dwyer, G., Marjoribanks, T., Morgan, F., and Farmer, J. (2025). “Bushfire public inquiries: From recommendations to hybrid emergency management arrangements.”Australian Journal of Public Administration, 1:40.
Funding and awards
- Chief Investigator and Inaugural Recipient of the Melbourne Public Humanities Initiative for Historical Fire Records as Community Data: Digitisation, Co-Design, and Climate Research, Faculty of Arts, The University of Melbourne.
- Joint recipient of the Climate Research Accelerator (CRX), Melbourne Climate Futures (MCF’s) funding scheme, 2023, in collaboration with the FLARE research group, University of Melbourne.
- Awarded Graduate Digital Research Fellowship, Queensland University of Technology, 2021.
- Research Partner, 2021 ARDC Grant, Time Layered Cultural Map of Australia: Dark Places.
Presentations and invited talks
- Digital Spatial Memories Panel. Panel presented by Fiannuala Morgan, Francesca Sidoti, Heather Ford, Claire Loughnan and Michael Falk at Digital Humanities Australasia 2025, Australian National University, Canberra, Roland Wilson Building, Seminar Room 3, 3 to 5 December 2025.
- Simple Algorithms, Big Discoveries: Using NLP to Unlock Digital Cultural Collections. Creative Technologist Lecture with SLV Lab, State Library of Victoria, Public Lecture, 9 October 2025.
- Mapping Histories and Writers: The Role of NLP in Enhancing Archival Work. NSW Branch of the Australian Society of Archivists, 3 April 2024.
- Historical Fires Near Me: (Re)Constructing Colonial Ecological Records.Lightning talk, Making Meaning 2024: Collections as Data, State Library of Queensland, 8 March 2024.
- Trove Research Webinar. Co-presented with Kate Ross, National Library of Australia, 31 August 2023.
- Latent Geographic Associations: Theorising Mapping in Journalistic and Fictional Accounts of Nineteenth-Century Bushfires. Conversations in HADES seminar series, The University of Melbourne, 19 May 2022.
- Space, Data, Place: Digital Tools for Australia’s Deep Past. Centre for Environmental History, Australian National University, 23 August 2022.
- Rethinking Settler (Un)Belonging: Reading Ecological Decline in Colonial Australian Literature. Coming to Terms, 30 Years On: The Mabo Legacy in Australian Writing, University of Tasmania, 4 July 2022.
- Bushfire Literature and Reporting: Mythology, Memorialisation and Omission. ResBaz Research Bazaar, The University of Queensland, 26 November 2021.
- Geo-locating Real and Fictional Place: Analysis of Bushfires in Australian Literature and Newspaper Articles. National School of Arts Winter Seminar Series, Teaching and Researching in the Digital Humanities, 24 June 2021.
Acknowledgements
This project draws on a range of digital infrastructures, data sources, and software tools. The resources listed below were essential to the development of the dataset and the methodological approach.
- Time Layered Cultural Map of Australia
- Trove, National Library of Australia
- GLAM Workbench Trove Harvester
- Stanford Named Entity Recognition (CRF-NER)
All source code used in this project is publicly available via the Historical Fires Near Me GitHub repository.