Emerging methods and technologies for tracking physical activity in the built environment

Full text

EMERGING METHODS AND TECHNOLOGIES FOR
TRACKING PHYSICAL ACTIVITY IN THE BUILT
ENVIRONMENT
Sean T. Doherty
Department of Geography & Environmental Studies
Wilfrid Laurier University
75 University Ave. West
Waterloo, Ontario, Canada N2L 3C5
sdoherty@wlu.ca
8TH INTERNATIONAL CONFERENCE ON SURVEY METHODS IN TRANSPORT
Annecy, France 25-31 May 2008
Resource Paper for Workshop on Understanding Relationships among Transportation
Infrastructure, Physical Activity and Health

Tracking Physical Activity in the Built Environment ii
Contents
Introduction ..................................................................................................................................... 1
Physical Activity Measurement ...................................................................................................... 3
Biochemical and Physiological Monitoring................................................................................ 4
Direct Bio-chemical Measurement ......................................................................................... 4
Heart-rate Monitors ................................................................................................................. 4
Biometric Monitoring Devices ............................................................................................... 5
Self Reported Physical Activity .................................................................................................. 6
Retrospective Questionnaires.................................................................................................. 6
Records/Diaries With a Physical Activity Focus .................................................................... 7
Records/Diaries with a Travel-Activity Focus ....................................................................... 7
Direct Observation of Physical Activity ..................................................................................... 9
Passive Motion Detection Devices ........................................................................................... 10
Pedometers ............................................................................................................................ 10
Accelerometers ..................................................................................................................... 11
Geo-Location Tracking ......................................................................................................... 12
Moving Forward: A Three-tiered Multi-Instrumented Approach ............................................... 14
Key Current Measurement Issues and Challenges ........................................................................ 18
Validation – An Immediate Need ............................................................................................. 18
Automated Physical Activity Detection Algorithms ................................................................ 19
Sorting out the Causation Dilemma .......................................................................................... 20
The Need for Cross Cultural Comparison ................................................................................ 20
Conclusions: New Opportunities at the Crossroads of Urban Studies and Health Sciences ........ 20
Acknowledgements ....................................................................................................................... 22
References ..................................................................................................................................... 23

Tracking Physical Activity in the Built Environment 1
“We do have a vision for a world in which people can walk to shops, school, friends’ homes,
or transit stations; in which they can mingle with their neighbors and admire trees, plants,
and waterways; in which the air and water are clean; and in which there are parks and play
areas for children, gathering spots for teens and the elderly, and convenient work and
recreation places for the rest of us.” (Frumkin et al., 2004, p. xvii)
Introduction
The built environment, in the form of land-use and transportation systems, has long been studied
with respect to impacts on human activity and travel patterns, accessibility, equity, safety,
environment, and congestion. More recently, we have become interested in how prevailing
trends in the built environment influence the physical and dietary activities we engage in, the
environmental hazards we face, the kinds of amenities we enjoy, and the resulting impacts on our
health. In particular, the sprawling low density suburbs that have dominated the landscape of
North American cities since the 1950’s, combined with new household technologies that require
less manual effort, appear to discourage physical activity, encourage auto dependence, limiting
opportunities for walking and cycling, limit accessibility to fresh food products for some,
encourage fast food consumption, limit social interaction, provide few natural amenities, and are
contributing to environmental problems; all of which have potentially adverse impacts on our
health. This is in stark contrast to a century ago when cities were compact, mixed-use, highly
walkable and accessible by public transit. Physical activity was embedded in everyday life, from
the chores we did to the walk trips to local food markets. It is no surprise then, that health
scientists and urban planners have turned increasing attention to the influence that the built
environment has on health in an attempt to establish designs that will return cities back to places
that encourage healthy behaviour.
Several recent books (Frank et al., 2003; Frumkin et al., 2004), committee reports (Dannenberg
et al., 2003; Humphrey, 2005; Transportation Research Board and Institute of Medicine of the
National Academies, 2005), special issues of journals (e.g. Heath et al., 2006, American Journal
of Health Promotion 2007 issue 21, American Journal of Public Health 2003 issue 93) and a
wide range of reviews (McCormack et al., 2004; Boarnet, 2005; Booth et al., 2005; Brownson et
al., 2005; Frank and Engelke, 2005; Handy, 2005; Davison and Lawson, 2006; Papas et al.,
2007), have firmly established the potential the built environment has to influence physical
activity patterns. However, it is widely recognized that the extent of this influence, and the
specific cause-and-effect relationships that exist, are relatively unclear, and depend to a large
extent on assessment methods and measurement issues. This state of affairs is no better
summarized than by Susan Hanson reacting to the ground-breaking National Research Council
report “Does the Built Environment Influence Physical Activity?: Examining the Evidence”
1
:
The committee’s findings can be summarized as follows: The built environment can facilitate or
constrain physical activity. The relationship is complex, however, and operates through many
mediating variables, such as socio-economic characteristics, personal and cultural variables, safety
and security in the built environment, and the individual’s decisions about time allocation.
Empirical evidence shows a linkage between the built environment and physical activity, but
causality has not been established, strengths of the relationships are not known, and the
characteristics of the built environment that are most closely associated with physical activity are
unknown. Also unknown are how the relationship between the built environment and physical
1
Produced by The Transportation Research Board (TRB) and Institute of Medicine (IOM) joint committee on
“Physical Activity, Transportation, and Land Use”.

Tracking Physical Activity in the Built Environment 2
activity varies by location (e.g., urban, suburban, rural) or by population subgroup (defined, for
example, by age, sex, race/ethnicity) or how important different characteristics of the built
environment are to total daily physical activity. The current literature reflects the lack of sound
theoretical frameworks to guide empirical research, inadequate research designs (e.g., most studies
are cross-sectional whereas longitudinal studies are needed to assess causality), and incomplete
data (e.g., national surveys on physical activity lack data on location, and national data on travel,
which do record location, neglect physical activity).
(Hanson, 2006, p. S259)
This extensive multi-disciplinary review clearly pits our lack of understanding of the impacts of
the built environment on physical activity on the lack of appropriate data collection, and focuses
the solution on the need to collect individual level data on all daily physical activities (not just
utilitarian travel) over long periods of time linked spatially to real-world characteristics of the
built environment in diverse settings (not just around the home), along with a wide range of
personal mediating variables (especially residential location preference).
The built environment is inherently distributed over space and connected by a transport network.
Wide scale availability of detailed land-use and transport network databases, combined with the
data management and analytical power of Geographic Information Systems (GIS), is allowing
clear progress to be made in capturing and analyzing objective data on the built environment (for
more details see Crane, 2000; Boarnet and Crane, 2001; Ewing and Cervero, 2001; Cromley and
McLafferty, 2002; Frumkin et al., 2004). GIS software allows points (landmarks, intersections,
etc.), lines (paths, walls, etc.), and areas (building footprints, natural areas, etc.) to be related to
each other spatially. Each feature can have an associated set of attributes documenting such
characteristics as types, sizes, purposes, flows, census data, etc. Aggregate features of
neighborhoods or zones such as land-use density, network connectivity, and walkability can be
calculated. To get a sense for what types of data are available and their accuracy, the layperson
need only check out on-line maps such as Google maps. Planning agencies, universities,
governments and other research institutes will have available additional data sets. However, a
key issue will be balancing use of these existing data sources with the desire/need for more
detailed or completely new features of the built environment at finer scales, including especially
more qualitative information (e.g. how well lit or safe streets are). Regardless of data source,
GIS represents the most promising technology for managing, analyzing and comparing
characteristics of the built environment.
The same cannot be said of physical activity, for which a more diverse history of data collection
methods exists, and continues to evolve owing to a variety of quickly emerging wearable
technologies. At present, no “gold standard” method has emerged for assessing physical activity
type and intensity under the real-world conditions of the built environment; in fact, most
methods have barely been tested outside of the lab, and those that have tend to experience
significant drops in accuracy and reliability. This resource paper provides a review of these
diverse methods and emerging technologies in an attempt to narrow the focus and identify the
most promising methods. Methods from both “urban studies” (geography, planning, civil
engineering, etc.) and the “health sciences” (kinesiology, physiology, public health, etc.) are
intentionally covered and contrasted, in particular to highlight means to learn from each other
and integrate the best of both worlds. Extensive illustrative examples are provided
2
as a resource
2
The author felt it was important at least for the initial conference draft for the varied methods to be illustrated in
detail as a “Resource” for readers. Any published version of this paper would have to be much more selective.

Tracking Physical Activity in the Built Environment 3
for those unfamiliar with each method and as a means to identify their differences and
similarities. Follow-up readings are provided in all cases for readers who want more detail on a
specific method. Based on this review, an integrated methodology drawing upon emerging
techniques from both fields is proposed as most viable for the future studies. Lessons learned
and key methodological challenges for the future are then posed in the conclusion.
PhysicalAct i vityMeasurement
Physical activity (PA) has been defined as “any bodily movement produced by skeletal muscles
that results in caloric expenditure” (Welk, 2002). It is most commonly characterized by:
1. Type
2. Duration
3. Intensity
Intensity can be a simple rating (low, moderate, high) or expressed more technically in terms of
energy expenditure (EE). EE is most often measured in metabolic equivalents (METS), wherein
one MET represents resting energy expenditure (about 3.5ml/kg/min of oxygen consumption)
and physical activity typically raises this by several magnitudes. EE can also be expressed in
terms of calories burned. EE can be measured directly, but is often estimated based on the
duration and type of physical activity using established classification schemes (most common,
that established by Ainsworth et al. 2000). However, as noted by Ermes et al. (2008, p. 20)
“energy expenditure is only one important aspect of physical activity ... a more detailed analysis
of physical effort can be obtained by activity recognition, i.e., by detecting the exact form of
activity the subject is performing.”
Whilst the importance of physical activity is well established, more accurate and practical
measures are still being sought by health professionals in order to better understand the specific
amounts/types that are needed for health benefits (Welk, 2002) and by a growing number of
urban planners/researchers in order better understand the impacts of the built environment.
Particularly challenging have been the attempts to develop accurate, valid, and cost-effective
techniques to quantify PA outside of the lab under free-living conditions. Overall, the more
accurate the measurement of physical activity, the more power researchers have to detect
changes or differences across various factors (which are many), and the smaller the sample
needed to test the relationships.
Based on review of methods from published studies of physical activity, five main types of
physical activity measurement methods are evident:
1. Biochemical and Physiological measurement of energy expenditure (METs) using
techniques such as doubly labeled water or indirect calorimetry, heart monitors, or skin
responses.
2. Self-reported activities and travel, wherein reported activity or travel types, intensities
and durations are converted to episodes of physical activity and/or estimated METs using
classification schemes.
3. Direct Observation, analogous to self-reported surveys but completed by an independent
observer.

