Page 113 - NGTU_paper_withoutVideo

Modern Geomatics Technologies and Applications

patterns as well as visualizing the variations over space and time are also the matter of issue. Digital images obtained from
smartphone-based dynamic maps include chronological histories of traffic congestion and facilitate different spatio-temporal
analysis.
In the present study, Tehran as the capital of Iran with more than 9 million population as the 25 most populous and the
th
27 largest city in the world with more than 2.5 million vehicles traveling through its network will be studied. In recent years,
th
new traffic solutions have been proposed by urban planners in order to reduce traffic congestion and its negative effects. Such
as Congestion Charging Zone (CCZ) Air Pollution Control Zone (APCZ) in which, movement of vehicles with no permission is
prohibited or restricted on weekdays during certain hours (except for police, emergency vehicles, public transport, and licensed
vehicles). This study aims to employ traffic images collected by smartphone-based applications and conduct spatio-temporal
analyses to evaluate the congestion pattern over a metropolitan area. Primarily, the spatial autocorrelations of traffic congestion,
using Moran's I indicator will be investigated. In the next step, Havmoler diagrams will be employed for the first time in
congestion analyses using voluminous data to explore the start time and duration of congestion within the time of study over the
study area on weekdays and weekends

2. Data Preparation and Congestion Index
Google Maps services allow to display current traffic conditions as well as chronological traffic state. Such applications
make traffic history data available to users and developers for different days and hours of the week, and update traffic conditions
over 15- minute intervals. Traffic maps are coded in a variety of colours, including green, orange, red and dark-red pixels on the
grid representing smooth, acceptable, slightly heavy, and heavy traffic conditions, respectively. Red and dark-red pixels are
important because they reflect poor traffic conditions (low traffic speed and high travel time).Considering the dynamic nature of
traffic condition, it is essential to identify the areas that continuously experience congestion. In the present study historic traffic
image data within 15-min intervals during weekdays and weekends from 18th April 2018-20th June 2018 have been collected.
Accordingly, 96 images during 24-hours of the day resulting in 7200 traffic data images have been obtained for the whole time
of study. According to administrative division, Tehran has been divided into 116 districts on which, congestion values on
freeways and arterial within 15-min intervals will be aggregated over. The congestion index in this study is the ratio of the total
number of red and dark-red pixels to the total number of pixels of each district within each 15-min interval. Also, due to the fact
that the travel pattern and consequently, the traffic congestion significantly differ on working and non-working days, so the
analyses will be conducted separately on weekdays and weekends as well as within CCZ, APCZ and out of both borders.

3. Spatial Analyses and Space-Time Analyses
Geographical areas often show specific spatial patterns. When dealing with spatial data, the value of a variable at a specific
location correlates with the values in proximate locations [14]. This phenomenon called spatial autocorrelation which can be
evaluated by various statistical indicators all based on the modelling of local relationships like proximity which can be measured
by spatial weight matrix. Among the indicators of spatial autocorrelation, Moran’s I can be considered as one of the most
frequently-used indicators. For any zone or area like i, the local Moran’s I statistic for any variable like x is defined as (1):

= ∑ (1)

Where zi and zj are the standardized values of x in zone i and zone j, and wij is the spatial weight matrix between zone i
and other zones like j, which quantifies the existing spatial relationships among zone i and other zones like j. Equation (1) is
used to draw a conclusion about whether the variable x is clustered or not. A positive value for I indicates that a feature has
neighbouring features with similarly high or low attribute values; this feature is part of a cluster. A negative value for I indicates
that a feature has neighbouring features with dissimilar values; this feature is an outlier. In either instance, the p-value must be
small enough for the cluster or outlier to be considered statistically significant. In the present study, local Moran’s I is applied to
examine the spatial dependency of traffic congestion for the time of study Accordingly, the null hypothesis postulates that there
is no significant difference between the distributions of the traffic congestion, while the alternative hypothesis proves a
significant difference of at least one of the samples. The result of this test indicates the rejection or acceptance of the null
hypothesis in the 95% confidence level. This is required to prove the correlation of congestion values during the given time of
study.
Although the spatial nature of traffic parameters highlights the key role of spatial analyses, most traffic data are associated
with a complex dynamic behaviour over space and time. That is since traffic situations indicates significant variations in space
and time, understanding the real behaviour can be conducted through a spatiotemporal analysis. Particularly dealing with large
space-time datasets has always been the challenge to researchers and therefore introducing the methods to facilitate
comprehending and visualizing the events and explore the latent patterns in a dataset can be appropriate. During the past years
different visualization techniques have been introduces particularly for large datasets [32-35]. A Hovmöller diagram is a common
way of plotting meteorological data to highlight the behaviour of waves, particularly tropical waves. The axes of Hovmöller

2