Histogram acc logacc

Cécile Nyffeler

and 2 more

Lugrin René, Nyffeler Cécile, Vieira Ruas MarcoEcole Polytechnique Fédérale de Lausanne IntroductionIn urban areas road networks have grown into a big factor of concern for public health as they become more complex to adapt to the efficiency required by their daily users. Yet within cities, not all populations are impacted in the same manner by the dangers and risk factors which ensue from transportation.As of now, many studies have been published, which have examined the correlation between a variety of road related nuisances and various levels of  social vulnerability. For example, in Montreal, \citet{Carrier_2016} highlighted the fact that lower income populations and visible minorities lived in areas more subject to higher road noise levels. In another study, \citet{lam_exposure_2012} pointed to the same sort of environmental inequity (see note 1: end of the paper) in Hong-Kong. Indeed, people of lower socioeconomic status were also found to be exposed to higher noise levels. Another factor to take into account when studying road hazards is accidents, as they represent one of the main causes of death worldwide \cite{Murray_1997}. A study by \citet{Aguero_Valverde_2006} demonstrated that regions inhabited by an indigent community or predominantly young (under 15) or older adult (over 65) populations were more often affected by car crashes. Age proves as a relevant variable when evaluating population vulnerability in correspondence to road risks, since children tend to underestimate the likelihood of pedestrian accidents \citep{Joshi_2018}. Thus, they usually lack the ability of recognizing risky situations and anticipating dangers as do the elderly, because the capacity to multitask decreases as people grow older \cite{verhaeghen_aging_2002}.  One of the elements associated with road networks often not considered is gas stations. These represent a source of air and ground pollution, which can also have health impacts. \citet{Morales_Terr_s_2010} investigated the influence gas stations had on air in their surroundings due to VOCs (see note 2: end of paper) emanating from them.  The paper concluded stations had an influence on close surroundings with high concentrations of benzene and n-hexane being measured. This influence was shown to depend highly on the characteristic of the area in the direct vicinity of the stations. When gas stations were closely surrounded by buildings the dispersion of VOCs was hindered. High concentrations of pollutants were therefore measured in these areas. In those regions of high pollution, youger and elder populations have also show a higher tendency to be affected healthwise \cite{Fairbairn1958}.The aim of the present paper is to regroup the above mentioned approaches in order to create a strong basis of analysis for the municipality of Vernier (Geneva, Switzerland). Vernier is highly diversified and although it makes for a small pool of experience, it might be representative of phenomena that happen at a larger scale. To establish the analysis, two indexes were created, which took into account the more relevant variables to assess both vulnerability and road hazards. The socioeconomic vulnerability index was defined in terms of income, age, unemployment rate, housing assistance and healthcare allowance. The danger index was created considering traffic noise and distance towards gas stations and accidents. The initial statement is that danger related to road hazards tends to be greater in regions inhabited by deprived populations. DataThe data used to perform the study came from various sources. The main tool of analysis, the hectometric raster grid, was built upon point coordinates. Those were contained in the demographic data of Switzerland collected from the Swiss Federal Office of Statistics (OFS). Also from the OFS, came the tables containing the age distributions within the municipality and the boundaries of Vernier in the form of a polygon shapefile.To establish the vulnerability, housing assistance and healthcare allowance data in the municipality of Vernier were used. Both these elements presented themselves as tables and were collected from the OFS as well. As for the other components of the index: income and unemployment data grids were found  in the the Inequality Analysis report, provided by the Center of Territorial Inequalities Analysis in Geneva \citep{cati-ge2014}. Then, data from the open data collection from the Geneva Territory Information System (SITG) was also used.  A polygon shapefile contained the GIREC neighborhoods of the municipality, whereas different point shapefiles displayed the addresses within Vernier and enabled to locate the sites of accidents and gas stations.Noise data in the form of a raster file was used as well. It was obtained from the Swiss Noise Database (sonBase, 2010). It contained values of noise predicted by models and calculations performed on noise measurements from traffic, urban industries and terrain configurations \cite{ingold2009}. The sound intensities were evaluated at night, since noise has a direct impact on sleep. Base and satellite maps of Vernier and its surroundings were obtained from Google Street Maps.  