Tracking Physical Activity in the Built Environment 4
4. Passive Motion Detectors, with varying capacities for providing objective measures of
activity type, intensity, and duration; includes pedometers, accelerometers and GPS.
5. Integrated/hybrid approaches, combining two or more of above.
Each of the above methods is reviewed in detail in the following sections; some more than
others, depending on their potential. In all cases, follow-up readings are provided as an
additional resource for readers with specific interests.
BiochemicalandPhysiologicalMonitoring
DirectBiochemicalMeasurement
Biochemical measurement involves measurement of chemical processes of the body. Physical
activity involves energy expenditure which is directly related to the consumption of hydrogen
and oxygen, carbon dioxide production, heat production, and other chemical affects. Several
techniques have been developed for measuring these outcomes, and represent the most direct
measurement of energy expenditure. The most common are through Doubly Labeled Water
(DLW) and Indirect Calorimetry (IC). The specifics of how these techniques work are covered
in Starling (2002). Briefly, IC involves analysis of respiratory gases captured through a
facemask, canopy or whole room analysis. DLW involves analysis of the elimination of
hydrogen and oxygen from urine samples. Whilst these methods are the most accurate measure
of physical activity energy consumption, they have tended to be too expensive and inconvenient
for use in free-living conditions (Dale et al., 2002; Boarnet, 2005), and typically do not measure
the qualitative nature of physical activity (such as type). For this reason, they are normally only
used in a lab setting with small samples of subjects as a benchmark or “gold standard” from
which to assess the validity the other measurement devices (Starling, 2002).
However, at least one new device is breaking cost, size and usability barriers - the MedGem
Indirect Calorimeter, shown in Figure 1. Users basically breath into the portable device for up to
10 minutes (much shorter than the average 30 minutes of past devices), which then measures
oxygen consumption and estimates EE in the form of calories burned. Although recent studies
have found it to be a reliable and
valid instrument for the measurement of EE (Rubenbauer et al.,
2006; Fields et al., 2006 ; Cereda et al., 2007; McDoniel, 2007), just as many studies appear to
question the devices validity and usability (Hlynsky et al., 2005; Reeves et al., 2005; Van Loan,
2007; Fares et al., 2008) suggesting that there is not enough evidence quite yet to support their
clinical usage. Additionally, there does not appear to be any published applications of this
device for assessing the types and duration of physical activity, which presumably would require
cumbersome repeated measurements. Thus, until such devices can provide continuous
monitoring capabilities, its application is likely limited to serving as a field-based validation
device, albeit more convenient and inexpensive than traditional IC methods.
HeartrateMonitors
The development of light-weight, continuous heart rate data recorders (as small as a watch) have
improved the ability to detect the intensity, duration and frequency of physical activity. Modern
heart monitors use the electrocardiogram (ECG) signal to detect heart rate, via a transmitter
attached to a chest strap with integrated electrodes (example shown in Figure 2). Heart rate has
been shown to increase in a predictable time-lagged linear fashion with moderate to vigorous

Tracking Physical Activity in the Built Environment 5
aerobic physical activity, but can be poorly correlated with light and sedentary activity (such as
TV watching or walking) or non-aerobic activity (for review, see Janz, 2002). Heart rate is also
known to be affected by many other confounding factors, such as type of physical activity,
physical fitness, hydration, caffeine usage, high temperatures, and emotional states. Because of
these shortcomings, many researcher either dismiss them as impractical (e.g. King et al., 2004)
or combine them with accelerometers to improve prediction of EE (see also Strath et al., 2000;
Brage et al., 2004).
BiometricMonitoringDevices
One of the latest technologies for estimating energy expenditure from physical activity is through
biometric/metabolic monitoring devices. The only commercially available device assessed
extensively in the literature appears to be the SenseWear Armband (SWA), shown in Figure 3.
This small wireless monitor worn on the arm for up to five days collects minute-by-minute: (1)
movements by a 3-axis accelerometer; (2) heat flux by a thermocouple array; (3) skin
temperature; and (4) galvanic skin response by two electrodes (an indicator of evaporative heat
loss). Using proprietary algorithms, the manufacturer’s software calculates and displays a
subject’s daily total EE (calories burned), number of steps, and duration of physical activity and
sleep. Although only daily totals have been reported in the literature
3
, it appears to have the
capacity to provide highly detailed minute-by-minute reports.
Based on an earlier report by the manufacturer (Liden et al., 2002) and information on their
current web site (www.bodymedia.com), the prediction algorithm appears to first identify the
subject’s “context” (sleeping, resting, walking, driving a car), followed by estimation of calories
burned (EE) in these contexts as a function of the accelerometer and heat sensing components
(plus “other” unreported parameters). This is deemed particular important for distinguishing
four key scenarios: 1) low motion high heat flux activity (e.g. weight lifting); 2) low motion low
heat flux (e.g. sitting a desk, watching TV); 3) high motion high heat flux (e.g. running); 4) high
motion low heat flux (e.g. driving a car). This addresses the shortcomings of using motion
detectors alone (e.g. accelerometer), as highlighted by the manufacturer:
“Other devices currently being used for free living energy expenditure monitoring are not able to
detect contextual differences, and, therefore, are not able to utilize such information in their
calculations. As described previously, motion detectors, pedometers, and accelerometers share the
disadvantages of being subject to the detection of false motion and the inability to accurately
detect nonambulatory physical activity. Particularly critical to the accuracy of accelerometer based
energy expenditure models are those times when subjects are traveling in motor vehicles. At these
times the rapid and continued motion of the accelerometer may equate to high-energy expenditure.
While the vehicle is expending high energy, the subject inside the vehicle is not. By incorporating
a “motoring” context into an energy expenditure algorithm such false detection can be identified
and corrected and, therefore, overall accuracy greatly improved. There are other significant
contexts such as sleeping, resting, and walking, which, if known, can improve the accuracy of
energy expenditure calculations.” (Liden et al., 2002, p. 12)
Further testing by Fruin and Rankin (2004) and Jakicic et al. (2004) showed that the first version
of the SWA provided reasonably reliable estimates of EE compared to indirect calorimetry, but
that some activities EE were significantly under estimated by as much as 17%, such as flat
walking, stepping and cycling exercise, whereas others were overestimated such as arm exercise
3
See http://www.sensewear.com/tutorial/WMS_tutorial_2008_HP.html for details on device operation capabilities.

Tracking Physical Activity in the Built Environment 6
(+29%). Other have found that the SWA can be significantly inaccurate for certain population
groups such as the obese (Papazoglou et al., 2006; Malavolti et al., 2007) and children
(Arvidsson et al., 2007; Dorminy et al., 2008). Jakicic et al. (2004) tested improved context-
specific algorithms developed by the manufacturer and found they produced equivalent EE
estimates to calorimetry for all activities. Similarly, Dorminy et al.(2008) found that a simple
adjustment for subject weight significantly improved the SWA-predicted EE.
Overall, like the MedGem Indirect Calorimeter, independent testing of this device has been
largely limited to its capabilities in assessing total EE rather than its ability to detect activity
types and duration (although it appears to have this ability). Additionally, one of the complaints
about the SWA from researchers (e.g. see Papazoglou’s response to the manufacturer’s editorial
in Andre, 2007) is that the details of the various algorithms and refinements are not revealed
4
.
Moving forward may require more open testing of the device.
SelfReportedPhysicalActivity
Self-reported instruments rely on subjects’ ability to recall the types, intensities, duration, and
patterns of their own physical activity. They have been used extensively in the health sciences
(for reviews, see Montoye et al., 1996; Sallis and Saelens, 2000; Matthews, 2002) and urban
studies (for reviews, see Richardson et al., 1995; Stopher and Jones, 2003). Although many
variations exist, two basic types are evident: retrospective questionnaires and diaries.
RetrospectiveQuestionnaires
Retrospective questionnaires pose a series of questions concerning a person’s “typical” physical
activity patterns in the recent past such as last week or month, by type, intensity, frequency, and
duration. An illustrative example is provided in Figure 4 drawn from the California Health
Interview Study. Note how each question systematically targets a specific physical activity and
intensity level followed by a query for duration. Other examples include the Behavioral Risk
Factor Surveillance System (BRFSS) which measures the frequency and duration of leisure-time
and occupational physical activity via phone interview; and the Youth Risk Behavior
Surveillance System (YRBSS) that measures number of days per week with at least 20/30
minutes of vigorous/moderate exercise
5
. A thorough “Collection of Physical Activity
Questionnaires for Health-Related Research” is provided in volume 29(6) supplement of the
Medicine & Science in Sports & Exercise journal. Other major national/regional surveys are
reviewed in Boarnet (2005), whereas detailed assessments of validity can be found in Sallis et al.
(2000) and Matthews (2002).
Overall, such studies tend to vary by target group (e.g. adult, children, elderly) and the physical
activity categories of interest (occupational, household, sports, school-related, conditioning,
transportation, recreational activities, etc.). The specific location of physical activities is not
normally obtained, likely in part because of the awkwardness of asking location for “typical”
behaviours. In lieu of this, aspects of the built environment surrounding a subject’s home
location are most often the focus of analysis.
4
An early report by the manufacturer (see Liden 2002) provides some conceptual level details on algorithm
structure and development, data collection, context detection, and accuracy.
5
Administered by the Centre for Disease Control in the U.S. See www.cdc.gov/brfss/about.htm or
www.cdc.gov/nccdphp/dash/yrbs/index.htm
.

