=Paper=
{{Paper
|id=Vol-1328/GSR2_Quinn
|storemode=property
|title=Spatio-Temporal Analysis of GPS Tracks of CODE RED: MOBILE an Experimental Mobile Scenario and Location Based Training Exercise
|pdfUrl=https://ceur-ws.org/Vol-1328/GSR2_Quinn.pdf
|volume=Vol-1328
|dblpUrl=https://dblp.org/rec/conf/gsr/QuinnC12
}}
==Spatio-Temporal Analysis of GPS Tracks of CODE RED: MOBILE an Experimental Mobile Scenario and Location Based Training Exercise==
https://ceur-ws.org/Vol-1328/GSR2_Quinn.pdf
Spatio-temporal analysis of GPS tracks of CODE RED: MOBILE an
experimental mobile scenario and location based training exercise
P. B. Quinn and W.E. Cartwright
RMIT Melbourne, Victoria, Australia, 3000.
pb.quinn@bigpond.com, william.cartwright@rmit.edu.au
Acknowledgements: The Mt Macedon Group of the Country Fire Authority (CFA) and the 33 firefighter
volunteers from its brigades, Newham RFB, Apple University Consortium, Macedon Ranges Shire Council,
Hanging Rock Reserve management and employees, RMIT University, and Chris Marmo and Monique
Elsley.
Abstract
As part of an ongoing research project, geovisualisations of bushfires were delivered at GPS-determined locations to volunteer
firefighters from the Country Fire Authority’s Macedon Ranges Group. The participants skill level ranged from basic wildfire
firefighter trained through to captain of brigade. The location-based scenario training exercise is called CODE RED: MOBILE.
Using information from the geovisualisations about a virtual bushfire at Hanging Rock, participants selected which houses would
likely burn down after a wind change. They were free to take any path to reach the virtual houses, indicated by markers on the
screen of an iPad New. They were asked to go to the virtual house location to observe the real landscape and to estimate where the
virtual fire would go. Most participants took about an hour to complete the exercise. A GPS device kept track of where they went.
A Fractal D score was assigned to participant’s tracks using Vilis O. Nams’ software: Fractal 5.20.0. Spatio-temporal analysis of
the GPS tracks using ArcMap 10 and Geotime 5.3 found that participants undertook the exercise by following unusual tracks.
Preliminary results showed that some of these participants, not following test procedure instructions closely, had sometimes
undertaken more direct tracks, shown by low Fractal D scores. However, they were able to choose the correct houses assigned to
visit. This type of analysis can assist in improving the design of mobile, location based exercises. It can also provide an additional
means of assessing and improving firefighter performance. This paper will outline the background behind the exercise, specify the
type of information that was sought and provides details of the results obtained through analysis.
Key words: Bushfires, Firefighters, Fractal D, Spatio-Temporal analysis, Heatmaps, Scenario based training, Mobile Learning.
Author biographies
Brian Quinn is a PhD candidate at RMIT University in the School of Geospatial and Mathematical Sciences, a member of the
Newham Fire Brigade, a volunteer fire mapper with the Country Fire Authority at the Gisborne Incident Control Centre. Fire
Behaviour Analyst Intermediate Level, DSE course completed in 2010.
William Cartwright; Professor of Cartography in the School of Mathematical and Geospatial Sciences at RMIT
University,Australia. He is Chair of the Joint Board of Geospatial Information Societies and Immediate Past-President of the
International Cartographic Association. His major research interest is the application of integrated media to cartography and the
exploration of different metaphorical approaches to the depiction of geographical information.
Introduction
30+ Country Fire Authority (CFA) firefighters, who belong to the mainly volunteer rural fire brigades of the Mt
Macedon Group, over several weekend days in May, 2012 took part in a training exercise utilising the 7scenes
application (7scenes.com) running on iPad New (see Figure 1) . Trimble Juno SB GPS handhelds were also carried by
the participants. Participants were in two groups. Group A received information by maps and annotated screenshots of
a bushfire undergoing a wind change, and Group B received maps and movies of the same fire. The screenshots for
Group A were taken from the same movies that Group B watched. The movies were made in the Crysis Wars,
Sandbox2 (crytek.com) computer game editor. The aim of the research was to see whether performance in the exercise
was affected by the different treatments Group A and B received. It also examined whether visualisations made in off
the shelf computer games are an effective means of providing information for a mobile scenario based training
exercise.
