Marta C. GONZALEZ is associate professor in City planning at Lawrence Berkeley National Laboratory. She combines human mobility, social science and computational social science in mobility studies as well as in city planning. In particular, her research exploits various sources of Location Based Service (LBS) data, including Mobile phone data (other sources include Satellite imagery, smart devices, various sensors). In her talk Professor GONZALEZ pointed some key challenges while processing and validating such versatile data sources : defining and labelling indivudal stays, infer attributes and characteristics of travelers (through the example of credit cards records), model calibration (for instance energy demand modelling through housing bills), etc.

She presented research results she obtained while collaborating with physicists. She underlined the complementarity of two approaches that are sometimes presented as opposite : trying to observe universalities in patterns, and trying to identify local specificities,. For instance, preferential returns in individual human mobility are somehow universal in the datasets Marta C. GONZALEZ studied (for any individual the most important location represents somehow half of her/his trip destinations), yet however the next destination is still very heterogeneous among people.

New research fields pointed at include time geography modelling framework, from sparse users to synthetic trajectories, urban behavior and social ties, peak load optimization, etc.

Catherine Morency is an engineer and she is professor of transportation and planning at Polytechnique Montreal, holder of “Mobilité” research chair as well as Canada research chair on personal mobility. She is involved in projects that gather a large variety of partners: provincial government, planning agencies, city authorities, transit operators, taxi / carsharing industry, nonprofit Organizations.

Her research in transport demand modelling (modelling travel behavior, factors, impacts, methods, data, models, formulation and estimation of scenarios) embraces a strong diversity of new data sources useful to transport modeling: GPS data; smartcard data; transaction data; mobile phone data ; social media data. The goal is to achieve methodological contributions: better tools to support decision making in collaboration with public / private stakeholders

She developed two examples : in the first one, by using taxi GPS data, she discussed the technical difficulties encountered when one wants to attach GPS points to road segments, an uneasy task when road crossing are involved, considering the precision of GPS technology. Direct applications of this problem include developing a dashboard for monthly strategic assessment: what are the strategies of successful taxi companies (number of hours worked, etc.), what policies shall be implemented to regulate this industry. Such data can provide a unique vision that would benefit to both taxi companies, public authorities and improve their work conditions.
In a second example, using bikesharing data in Montreal (6 years of data), Catherine MORENCY showed how a wide variety of problems could be solved using appropriate algorithms, including the problem of full stations (as an order of magnitude, 15% of stations were full before the implementation of measures, and decreased down to 3% after a few years of dynamic reallocation of bikes in stations).

She also listed some usual problems encountered when working with data. Their volatility (data / metadata tend to disappear after several months of collaborative projects), ageism (using very old data because they are still the best available ones), public / private injustice, especially about the access to data, among others. Travel surveys are still being used: learning how to efficiently combine datasets seems an important issue in the transportation research agenda.

Overall her position is rather to make research usable in practice. She gives keys for long lasting collaboration with stakeholders: contribute to discussion, be sensitive to operational problems, and add value to available datasets. However she insisted on the need to keep in mind traditional issues in decision making tools: representativeness is still an issue, would it be for the associated equity issue for planning.

Remko UIJLENHOET is Professor at Wageningen University in Geosciences and environmental issues. After a PhD in Hydrometeorology 1999, he became in 2007 full professor and chair of the Hydrology and quantitative water management group.

His contribution to this summer school is about bridging gaps in science, with a focus on what he called “ahah” moments : when the scholar actually realizes that two fields can be linked with interesting applications. In the case he presented, such an application involved the use of data from the mobile phone industry to monitor rainfall at the national level, a rather unexpected usage of cell phone technology.

In Remko UIJLENHOET’s research laboratory scholars are using observations and models to understand hydrological processes, and especially rainfall variability over a range of scales (ex. River catchment for flash flood early warning systems). They also conduct research about water management in urban design. In his presentation Professor UIJLENHOET showed how measures from telecom antennas can be derived into dynamic rain maps. This requires some notions about hydrology. The observation of optical extinction caused by rainfall can be put in equations, optical extinction being a function of the raindrop size distribution. Applying it to signal emitted by radio antennas, empirical observations were gathered to confirm the theory. This was the first “ah ah” moment. Interestingly, this was already found in separate research conducted by telecommunication engineers in the 1960’s.

The second “ah ah” moment was very close: microwave links are omnipresent, therefore they can be used to monitor rainfalls. More specifically, it is possible to map rainfall through telecom link attenuation, using variograms and interpolation methods. The methodology and empirical experiment to demonstrate this use case has been detailed ; a crucial aspect is to sign agreements with telecom companies to get access to receiver signal data (note that received signal level data are stored operationally by telecom companies and already commercialized in more than 10 countries). A R package called RAINLINK has been developed around this project.

Marc BARTHELEMY is a physicist, he holds a research director position at CEA (Commissariat à l’Energie Atomique et aux énergies renouvelables), where he is a member of the IphT (Institut de physique Théorique) laboratory. As a second affiliation he also holds a position at the EHESS in the CAMS laboratory (Centre d’Analyse et de Mathématique Sociale). Marc BARTHELEMY delivered a comprehensive presentation of state-of-the-art knowledge about spatial networks. His talk covered both empirical results and theoretical models, and was pedagogical enough to be understable and expressive for the broad audience of participants which included mainly geographers, but also urbanists and physicists. He developed a morphodynamical approach of spatial networks with a particular emphasis on infrastructure networks such as streets, roads and transportation networks (subway, train). He presented some mathematical tools needed to characterize these structures and how they evolve in time. He discusses some important empirical results and stylized facts, and presented some of the most important models of spatial networks. He ended his talk with considerations on scaling in the case of transportation networks.

Taylor SHELTON is professor of geography at Mississippi State University and Matthew ZOOK is professor of geography at the University of Kentucky. They began their lesson by insisting on the fact that, while the use of “big data” to better understand urban questions is an exciting field with challenging methodological and theoretical problems, it is also troubling when such data are particularly those derived from social media – are applied uncritically to urban governance via the branding of “smart cities”. Drawing upon examples they illustrated that one must be ever mindful that metrics don’t simply measure; in the process of deciding what is important to measure these urban data are simultaneously defining what cities are. To illustrate their critical examination of these big data-driven urban studies, they presented a study of their own with based on geolocated tweets collected during several months in Louisville, Kentucky, a city than they know well since they conducted some fieldwork there. The purpose of this study was to show how Twitter data can document the spatial imaginaries and processes of segregation and mobility at play in the notion of the ‘9th Street Divide’, a large road in the city of Louisville, Kentucky. By analyzing the everyday activity spaces of different groups of Louisvillians through their geotagged Twitter data, they argue for an understanding of these neighborhoods as fluid, porous and actively produced, rather than as rigid, static or fixed. Through this example they conveyed a methodological framework proposal useful for the analysis of social media data that is more attentive towards the multiplicity of socio-spatial relations embodied in such data.