Tracking Physical Activity in the Built Environment 7
Records/DiariesWithaPhysicalActivityFocus
Rather than “typical” behaviours, diaries are meant to capture a subjects “actual” episodes of
physical activity as they occur during the day (or record them at regular intervals such morning,
noon, and night), by types, intensity, start/end time and/or duration. They tend to capture more
detail about a diverse range of physical activities in daily life compared to retrospective
questionnaires. They can also be easier for subjects to recall without the cognitive effort
involved in generalizing about “typical” behaviours over long periods.
There are a variety of different diary formats, some with more/less detail, as shown in Figure 5
through 9. Stel et al.’s (2004) seven-day Activity Diary (Figure 5) was completed every evening
by the respondents, and covered the duration of walking inside and outside, bicycling, gardening,
light household activities, heavy household activities, and sport activities. Weston et al.’s (1997)
Previous Day Physical Activity Recall (Figure 6) involved recall of the previous day's activities
and subjective intensities in 30-min intervals by primary activity type (eating, sleeping/bathing,
transportation, work/school, spare time, and physical activities) and sub categories. Bouchard et
al.’s (1983; 1997) 3-Day Physical Activity Diary (Figure 7) is similar, but broken into 15 minute
intervals and using a different activity categorization scheme. Wormald et al.’s (2003) physical
activity diary (see Figure 8) asks participants to record everything they do from getting up in the
morning to going to bed at night, for seven days. The actual start and finish time of each activity
is recorded and a descriptive measure of physical activity intensity is indicated by circling a
marker. Participants were instructed to fill in the diary four times per day. A simple diary for
home use suggested by the Heart and Stroke foundation is provided in Figure 9. Methods for
operationalizing diaries of this type on hand-held computers or smart-phones will most likely be
the next generation of such designs, especially as it could simply presentation of the often long
categorical lists of physical activity types as fly-out or drop-down lists.
In health studies, the activity types/intensities and durations collected from the above surveys are
used to estimate energy expenditure using established activity-specific rates of energy
expenditure multiplied by the duration. The most common and extensive coding scheme appears
to be the compendium provided by Ainsworth et al. (2000), which includes a very highly
detailed listing of 650 specific physical activity types in 21 major categories (see Figure 10).
However, even these authors acknowledge that such tables do not take into account individual
differences in body mass, body fat percentage or fitness level, or geographic environment that
are known to alter energy expenditure, and that no such general correction is available.
Records/DiarieswithaTravelActivityFocus
More recently, attempts have been made to elicit physical activity from daily activity and travel
diaries deployed in urban and time-use studies. An extensive literature on such methods exists,
exploring the variation in methods, administration, conduct, data quantity/quality,
standardization, and use of emerging technologies (Richardson et al., 1995; Stopher and Jones,
2003). Generic to such techniques is the capturing of a listing of self-reported daily events
(activities, trip) for one or more days plus their attributes. Attributes normally include at least
activity type (by main and sub-categories which vary widely), travel mode (auto, transit, walk,
bike, etc.), start/end time, and location and. A wide range of additional attributes tailored to the
particular study purpose are also collected (involved persons, costs, frequency, when planned,
flexibility, etc.). An example trip diary is shown in Figure 11, as compared to the activity and

Tracking Physical Activity in the Built Environment 8
travel diaries shown in Figure 12, and an activity and time-use diary in Figure 13. Note that
most recent large scale surveys of this kind are often conducted via computer assisted telephone
interview (CATI).
The capturing of location characteristics is what distinguishes these diaries from those adopted in
the health sciences, often including specific placenames and geo-coordinates (longitude and
latitude). For this reason, daily events can be precisely associated with characteristics of the
built environment. However, in practice, complete details on the types, intensity and duration of
physical activity have not been captured in as much detail as in health studies; although
theoretically they could in the form of more specific activity types and additional attributes.
What can typically be identified is utilitarian walking and bicycling trips by duration (but not
usually intensity), along with some capturing of recreationally active activities by duration (e.g.
dog walking, jogging). Much depends on the detail in activity/trip-purposes types captured, and
to a lesser extent the location. As found by Forsyth et al.(2007) focus only on walking travel
without full accounting of leisure/recreational walking and other exercise may unjustly lead to
associating land-use attributes and exercise.
A recent example that that goes beyond use of just walk trips is by Copperman and Bhat (2007).
They used the 2000 San Francisco Bay Area Travel Survey (BATS) to examine how physical
activity of children is related to the built environment. Whilst it was straightforward to identify
utilitarian trips involving physical activity by mode type (i.e. walking, cycling vs. auto modes),
they also identified two other types of physical activities through a combination of activity type
and location:
Recreationally physically active activities – recreational coded activities that occurred at
locations assumed to be associated with physical activity. Of the 450 location categories,
sixteen were identified as active locations (Golf Courses, in/outdoor Sports Facility,
Dance Studio, Miniature Golf, Bowling Alley, Ice Rink, Public Park, Health Club,
Gymnasiums, Ski Resorts, Campgrounds, Sports Camp, Skate Park, Martial Arts Center,
Swimming Pool).
Physically active recreational travel - non-motorized travel episodes that begin and end
at home without any stops in-between (for example, walking or bicycling around the
neighborhood).
Whilst 19% of children in their sample participated in utilitarian activity travel, a further 16%
and 3% (respectively) participated in these latter categories, effectively doubling the amount of
data available for analysis. The authors recognize that “a limitation of our classification
procedure is that it is solely based on the location type of activity participation, and does not
consider any measure of the physical intensity level or nature (structured versus unstructured) of
activity participation” (Copperman and Bhat, 2007, p. 69) and that such an approach ignores
physical activity associated with other types of activities, such as during free play, because they
cannot be identified in the survey data.
Recognizing these limitations, and the significant impact they have on validity, others are turning
to time-use studies that provide much more specific detail on activity types. For example, Sener

Tracking Physical Activity in the Built Environment 9
et al. (2008), used the Time-use Diary shown in Figure 13 that detailed 28 types of “Active
Leisure, Sports, and Exercise”, ten types of “Attending sports events” (so as not to confuse them
with activity sports), 12 types of game playing, 24 types of passive leisure, 12 types of household
chores, and a range of travel codes (Walking, Hiking, Jogging, etc)
6
. Sener et al. aggregated
these into 8 distinct activity categories for analysis, by location (in/out-of-home), day of week
(weekend/weekday), structure (organized or unstructured), and intensity of physical activity
(active or passive). Note how physical activity is measured in only two categories, in parallel to
the focus in health studies on estimating EE more explicitly.
Overall, existing activity-travel diaries used in urban planning arena may be useful for assessing
the physical activity involved with utilitarian walking and cycling trips, and with some
assumptions, some physically activity recreational activities; however, it should be noted that the
vast majority of American children and adults report very few (if any) walk/cycling trips to begin
with, thus providing a very limited perspective on their overall physical activity levels. The type
and intensity of physical activity assumed associated with other activity types (e.g. recreational
activity) has obvious limitations with respect to actual intensity and duration; even further, the
physical activity associated with the many other activities people conduct on a daily basis is
completely missed, such as in-home or at school. The sum of these concerns suggests that a
priority in the near future would be to assess the validity of such an approach for assessing
overall physical activity levels; until then, sophisticated modeling using such data and the
drawing of strong conclusions about the association to the built environment from such studies
should be viewed with some caution.
A third less obvious methodological contribution from urban studies is the recent evolution of
activity-scheduling surveys that target both observed activity-travel patterns and underlying
scheduling decisions processes (e.g. Doherty and Miller, 2000; Doherty et al., 2004; Lee and
McNally, 2006; Ruiz and Timmermans, 2006; Zhou and Golledge, 2007). Using computerized,
internet, and hand-held technologies, these surveys attempt to capture when, how, and why
activities are planned, modified and executed over time, space and across individuals, leading up
to their observed outcome. On the surface, they appear much like a standard activity diary to
subjects, but involve more in-depth querying and decision tracking typically over multiple
survey days. Including a focus on physical activity in such surveys could provide wholly new
insights into how it is planned and managed within a person’s everyday life, barriers to its
completion, and identification of ways to “Fit physical activity into your schedule” (see Figure
9).
DirectObservationofPhysicalActivity
Direct observation involves a second party shadowing/following subjects and making records of
their behaviour. Although regarded as labour intensive and tedious, direct observation is a low-
tech method that exceeds other measures of physical activity in providing contextually rich
quantitative and qualitative data, especially with regard to the what, when, where and with whom
it occurs (McKenzie, 2002). In the health sciences, this method has been almost exclusively
applied to the study of children and adolescents in specific locals, including at school, during
education classes, leisure time, and at home. In a lab setting this method is often used to provide
6
A particularly unique interactive activity type tree diagram is at
http://psidonline.isr.umich.edu/CDS/time_diary/activity_tree_view.html