Figure 1: View of the CODE RED: MOBILE markers in the 7scenes application on the iPad.
Participant’s location marked by the blue bubble. Hanging Rock is near the centre of the image.
The scenario in 7scenes was typical for a hot windy summer’s day in Central Victoria, Australia. It simulated a
bushfire burning southwards through lightly treed areas of the local racecourse that is situated at the eastern side of the
Hanging Rock Reserve. An aerial view can be seen in Figure 1. The markers in blue are in the CODE RED: MOBILE
exercise area to the right and east. To the left and west is Hanging Rock. Figure 3 shows the exercise area from the
north east in the Sandbox2 editor. Later in the day a wind change turned the fire eastwards. The visualisations and
other information were delivered to the device when the player entered a GPS defined location. This is characterised
as a form of location based service. The visualisations for CODE RED: MOBILE were created in Sandbox2, the editor
that is included with the Crysis Wars computer game (crytek.com). This game editor and game has been
acknowledged as an excellent means of creating geovisualisations. Bishop et al (2011) presented a house fire
simulation game built using the Sandbox2 editor in the Crysis game (crytek.com). Quinn and Cartwright (2011) have
described how visualisations about bushfires for a mobile device, can be made in the Sandbox2 editor.
Background to the study
Conversation Theory
Sharples (2005) presented a framework for learning in a mobile communication setting that recognised the social
dimension of learning via conversations. This partly emanated from the work of Pask (1976) whose Conversation
Theory derived from Cybernetics and Second Order Cybernetics where a machine- like element in a system can be
understood in a context by observers, who then become participants in the system. In a sense they study the system
through questioning it, which is a form of conversation. Laurillard (2002) amended Pask’s ideas by dividing learning
through conversations into a Level of Actions, which is a conversation between people about a system or some
learning they are engaged in; and a Level of Descriptions, where they consider why things happened and what that
means for those involved.
CODE RED: MOBILE communicates information to participants about a virtual bushfire at the real world context of
Hanging Rock via the 7scenes app (7scenes.com) on an iPad New (apple.com). The older paradigm of learning in a
classroom can now also be learning in an outdoors context using mobile devices (Sharples 2005). The iPad New and
7scenes application mediates a conversation, in Pask and Laurillard’s term, between participating firefighters and the
information in the visualisations, at the real world context of the terrain, vegetation and buildings, in which the
visualisations are set. Pervasive wire- based computing has changed in the last few decades into a near ubiquitous
mobile dimension that enables conversations for learning, in and about the outdoor dynamic world.
Cognitive Artefacts
Barbara Tversky (2000) wrote that humans are separated from other creatures by the intellectual achievement of
cognitive artefacts. These are devices external to the human mind that extend our cognitive abilities. They increase our
efficiency of thinking by storing the knowledge required by a task to the artefact, thus reducing working memory’s
cognitive load. They can assist with the calculations necessary for a task. Cognitive artefacts can be considered an
extension of Pask’s Conversation Theory as they are a formalisation of a set of concepts into a connected interacting
symbolic or machine like form that can be stored, accessed, learnt from and promulgated.
‘Scaffolding’, summarised by Andrews et al (2007), is a technique where phenomena are converted to cognitive
artefacts for pedagogical purposes. This can be done by: ‘reducing the degrees of freedom...accentuating relevant
features of the task...and modelling solutions.’ (p. 258). de Jong (2005) found that individuals learned more using
multimedia if there was guided discovery, sometimes referred to as ‘scaffolding’. Scaffolding is also the structuring
and problematising of a problem or domain of knowledge (Reiser 2004 p. 287). The structuring breaks down the
problem or what is to be learned; problematising directs learners to issues and tasks by modelling the conversations
with devices and other persons. They will reason more realistically about the issue because they have been engaged
affectively. Software tools can assist in this. Training can utilise the types of software and equipment including mobile
devices that participants would use in the task being modelled. The CFA in 2011 (CFA 2011) began issuing iPads to
members. They are loaded with maps of the State of Victoria that are currently provided on all fire trucks. CODE
RED: MOBILE also uses iPad New and models what may be an important future tool for firefighting.
