MAPHIS: Mapping History--What Historical Maps Can Tell Us About Urban Development

Lead Research Organisation: University of Bristol
Department Name: Economics

Abstract

Little is known about the patterns of city development during the structural transformation of economies. This project will systematically process high-resolution and manuscript historical maps to make a dormant body of information about our cities' and regions' past accessible.

The proposed research will advance our understanding of long-run urban growth through the development of three innovative methodologies, which will overcome practical limitations of historical data sources: 1) A technique to extract land use patterns from historical colour maps applied to France (1750-1950); 2) A recognition algorithm to detect, tag and geo-locate points of interest in historical high-quality maps of the 70 largest urban centre in England and Wales; 3) An algorithm to geo-locate address information from Micro-censuses and trade registers.

We have identified four main research questions that will be developed in the following separate research projects. In Project 1, the main question is: what are the long-term empirical patterns of urban development, most notably the persistence of the spatial organisation of economic activity and the role of building infrastructure in shaping such persistence? In Project 2, the main question is: How do environmental disamenities and their unequal distribution within cities affect the spatial organisation of consumption amenities and production? In Project 3, the main question is: Do cities grow towards their bad parts, their neighbourhoods with the lowest environmental amenities? In Project 4, the main question is: How does vertical growth and advances in building technologies affect the spatial organisation of cities?

To address these research questions, we will organise our workflow in six inter-connected work packages (WP):

WP1--Classification of land use in France (1750-2015): The objective of WP1 will be to recover land use information at a fine scale from digitised maps using state-of-the-art machine learning techniques;

WP2--Digitisation of micro-features embedded in Ordnance Survey (OS) city maps of England and Wales (1870-1960);

WP3--Geo-localization of residents and production units in England and Wales (1851-1911);

WP4--Dynamic model of city growth with persistent building stock: WP4 builds a general equilibrium model of spatial economic activity that embeds the durability of housing and infrastructure and exploits the three hundred years of population settlement data produced in WP1;

WP5--Pollution and the long-run development of cities: WP5 builds on WP2,3 and proposes to study the joint dynamics of residential sorting and the location of production within cities to understand how a major environmental disamenity-industrial pollution-affects the spatial organisation of cities in the longer-run;

WP6--Horizontal and vertical urban growth in Montreal and Toronto: WP6 will bridge between the previous working packages WP1, WP2, WP4 and WP5, and study--empirically and theoretically--horizontal and vertical urban growth.

The project will be jointly led by three teams. The French team will be composed of Gobillon (PI), Combes (CoI) and Duranton (TM) who have contributed to the development of major theoretical approaches in urban economics. The Canadian team will be led by Heblich (PI), who is a lead researcher in urban economics/economic history, and Fortin (Co-I), a lead in GIS analysis. The UK team will be led by Zylberberg (PI), who is an economist specialist in data extraction form historical sources and remote sensing. Shaw-Taylor and Schürer, advisory board, will help design the analysis of the population micro-censuses between 1851 and 1911 (WP3). The collaboration partner, Redding (TM), involved in the design of WP3 and the implementation of WP6, is one of the World lead researchers in urban economics.

Outputs will include articles in top economic journals, and detailed algorithms to extract relevant spatial information from manuscript maps.

Publications

10 25 50
publication icon
Combes P (2021) Urban economics in a historical perspective: Recovering data with machine learning in Regional Science and Urban Economics

publication icon
Gobillon L (2022) Introduction to the Special Issue: "Urban economics and history" in Regional Science and Urban Economics

publication icon
Heblich S (2021) East-Side Story: Historical Pollution and Persistent Neighborhood Sorting in Journal of Political Economy

 
Description Floods, risk diversification and livelihoods in the Central Highlands of Vietnam
Amount £41,146 (GBP)
Funding ID TGC\200149 
Organisation The British Academy 
Sector Academic/University
Country United Kingdom
Start 01/2021 
End 12/2023
 
Title City recognition 
Description We developed a simple machine learning algorithm to recognise city boundaries on very early historical maps from 1790--1820 across England and Wales. This method allowed us to create a dataset of early cities at the onset of the nineteenth century. 
Type Of Material Data analysis technique 
Year Produced 2022 
Provided To Others? No  
Impact The research will contribute to the nascent research on feature extraction from historical maps (see the research output "Urban economics in a historical perspective: Recovering data with machine learning"). 
 
Title Feature extraction algorithm 
Description The model uses visual recognition to extract features from handwritten, historical maps. More specifically, it proceeds in the following steps: - the image is cleaned and we produce various derivatives; - we designed a contour detection algorithm to isolate contours and recognise shapes ("features", e.g., a workshop, a chimney, a tree), we adapted an off-the-shelf OCR augmented with a neural net to recognise handwritten text; - we designed an iterative algorithm to classify micro features (and macro neighbourhoods) using (i) feature characteristics (shape, surrounding features/labels) and (ii) the classification of their macro neighbourhood. The idea is that a feature can be better classified (e.g., as a chimney) if we can add information about its neighbourhood. A neighbourhood can be better classified (e.g., as an industrial neighbourhood) if we can precisely characterise micro features (e.g., chimneys). 
Type Of Material Computer model/algorithm 
Year Produced 2022 
Provided To Others? No  
Impact This code has allowed to extract features from a subset of historical maps that were never digitised (i.e., they were scanned, but the information was never systematically extracted from these maps). 
 
Description Article in the Economist 
Form Of Engagement Activity A press release, press conference or response to a media enquiry/interview
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Public/other audiences
Results and Impact The Economist covered the main results of the research article "East Side Story", and we collaborated to produce clear visual evidence that would be easier to understand for a general audience.
Year(s) Of Engagement Activity 2021
URL https://www.economist.com/graphic-detail/2021/05/08/the-legacy-of-victorian-era-pollution-still-shap...
 
Description Public workshop 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach Local
Primary Audience Public/other audiences
Results and Impact We organised a workshop aimed at describing the long-run evolution of neighbourhoods with a focus on the city of Bristol. The stakeholders were the general public, and were reached through various channels (e.g., through facebook pages of community groups in poor neighbourhoods).
Year(s) Of Engagement Activity 2021
 
Description Short video and blog entry 
Form Of Engagement Activity Engagement focused website, blog or social media channel
Part Of Official Scheme? No
Geographic Reach National
Primary Audience Undergraduate students
Results and Impact We recorded a video and described the project in a blog entry.
Year(s) Of Engagement Activity 2021
URL https://www.bristol.ac.uk/economics/research/major-grants-/mapping-history/