Tracking Physical Activity in the Built Environment 10
objective baseline measures of physical activity for validation or algorithm development
purposes. For instance, Ermes et al. (2008) had research assistants use the hand-held computer
program shown in Figure 14 to record the activities of subjects doing a specified set of indoor
and outdoor physical activity tasks (sitting, stationary exercise, walking, running, cycling,
playing sports, etc.). Ermes et al. also experimented with having subjects use the same device to
self-report their activities under free-living conditions. The result is not dissimilar to what would
be produced from the diary methods reviewed above. The difference is in intended application –
in this case, the physical activity events were plotted on a time line along side accelerometer and
GPS speed data (see Figure 21) for comparison and algorithm development purposes (see also,
Moving Forward: A Three-tiered Multi-Instrumented Approach section below) rather than for
large sample assessment of physical activity.
Although applied infrequently, direct observation is an important method to consider in context
of this review. Firstly, it demonstrates the depth to which researchers will often go to obtain
objective data on physical activity, especially for validation purposes – a crucial step in the
application of new technologies and development of automated detection algorithms. Perhaps
more importantly, an observer can make detailed notes not only on the location of a person’s
physical activity, but on significant interactions with the built environment, such as with
facilities, equipment, playgrounds, sidewalks, natural features, etc. Additionally, an observer can
make detailed notes on the quantity and quality of the social context of physical activity - beyond
just “who with” type questions - including observation of strong and weak ties, involved person
characteristics, the nature of interaction, etc. And unlike all previous methods that focus on
individuals, direction observation can be used to study groups of people in specified locals– such
as in specific neighborhoods. For instance, Figure 15 shows a form used in the SOPLAY
(System for Observing Play and Leisure Activity in Youth) to record youth activities in specified
time intervals (McKenzie et al., 2000; McKenzie, 2002). An updated version can be found in
McKenzie et al. (2006).
PassiveMotionDetectionDevices
A vast array of passive wearable monitoring devices attempt to provide more objective
measurement of physical activity including under free-living conditions. They are capable of
measuring one or more of motion, acceleration, position, and speed which are then analyzed to
estimate aspects of physical activity including type, intensity, durations, and energy expenditure.
For the most sophisticated devices, they can even be used to analyze posture and gait.
Pedometers
Pedometers are small inexpensive devices that have gained widescale acceptance among
physical activity researchers for measuring steps from walking or running (Bassett and Strath,
2002). Normally worn on the waistband, pedometers contain a horizontal spring-suspended
lever arm that moves up and down in response to vertical acceleration, as shown in Figure 16.
Estimates of distance and speed can be made if stride length is known. Extensive validity and
reliability tests have been conducted on a wide range of devices (e.g. Bassett and Strath, 2002;
Dale et al., 2002; Marsh et al., 2007). The best brands are fairly accurate at counting steps, but
are less accurate for estimating distance and intensity/energy expenditure. Like other waist-
mounted devices, they cannot detect other types of physical activities such as arm movements or
external work, and may incorrectly count steps when users are in a moving vehicle. They also

Tracking Physical Activity in the Built Environment 11
do not work well for certain groups, such as the frail who walk at slow speeds or obese subjects.
For these reasons, their application is limited.
Accelerometers
Accelerometers are one of the most popular devices for assessing physical activity in free-living
conditions owing to their small size (no larger than pedometers), noninvasive nature, low
respondent burden, and ability to provide objective measures of both physical activity
intensity/EE, duration and type over long periods of time. A good overall introduction to these
devices and their validity is provided in Welk (2002); the physics and measurement principles
are covered well in Chen and Bassett (2005); comparison of some of the most popular devices is
covered in (King et al., 2004). In general terms, an electronic component in the device measures
acceleration of the body-part it is attached to in one or more specific directions (up/down, side-
to-side, back-forth) at frequent sub-second intervals. Acceleration is defined as the change in
speed with respect to time; if information on stride length is available, the data can be post-
processed into speed and distance values. “Uniaxial” accelerometers measure acceleration in the
vertical plane, whereas “biaxial” and “triaxial” accelerometers measure movements in two and
three dimensions respectively. Technically, note that accelerometers do not measure overall
body acceleration, only acceleration of specific body part it is attached to (typically hip, but also
arms, legs and feet) and that signals during steady-states movements (such as in a car) will
typically be lower in magnitude. If worn on only one body part (e.g. hip), it will be incapable of
detecting all body movements.
A wide range of literature reports on ability of accelerometers to detect the type and intensity of
physical activity (see reviews in Welk, 2002; Aminian and Najafi, 2004; King et al., 2004;
Mathie et al., 2004; Chen and Bassett, 2005). A significant number of laboratory studies have
demonstrated accurate linear relationships between accelerometer “counts” and EE during
physical activities such as walking and running, but frequent underestimation of EE (as high as
67%) during free-living conditions (Welk, 2002; King et al., 2004). The development of
mathematical models of energy expenditure from accelerometer data is an on-going effort (Chen
and Bassett, 2005). In terms of activity detection, Ermes et al. (2008) points out that
accelerometer data has been used to automatically detected physical activity types in a lab
setting, but that long-term out-of-lab monitoring potential is not as clear or well tested.
Recognizing the various limitations of using single accelerometers, especially for free living
conditions, emerging approach are involving application of multiple sensors (e.g. on hip and
other limbs), or combination with other physiological measurements (e.g. heart rate). In a short
out-of-lab experiment with 24 subjects, Foerster et al. (1999) placed accelerometers on subjects’
sternum, wrist, thigh, and lower leg and had an observer take notes on nine physical activity
patterns conducted by subjects in a lab and then under free-living conditions (sitting, standing,
lying, sitting and talking, sitting and operating PC, walking, stairs up, stairs down, and cycling).
Based on this data, they developed an algorithm for predicting activity type obtaining an overall
accuracy 96% prediction in the lab, but only of 67% under free living conditions. Similarly,
Parkka (2006) developed a detection algorithm capable of detecting eight physical activities
(lying, rowing, cycling, sitting , standing, running, Nordic walking, and walking) in a supervised
free-living scenario, obtaining accuracy of 58-97%. Two key challenges to detecting physical
activity type relate to the increased complexity/diversity of physical activity in daily life, and less

Tracking Physical Activity in the Built Environment 12
control over being able to accurately “annotate” actual physical activity for verification and
algorithm detection purposes.
The most advanced off-the-shelf multi-accelerometer device appears to be the IDEEA
(Intelligent Device for Energy Expenditure and Activity) portable monitor which uses five
integrated thumbnail sized sensors attached with tape to the chest, midthigh of both legs, and on
the soles of both feet (see Figure 17). The sensors measure angles of body segments and
accelerations in two orthogonal directions 32 times per second. Software that includes a neural
network model outputs estimated EE, speed and distance, and an activity code that specifies the
body position and activity being performed. This included 5 primary postures (Sitting, Standing,
Leaning, Lying down, Limb movement), 27 secondary postures and limb movements, and 5 gait
types (Walking, Running, Climbing stairs, Descending stairs, Jumping). Zhang et al (2003;
2004) demonstrated breakthrough accuracy in a lab setting, finding that the IDEEA can detect
the type, onset, duration, and intensity of most fundamental movements with 98% accuracy, near
perfect speed measurement, and EE estimates within 95-99% accuracy compared to bio-chemical
measurement (calorimeter and a metabolic chamber). Similar results have been found in
subsequent studies under free-living conditions (Huddleston et al., 2006; Gardner et al., 2007;
Maffiuletti et al., 2008). In comparison to the SWA armband monitor (see above) Welk et al.
(2007) concluded that both are well suited for detecting free-living activity types and EE, but the
SWA was less invasive and cumbersome for long-term monitoring compared to the multiple
sensors and wires required of the IDEEA monitor.
GeoLocationTracking
In the last decade, several geo-location tracking technologies have become available that are
small enough to be worn by a person, and provide accurate and frequent enough information on
their location (typically longitude, latitude, altitude) to allow analysis of motion and speed. The
most popular are Global Positioning System (GPS) devices that receive information from a
system of 24 satellites orbiting at high altitude allowing location to be calculated using
triangulation (for a basic overview, see Bajaj et al., 2002). Others include: land-based tracking
systems (e.g. Time Difference of Arrival – see Shoval 2006); cellular phone network-based
location estimation (Asakura and Hato, 2004; Sayed et al., 2005); WiFi fingerprinting (Retscher,
2007); and to a limited extent RFID tags (radio-frequency identification) that have shown some
potential for indoor navigational aid (Szeto and Sharma, 2007). Various combinations of these
technologies are also being experimented with to improve accuracy in and outdoors, such as
Assisted-GPS (GPS + Network-based location), GPS + WiFi (Retscher, 2007), and GPS + dead-
reckoning via accelerometer. The most widely applied of these technologies to date is GPS,
owing to its small size, accuracy, tracking frequency, low cost, and wide-scale availability.
Many small off-the-shelf GPS receivers or GPS-enabled smartphones are now available that can
track and store location to within 2-3 meters accuracy on a second-by-second basis over multi-
day periods.
Early studies of the use of GPS in assessing physical activity focused on speed, distance, and
gait. GPS receivers were found to provide highly accurate measure of the speed of walking,
running and cycling under open-sky test conditions to within about 0.08 kms/h (Schutz and
Chambaz, 1997; Schutz and Herren, 2000; Witte and Wilson, 2005; Townshend et al., 2008), and
distance to within about 3 meters (Rodriguez et al., 2005). Terrier et al. (2001; 2005) found that

Tracking Physical Activity in the Built Environment 13
differentially corrected GPS with sub-centimeter accuracy is even capable of recording small
body movements and thus a promising tool for gait analysis.
The further potential of geo-location technologies for assess physical activity becomes apparent
when examining traces of personal or vehicular movements over long periods of time (e.g.
Murakami and Wagner, 1999; Wolf et al., 2003; Lee-Gosselin and Harvey, 2005; Stopher and
Greaves, 2007). When viewed on a map (see Figure 18 through 20 for examples), physical
activity types/intensities and their location become much more apparent. For example, Elgethun
et al. (2003; 2007) overlaid children’s GPS traces with aerial photos in a GIS and found they
could manually determine subjects’ location (inside, outside, in-building) and distinguish a
variety of human activity types such as entering a retail store, walking on a sidewalk, traveling
by car or bus, playing on a schoolyard, or playing in and around a house. Interestingly, when
compared to parent-reported child activities via a manual diary, they found that parents
misclassified 48% of their child’s time. Rodriguez et al. (2005, p. s579) adopted a similar
approach but used land use, road network, building footprint, and aerial photos, as shown in
Figure 18. They were able to develop automated rules that determine the location of “bouts” of
continuous physical activity (measured via accelerometer), classed as indoors, outdoors in the
neighbourhood, or outdoors out of the neighbourhood. They found that 40.7% of bouts of
physical activity had no GPS data, 14.4% had less than 30% of data, and 44.9% had more than
30% GPS data sufficient for accurately classifying the location. They suggest that increased
accuracy may be gained by examining first and last successfully recorded GPS coordinates
around a bout of physical activity to make inferences about its location.
One of the first attempts to explicitly validate GPS as a potential means to automatically detect
walking versus resting episodes was reported by Le Faucheur et al. (2007). Subjects were given
prescribed walking and rest routines in a park dictated to them orally via earphones, whilst being
tracked with an off-the-shelf GPS receiver/logger with 1-3m accuracy. They found initially that
the GPS provided artificially high speeds on occasion (at start of walking, at other random times)
and seldom provided true zero speeds during rest. However, with some automated data
processing to correct these artifacts, 90% of walking and resting bouts could be automatically
detected. A manual post-processing methodology using graphic analysis reached an accuracy of
97%.
Although not explicitly focused on physical activity, emerging from urban studies are additional
attempts to use GPS to detect walking and cycling versus auto modes (Chung and Shalaby, 2005;
Stopher et al., 2005; Stopher et al., 2008). Chung and Shalaby (2005) collected GPS traces for
60 simulated trips by bus, car, and walking, and kept a separate record of actual start/end times,
mode, and route. After filtering out invalid or poor quality data points, they developed
sophisticated link-matching techniques in a GIS to identify routes, and automated rules for mode
detection based on speed, speed range, proximity of trip ends to transit stops. They report 79%
classification accuracy on routes, and 92% of modes, using the same calibration data as test data.
Recent work by Tsui and Shalaby (2007) has refined these techniques for incorporation into a
web-based prompted recall diary (see also Figure 23b), reporting 100% detection of stationary
activities, and 91% of travel mode (based on a sample of 9 people over 58 days of real world
GPS tracking, plus self-reported activities and trip annotation). However, details on the
validation testing were not reported. Stopher et al. (2005) also used speed, acceleration and