Students may learn to use the cognitive artefacts of others, for example a model of the effects of a wind change on a
bushfire. In engaging with that model they may then understand how a bushfire alters after a wind change. Students of
any topic or interest can assemble, through conversations with reality and with other cognitive artefacts on mobile or
other devices, their own particular cognitive artefacts. These new cognitive artefacts may be examples of scientific
method, Mathematical Game Theory, computer games, art or fiction writing.
Game Theory
One way of problematising an issue is to make a game of it. Chess can be said to be a problematising of war,
especially the tactical aspects of it. Civilization created by Sid Meiers (civilization.com) could be said to problematise
the issue of progress and the development of civilized society.
Mathematical Game Theory as defined by Myerson (1991 p.1) is ‘the study of mathematical models of conflict and
cooperation between rational decision makers'. Mathematical Games can also be between people and nature. The
opponent although inanimate seems often to be given a name such as Cyclone Tracy, Black Saturday Bushfires and so
on. The very personalisation seems to make us more alert to their machinations.
Playing the stock market has been described as a zero-sum, multi-player game. Zero sum means that there is no
overall net change so all the losses on the stock market equal the total gains, but some individuals do gain or lose.
With the game of chess there is a winner and loser and the net effect is no change; it is a zero sum, two player game. A
firefighter, engaging in a scenario training exercise, is trying to ameliorate the effects of a bushfire to the advantage of
people; nature has adapted to fire in Australia and the health of the bush requires fire, so nature in a sense is the
opponent. Either the firefighter or nature will win, so it can be thought of as a zero sum two player game. The game is
not fair as the firefighter loses when almost any damage is done.
Location based games and learning
Fetter et al (2007 p.1) on location-based games noted:
‘Location-based Games are one of the many areas where the concepts of pervasive gaming come to life.
Thereby single players or teams perform tasks in specified scenarios using mobile computers like laptops,
personal digital assistants or mobile phones in combination with wireless communication and location-
sensing technologies, having the real world as their game board’.
Quinn and Cartwright (2011) described how visualisations and other information were delivered to a mobile device
when the player entered a GPS defined location in a game about bushfires. This can be characterised as a form of
location based service. WebPark (Edwardes et al 2005) is a location based service for the Texel Dunes National Park
(the Netherlands) and the Swiss National Park in eastern Switzerland and has been described by Dias (2007) as a
context-aware, location-based service (LBS). Dias argued that main advantage of the service was the delivery of
multimedia and information about the park rules and advisories at a location for visitors to the park, as well as in the
management of the park. Brown et al (2010) designed Routemate an application on an Android (google.com) based
mobile phone that is designed to reduce cognitive load for intellectually disabled and other disabled participants who
are learning a route to work or school. However the vital personal cognitive maps may not be constructed by users if
there is too much support from the device. A game based approach may assist the development of personal cognitive
maps. Testing of participants included testing their skills in planning a route, doing a practice run and then
independently navigating the new route.
Media for location based learning
Dynamic-static visualisations incorporate a static sequence spliced into the animation. The static section can show text
for enough time that it is easily understood (Pfeiffer et al 2009). They showed that dynamic-static visualisations
overcome some of the problems of cognitive load with simple animations (dynamic visualisations) and were superior
for learning to strictly static visualisations. Pfeiffer et al. (2011) conducted a field trip for students learning to use fish
species identification keys. Fish animations showed movement of fins and other behaviour variations of several
species. They proposed that dynamic-static visualisations supported concrete thinking relying less on abstract
reasoning. Concrete thinking being more childlike thinking about the here and now and abstract more adult and
concerned with abstractions such as beauty and justice.
Fractal Analysis
Fractal analysis has been used for studying searching, dispersal, orientation and navigation, especially in animals from
insects to whales. Mandelbrot (1983) suggested that fractals would be one way of analysing their tracks and Milne
(1997) that tortuosity, in the form of the Fractal Dimension (D), can measure an animal’s movements in relationship to
environmental and behavioural factors (Nams
and Bourgeois 2004).
A Fractal D score is a measure of the tortuosity of a person or animal’s track and can be between 1 and 2. A score of 1
indicates a straight line and a Fractal D of 2 is a line that is an exceedingly tortuous track. The spatial scale of the track
reveals the Fractal D at various lengths of movement. For example at the length of movement of 10 metres the Fractal
D might be 1.5.