Finally, all vector and raster layers used were projected according to the Swiss coordinates system: EPSG21781.Methods Treatment of the data The investigations concerning the correlation between road exposure and vulnerability were done using the QGIS and GeoDa software, as indicated by the  "QGIS User Guide" \cite{sherman_quantum_2004}  and the "GeoDa User Guide" \cite{anselin_geoda_2003}  respectively. The variables adopted to define vulnerability were acquired through different means. The address point data vector file was imported in QGIS and intersected with the municipality borders in order to keep the information relevant to Vernier only. This address data file containing both an identifier for each address ( IDPADR ) and one for each sub-sector of Vernier (ID_GIREC) enabled to combine different files containing valuable information, such as allowances, revenue and unemployment rates. A vulnerability grid, containing initially only the identifiers of the cells (100 m x 100 m) of the inhabited regions of Vernier, was generated. The excel data sheets containing the addresses of the households receiving aids (housing assistance and healthcare allowance) were then imported as tables in QGIS. Using the common identifier IDPADR between the allowance tables and the address vector file, a merging action was performed in order to have the geographical locations of the people needing financial help. A count of the number of people receiving healthcare allowance per cell of the vulnerability grid was then performed. The sum of the people receiving housing assistance was calculated in a similar way. In order to be more rigorous, these sums were converted to percentages by dividing the absolute numbers by the number of inhabitants per cell. The revenue and unemployment rate data, added as tables in QGIS, were then joined to the vulnerability layer using the common identifier ID_GIREC. Using once again the joint property of QGIS, the age distribution of the citizens per hectometric cell was added to the vulnerability grid, with the id of the cell this time as common identifier. The vector file containing the accident points was then imported in QGIS. A distance matrix analysis was thereupon performed, seeking for the minimal distance between the centroids of the vulnerability grid and the accident points. The same procedure was applied to the vector file containing the locations of the gas stations within and around Vernier. Finally, a grid containing the average night noise per inhabited cell was loaded on QGIS. The noise data, given in decibels, was previously logarithmically averaged over each cell of the grid, meaning values were primarily elevated to the power of ten, then the zonal statistics tool could be used to compute the averages before converting those averages back into decibels using a logarithm. The distance matrix tables and the noise grid were eventually joined to the vulnerability grid, using the cells' id as their shared identifier.  These different operations allowed to get one final file containing both the data of Vernier's inhabitants vulnerability and their exposure to road related hazards, such as accidents, pollution and noise.     IndexesOnce the table containing all the necessary data was computed, the actual purpose of this paper, relating vulnerability to road associated threats, could then finally be pursued. The first step consisted in establishing how vulnerability depended on different factors. As indicated by previous studies and stated in the introduction, the economical situation of a person, as well as its employment status, its reliance on governmental aid and its age all influence exposure to traffic hazards. These factors thus all participate in defining one's vulnerability to road incidents. Based on that conclusion, and the data available, the vulnerability index was defined in the following manner: Vulnerability = f(Revenue, Unemployment, Age, State Financial Support)In order to get a final index describing the precariousness of the inhabitants of the municipality of Vernier, the different selected indicators had to be formatted. The decision was therefore taken to use a linear regression approach to create a sub-index for each of the factors inducing vulnerability. This means that the values of revenue, unemployment rate, housing aid rate, healthcare allowance rate and age distribution were all scaled as to have a value in between 1 and 3. A person being more vulnerable for a category would consequently have a higher score in that category. The linear regression equations used in GeodDa to format the revenue, unemployment rates and allowances indicators are presented in the following Table 1.  Table 1 - Linear regression equations allowing to get the indicators' sub-indexes of vulnerability