Tracking Physical Activity in the Built Environment 14
route information along with a detail GIS of the road and transit network to predict travel mode.
They first used extensive automated rules to process artifacts in the data, such as removing
inaccurate GPS points, convert near-zero speed records to stationary points, and imputing values
during cold starts and other signal losses such as in “Urban Canyons” and indoors. A series of
rules were then applied to detect “trip ends” (no movement for 120 seconds or more) and mode
of travel (car, walk, bicycle, bus, train). An example results is shown in Figure 19. Although
they report a success rate of about 95% in detecting modes, and 90% of trip ends, the data and
validation method are not reported.
Overall, whilst GPS appears to offer considerable potential for detecting physical activity under
real-world conditions over long periods, few explicit studies of physical activity detection have
been conducted, and the accuracy of detection algorithms has not been tested nearly adequately
as other physical activity technologies. In addition, significant data quality challenges remain,
such as well known signal outage issues indoors and in urban canyons. However, a unique
advantage of GPS over other devices is the ability to obtain context-specific measures such as
where and when the activity occurred, which obviously represents a key link to the built
environment. This makes it particular valuable as a supplemental device to other technologies
(e.g. accelerometers, biometric monitors, heart monitors) that are more amendable to tracking a
wider range of physical activities. Additionally, as GPS accuracy improves, especially indoors,
finer patterns of movement may become detectable, such as moving about in the kitchen or
garden or working out on stationary equipment. The example set by Terrier et al. (2001; 2005)
using GPS with sub-centimeter accuracy to examine a person’s gait, suggests that potential exists
on this front.
MovingForward:AThreetieredMultiInstrumentedApproach
It seems clear from the review above, that most individual methods of assessing physical activity
have significant limitations with respect to their ability to detect the diverse types and intensity
of physical activity, especially under free-living or real-world conditions. This would explain
why many recent studies are experimenting with multiple instruments. Numerous studies in the
literature have measured physical activity in lab setting using more than one instrument (Treuth,
2002; Chen and Bassett, 2005). Given the focus on the built environment, and the clear
consensus reached by researchers from urban studies and health sciences over the need for a cost
effective method for comprehensively estimating physical activity in larger samples over longer
periods of time under diverse real-world built environment conditions (as identified in the
Introduction), a more limited combination/integration of methods appear viable.
The following three key generic components are thus proposed:
1) Passive Location tracking: some form of person-based continuous location tracking
technology appears as a necessity in order to establish meaningful, specific and accurate
links to environmental features encountered under free living daily-life conditions over
multiple days; GPS currently offers this ability, although continued improvements in
indoor tracking sensitivity (via combination with other technologies such as WiFi or cell
network towers) and artifact data processing is needed. Such technologies can also be
used to assist with automatically detecting certain types of physical activity and

Tracking Physical Activity in the Built Environment 15
intensities involving movement over space, such as walking and cycling (especially
outdoors, but increasingly indoors).
2) Passive Motion/bio-metric tracking: A full accounting of physical activity will require
the addition of some form of person-based continuous logging of energy expenditure or
body-motion, in particular to detect stationary and indoor physical activities. Multiple –
accelerometers (e.g. IDEEA, see Figure 17), or single accelerometers combined with bio-
metric sensors (e.g. SenseWear in Figure 3), appear most viable, although the latter
device offers the advantage of being a single unit. Accelerometers can also be used to
assist with location tracking by allowing dead-reckoning during short periods of GPS (or
otherwise) signal loss.
3) Limited Self-reporting: An additional concern with the built environment is the desire to
differentiate utilitarian and recreational types of physical activity, and capture other
correlates of physical activity such as involved persons – attributes that most likely
require some form of self-reporting. Keeping these to a minimum is key. Integrating
self-reports and passively tracked data is a viable option, wherein passively detect
patterns are displayed to users so additional information can prompted for either in-situ or
after-the-fact.
In addition, early in application of these methods, additional objective direct observation
methods may be needed to validate the method, or develop algorithms for processing the data
(e.g. for activity type detection).
As it turns out, several recent studies from both the health sciences and urban studies realms
reflect this 3-tiered ideal, and perhaps serves as an indication of how best to integrate and move
forward.
Ermes et al. (2008) combined three specific methods: 1) 3-axis accelerometers located on wrist
and hip; 2) GPS logging every 20 seconds; and 3) Direct observation of physical activity type
and start/end time (using PDA software shown Figure 14). They demonstrated how such data
could be used to develop an algorithm for detecting physical activity type and duration during
prescribed “supervised” itineraries (lying down, sitting/standing, walking, running, Nordic
walking, rowing machine, stationary exercise bike, cycling with real bike, and playing football)
and free-living conditions (only lying down, sitting/standing, walking, and cycling were
observed), using a variety of explanatory variables derived from the combined sensors including
peak frequency/range/spectral-entropy of up-down acceleration and GPS speed (see Figure 21).
Using a combination of decisions trees and artificial neural network classification models, they
were successful in predicting 90% of prescribed physical activity types, and 72% of free living
physical activities - a significant improvement over past methods. Variation in classification
accuracy was evident between activity types. They note in particular, how GPS improved
detection of cycling outdoors and how placement of accelerometer affects results (e.g. lack of leg
sensor made differentiation of sitting and standing difficult). The suggest that the main
challenge for the future is fine-tuning the activity-detection algorithms using more free-living
data so that a wider spectrum array of real-life physical activities can be detected. Although not

Tracking Physical Activity in the Built Environment 16
discussed, the existing use of GPS could be extended to allow aspects of the built environment to
be examined and to assist with identifying additional explanatory variables for classification.
A similar but more extensive daily life study of children with explicit focus on the built
environment and role of parents has recently been completed by Mackett et al. (2007; 2007;
2008). They combine: 1) a 3-axis accelerometers recording at 1 minute intervals); 2) a GPS
recording about every 10 seconds; and 3) a self-reported activity diaries (see Figure 12b).
Methodologically, the authors highlight some of the complexities of combining such data,
especially reconciling separately collected self-report diaries with GPS data. Two key problems
occurred: un-reported trips in the diary were detected by GPS, and missing GPS data due to
signal loss and battery failure. Conceptually, the authors demonstrate some innovative new ways
to display the data in space, as shown in Figure 20. In particular, note how intensity of EE of
physical activity (in calories) is mapped over space in Figure 20a for a single event, versus
mapping of multiple physical activity types and locations in Figure 20b. The authors were able
to subsequently explore EE in varying specific environments, including on land versus roadways
(but not in private spaces such as indoors), and during play, in clubs, or during walking. The
authors note that much more work is to be done, including larger datasets, more specific land-use
classification, and analysis of wider variety of environments.
One way to overcome the methodological complexities of combining GPS, accelerometer, and
diary data identified by Mackett et al., is to integrate the passively detected activities and
attributes (from GPS/accelerometer data) into the diary prior to subjects completing it, rather
than collecting them in isolation. In the urban studies field, these hybrid surveys are known as
“Prompted Recall Diaries”. There are several distinct methods/interfaces for conducting
prompted recall surveys, including in-situ, spatial, temporal/tabular, or some combination of
these. They are reviewed in depth here as they represent a key way to more fully integrate the
three 3-tiered multi instrument approach identified above.
In-situ prompted recall would involve processing data in real time on the device allowing
opportunities for more “interactive” survey approaches. For instance, algorithms could be
developed to automatically detect episodes of physical activity based on live streaming sensor
data (be it GPS, accelerometer, pedometer, heart rate, biochemical, etc.). This could then
“trigger” interactive queries for more qualitative data on the physical activity (intensity, type,
pain, emotions, etc.) and contextual data (involved persons, natural/built environment) that is
better captured in situ. Modern smartphones offer efficient means for capturing such data via
textual prompts, voice recordings, or even via pictures and video of the surrounding
environment. At this point however, automated detection algorithms are likely not sufficiently
robust to support such an approach. A less ideal alternative would be to ask subjects to manually
annotate their behaviour immediately prior-to or following a specified set of activities or events,
using an interface such as that in Figure 14. One of the few large sample examples of this
approach was conducted by Murakami and Wagner (1999) who used GPS along with a hand-
held computers to prompt for trip purpose and passenger information immediately prior to
automobile trips. More recently, Zhou (2007) pilot tested a person-based system involving
manual coding of activity and trip onset on a hand-held computer, followed by GPS tracking.

