Project aims.

Project aim is the feasibility study of using the geo-coded data on bicycle accidents from cyclestreets.net and spatial modelling within Bayesian paradigm to study road safety, on the example data from Manchester area. We will first review the recent literature on the methods used in spatial modelling of transportation systems and road accidents especially involving cyclists; second, drawing on the literature review we will develop a statistical model that will explore the risks involved in cycling such as the type of road, availability of bicycle paths, presence and type of junctions, time of travel, day of travel.

Bicycling is an often neglected means of transportation, both for commuting and for recreation and, in recent years, has gained in popularity (Kim et al. 2007). It is also an environmentally friendly, economical and healthy activity, though statistics show that in last few decades cycling safety has been neglected comparing to driving safety (Dozza & Werneke 2014).

Methods.

Various statistical and mathematical methods can be used to analyse risks involved while cycling. In the proposed project we will concentrate on models typical in social statistics such as logistic and Poisson regressions to analyse causes and risk factors involved in cycling and accidents. We will then supplement these data with the available spatial information by borrowing from spatio-temporal modelling techniques (e.g. Diggle 2013), conditional autocorrelation used in modelling spatial interactions (Dong et al. 2016) and GIS modelling (e.g. Maguire et al. 2005).

The study area

Greater Manchester is a metropolitan county in North West England, with a population of 2,798,800. It encompasses one of the largest metropolitan areas in the United Kingdom and comprises ten metropolitan boroughs: Bolton, Bury, Oldham, Rochdale, Stockport, Tameside, Trafford, Wigan, and the cities of Manchester and Salford. Greater Manchester was created on 1 April 1974 as a result of the Local Government Act 1972; and designated a City Region on 1 April 2011. Greater Manchester spans 493 square miles (1,277 km2), which roughly covers the territory of the Greater Manchester Built-up Area, the second most populous urban area in the UK.

Data.

We will utilise real data https://data.gov.uk/dataset/road-accidents-safety-data https://www.cyclestreets.net/. The statistics relate only to personal injury accidents on public roads that are reported to the police, and subsequently recorded, using the STATS19 accident reporting form. The Ordnance Survey data on transportation links (types of roads, junctions) will be used as covariates. Below we showed the first six observations of the collected data.

##   Accident Index                  Timestamp severity Casualties
## 1  201606E024792     1:10pm, 22nd July 2016    fatal    Cyclist
## 2  201506E096003  7:36am, 13th January 2015    fatal    Cyclist
## 3  201506M006977   5:03pm, 2nd January 2015    fatal    Cyclist
## 4  201406A089515     7:45am, 24th July 2014    fatal    Cyclist
## 5  201406E083033 3:25pm, 11th February 2014    fatal    Cyclist
## 6  201306E075692    3:00pm, 3rd August 2013    fatal    Cyclist
##   Number of Casualties
## 1                    1
## 2                    1
## 3                    1
## 4                    1
## 5                    1
## 6                    1
##                                                              url Latitude
## 1 https://www.cyclestreets.net/collisions/reports/201606E024792/ 53.45775
## 2 https://www.cyclestreets.net/collisions/reports/201506E096003/ 53.46718
## 3 https://www.cyclestreets.net/collisions/reports/201506M006977/ 53.38627
## 4 https://www.cyclestreets.net/collisions/reports/201406A089515/ 53.48164
## 5 https://www.cyclestreets.net/collisions/reports/201406E083033/ 53.44317
## 6 https://www.cyclestreets.net/collisions/reports/201306E075692/ 53.47037
##   Longitude
## 1 -2.243951
## 2 -2.229378
## 3 -2.346516
## 4 -2.194447
## 5 -2.218615
## 6 -2.247936

Summarising the data

We start by summarising the different types of accidents across different years. The table below showed the distribution of the different types of accident from 2005 to 2016.

Accidents Per Years - Greater Manchester
Years 2005 - 2010
2005 2006 2007 2008 2009 2010
fatal 3 4 4 2 2 1
serious 57 54 46 53 56 73
slight 439 439 417 466 457 430
Years 2011 - 2016
2011 2012 2013 2014 2015 2016
fatal 4 1 1 2 2 1
serious 68 70 57 65 43 51
slight 465 331 336 363 242 209

Calendar Plot

##  [1] "Accident Index"       "Timestamp"            "severity"            
##  [4] "Casualties"           "Number of Casualties" "Number of Vehicles"  
##  [7] "url"                  "apiUrl"               "Latitude"            
## [10] "Longitude"            "new.years"            "new.weeks"           
## [13] "new.dows"             "new.months"           "new.w.months"        
## [16] "Hour"                 "Dater"                "DateRecode"
## [1] 5314   18

Seasonal Plot

Data Visualization.