“A person walking down a corridor, but turning every now and then to investigate something, would have a low
Fractal D (it approaches 1.0) at large spatial scales and a high Fractal D (it approaches 2.0) at small spatial scales”
(Kearns et al. 2010 p. 592).
In Figure 2 the track for a wombat is quite tortuous and approaches a Fractal D of 2 while it is digging up roots in
small patches i.e. at a short spatial scale. At a length of movement of perhaps a 100 meters (i.e. the spatial scale), the
Fractal D for the wombat might be 1.1 which is a fairly straight movement. This might show the wombat has to walk,
generally speaking, something like 100 metres in fairly straight lines over the barren ground between its favourite
patches of food.
Figure 2: Variations in tortuosity of a fictional wombat’s track as described by a Fractal D score at various spatial scales. The green circled
areas show the tortuous track with a high fractal score where the animal is foraging for food. The less tortuous line, where the animal is
travelling to where it knows the forage is better, has a low fractal score.
Kearns et al (2010) investigated the relationship between Fractal D and the prognosis of dementia for a group of
elderly residents in a long term care facility. Their research showed that residents with high Fractal D scores, were
more likely to have dementia and consequently an increased chance of falls. The Fractal D of these residents shows
that the cognitive impairment is probably at least partially, in navigation and decision making abilities. Thus testing
firefighters for their Fractal D score at different scales may reveal something about their navigation skills in the
exercise.
Webb et al (2009) found that rutting male white tailed deer usually had a low Fractal D, but in spring and summer a
higher Fractal D. This showed that the male deer move in straighter lines when rutting, presumably to contact females,
but take a generally more tortuous path the rest of the year, when they are looking for fresh grazing. Where there is a
change in domains of scale of movement, animals are often moving from one vegetation type to another, one perhaps
for food the other not (Wiens 1989).
Summing Up
Conversation Theory with theories about learning from Cognitive Science lead to the idea that visualisations for
CODE RED: MOBILE are cognitive artefacts. These visualisations showed the progress of the fire from its point of
origin to the north of the Racetrack to a point to the east of the central lake where a wind change turns the fire from
burning south on a narrow front to one that is burning east from the whole long eastern flank. These visualisations and
accompanying maps which were delivered on the iPad New on location at Hanging Rock and then used by the
firefighters to form a mental model of the fire and to predict which houses would burn. This occurred in a form of
location based game framework. It is also a type of location based service. The delivery of learning about the fire
using visualisations and the decision making task in CODE RED: MOBILE using these frameworks provided
scaffolding for learning about the effects of a wind change on a bushfire.
Fractal D was used to quantify the participants tracks during the second half of the exercise in the decision making
phase. It was hypothesised that there might be a correlation with scores on the correct choices of houses which would
or would not burn. However preliminary results show no relationship with the scores on the tasks. As firefighters must
be cognitively and navigationally adept, this is not a surprise. However there are many complicating factors such as
expertise levels and also that some may ‘cheat’ and ignore rules whilst others may get tired and guess rather than
reason. They may also skip a task.
The CODE RED: MOBILE training experiment and training exercise
CODE RED: MOBILE incorporates many of the above ideas. Visualisations are conceived of as cognitive artefacts
and engaging with them is a form of conversation with the embedded, structured ideas and information, the mobile
technologies that convey them and the real world location of the virtual bushfire at the Hanging Rock Reserve. The
exercise for firefighters delivered information about a bushfire and wind change in a first phase of the exercise
followed by a second phase where decisions were made based on that information.
Two Groups of firefighters with separate treatments
For each session, the firefighters were divided into two groups: Group A, who received maps and the visualisations of
the virtual bushfire as annotated screenshots and Group B, who saw the same maps and the virtual bushfire recorded
as a movie with the annotated screenshots spliced into it. Thus Group A saw static media and Group B saw dynamic-
static media (Pfeiffer et al 2009). These two treatments were compared in the experiment with the scores participants
achieved on various tasks and to the fractal scores of their GPS tracks. Movies and screenshots were made from a 3D
model created in Sandbox2 (crytek.com) as shown in Figure 3 to the left. The right hand image shows aerial view of
the Hanging Rock Reserve. The buildings to lower left were a temporary stage for a concert.
