Geoscience Reference
In-Depth Information
FIGURE 10.6
Examples of data generated by the GeoCrimeData project. (© Crown Copyright/database
rights 2011. An Ordnance Survey/EDINA supplied service.)
0.5
Model performance
0.45
0.4
Impact of the charge (stated)
Not affected by the charge
0.35
9%
0.3
0.25
No change in behaviour
Travel at a different time
Switch to public transport
Work or shop at new location
14%
14%
17%
46%
0.2
0.15
0.1
0.05
0
1
2
3458161738
Number of iterations
39 79 77 85 86 95 96100
(a)
(b)
(c)
(d)
FIGURE 10.7
Calibrating a model of congestion charging using crowdsourced data. (a) Behavioural change
from the congestion charge. (b) Model goodness of fit. (c) Areas affected by the charge. (d) Modal split for
baseline and two scenarios.
survey to define a study area (Figure 10.7b). We then applied a SIM to estimate these behaviour pat-
terns, and we used a genetic algorithm (GA) to calibrate the values of the parameters in the model to
the data supplied online (Figure 10.7c). The GA required tens of thousands of runs of the embedded
modelling equations and was implemented using the Arc1 Grid system at Leeds. After calibration,
the model was able to reproduce the behaviour of the survey respondents. In this way, it was possible
