Geography Reference
In-Depth Information
estimates of arrival times, therefore the variation between arrival times and GIS
estimates of arrival times already contains some subjective assumptions. These
assumptions need to be considered when determining if an arrival time with a
different value from the GIS estimates is an error, or on the other hand, the
definition of error in temporal events is influenced by the nature of the stops
involved. Certainly, the acceptable levels of consistency of arrival times have been
given a higher tolerance for overnight stops than for short stopovers. Finally,
taking the above factors into consideration, for overnight stops and stops with a
duration of over 4 h, only the difference between arrival times and model estimates
that are [1 h are identified as errors for further modelling. By contrast, differences
of only half an hour are defined as anomalies for other stops. At the end of the
consistency checking, the arrival times were qualified as two types: acceptable
events (within anticipated parameters) and error/anomalous events which sug-
gested a need for action. Error events were consequently replaced by GIS esti-
mates of arrival times and/or subject to further analysis.
As the newly assigned arrival times can cause new inconsistency in the dataset
the process has to be recursive and repeat the checking and updating. When
complete the rectified arrival times are stored in a new field in the attribute table
and a comment field is necessary to record the actions. Finally, there are about 2 %
records out of the total WCTFS whose arrival times are identified as errors and
assigned new arrival times in the dataset.
4.3 Evaluation of Enhancement
In order to verify the enhancement process and assess the validity of the meth-
odologies, an evaluation of the enhancement is undertaken by applying the same
space-time algorithms on a sub-sample of complete movement records that have
had random arrival times removed. The result of this reconstruction provides a tool
to assess the validity of this approach.
Figure
11
shows the variation between evaluation results and original reported
values. There are approximately 70 % of the evaluation results have less than
about half an hour difference from reported values, and approximately 84 % are
within 1 h variation from the original reported temporal events.
The evaluation results above indicate that the enhanced events largely match
the reported values. Although there are noticeable variations between them, in
considering the fact that original reported events by respondents might contain
errors and affect the accuracy of the evaluation, this variation is in line with
expectations and the general accuracy of reporting by respondents.
