UrbanNet 2016: Smart Cities, Complexity and Urban Networks  (U2SC) Session 1

Abstract: Urban mobility can be modeled as a multilayer network, where each layer represents different modalities such as pedestrians, cyclists, buses, trams, metros, taxis, transportation network companies, private, logistics, freight, and emergency vehicles. The technology required to integrate different transportation modes already exists. However, there are several challenges beyond technology for achieving truly integrated transportation networks. Even within a single transportation mode, different actors may fail to coordinate for different reasons: economic, social, or political. These reasons extrapolate and limit the coordination between layers. After reviewing the limitations, I suggest possible scenarios which could overcome them. As an incentive for doing so, I outline a way of quantifying the benefits on integrated transportation networks. Close

Abstract: The problem of identifying natural, socio-economically defined neighborhoods arises in applied contexts including Census reporting, measurement of segregation, and dimension reduction in urban computing. This problem is also of interest for urban theory, since the difficulty of identifying neighborhoods may be viewed as a measure of socio-spatial complexity. We develop a rigorous, information-theoretic approach to this topic, using open data on race in American cities as a case study. First, we formulate the mean local information J(X, Y ) as a localization of the mutual information between spatial and socio-economic variables. The measure J(X, Y ) is closely related to the Fisher information of the underlying joint distribution, and is therefore a measure of the intrinsic spatial complexity of an urban phenomenon. Unlike standard global information measures, the mean local information clearly distinguishes between cities like Detroit?which is dominated by a few huge, monoracial superclusters?and cities like Philadelphia?which is an intricate patchwork of small, racially-distinct neighborhoods. Second, we provide a practical algorithm for identifying natural neighborhoods through greedy information maximization, and relate this algorithm?s behavior to the mean local information. Questions raised by this work include the social, economic, and policy determinants of socio-spatial complexity in cities, and the potential use of spatial information measures in quantifying temporal changes in socio-economic structure, on time scales ranging from days to decades. Close

Abstract: Everyday massive amounts of geolocated data are passively generated by individuals devices like smart phones, credit cards, GPSs, RFIDs or remote sensing devices. This deluge of information growing at an astonishing rate represents an unprecedented opportunity for researchers, to solve challenging problems and unveil fundamental insights on our society. Many disciplines are concerned, ranging from mathematics, physics and computer science for the analysis and management of research data, to applications in astronomy, medicine, geography and social science. Although data passively generated by the use of Information and Communication Technologies (ICT) have the advantage of the large size of the samples (millions of observations) with a high spatio-temporal resolution, they raise many new challenging issues, related to their storage, transport, management and processing. In particular, they may suffer from hidden biases, and therefore observing the world through the lens of big datasets can lead to possible distortions which may lead to erroneous conclusions. It is thus crucial to develop statistical tools and methods to assess the uncertainty in ICT data, notably by comparing the results obtained with different data sources. In the following we present two examples of such uncertainty analysis on results obtained with mobile phone data recorded in Senegal in 2013. We concentrate on two information-retrieval tasks: first, we evaluate the uncertainty when inferring land use from the rhythms of human activity, and second, we study the uncertainty when identifying individuals? most frequented locations. We conclude by mentioning possible future steps to clearly assess the relevance of various ICT data sources for studying different phenomena. Close

Abstract: As social beings, we create structures that ensure that interactions between and within communities take place. These structures have been changing over time, but they have left footprints that can be identified as patterns that translate into hierarchies of regions and social divisions that are the outcome of a historical process. In this work we use different clustering methodologies and percolation theory to uncover the different communities that can be identified as the outcome of this process, and as the emergence of new kinds of interactions Close

Abstract: The distribution of population in urban settlements has been extensively characterised in the literature by using Zipf's law but there exist well known deviations from this power-law distribution in the upper and lower tails of the spectrum. In this work we use the definition of cities proposed in a previous paper using a percolation approach on the road network, and show that the same type of power-law distribution exists as well for the number of intersections of a city and that this distribution is fitted with a greater precision presenting less deviations from the theoretical power-law. We also show that it is possible to derive the population of a settlement from this number of intersections and that the existing variability within this approximation can be partially explained by quantifying the position in the regional hierarchy of each settlement. This result gives us another insight into why some cities are over/under-populated with respect to its expected position in the Zipf's law and at the same time renders possible to extract an approximated population of each settlement having as sole source of data the road network of the system. Furthermore, we show that by combining both distributions we can find a clear cut between large and small settlements that can be used to quantitatively define a threshold between urban/rural settlements. Close