Tracking Physical Activity in the Built Environment 17
A Temporal/tabular prompted recall diary would involve some form of time-ordered after-the-
fact display of passively-detected stationary activities and movements and their attributes (type,
start and end times, location, route, etc.). Subjects could then be asked to correct, confirm,
and/or add supplemental information to such a display. A example of such a system is shown in
Figure 22 (Doherty et al., 2001; Doherty et al., 2006) pilot tested recently by Clark and Doherty
(in press). GPS-enabled BlackBerry smartphones were used to capture second-by-second data
transferred to a central server for processing. An automated detection algorithm reads in GPS
data, and outputs an event listing of stationary activities and movements/trips, including their
type (travel mode in case of movement, indoor/outdoor in case of activity), and location
(mapped, along with label derived from nearest land-use and roadway). Interaction with the
diary involves scrolling up/down in time and clicking on individual attributes to confirm, update,
or add information. Event types are specified via fly-out list, locations via an interactive map,
and time via +/- controls or stretching. Other attributes such as involved persons form additional
columns. Current analysis is focusing on the accuracy and refinement of algorithms.
A GIS/spatial interface for prompted recall would involve generation of a map showing a
person’s trip routes, activity stops/location, and an array of text boxes or other map attributes
depicting such items as mode, speed, location name, start/end times, trip and activity sequence
(in the day). All this overlaid on the road and land-use network for context. Again, the task of
the user would be to correct, confirm, and/or add supplemental information either directly on the
map or via some alternative input means. This could be done using a paper-and-pencil map, or
interactively on a computer screen. An early unpublished example is by Marca (2002) who used
a simple map generator coupled with an HTML form for input of attributes such as destination
name and involved persons for each trip segment displayed in the map. More recent applications
are reported by Stopher et al. (2005; Stopher et al., 2007) and Li and Shalaby (2008) who used
both maps and a tabular (temporal) presentation of GPS detected trips and stops and then
prompted for additional information such as passengers, trip purposes, and travel costs (see
Figure 23). Stopher et al. used partially automated routines to initially detect trips, followed by
manual preparation of tables and maps mailed to subjects. Tabular displays listed trips by end
time, travel time, distance, location of stop, and a reference to a map depicting the trip. The GIS-
produced maps display color-coded trips by route, direction of travel, and stop location, overlaid
on the transport and land-use network. An eight-page questionnaire was then used to
systematically prompt for stop purposes and travel modes for each trip detected, as well as
queries for undetected stops and/or incorrectly identified stops. Alternatively, Li and Shalaby
used a web-based interface to display maps and tabular information on automatically detected
activities and trips.
Although no explicit interaction with subjects is included, De la Torre and Agell (2007) recently
report one of the only studies combining the Sensewear biometric monitor (see Figure 3 and
section “Biometric Monitoring Devices”) with a wearable GPS tracking device to create an
integrated interface for examining daily activity, worthy of note here. Interestingly, they also
used a video camera and computer software usage tracker to monitor specific work-related
activities, as shown in Figure 24. Although based on only 2 subjects for 3 days, they developed
an automated clustering algorithm to classify activities on a minute-by-minute basis into eight
types of activity: sleeping, walking, away (inside the building), away (outside the building),
working (no PC), internet surfing, working on computer and talking. They report achieving

Tracking Physical Activity in the Built Environment 18
accuracies ranging from over 90% for walking, working on PC, and away-outside, to only 60%
for talking and sleeping. They then display results in a unique multi-data interface showing all
the various streams of passively collected data as they change over time, along with the activity
type and location, as shown in Figure 25.
Overall, prompted recall interfaces offer several key advantages for thoroughly exploring
physical activity in the built environment:
- Lower respondent burden (passive tracking, limited self-reporting, memory jogging)
- More accurate and detailed than self-reports (owing to passive location and body motion
tracking technologies, assuming automated/manual detection is accurate and validated)
- Allows additional attributes of interest to be optionally added manually by subjects.
- Allows any cases of missing data or undetected activities to be correct/completed by the
real “expert” (the subject), rather than imputed, allowing more thorough assessment.
It can be said, however, that such interfaces are still very much in development, as few large
samples or thorough assessments of accuracy, validity and reliability have been conducted. The
key challenge will be the development of valid and reliable automated detection algorithms that
incorporate the various streams of passively collected data for detecting the type, intensity,
duration and location of all physical activities conducted during daily life. Equally important is
development of user-friendly interfaces for displaying temporal and spatial patterns of activity
back to subjects for interaction purposes. Wireless transmission of the data to a central server
may be needed to accomplish this (such as through cellular phone wireless network), both to
allow easy access to interfaces via the web, and to be able to effectively utilize large GIS
databases or computationally intense data processing techniques beyond the capabilities of
wearable devices.
KeyCurrentMeasurementIssuesandChallenges
The following issues represent a prioritized list of challenges with respect to measuring physical
activity in the built environment using the emerging 3-tiered multi-instrumented methods
outlined above.
Validation–AnImmediateNeed
Even integrated methods that combine the best available technologies are still in need of truly
objective data to test their validity; i.e. the degree to which a measure truthfully reflects what it
is intended to measure (in this case, the types, intensity and location of physical activity).
Simply put, if the instrument cannot be trusted as a result of poor reliability or validity, then it
could very well give inaccurate or misleading information and cause one to draw invalid
conclusions (Marrow, 2002). Despite their often high success rates under lab or simulated free-
living conditions, all the devices reviewed here tend to perform much less accurately under real-
world conditions, if tested at all. Additionally, most attempts at validation under real-world
conditions are based on limited sample sizes, a limited set of physical activity types, pertain to a
limited set of attributes, and/or do not include adequate details on validation methods. Many
real-world tests also suffer from common data problems/complexities: equipment failures, user
errors, signal complexities/outages, etc.; most of which are not well documented or assessed.

Tracking Physical Activity in the Built Environment 19
Thus, a key immediate issue for emerging physical activity measurement methods, prior to
drawing strong conclusions about their relationship to the built environment, is proper validation.
Indeed, the lack of proper validation may be the single most important force hindering our ability
to make solid conclusions thus far. For this, we should take our queue from the rigourous
validation methods deployed in the health sciences. Extensive efforts have been made to test and
compare the reliability and validity of any given physical activity measurement method in the
health sciences (Mahar and Rowe, 2002; Marrow, 2002). Despite this, there is still a surprising
amount of variability in the validity of all methods (for instance, energy expenditure is often
over/under estimated by 30 to 70% Montoye et al., 1996; Dale et al., 2002; Morrow, 2002),
leading Dale et al. (2002, p. 25) to conclude that “a major challenge in physical activity
assessment research is the lack of a true gold standard of measurement”. In comparison, in urban
studies, we’ve been using self-reported travel diaries as the main source of data for decades, yet
only recently has their validity been seriously examined using objective data derived from GPS,
finding significant trip under-reporting rates (Wolf et al., 2003).
If we take anything from this experience, it should the importance of not putting blind faith in
any given method. Applying such rigour to testing the validity other aspects of physical activity;
including especially emerging methods attempting to detect activity type, duration, and location,
is a critical development issue. This will mean that in the short term, we need to invest
strategically in a reliable fourth component to the emerging 3-tiered multi-instrumented approach
(Passive Location tracking; Passive motion/bio-metric tracking; Limited Self-reporting) that
simultaneously provides truly objective data on physical activity types, intensities, and location.
Techniques such as “direct observation” and “direct bio-chemical measurement” reviewed above
(see sections of same name), often ruled out as too tedious or expensive, should be deployed in
the early stages at least for sub-samples. In lieu of such direct measures, self-reported (or
“annotated”) physical activity in-situ, may be a viable alternative if implemented carefully.
Hand-held devices and simple software would appear to offer potential on this front; getting
consistent results under real-world conditions, whilst minimizing respondent burden will be a
key challenge.
AutomatedPhysicalActivityDetectionAlgorithms
On a related note, validation efforts have a side benefit: they provide the “true” baseline
measures of physical activity that are an essential ingredient to development of automated
detection algorithms of the types, intensities and location context of physical activity.
Development of such is very much in its infancy, both in the urban studies and health science
disciplines, as shown above. In the health sciences, the focus is often more on detecting the
intensity of physical activity in the form of energy expenditure; for making the connection to the
built environment, location and activity type are more critical, as specific features of the
environment are known to impact differentially on utilitarian and recreational types of physical
activity. It turns out that there are also interesting synergies – detecting the type of physical
activity first, allows more accurate estimation of energy expenditure; intensity of activity also
assists in detecting type. Combining multiple types of sensors provides new opportunities for
identifying explanatory factors that increase the accuracy of these algorithms; but we have barely
scratched the surface of this potential. An exception is the work of Ermes et al. (2008) who used
four explanatory variables derived from an acceleromter and one from GPS (speed) to detect
activity type; those with more recent GPS experience (e.g. Chung and Shalaby, 2005; Stopher et

Tracking Physical Activity in the Built Environment 20
al., 2005; Li and Shalaby, 2008; Stopher et al., 2008) would suggest that many more GPS-
derived variables are possible (e.g. post processed proximity to roadway types). Bringing these
efforts together would appear to be a key development opportunity.
SortingouttheCausationDilemma
Existing research cannot disentangle, for example, if the observed association between certain
neighborhood characteristics (e.g., high population density, mixed land use, good sidewalks) and
higher-than-average levels of walking reflect the effects of the built environment on physical
activity or the residential preferences among people who enjoy walking which influence their
decisions to live in such neighborhoods. The ability to answer this puzzle is important for policy:
if the observed association between the built environment and physical activity is due to self-
selection, then changes to the built environment, which can be very costly, are unlikely to yield the
desired significant increase in physical activity.
(Hanson, 2006, p. s259)
As so clearly pointed out by Susan Handy, and widely recognized by others (e.g. Bhat and Guo,
2007)), an association between a built environment attribute and health-related characteristic
does not imply causation, given the complicating residential self-selection bias. More simply
put “people who like to walk choose walkable neighborhoods” (Frank et al., 2005, p. 14). Two
ways for deal with including assesses residential self selection bias through questionnaire, and
controlling for it during analysis by using a variety of quasi-longitudinal designs (Frank et al.,
2007; Mokhtarian and Cao, 2008); or conducting true longitudinal studies before-and-after a
person experiences a change in the built environment.
The implication for physical activity measurement is twofold: 1) more accurate and valid
measurement would allow for smaller samples; and 2) more passive and automated detection
algorithms would support the longer observation periods required of longitudinal studies.
TheNeedforCrossCulturalComparison 
No matter how valid and accurate the measurement, if significant variance in built environments
and/or physical activities of the sample does not exist, there will be little opportunities for
identifying healthy environments and peoples. The large number of correlating factors being as
they are, disentangling subtle differences in built environments that exist across sprawling North
American cities is surely a challenge; this is compounded by low rates of walking behaviour in
the first place. It would seem prudent to launch international cross-cultural studies, in particular
in cities with contrasting built environments where physical activity continues to be woven into
everyday life. Passive collection of physical activity data will assist in applying methods to
diverse cultural groups and in different languages.
Conclusions:NewOpportunitiesattheCrossroadsofUrban
St udiesandHealthSciences
This resource paper draws upon methods from two loosely defined fields of inquiry – urban
studies and health sciences. On close examination, some of the methods are not all that different
(e.g. diary methods), whereas in other ways they contrast significantly, such as the focus on
accelerometers in health studies and GPS in urban studies. In the end, it appears that researchers
in both fields have arrived at the same conclusion – multiple methods are needed to assess