Figure 3: On the left is a view of the bushfire point of origin at Hanging Rock Reserve, created in Sandbox2 (crytek.com). To the right
is an aerial photo. The photo is with the permission of Mr Bruce Hedge.
Participants received a plain language statement of what was involved and signed enrolment forms. This was followed
by 45 minutes of instruction on how to use the iPad New and 7scenes. They received a booklet summarising these
instructions and carried it with them during the exercise. The field part of the experiment took up to about an hour and
a half. Originally there were more tasks to perform but these had to be eliminated as earlier trials showed they took too
long and were too exhausting especially in wet and windy weather. The actual experiments in May, 2012 all occurred
on drier and less windy days, somewhat of a rare occurrence in late autumn at Hanging Rock.
Hanging Rock itself can be seen throughout the exercise and is a prime means of finding one’s location. The images,
both static and dynamic, featuring the bushfires include iconic features such as Hanging Rock, Mt Macedon, the
fishing pier and race callers tower (Figure 4), the dam at the centre of the racetrack and the racetrack fences. These
provide enough spatial information for the participant to locate themselves and the virtual bushfire in relation to the
markers on the map and to the real world. The information phase of the experiment instructs about the bushfire but
also orients the participant in the Hanging Rock Reserve.
Figure 4: Left to Right: Hanging Rock; race callers tower to right of middle image and the Fishing Pier.
Phase 1: Information
Group A and B took separate routes around the racetrack and looked at separate sets of information media. Group A’s
route was to the west of the lake, Group B’s to the east (see Figure 5). Each group had to find their group’s three
markers, overlain on the Google Map of the Hanging Rock Reserve shown on the iPad New screen in 7scenes. The
Blue and Yellow bounded areas are where Phase 1 information is received at the markers for Group A and B
respectively. The first marker at the top shows the fire’s origin, the second its progression to the east of the lake and
finally at the third marker the wind change turns and the fire then heads to the east, instead of southwards.
At the shared green marker the participants are asked to predict which house(s) will burn after the wind change. The
orange bounded area is Phase 2 and the four House Markers can be seen from top right to bottom left (numbers 1-4).
The BBQ area which is the start and finish is marked by a blue sphere next to House Marker 4. The lake at the centre
of the racetrack shows as ochre colour on the satellite image dated from before the end of the last drought. During the
exercise it was full (see Figure 3 and 4).The slightly darker green of the racetrack can be seen around the yellow and
orange marker areas.
Figure 5: Overview of exercise area at 7scenes website.
The participants were asked to get as close as they could to the point at the base of the virtual markers, (Figure 5) in
order to see the media at the location in which the 3D scenes had been set in the Sandbox2 editor. They were expected
to orient themselves to the real world view of what they could see in the virtual view, whether that media was a
screenshot of the movie i.e. static view or the dynamic view in the movie itself. Some participants did not get close to
the specified markers, despite being asked to do so. However, participants were able to open markers from long range
after they had visited the location where the information was supposed to be viewed, enabling them to review
previously seen information again.
This first phase of CODE RED: MOBILE: the Information Phase, prescribed where and in what order markers were
visited. After the first day’s two sessions, several participants’ results had to be eliminated as they had viewed the
other group’s media. Subsequently participants’ carried a map explicitly showing Group A and Group B’s separate
three information markers and the order in which to visit them.
Phase 2: Decision Making
In the second half of the exercise in the Decision Making Phase participants had more freedom to decide their route as
the next set of markers were common to both groups. Their first common marker was a task marker which asked them
to go to the four markers representing virtual houses. These four virtual houses were located on a map of the Hanging
Rock area in the 7scenes application (7scenes.com). The houses were represented by the standard 7scenes markers.
Participants, reaching the real world locations indicated by a House Marker on the map, were presented with a
question on the iPad New, asking them if that house would most likely be burned down by the fire or not. Participants
typed in their answer, which was recorded to the 7scenes server, and retrieved by the researcher later. Participants
instantly received a score of 10 if they were correct, and then given a second chance if incorrect. They could then gain
a score of 5 if correct at the second attempt. However, in fact only the first 10 was counted for the purposes of the
experiment. In a final task participants were given a randomly sorted list, on paper, of the main events in the scenario.
They were asked to write next to the events listed, the actual numerical order of occurrence. This was scored for
accuracy.
