Slides presented at the event Accessibility in urban modelling: from measurement to policy instruction co-organized by NECTAR Cluster 6 and Urban Europe Research Alliance (UERA).
Lyon, June 18th-20th, 2018
Separation of Lanthanides/ Lanthanides and Actinides
The impact of temporal resolution on the precision of accessibility measurement
1. The impact of temporal resolution on the
precision of accessibility measurement
CALCULUS
This project has received funding from the European Union’s Horizon 2020 research and
innovation Programme under the Marie Sklodowska-Curie Grant Agreement no 749761
Marcin Stępniak
John P. Pritchard, Karst T. Geurs & Sławomir Goliszek
19/06/2018
2. CALCULUS
This project has received funding from the European Union’s Horizon 2020 research and
innovation Programme under the Marie Sklodowska-Curie Grant Agreement no 749761
CAlCULUS project
Causes and Consequences of low urban accessibility.
Defining proper policy responses
CAlCULUS project
Causes and Consequences of low urban accessibility.
Defining proper policy responses
Introduction
19/06/2018
3. CALCULUS
This project has received funding from the European Union’s Horizon 2020 research and
innovation Programme under the Marie Sklodowska-Curie Grant Agreement no 749761
Temporal dimension
Travel time varies temporally -> temporal variability in accessibility
• Travel time to city centre
• Morning peak-hours
(6:30 – 8:30)
• Starting time: every 5 minutes
19/06/2018
4. CALCULUS
This project has received funding from the European Union’s Horizon 2020 research and
innovation Programme under the Marie Sklodowska-Curie Grant Agreement no 749761
Temporal dimension
Travel time varies temporally -> temporal variability in accessibility
• Dynamic accessibility
• Time-continuous accessibility
(Owen & Lewinson, 2017)
(Moya-Gómez et al., 2017)
Owen & Lewinson (2014): improvement
of the [accessibility] model due to
application high-temporal resolution
Boisjoly & El-Geneidy (2016): high
correlation between all tested models
no significant improvement of a model
≠
Kaza (2015): advocates for systematic study of the time resolution
?
19/06/2018
5. CALCULUS
This project has received funding from the European Union’s Horizon 2020 research and
innovation Programme under the Marie Sklodowska-Curie Grant Agreement no 749761
Temporal dimension of accessibility
Temporal dimension
Temporal variability Temporal resolution
19/06/2018
6. CALCULUS
This project has received funding from the European Union’s Horizon 2020 research and
innovation Programme under the Marie Sklodowska-Curie Grant Agreement no 749761
Literature review: temporal resolution
Temporalresolution
Single
departure
time
Several
departure
times
Temporal
resolution
• El-Geneidy et al. 2016
• Guthrie et al. 2017
• Widener et al. 2015
Variability
Average
• Jäppinen et al. 2013 (9:00, 20:00)
• Boisjoly & El-Geneidy 2016 (8:00, 9:00, 10:00, 11:00)
• Kerkman et al. 2017 (8:00, 8:07, 8:18 am)
• Salonen & Toivonen, 2013 (2 peak + 2 peak-off)
• El-Geneidy et al. 2015 (1-hour resolution)
• Legrain et al. 2016 (1-hour resolution)
• Stępniak & Goliszek 2017 (15´ resolution)
• Järv et al. 2018 (12’ resolution)
• Fayyaz et al. 2016 (10´ resolution)
• Fransen et al. 2015 (5´ resolution)
• Farber et al. 2014 (1´ resolution)
• Owen & Lewinson 2015 (1’ resolution)
19/06/2018
7. CALCULUS
This project has received funding from the European Union’s Horizon 2020 research and
innovation Programme under the Marie Sklodowska-Curie Grant Agreement no 749761
Research questions
1. What is an impact of the decrease of temporal resolution on the
precision of travel time measurement?
2. How (to what extent) the decrease of temporal resolution
influences results of accessibility analysis?
3. To what extent the decrease of temporal resolution distorts
conclusions derived from accessibility analysis?