Tracking Physical Activity in the Built Environment 21
physical activity in the built environment. This paper has identified three necessary components
of this integrated method, building on the latest studies. It is fitting to end this paper with a look
at some of the key lessons learned and new opportunities that have emerged at the crossroads of
Urban Studies and Health Sciences.
Lets’ validate our methods before we draw conclusions: re-iterating the discussion above, the
number one lesson from the health sciences is the priority they place on validation. No
measurement device is ever accepted as a means to draw conclusions about health until it has
been repeatedly validated with respect to what it is intending to measured, and is deemed reliable
across varying people and situations. Despite the plethora of techniques reviewed here, no “gold
standard” exists, even from the health sciences. Drawing strong conclusions about the impacts
of the built environment on health should be viewed with considerable caution until we establish
clearly establish the validity of the necessary emerging integrated methods. This will require
more concerted effort in the urban studies, larger tests samples, and more consistent assessment
means.
More accurate assessment means smaller sample sizes and less chance of erroneous
conclusions: both fields of inquiry have aggressively sought more objective, passive, automated
technologies for assessing physical activity recognizing the limitations of self-reports. These
devices are surely only going to get smaller, less power hungry, more sensitive, and offer more
integrated capabilities. Even bio-chemical measurement has emerged from the lab in the form of
a wearable device (e.g. see Figure 1), just as heart-rate monitors did decades before. We need to
cautiously, but aggressively, keep up with these methods, initially with small sample
experimentation, but quickly followed by validation studies and applications.
There is more to physical activity than just walking: Current urban studies are often limited to
examining walking as an indicator physical activity; few successful attempts have been made to
examine recreational and stationary physical activities. Adopting methods from the health
sciences that focus on assessing all physical activity types (moving and stationary) will surely
expand this, and tie results more holistically to health.
There is more to daily life and health than physical activity: Ok, you don’t have to tell health
scientists that there are other correlates of health, but the activity-based methods emerging from
urban studies would surely broadening the scope of analysis beyond physical activities in
isolation. This includes especially social activities that affect mental health, and dietary
activities (eating in-home, preparing foods, eating out, grocery shopping) that affect weight and
well-being, both of which have strong ties to the built environment and accessibility. Extending
our observation to include these activities will provide for a more holistic assessment of health,
especially in built environments with low walkability, poor access to fresh foods (or too many
fast food restaurants), and isolating social networks.
Similarly, there is more to the built environment than just the home: the field of urban studies
has a long history of examining the location of activities, whereas in health sciences, the location
“context” within which physical activity takes place has received only recent attention. Building
on this, we need to refine the scale and extent of locations considered to impact physical activity
using increasing accurate and longitudinal datasets, especially considering those around multiple

Tracking Physical Activity in the Built Environment 22
anchor points of one’s life beyond just the home. Additionally, a key challenge will be acquiring
accurate and up-to-date data on the key features and attributes within these contexts that affect
physical activity. Whilst current GIS databases provide a reasonably accounting of the
“quantity” of these attributes (e.g. densities, lengths, connectivity), the same cannot be said of
the environmental “quality” attributes that can have a strong affect on behaviour such a lighting
and safety, which may require more direct observation.
The built environment is not the only constraint on physical activity: whilst health scientists
often study physical activity at discrete prescribed moments in the lab, its occurrence in the real
world is part of a continuous decision making process constrained not only by the external built
environment, but by personal constraints. Evolving from urban studies are new methods that
provide more holistic insights into the spatial, temporal, and interpersonal scheduling constraints
on daily activities. Applying these approaches to physical activity is likely to provide wholly
new insights into how it is planned and managed within a person’s everyday life and barriers to
its completion.
Acknowledgements
The author would like to acknowledge the financial support received from the Canadian
Institutes of Health Research (CIHR), The Health Technology Exchange (HTX), the Social
Sciences and Humanities Research Council of Canada, and from the GEOIDE (Geomatics for
Informed Decisions) Network of Centres of Excellence Program of the Canadian federal research
councils. Special thanks also go to Johanna Zmud for her patience, to Andrew Clark who
assisted with the literature review, and to colleagues and reviewers who generously provided
advice and copies of their recent work.

Tracking Physical Activity in the Built Environment 23
References
Ainsworth, B. E., W. Haskell, M. Whitt, M. Irwin, A. Swartz, S. J. Strath, W. O'Brien, D. R. J.
Bassett, K. H. Schmitz, P. Emplaincourt, D. Jacobs and A. Leon (2000). Compendium of
physical activities: An update of activity codes and MET intensities. Medicine & Science
in Sports & Exercise 32(9): S498 - S516.
Aminian, K. and B. Najafi (2004). Capturing human motion using body-fixed sensors: outdoor
measurement and clinical applications. Computer Animation and Virtual Worlds 15(2):
79-94.
Andre, D. (2007). Letters to the Editor: Assessing Resting Metabolic Rate Using a Multi-sensor
Armband. Obesity 15 (5): 1337-8.
Arvidsson, D., F. Slinde, S. Larsson and L. Hulthen (2007). Energy cost of physical activities in
children: validation of SenseWear Armband. Med Sci Sports Exerc 39(11): 2076-84.
Asakura, Y. and E. Hato (2004). Tracking survey for individual travel behaviour using mobile
communication instruments. Transportation Research Part C 12(3-4): 273–291.
Bajaj, R., S. L. Ranaweera and D. P. Agrawal (2002). GPS: location-tracking technology.
Computer 35(4): 92 - 94.
Barnard, P. (1986). Use of an activity diary survey to examine travel and activity reporting in a
home interview survey: an example using data from Adelaide, Australia. Transportation
13(4): 329-357.
Bassett, D. R. and S. J. Strath (2002). Use of Pedometers to assess physical activity. Physical
Activity Assessments for Health-related Research. G. J. Welk. Champaign, IL, Human
Kinetics Publishers: 163-177.
Bhat, C. R. and J. Y. Guo (2007). A comprehensive analysis of built environment characteristics
on household residential choice and auto ownership levels. Transportation Research Part
B: Methodological 41(5): 506-526.
Boarnet, M. and R. Crane (2001). Travel by Design: The Influence of Urban Form on Travel.
New York, Oxford University Press.
Boarnet, M. G. (2005). The Built Environment and Physical Activity: Empirical Methods and
Data Resources. Washington, D. C., Transportation Resarch Board and the Institute of
Medicine Committee on Physical Activity, Health, Transportation, and Land Use. TRB
Special Report 282.
Booth, K. M., M. M. Pinkston and W. S. C. Poston (2005). Obesity and the Built Environment.
Journal of the American Diabetic Association 105(5 Sup): 110-117.
Bouchard, C. (1997). Bouchard Three-Day Physical Activity Record. Medicine & Science in
Sports & Exercise 29(6): 19-24.
Bouchard, C., A. Tremblay, C. Leblanc, G. Lortie, R. Savard and G. Theariault (1983). A
method to assess energy expenditure in children and adults. American Journal Of
Clinical Nutrition 37: 461-467.
Brage, S., N. Brage, P. W. Franks, U. Ekelund, M. Y. Wong, L. B. Andersen, K. Froberg and N.
J. Wareham (2004). Branched equation modeling of simultaneous accelerometry and
heart rate monitoring improves estimate of directly measured physical activity energy
expenditure. Journal of applied Physiology 96(1): 343-51.
Brownson, R. C., T. K. Boehmer and D. A. Luke (2005). Declining Rates of Physical Activity in
the United States: What Are the Contributors?
Annual Review of Public Health 26: 421-
443.

