Geography Reference
In-Depth Information
attractions have fewer gaps and errors in the related touring events, occasional
stops are missed out altogether, temporal events get overlooked at stops with a
large number of visits, and enter/exit stops have lower completeness but more
accurate recorded arrival times due to the nature of the experience of entering or
leaving the Coast. Against this the profiles of individual tourists have lesser
impact. In a number of contexts higher completeness does not result in higher
consistency. Tourists tend to round the temporal events into half hour segments
and combine with the cumulative impact of slowing down for views and taking
micro stops perceived as less than 5 min and so not recorded, this appears to result
in reported arrival times that are later than those modeled. Clearly accuracy in a
number of forms varies systematically and in a complex way across space, time
and circumstance, and although this chapter cannot hope to resolve the issues
raised, it can still suggest a way to counter at least some of the issues around
absence and anomalous reporting.
4 Enhancing Spatiotemporal Data Quality of Movement
Survey
4.1 Filling Temporal Gaps of Movement Data
by Cross-Inference
4.1.1 Interpolate/Extrapolate Missing Temporal Events from
Known Variables
In any movement dataset temporal processes (and individual tourists) should be
represented as moving forward continuously, which is in contrast to the stop-
oriented representation from survey data, and for that matter from GPS, but at a
finer scale. The original stops recorded on the WCTFS framework are certainly
point based with significant gaps, gaps we would like to reduce. We can do this
crudely and inaccurately by linking sequential points with straight lines, but this
very crude technique can be replaced where movement is channelled within a
network such as a road network. Here we can trace possible paths along a con-
tinuous line, interpolating back to a finer pattern of intermediate points if we need
this for further analysis. At the same time we can model characteristics of the
network links to allow estimations of typical transition times for different forms of
travel. In this way incomplete data on arrival times and durations as well as
anomalies can be augmented by interpolating/extrapolating from known events as
a simple formula below explains. Arrival time (T
i
), duration time (T
i
), and travel
time (T
i
) values were formatted into decimal times to enable the calculations.
Technically, filling gaps by cross-inference is an iterative process designed to
eventually achieve maximum retrieval of missing events.