Manchester and the surrounding areas are now well served by a number of attractive cycle routes. This includes National Cycle Networks routes 6 & 60 which pass through Manchester and are an ideal way to see some of the cities highlights. We start by summarising the most frequent bikes routes in Manchester.

Manchester Limit Speed

On some roads, different speed limits apply to different classes of vehicle for safety reasons. We are able to change speed limits if we feel that they are inappropriate for some of the roads which they cover.

There is a direct link between vehicle speeds and casualties. On roads with lower vehicle speeds the number of collisions will be less and the severity of the collision will be reduced.

Where are crashes most likely to happen?

The extent to which bicycle crashes are concentrated in urban areas. Preliminary results shows that crashes are more likely to happen around the city center. This pattern sets the scene for the geographically aggregated statistics presented below.

setwd("~/Dropbox/Data Science Small Project 2017/Data")

load("Collisions_data_clean.RData")

data <- collisions_data_clean

#######################################################
# Map 
#######################################################

leaflet(data) %>% addTiles() %>%  addCircles(~Longitude, ~Latitude, popup=data$severity, weight = 3, radius=40, 
                                             color="black", stroke = TRUE, fillOpacity = 0.8) 
leaflet(data) %>% addTiles() %>%
  addMarkers(lat = ~Latitude, lng = ~Longitude ,
             clusterOptions = markerClusterOptions() )

The extent to which bicycle crashes are concentrated in urban areas. We investigate the performance using a bivariate Kernel Density Estimate (KDE). This is presented in the picture below, which shows the extent of the clustering of incidents involving cyclists. This pattern likely reflects the high rate of cycling and around Manchester city centre.

Below, we showed the clustering of incidents involving cyclists. Of course, this pattern likely reflects the high rate of cycling and around Manchester city centre.

setwd("~/Dropbox/Data Science Small Project 2017/Data")

load("Collisions_data_clean.RData")

Data_Collision <- collisions_data_clean

#table(Data_Collision$new.years)

Data_Collision_list_Df = split(Data_Collision,Data_Collision$new.years)

leaf <- leaflet() %>%
  addProviderTiles(providers$CartoDB.Positron)

purrr::walk(
  names(Data_Collision_list_Df),
  function(new.years) {
    leaf <<- leaf %>%
      addHeatmap(
        data = Data_Collision_list_Df[[new.years]],
        layerId = new.years, group = new.years,
        lng=~Longitude, lat=~Latitude,
        blur = 20, max = 0.05, radius = 15)
  })

leaf %>% addLayersControl(
    baseGroups = names(Data_Collision_list_Df),
    options = layersControlOptions(collapsed = FALSE)
  )

The overall spatial distribution of incidents involving cyclists follows the road network, as do incidents with no cyclist casualties. The relative density of cyclist crashes is clearly higher in urban centres, however, and is clearly absent

Further Research

There is much important research that needs to be done to help make the transport systems in many cities safer. We hope that the data provided by stats19 data can help safety researchers develop new methods to better understand the reasons why people are needlessly hurt and killed on the roads.

References

Kim, J. K., Kim, S., Ulfarsson, G. F., & Porrello, L. A. (2007). Bicyclist injury severities in bicycle–motor vehicle accidents. Accident Analysis & Prevention, 39(2), 238-251.

Dozza, M., & Werneke, J. (2014). Introducing naturalistic cycling data: What factors influence bicyclists’ safety in the real world?. Transportation research part F: traffic psychology and behaviour, 24, 83-91.

Diggle, P. J. (2013). Statistical analysis of spatial and spatio-temporal point patterns. Chapman and Hall/CRC.

Dong, G., Ma, J., Harris, R., & Pryce, G. (2016). Spatial random slope multilevel modeling using multivariate conditional autoregressive models: A case study of subjective travel satisfaction in Beijing. Annals of the American Association of Geographers, 106(1), 19-35.

Longley, P. A., Goodchild, M. F., Maguire, D. J., & Rhind, D. W. (2005). Geographic information systems and science. John Wiley & Sons.