The GPS Tracks
GPS tracks were recorded by Trimble Juno SB GPS handhelds and processed in Fractal 5.20.0 (authored by Vilis O.
Nams, 2010), using the FractalMean application to determine the Fractal D score. This score can assist in analysing
and evaluating participants’ and the exercise’s performance. Preliminary results are presented here.
The GPS tracks were prepared for FractalMean and for ArcMap 10 and Geotime 5.3 using Excel2007. Several of the
GPS had not worked properly or were accidentally turned off so some tracks were missing. 7scenes also records GPS
tracks for participants on the 7scenes server, using the GPS internal to the iPad New but this seemed to cut out more
than the Juno SB. Several participants had scores on the tasks but no usable tracks for one or both GPS recording
systems. The Juno SB GPS were turned on just as the participants left the start point but on several occasions people
were held back and tracks were recorded for several minutes, similarly at the end some GPS were not turned off until
several minutes after participants had finished. However the variations tended to be less than approximately 10 m as
they reflect the natural variation of GPS tracking. Fractal 5.20.0 was set to ignore GPS events at a scale of less than
10 m.
The exercise had to change start location at the last minute. The new location at the South East BBQ shelter,
unfortunately was very close to the House 4 marker and the GPS tracks do not distinguish the search for House 4 from
finishing at the BBQ shelter.
Fractal D
These analyses of the Fractal D score and spatial scale are preliminary findings for a limited set of data from the
CODE RED: MOBILE exercise.
The track for participant crm6 is formed of long straight tracks with occasional sharp changes of direction. This
produces an overall relatively high Fractal D of 1.3988 (V Nams 2012, pers. comm., 30 Oct.)
The Fractal D scores for many participants in CODE RED: MOBILE, were higher over the 100m spatial scale, thus
they walked in a meandering fashion on a large scale (see Figure 6 and 7). Below a hundred metres in spatial scale the
Fractal D was lower, thus on the small scale they mostly walked in less tortuous paths. This is in contrast to Kearns et
al.’s (2010) example of the long corridor, where a gallery visitor’s tortuous part of their track occurs while looking in
detail at the exhibits and the less tortuous path is along connecting corridors.
Nams and Bourgeois (2004) showed that the American Marten displays different Fractal D below and above a spatial
scale of 3.5 m. Below 3.5 m. the Marten went in straighter lines and above in more tortuous paths. This was directly
related to habitat. The A merican Marten went in straighter tracks where conifers were more plentiful in the
understory and where there was lesser canopy closure in the overstorey (p. 1744). The authors did not offer a reason as
to why the Martens’ tracks were more tortuous over a scale of 3.5m.
Figure 6: Fractal D of 1.3988 for participant: crm6.
Similarly, at least for the tortuosity of tracks for CODE RED: MOBILE, it would appear that in small areas
participants are not wandering much from side to side. At the small scale, participants were walking in straighter lines.
In contrast, with large areas and spatial scales, participants are walking in a more tortuous pattern. How do we account
for that?
Figure 7: In the bottom chart of the three, for participant crm6, the Fractal D score increases markedly above scales of 100 metres. The
horizontal scale is in metres. The Fractal D vertical scale is from 1 to 2.
It is proposed that this is perhaps like searching for a lost car in a very large car park. You walk to where you think the
car is and search there in detail, but if it is not there, you start searching much larger areas. The searcher has to go a
long way whilst undertaking a fairly detailed but as random as possible pattern of searching. The searcher will be
travelling at a large scale with a tortuous path thus a high Fractal D.
In CODE RED: MOBILE this tortuous long range searching may be due to the House Markers being in areas to the
east and south of the more open racetrack area. The area to the east with two House Markers consists of open fields
and can be relatively easily negotiated but gates through fences had to still be found. The final area to the south of the
racetrack contains the last two markers and is an area that is wooded and traversed by a small river. It contains several
dams or small lakes. It is a confusing area. Some of the paths are not well demarcated.
In the second decision making phase of the experiment there are long walks from one virtual House Marker to
another, but these four virtual houses are located in fenced off areas separate from the open area at the centre of the
Racetrack which includes the cricket oval at the south. As participants try to find the virtual House Markers, they have
to move from the more open area inside the racetrack to smaller fields and wooded areas. In order to do this they must
find gates or gaps in the fences. Sometimes these are hidden in the shade of trees. It is hard see gaps in wire fences at a
long distance away. Hence people sometimes have to walk a long way to find a suitable gap or gateway to get to
where they think the House Markers may be located.