Abstract: The integration of Information and communications technology (ICT) data sources to generate activity-travel information opens interesting opportunities for feeding agent-based transportation models, whose practical implementation is often hindered by the lack of sufficient data. In this talk we present a module developed to generate activity-travel diaries needed to feed the MATSim simulation framework. Activity diaries are generated by merging data from mobile phone records and census data. We discuss of the process followed for the creation of synthetic agents, from the extraction of mobility patterns to the expansion of the sample data to the total population: the dataset provided by the mobile network operator provides the age and gender of the users, while residence location and daily trips are obtained by analysing the mobile phone records; the information obtained from mobile phone records is then expanded to cover the whole population by using census data at census tract level. Finally, some lines of future research for the improvement of the current methodology are discussed. The resulting synthetic population is validated with the EMEF survey results for Barcelona, The comparison shows a good fit in number of trips by gender and age and major discrepancies in trip pourpose assignation. The resulting activity diaries are used to Feed and Calibrate the MATSim simulation model. Close

Abstract: Residential location choices are influenced by a series of factors whose influence varies across space. We aim to improve one of Land Use and Transport Integrated (LUTI) models MARS model introducing the different impact of those factors on residential choice by computing local coefficients. In particular, this research explores the methodology of integrating the public choice model into the MARS (Metropolitan Activity Relocation Simulator) model using a general accessibility indicator, thus creating a new approach to estimate the coefficients of each public service section with the use of Geographically Weighted Regression (GWR). The MARS model includes a transport model which simulates the travel behaviour of the population related to their housing and workplace location, a housing development model, a household location choice model, a workplace development model, a workplace location choice model. The method to embed the public services location model into MARS is to re-develop the accessibility indicator which is the key connection between the transport submodel and housing, workplace loction submodel, not only considering the capability to reach workplaces but also involving the ability to access certain public services. Accessibility plays a major role to influence where people to live and work (Wang, et al., 2015). It is one of the outputs of the transport sub-model in year n as well as the input to the land use sub-models in the year n+1. The new accessibility indicator is calculated by integrating a series of travel motives, and then weighting each of them using the results from a GWR. In this manner, the accessibility is evaluated as the key location factor to express the level of public services in land use, which in turn attracts travel demand. We applied GWR (Fotheringham et al., 2002) to generate local models in which specific coefficients are computed for each observation (i.e. spatial unit) and for each significant variable. The calculation of these local coefficients is based on the values of the corresponding variable in nearby locations, giving more weight to close locations, thus establishing an inverse distance relationship. An origindestination matrix through the road network was computed with ArcGIS Network Analyst extension. This matrix was used as input in GWmodel R package in order to consider network distance in local correlations and GWR. The model update and extension of MARS are all based on the Region of Madrid, Spain. The external scenario update is based on the zoning of MARS, which aggregates the 199 municipalities of Madrid Region into 90 modelling zones. The zoning was carried out following homogeneity parameters of socioeconomic characteristics and mobility, plus correspondence with transport zones and regional rings. Data collected at different levels were aggregated to the MARS zoning. The GWR was based on the most disaggregated spatial units that were available (i.e. census tracts). Data sources include INE, Nomecalles, DUAE and TomTom. We computed local correlation statistics between population data and the highest correlated variables for each topic. Based on previous work, we applied a 5km bandwidth to incorporate the values of nearby locations using a Gaussian function. We then performed a model selection process based on the spatial relationships of the observations within 5 km, after which we estimated the best bandwidth in terms of distance (fixed) and number of neighbors (adaptive) both for distance along the road network and for private motor vehicle travel time. In all cases, a 5-neighbor adaptive bandwidth provided the best fitted models, with network distance performing better than travel time. The best-fitted model is the one considering the number of workplaces, education centres and retail. Figure 1 shows the standardized local coefficients once aggregated at the MARS zoning. The intensity and sign of the relationship between each factor and population location have a great variety across space, especially in the case of education centres. Population figures predicted with our model are consistent with real figures, with discrepancies below +- 0.5 standard deviations in the most populated areas. Close

Abstract: The last few years have seen the emergence of the so-called sharing economy (also known as collaborative consumption) that has been driven by the development of Internet platforms that facilitate peer-to-peer relations. One of the fields in which collaborative consumption has burst onto the scene with greater intensity is that of tourism, both in the travel sector (car-sharing) and that of accommodation. Airbnb is the most successful P2P platform in the field of accommodation, offering more than 2,000,000 listings in 190 countries. The potential impacts of Airbnb on local economies are complex and difficult to measure. The results of the study by Fang et al. (2006) suggest that the entry of sharing economy benefits the entire tourism industry by generating new job positions as more tourists would come due to the lower accommodation cost. From the perspective of the spatial distribution of the Airbnb impacts within the cities, it has been argued that Airbnb listings are more scattered than hotels, so Airbnb guests may be especially likely to disperse their spending in neighbourhoods that do not typically receive many tourists (see Guttentag, 2014). Nevertheless, this possible dispersion may be compatible with a particular concentration of listings in the central areas of the cities, including areas not covered by hotels. This fact could aggravate the problems of crowding and tourism gentrification that some of these areas have to support in certain heritage cities (Russo, 2002; Neuts and Nijkamp, 2012). This article analyses the spatial patterns of Airbnb in Barcelona and compares them with hotels and sightseeing spots. Close