Mapping History -- What Historical Maps Can Tell Us About Urban Development? – MAPHIS (ES\T010371\1)

The proposed research develops novel, interdisciplinary methods that facilitate the extraction of information from historical maps to study the evolution of land use patterns and urban growth. Specifically, we will (i) employ insights from imagery techniques and photogrammetry to extract colour-coded information on land-use and (ii) use recognition algorithms to detect points of interest, street names and labels on maps that can be matched to data sources like population censuses. Together, this will help draw a uniquely detailed picture of changes in the socio-economic structure and organisation of cities.

A few recent studies show that high-resolution maps from archives provide numerous pieces of information. The microfeatures embedded in historical maps can be systematically exploited with the help of (a) recent advancements in machine learning, (b) novel methods to delineate and classify urban areas (de Bellefon et al., 2019), and (c) the recent development of analytical tools in economic geography (see Redding and Rossi-Hansberg, 2017, for a summary). To our knowledge, this proposal marks the first attempt (i) to digitise historical maps with such time and spatial coverage, (ii) to demonstrate the scope of recognition methods on historical maps, and (iii) to draw an almost complete picture of the urban structure and socio-economic composition of cities, regions and countries in the context of three distinct historical environments: France, England and Wales, and North America (e.g., Toronto, Montreal) over the past centuries.

Our research will advance current knowledge in social sciences through the development of three innovative methodologies:

I. A machine learning approach to extract land use patterns from historical colour maps (1750-1950), adapted from state-of-the-art imagery techniques;

II. A recognition algorithm to detect, tag and geo-locate points of interest in high-quality scans of urban centres (22,000 Ordnance Survey maps covering the 70 largest industrial areas in England and Wales, 1870-1960, and fire maps, Toronto/Montreal);

III. A location algorithm which geo-locates entries from Micro-censuses (England and Wales, 1851-1911) and trade registers, and combines this source of information with maps.

These newly developed methods will overcome practical limitations in the use of hand-drawn historical maps and help exploit this mostly unexploited source of historical information. This will advance our knowledge on long-run urban development and generate the following output:

IV. High-resolution vectorised maps of France (1750-1950), cities of England & Wales (1870-1960), Toronto and Montreal (1876-1975);

V. A dynamic model brought to the data of the physical development of cities accounting for persistent building stock and changing drivers of location over time from agriculture to manufacturing to services--informed by novel stylised facts covering 250 years of city growth in France;

VI. An analysis of the long-run impact of atmospheric pollution on urban centres and their outskirts in England and Wales (1870-1960);

VII. An empirical and theoretical investigation of the horizontal and vertical growth of Toronto and Montreal; the identification of drivers of the physical development of two cities that have experienced English or French influences.