It is proposed that participants at a large scale are, to some extent, walking tortuously in order to find their way
through gateways in fence lines. These are very hard to see on the Google Map satellite image in 7scenes and some
can be quite hard to find if you don’t know the Hanging Rock Reserve. In addition the racetrack fence and the
Hanging Rock Reserve fences provide a boundary to the exercise and participants have to turn back when they reach a
boundary fence. If a participant gets lost they will eventually run into a fence, here they realise they have to go back
again. The sharp turns add to the total tortuosity of the track. Larson-Praplan (2010 p. 51) found that cattle showed a
Fractal D change at a spatial scale of 200m showing the animals movements were constrained by the presence of a
fence.
At the small spatial scales of less than 100 metres, the tortuosity was low. The participants in the exercise did not have
to search very hard for the markers because they already knew where the markers were from the map on the iPad New
and could walk more or less directly to them. The majority of participants did try to get close to where the virtual
houses real world locations were indicated, but some did not. Participants could relatively easily find where the virtual
markers, marked on the screen, were in the real world and thus see at the real location where the bushfire was in
relation to that virtual house. They did not have to hunt for real houses visually as they did for the gateways. They had
the clues on the map for the markers but few clues about the fences and gates.
Fractal analysis may thus useful in understanding behaviours of the experimental trainees and also in finding aspects
of the exercise that can be improved. Additionally we can look at the tracks in Geotime 5.3, a Space- Time Cube or
spatial temporal analysis application (Hägerstrand 1970), to further understand and improve mobile training exercises.
Nara and Torrens (2007) used fractal analysis in combination with spatial temporal analysis of pedestrian flow.
Geotime 5.3
Geotime 5.3 has also proven useful in the analysis of participants’ performances and the overall exercise evaluation.
Figure 8: The blue to white shades shows a Heatmap of the tracks overlain by the GPS tracks.The non pink icons are participants who did not
go through the red shaded area which contained House marker 3.
The kernel density heatmap shown in Figure 8 was produced in Arcmap 10 and overlain with the original tracks as
black lines and then used as the base map for Geotime 5.3. Heatmaps show “hotter” colours where activity has been
more intense, white being the highest density, in this case obscured by the black line trace of the GPS tracks. This
overcomes one of the disadvantages of heatmaps in that individual participant’s tracks can still be seen especially
away from the main mass of participants. However, having tended to follow a common path; the black lines coalesce
somewhat obscuring the heatmap. The view in Figure 8 is from overhead in Geotime 5.3, and is termed the overhead
view. A 3D view is also possible.
The icons in Figure 8 representing the firefighters are all set to start at the same time. In reality they started at different
times and on various days. The icons are animated and a time slider allows the viewing of progress at one second
intervals. The red trapezium in Figure 8 is an Active Zone selection, a facility of Geotime that allows you to select
participants who were in an area for a set length of time. This also reveals those who did not go through that area. This
is useful for finding participants who did not go close to the House Markers, though they were asked to do so. The
scores of the participants on their predictions of which houses would not burn and on the recall of the sequence of
events can thus be compared to compliance with the rules.
In Figure 8 the three non-pink participants did not go to all four House markers. The grey (non- pink) participant
May13crm5 at the bottom left did not go to any of the house markers. However this firefighter scored very well in the
exercise tasks. May13crm5 also had a low fractal score. This firefighter possibly had weighed up the situation very
quickly, found and exploited the maximum distance at which one can access the markers, and finished the exercise
with a low expenditure of energy.
CODE RED:MOBILE delivered bushfire information to CFA volunteer firefighter participants in a mobile location
based scenario exercise so that most participants could successfully complete the exercise. Fractal analysis found that
participants’ movements at short spatial scales were different from movements at long spatial scales. It was proposed
this was due to the difficulties participants had in negotiating gateways and fence lines as shown by high fractal D
scores at large spatial scales. Low Fractal D scores for small spatial scales it was proposed was due to the ease of
finding the virtual markers which could be located using the map on the iPad New’s screen. Geotime 5.3 was very
useful for examining participant’s performance and experiences in the exercise. Participants who had taken short cuts
could be detected. Parts of the exercise that were not creating a great user experience could be found and provided
information that can be used to improve the exercise.