Temporal resolution
Public transport (GTFS)
car
public transport
19/06/2018
8. CALCULUS
This project has received funding from the European Union’s Horizon 2020 research and
innovation Programme under the Marie Sklodowska-Curie Grant Agreement no 749761
Case study: Szczecin (Poland)
0.5
0.6
0.7
0.8
0.9
1
0 10 20 30 40 50 60
Low regularity of public transport schedules
Correlations of travel times between subsequent departure times
Szczecin
The Netherlands
19/06/2018
9. CALCULUS
This project has received funding from the European Union’s Horizon 2020 research and
innovation Programme under the Marie Sklodowska-Curie Grant Agreement no 749761
Data: schedule-based travel time information
• calendar_dates.txt
• fare_attributes.txt
• shapes.txt
• frequencies.txt
• transfers.txt
ArcGIS®
Network Analyst
+
ArcGIS® Network
19/06/2018
10. CALCULUS
This project has received funding from the European Union’s Horizon 2020 research and
innovation Programme under the Marie Sklodowska-Curie Grant Agreement no 749761
Data: GTFS (public transport)
ArcGIS
Network
2145 OD
nodes
PT & walking
network
287 026 edges
Public transport network
• 88 PT routes:
• 12 tramway routes
• 16 night-bus routes
• 7 express routes
• 1395 stops
• Schedule for typical week-day (June, 2015) -> real
travel time
study
walking
speed (km/h)
comment
(Reyes et al., 2014) 3.2 Minimum typical speed for children aged 5-11
(Fransen et al., 2015) 4.0 Adult's average
(Ritsema van Eck et al.,
2005)
4.0 Distance as the crow flies
(Hadas, 2013) 4.0 -
(Nettleton et al., 2007) 4.8 -
(Farber et al., 2014) 4.8 -
(Willis et al., 2004) 5,3 mean walking speed of individuals
(Reyes et al., 2014) 5.4 Maximum typical speed for children aged 5-11
(Krizek et al., 2012) 5.4 average walking speed for 14-64 year old
Walking speed: 4.5 km/h
19/06/2018
11. CALCULUS
This project has received funding from the European Union’s Horizon 2020 research and
innovation Programme under the Marie Sklodowska-Curie Grant Agreement no 749761
Travel time measurement & aggregation
Benchmark value: 1-minute temporal resolution
1-hour long period of time (61 measurements): average / harmonic mean aggregation
4 cases: different frequency:
1) Very low frequency – during the night (2-3am);
2) High frequency during the morning peak hours (7-8am);
3) Daily low frequency during peak-off period (10-11am);
4) Evening low frequency (10-11pm).
Applied lower temporal resolutions:
- Every 2, 3, 4, 5, 10, 15, 20, 30 & 60 minutes
19/06/2018
12. CALCULUS
This project has received funding from the European Union’s Horizon 2020 research and
innovation Programme under the Marie Sklodowska-Curie Grant Agreement no 749761
• Relation between frequency and precision
• No visible change up to 5 minutes
(MAE <0.8 min., MAPE < 1.8%)
• Low impact up to 15 minutes
(MAE < 2 min., MAPE < 3.5%)
• High correlation (0.99 at 15 minutes)
RQ1: Travel time comparison
19/06/2018
13. CALCULUS
This project has received funding from the European Union’s Horizon 2020 research and
innovation Programme under the Marie Sklodowska-Curie Grant Agreement no 749761
RQ2: Accessibility comparison (1)
Different weight of travel time:
• travel-time-to-the-closest-facility
• Cumulative opportunities
• Potential accessibility
Public services of different level of centrality:
• High centrality: low number of providers, highly concentrated
• Low centrality: high number, evenly distributed across the space
19/06/2018
14. CALCULUS
This project has received funding from the European Union’s Horizon 2020 research and
innovation Programme under the Marie Sklodowska-Curie Grant Agreement no 749761
RQ2: Accessibility comparison (2: services)
Type of nodes Spatial
accuracy
Accessibility
measure
Centrality level Data
Origins* Centroid OD travel time NA Census tracks (1745)
Destinations
Address
points
travel time to
the nearest
provider
High City council (1)
Low Nurseries (30)
Isochrone
measure
High Theatres (21)
Low
Specialized health care
(169)
Potential
accessibility
measure
High Hospitals (9)
Low Secondary schools (68)
19/06/2018
15. CALCULUS
This project has received funding from the European Union’s Horizon 2020 research and
innovation Programme under the Marie Sklodowska-Curie Grant Agreement no 749761
RQ2: Accessibility comparison (3: measures)
MAPE
(%)
MAE
max
diff.
Corr.
MAPE
(%)
MAE
max
diff.
Corr.
MAPE
(%)
MAE
max
diff.
Corr.
