Geography Reference
In-Depth Information
Fig. 11 Proportions of the
variation between evaluation
results and original reported
values
5 Applications and Conclusions
This paper has sought to investigate the accuracy properties of a fairly typical rich
movement dataset collected through a non-digital, field based methodology. In
small part this has entailed basic stock-taking on non-responses to questions, but
the major contribution has been to view these data in the context of the constraints
and influences of space, time and process. This has involved the use of GIS in
transforming the survey's spatial-temporal information from a discrete, point-
oriented representation to a continuous trajectory reflecting actual transport con-
ditions and norm travel times. This has two positive outcomes, it allows both the
identification of travel-time anomalies in the dataset and the development of a
framework for estimating missing values consistent with space-time constraints
and behavioural norms at tourist attractions. In short, we have enhanced insights
on the quality of the data (in both time and space) and have gained the option of
working with an augmented dataset that amongst other things is formatted in a way
that is supportive of sophisticated geovisualisation and query.
The literature on quality assessment of tourist movement datasets is fairly
sparse, and close comparison with other research is hard to achieve, but our results
are probably not atypical of field interviews constricted by a limitation on delivery
time. Certainly the findings confirm expectations on factors affecting response
levels and accuracy, and show interesting differences across the two survey
instruments. The literature on augmentation of space-time datasets is also fairly
brief. The fairly simple model utilised in this paper adds substantial value, and
opens the way to use geovisual analytics as a means to interrogate the data in a
semi-qualitative way (Zhao et al.
2011
). Future research could expand the ability
to extract more value by enhancing the models used to allocate activities or
timings to blank entries. We cannot hope to recapture the information totally, but
as Tobler and Wineberg (
1971
) speculated on Cappadocia, cross-referencing data
and distance can sometimes restore the lost information. At present, although some