Conclusion
CODE RED: MOBILE was a successful exercise and fractal analysis together with Geotime 5.3 provided a means of
detecting and characterising detailed movements of participants together with some idea of their experience with
regard to navigation and decision making at Hanging Rock. The temporal component of geographic analysis was seen
to be well supported by these applications. It is likely that a wide range of activities in outdoor areas could benefit by
these kinds of analyses using Fractal D and Geotime 5.3, from team games to war games and traffic flow to pedestrian
movements.
References
Bishop, ID, Handmer, J, Winarto, A & McCowan, E 2011, Survival in Dangerous
Landscapes - A Game Environment for Increasing Public Preparedness. In: Buhmann, E,
Palmer, S, Tomlin, D & Pietsch, M (eds.), Preliminary Proceeding of Digital Landscape Architecture 2011. May 26-
28, Dessau, pp. 222-229.
CFA, 2011, viewed October 12, 2012 http://www.cfaconnect.net.au/news/cfa-trialling-new-digital-mapbook.html
de Jong, T 2005, ‘The guided discovery principle in multimedia learning’, in Mayer, RE (ed.), The
Cambridge Handbook of Multimedia Learning , Cambridge University Press, New York, pp. 215–228.
Hägerstrand, T 1970 ‘What About People in Regional Science’? Papers inRegional Science, vol. 24, no. 1, pp. 7-24.
Larson-Praplan, S 2010, ‘Modeling animal movement to manage landscapes’ PhD Dissertation, Oregon State
University, viewed October 15 2012 at
http://scholarsarchive.library.oregonstate.edu/xmlui/bitstream/handle/1957/14341/StephanieLarson2010All.pdf?seque
nce=1
Laurillard, D 2002, Rethinking university teaching: A framework for the effective use of educational technology (2nd
ed.),Routledge, London.
Mandelbrot, BB 1983, The fractal geometry of nature. W. H. Freeman and Company, San Francisco, California.
Milne, BT 1997, ‘Applications of fractal geometry in wildlife biology’, in Bissonette, JA, (ed) Wildlife and landscape
ecology: effects of pattern and scale, Springer-Verlag, New York, pp. 32–69.
Myerson, RB 1991, Game Theory: Analysis of Conflict, Harvard University Press, p. 1.
Nams, VO & Bourgeois, M 2004, ‘Fractal analysis measures habitat use at different spatial scales: an example with
American marten’, Canadian Journal of Zoology, vol.82, pp.1738–1747.
Nara, A & Torrens, PM 2007,'Spatial and temporal analysis of pedestrian egress behaviour and efficiency', In Samet,
H, Shahabi, C, Schneider, M (eds.) Association of Computing Machinery (ACM) Advances in Geographic Information
Systems, ACM, New York, pp. 284-287.
Pask, G 1976, Conversation theory: Applications in Education and Epistemology,Elsevier, Amsterdam and New
York.
Pfeiffer, VDI, Scheiter, H, Kühl, T & Gemballa, S 2011, ‘Learning How to Identify Species in a Situated Learning
Scenario: Using Dynamic-Static Visualizations to Prepare Students for Their Visit to The Aquarium’, Eurasia
Journal of Mathematics, Science & Technology Education, vol. 7, no. 2, pp.135-147.
Quinn, B & Cartwright, WE 2011, ‘Location based mobile games for learning and decision making’, Proceedings of
the 24th International Cartographic Conference, Paris, France: International Cartographic Association, Springer-
Verlag, Heidelberg, pp. 8.
Reiser, BJ 2004, ‘Scaffolding complex learning: The mechanisms of structuring and problematizing student work’,
Journal of the Learning Sciences: vol. 13, no.3, pp. 273-304.
Sharples, M 2005, Learning As Conversation: Transforming Education in the Mobile Age,
in: Proceedings of Conference on Seeing, Understanding, Learning in the Mobile Age, Budapest, Hungary, pp. 147-
152.
Webb, SL Riffell, SK Gee, KL & Demarais, S 2009, ‘Using fractal analyses to characterize movement paths of white-
tailed deer and response to spatial scale,’ Journal of Mammalogy, vol. 90, no. 5, pp. 1210–1217.