Temporalresolution
Highcentrality
Administration centre
travel-time-to-the-closest-
facility
Theatres
contour measure
Hospitals (no. of beds)
potential accessibility
5 min 1.3 0.5 5.6 0.999 3.7 0.2 7.0 0.995 0.9 10.8 70.3 0.999
10 min 2.3 0.8 12.1 0.997 6.6 0.4 16.0 0.988 1.7 19.8 143.2 0.996
15 min 2.7 1.1 14.2 0.996 8.6 0.5 16.0 0.980 2.1 24.6 177.2 0.994
20 min 3.8 1.5 22.6 0.993 12.1 0.8 17.0 0.971 2.8 32.1 217.8 0.991
30 min 4.3 1.7 25.5 0.990 13.9 0.9 17.0 0.964 3.5 40.8 276.7 0.985
60 min 6.0 2.5 39.7 0.980 20.3 1.1 17.0 0.947 5.0 55.6 383.9 0.974
Lowcentrality
Nurseries
travel-time-to-the-closest-
facility
Specialized health care
contour measure
Secondary schools (no. of
classes)
potential accessibility
5 min 0.8 0.2 6.8 0.999 4.4 1.7 28.0 0.996 1.0 2.1 13.7 0.999
10 min 1.4 0.3 13.6 0.997 9.8 2.8 42.0 0.992 1.7 3.8 26.5 0.997
15 min 1.6 0.4 15.1 0.997 10.4 3.5 41.0 0.990 2.3 5.0 35.3 0.995
20 min 2.0 0.5 21.0 0.995 16.2 4.6 53.0 0.985 2.9 6.3 43.7 0.993
30 min 2.5 0.6 22.1 0.991 16.8 5.2 58.0 0.981 3.8 8.1 65.2 0.988
60 min 3.0 0.7 19.9 0.989 23.6 7.0 73.0 0.967 5.4 11.5 97.8 0.978
19/06/2018
16. CALCULUS
This project has received funding from the European Union’s Horizon 2020 research and
innovation Programme under the Marie Sklodowska-Curie Grant Agreement no 749761
RQ2: Accessibility comparison (4: measures)
• The most affected: cumulative
opportunity
• Differences between resolutions:
10’-15’ << 15’-20’
• No visible impact of the level of
centrality of services
• Very high max differences (risk
when looking for extreme cases)
• High correlation (~0.99)
19/06/2018
17. CALCULUS
This project has received funding from the European Union’s Horizon 2020 research and
innovation Programme under the Marie Sklodowska-Curie Grant Agreement no 749761
RQ2: Accessibility comparison (5: frequency)
• Closest facility: low frequency (night period) is
biased by walking time
• Potential accessibility: direct relation between
frequency and precision
• Cumulative accessibility: low centrality biased
by walking time; high centrality: direct relation
between frequency and precision
19/06/2018
18. CALCULUS
This project has received funding from the European Union’s Horizon 2020 research and
innovation Programme under the Marie Sklodowska-Curie Grant Agreement no 749761
RQ3: Interpretation (1: Ethical theories)
Egalitarianism:
Gini
Sufficientarianism
threshold
19/06/2018
19. CALCULUS
This project has received funding from the European Union’s Horizon 2020 research and
innovation Programme under the Marie Sklodowska-Curie Grant Agreement no 749761
RQ3: Interpretation (2: Gini)
5 10 15 20 30 60 GINI
Administration 0.000 0.000 0.000 0.001 0.001 0.003 0.287
Nurseries 0.000 0.000 0.000 0.001 0.002 0.003 0.396
Theatries 0.002 0.006 0.005 0.009 0.010 0.015 0.241
Health Care 0.000 0.001 0.001 0.003 0.000 0.001 0.444
Hospitals 0.001 0.001 0.001 0.001 0.002 0.005 0.130
Schools 0.000 0.001 0.001 0.001 0.003 0.007 0.151
• No significance difference
19/06/2018
20. CALCULUS
This project has received funding from the European Union’s Horizon 2020 research and
innovation Programme under the Marie Sklodowska-Curie Grant Agreement no 749761
RQ3: Interpretation (3: threshold)
• 15’ resolution – error ≤ 2.5%
• No clear pattern
Public service
Accessibility
measure
Threshold
Temporal resolution
5 10 15 20 30 60
Administration Travel-time-
to-the-closest-
facility
30 minutes
21 40 40 47 67 94
Nurseries 7 14 16 18 16 18
Theatries No facilities
within 30
minutes
0 (no access
within 30
minutes)
7 15 12 21 28 25
Health Care 17 25 40 33 49 57
Hospitals
potential
accessibility
20% the lest
accessible units
16 33 37 39 47 65
Schools 26 34 44 47 46 60
19/06/2018
21. CALCULUS
This project has received funding from the European Union’s Horizon 2020 research and
innovation Programme under the Marie Sklodowska-Curie Grant Agreement no 749761
RQ3: Interpretation (4: spatial distribution)
• Proximity & high centrality:
clear transition zone
• No clear spatial pattern
(random distribution)
• Different patterns for all
scenarios – random distributions
19/06/2018
22. CALCULUS
This project has received funding from the European Union’s Horizon 2020 research and
innovation Programme under the Marie Sklodowska-Curie Grant Agreement no 749761
Conclusions
• Very high correlations between benchmark and tested values (travel times & accessibility)
• Frequency has an impact on the loss of the precision.
• Accessibility measures:
• Significant impact on contour measure;
• hardly visible impact in case of proximity and potential accessibility.
• Up to 5 minutes of temporal resolution – no significant impact on precision.
• 15 minutes temporal resolution: seems to have the best trade-off between computational
and precision:
• < 1.8% in case of travel time;
• <2.5% in case of accessibility measures (except contour measure);
• < 1% in case of Gini and ~2.5% on average of thresholds
• Hardly affected interpretation of accessibility analysis.
• Except: extreme cases (maximum differences).
19/06/2018
23. CALCULUS
This project has received funding from the European Union’s Horizon 2020 research and
innovation Programme under the Marie Sklodowska-Curie Grant Agreement no 749761
Thank you for your attention!
19/06/2018
marcinstepniak@ucm.es
@marcin_stepniak ● @tGIS_ucm