Visualising field trips
Parties working in the field (away from the station) on Macquarie Island are required to radio in each night, to let the station know that they are safely in the hut and also what they intend to be doing on the following day. These sitreps ("situation reports") are kept on file, and we can use them to reconstruct the movements of the field parties around the island over time. This information can help with the management of the island's tracks and huts.
About
Parties working in the field (away from the station) on Macquarie Island are required to radio in each night, to let the station know that they are safely in the hut and also what they intend to be doing on the following day. These sitreps ("situation reports") are kept on file, and we can use them to reconstruct the movements of the field parties around the island over time.
The sitreps are filled out by the radio operator on station using a standard template. The content of the sitrep is a mixture of standardised names and acronyms, and free text. A typical sitrep might be as follows:
MACQUARIE ISLAND FIELD SITREP
Date: 8 MAR 2009 Time: 2000 (Local)
| Hut | Names | Intentions | Destination |
|---|---|---|---|
| HP | Annie, Bertie | OLT>Tio tk>OLT | GG |
| BB | Charlie | Wk xct Sth of BB & west of Prion lake>Mt Harrison>Featherbed | BB |
The interpretation of the first one is pretty straightforward: Annie and Bertie filed their sitrep from the Hurd Point (HP) hut. Tomorrow they are walking to the Green Gorge (GG) hut by way of the Overland Track (OLT), then a detour on the Tiobunga Track (Tio tk), and then back onto the Overland Track.
Charlie's sitrep is a little more difficult to process. He is at Bauer Bay (BB) hut, and will be working cross-country (Xct) south of Bauer Bay and west of Prion Lake, then proceeding to Mount Harrisson and the Featherbed and back to the Bauer Bay hut tomorrow night.
Processing
The broad analysis method is as follows:
- Construct a matrix which, for selected locations, contains the most likely route between those locations. For example, the most likely route from Bauer Bay to Brothers Point is BB-BBT-SBT-BP (Bauer Bay - Bauer Bay Track - Sandy Bay Track - Brothers Point).
- For each day, parse the sitrep document and extract each party's hut, names, destination, and intentions. Search for all recognised track, hut, and place names (or aliases) in the intentions text. This gives us a sequence of locations (starting at the hut, via the recognised features in the intentions text, and ending at the destination).
- We now try and supplement this list of locations, by finding non-adjacent locations and inserting intermediate locations between them. We check consecutive pairs of locations, and if we have a "most likely" route between those locations (see step 1), we insert that route into our sequence of locations. If we don't have a "most likely" route for a particular pair of locations, we assume that the party went cross-country (didn't use a track).
- We now wish to try and construct the actual path taken by the party between these locations. Tracks, named features, and huts are all relatively easy, because we have the coordinates in our GIS shapefiles and gazetteer. Cross-country is more difficult — at the moment we use fairly simple interpolation method to pick a smooth path between points.
- Now we have the path taken by each field party over time, and we can generate visualisations in the form of images and animations.
Interesting issues:
The processing method overview glosses over many of the difficulties which prevent the analyses from working well in practice. Some of these include:
- multiple names for the same thing. For example, Petrel Peak is variously entered as "PP", "PPeak", or "Petrel Pk". We get around this by constructing a dictionary of aliases for each place name. Spelling errors are handled in the same way. It would be more elegant to use a fuzzy text-matching search to find spelling errors. (For a spelling error example, in Charlie's sitrep above, Mount Harrisson was incorrectly spelled as Harrison).
- the sitreps represent the intentions of the field party. If the party changes their intentions, they are required to advise the station, but these changes do not always make it into an updated sitrep file. This situation is not currently handled very well, and contributes to some of the glitches you can see in the animation.
- sitreps are only filed by parties in the field - there is no record of their route from the station to the hut from which they file their first sitrep of the trip. We have to make a guess as to the tracks they used for the outward journey.
- cross-country trips are handled with a simple interpolation that does not take into account the terrain being crossed. Thus, you will sometimes get parties going through lakes or bays. A better solution would be a cost-weighted interpolation, that penalises difficult terrain (like underwater) and so avoids it.
