isparities between modelled routes and those actually taken.
We illustrate the magnitude of observed errors in GIS modelling with some examples. Taking one study participant who travelled
20 km home from work by car on one trip, the shortest route modelled using GIS software underestimated their actual distance by 79%
and overlapped the actual route by only 15%, with the proportion of the actual route that involved busy roads being half of that predicted.
For a second participant who walked a distance of 2.2 km from work to home, their route was more similar in length to that predicted (just
16% shorter) but only 19.2% of the route overlapped the modelled route. Further, their actual route passed no destinations and involved no
busy roads, whereas the model indicated that they would pass 14 leisure, food, or retail attractions and spend 41% of the journey on ‘A’ or
‘B’ roads or motorways. Researchers from public health backgrounds using geographical analysis to estimate the health implications of
exposure and route choice during the commute must be aware of the potential impact of using these methods