Tracking Physical Activity in the Built Environment 24
Cereda, E., M. Turrini, D. Ciapanna, L. Marbello, A. Pietrobelli and E. Corradi (2007).
Assessing Energy Expenditure in Cancer Patients: A Pilot Validation of a New Wearable
Device. Journal of Parenteral and Enteral Nutrition 31(6): 502-507.
Chen, K. Y. and D. R. J. Bassett (2005). The Technology of Accelerometry-Based Activity
Monitors: Current and Future. Medicine & Science in Sports & Exercise 37(11S): 490-
500.
Chung, E.-H. and A. Shalaby (2005). A Trip Reconstruction Tool for GPS-based Personal Travel
Surveys. Transportation Planning and Technology 28(5): 381-401.
Clark, A. and S. T. Doherty (in press). Examining the Nature and Extent of the Activity-travel
Preplanning Decision Process. Journal of the Transportation Research Board.
Copperman, R. B. and C. R. Bhat (2007). An analysis of the determinants of children's weekend
physical activity participation. Transportation 34: 67-87.
Crane, R. (2000). The Influence of Urban Form on Travel: An Interpretive Review. Journal of
Planning Literature 15(1): 3-23.
Cromley, E. K. and S. L. McLafferty (2002). GIS and Public Health. New York, Guilford.
Dale, D., G. J. Welk and C. E. Matthews (2002). Methods for assessing physical activity and
challenges for research. Physical Activity Assessments for Health-related Research. G. J.
Welk. Champaign, IL, Human Kinetics Publishers: 19-34.
Dannenberg, A. L., R. J. Jackson, H. Frumkin, R. A. Schieber, M. Pratt, C. Kochtitzky and H. H.
Tilson (2003). The Impact of Community Design and Land-Use Choices on Public
Health: A Scientific Research Agenda. American Journal of Public Health 93(9): 1500–
1508.
Davison, K. K. and C. T. Lawson (2006). Do attributes in the physical environment influence
children's physical activity? A review of the literature. International Journal of
Behavioral Nutrition and Physical Activity 3(19): 1-17.
De la Torre, F. and C. Agell (2007). Multimodal Diaries. 2007 IEEE International Conference
on Multimedia and Expo, July 2-5, 2007, Beijing, China.
Doherty, S. T. and E. J. Miller (2000). A Computerized Household Activity Scheduling Survey.
Transportation 27(1): 75-97.
Doherty, S. T., E. Nemeth, M. Roorda and E. J. Miller (2004). Design and Assessment of the
Toronto Area Computerized Household Activity Scheduling Survey. Journal of the
Transportation Research Board 1894: 140-149.
Doherty, S. T., N. Noël, M. E. H. Lee-Gosselin, C. Sirois and M. Ueno (2001). Moving beyond
observed outcomes: Integrating Global Positioning Systems and interactive computer-
based travel behaviour surveys. Transportation Research Circular E-C026: 449-466.
Doherty, S. T., D. Papinski and M. Lee-Gosselin (2006). An Internet-based Prompted Recall
Diary with Automated GPS Activity-trip Detection: System Design. 85th Annual Meeting
of the Transportation Research Board, January 22-26, 2006., Washington, D.C.
Dorminy, C. A., L. Choi, S. A. Akohoue, K. Y. Chen and M. S. Buchowski (2008). Validity of a
Multisensor Armband in Estimating 24-h Energy Expenditure in Children. Med Sci
Sports Exerc 40(4): 699-706.
Elgethun, K., R. A. Fenske, M. G. Yost and G. J. Palcisko (2003). Time-location analysis for
exposure assessment studies of children using a novel global positioning system
instrument. Environ Health Perspect 111(1): 115-22.

Tracking Physical Activity in the Built Environment 25
Elgethun, K., M. G. Yost, C. T. Fitzpatrick, T. L. Nyerges and R. A. Fenske (2007). Comparison
of global positioning system (GPS) tracking and parent-report diaries to characterize
children's time-location patterns. J Expo Sci Environ Epidemiol 17(2): 196-206.
Ermes, M., J. Parkka, J. Mantyjarvi and I. Korhonen (2008). Detection of Daily Activities and
Sports With Wearable Sensors in Controlled and Uncontrolled Conditions. IEEE
Transactions on Information Technology in Biomedicine 12(1): 20 - 26.
Ewing, R. and R. Cervero (2001). Travel and the Built Environment - A Synthesis.
Transportation Research Record 1780: 87-114.
Fares, S., M. Miller, S. Masters and M. Crotty (2008). Measuring energy expenditure in
community-dwelling older adults: are portable methods valid and acceptable? Journal of
the American Dietetic Association 108(3): 544-8.
Fields, D. A., J. T. Kearney and K. C. Copeland (2006 ). MedGem Hand-Held Indirect
Calorimeter Is Valid for Resting Energy Expenditure Measurement in Healthy Children.
Obesity 14: 1755-1761
Foerster, F., M. Smeja and J. Fahrenberg (1999). Detection of posture and motion by
accelerometry: a validation study in ambulatory monitoring. Computers in Human
Behavior 15(5): 571-583.
Forsyth, A., J. M. Oakes, K. H. Schmitz and M. Hearst (2007). Does Residential Density
increase Walking and Other Physical Activity? Urban Studies 44(4): 679-697.
Frank, L., B. E. Saelens, K. E. Powell and J. Chapman (2007). Stepping towards causation: Do
built envronments or neighborhood and travel preferences explain physical activity,
driving, and obesity? Social Science & Medicine 65(9): 1898-1914.
Frank, L., J. Sallis, B. Saelens, W. Bachman and K. Washbrook (2005). Travel Behavior,
Environmental, & Health Impacts of Community Design & Transportation investment: A
Study of Land Use, Transportation, Air Quality and Health in King County, WA. King
County, Washington.
Frank, L. D. and P. Engelke (2005). Multiple Impacts of the Built Environment on Public Health:
Walkable Places and the Exposure to Air Pollution. International Regional Science
Review 28(2): 193-216.
Frank, L. D., P. O. Engelke and T. L. Schmid (2003). Health and community design.
Washington, Island Press.
Fruin, M. L. and J. W. Rankin (2004). Validity of a multi-sensor armband in estimating rest and
exercise energy expenditure. Med Sci Sports Exerc 36(6): 1063-9.
Frumkin, H., L. Frank and R. Jackson (2004). Urban sprawl and public health: Designing,
planning, and building for healthy communities. Washington, Island Press.
Gardner, M. J., J. U. Barker, S. M. Briggs, S. I. Backus, D. L. Helfet, J. M. Lane and D. G.
Lorich (2007). An evaluation of accuracy and repeatability of a novel gait analysis
device. Arch Orthop Trauma Surg 127(3): 223-7.
Handy, S. L. (2005). Critical Assessment of the Literature on the Relationships Among
Transportation, Land Use, and Physical Activity. Washington, D.C., Transportation
Research Board and the Institute of Medicine Committee on Physical Activity, Health,
Transportation, and Land Use. TRB Special Report 282.
Hanson, S. (2006). Active Living Research in Light of the TRB/IOM Reports. Journal of
Physical Activity and Health 2006, 3, Suppl 1,(3(Suppl 1)): S258-S266.
Heath, G. W., R. C. Brownson, J. Kruger, R. Miles, K. E. Powell, L. T. Ramsey and Task Force
on Community Preventive Services (2006). The Effectiveness of Urban Design and Land

Tracking Physical Activity in the Built Environment 26
Use and Transport Policies and Practices to Increase Physical Activity: A Systematic
Review. Journal of Physical Activity and Health 3(Suppl 1): S55-S76.
Hlynsky, J., C. L. Birmingham, M. Johnston and S. Gritzner (2005). The agreement between the
MedGem indirect calorimeter and a standard indirect calorimeter in anorexia nervosa. Eat
Weight Disord 10(4): e83-7.
Huddleston, J., A. Alaiti, D. Goldvasser, D. Scarborough, A. Freiberg, H. Rubash, H. Malchau,
W. Harris and D. Krebs (2006). Ambulatory measurement of knee motion and physical
activity: preliminary evaluation of a smart activity monitor. Journal of NeuroEngineering
and Rehabilitation 3(1): 21.
Humphrey, N. P. (2005). Does the Built Environment Influence Physical Activity? Examining
the Evidence. TR News 237: 31.
Institute for Social Research (2002). 2002 Time Diary Questionnaires from the Child
Development Supplement (CDS) to the Panel Study of Income Dynamics, Institute for
Social Research, University of Michigan.
Jakicic, J. M., M. Marcus, K. I. Gallagher, C. Randall, E. Thomas, F. L. Goss and R. J.
Robertson (2004). Evaluation of the SenseWear Pro Armband to assess energy
expenditure during exercise. Med Sci Sports Exerc 36(5): 897-904.
Janz, K. F. (2002). Use of heart rate monitors to assess physical activity. Physical Activity
Assessments for Health-related Research. G. J. Welk. Champaign, IL, Human Kinetics
Publishers: 143-161.
King, G., N. Torres, C. Potter, T. Brooks and K. Coleman (2004). Comparison of Activity
Monitors to Estimate Energy Cost of Treadmill Exercise. Medicine & Science in Sports
& Exercise 36(7): 1244-1251.
Le Faucheur, A., P. Abraham, V. Jaquinandi, P. Bouye, J. L. Saumet and B. Noury-Desvaux
(2007). Study of human outdoor walking with a low-cost GPS and simple spreadsheet
analysis. Medicine & Science in Sports & Exercise 39(9): 1570-8.
Lee-Gosselin, M. E. H. and A. S. Harvey (2005). Travel survey applications of non-web
technologies. Developments in Travel Survey Methods. P. R. Stopher. Oxford, Elsevier
(forthcoming).
Lee, M. and M. G. McNally (2006). An empirical investigation on the dynamic processes of
activity scheduling and trip chaining. Transportation 33(6): 553 - 565.
Li, Z. J. and A. Shalaby (2008). Web-based GIS System for Prompted Recall of GPS-assisted
Personal Travel Surveys: System Development and Experimental Study. 87th Annual
Meeting of the Transportation Research Board, , , January 13-17, 2008, Washington D.C.
Liden, C. B., M. Wolowicz, J. Stivoric, A. Teller, S. Vishnubhatla, R. Pelletier and J. Farringdon
(2002). Accuracy and Reliability of the SenseWear Armband as an Energy Expenditure
Assessment Device. Pittsburgh, PA BodyMedia, Inc.
Mackett, R. (2008). Understanding the impacts of the travel behaviour and activity patterns of
children on their physical activity and health. 8th International Conference on ‘Survey
Methods on Transport: Harmonisation and Data Comparability, May 25-31, 2008,
Annecy, France.
Mackett, R. L., B. Brown, Y. Gong, K. Kitazawa and J. Paskins (2007). Children’s Independent
Movement in the Local Environment. Built Environment 33(4): 454-468.
Mackett, R. L., Y. Gong, K. Kitazawa and J. Paskins (2007). Children’s local travel behaviour -
how the environment influences, controls and facilitates it. 11th World Conference on
Transport Research, June 24-28, 2007, Berkeley, California.

Tracking Physical Activity in the Built Environment 27
Maffiuletti, N. A., M. Gorelick, I. Kramers-de Quervain, M. Bizzini, J. P. Munzinger, S.
Tomasetti and A. Stacoff (2008). Concurrent validity and intrasession reliability of the
IDEEA accelerometry system for the quantification of spatiotemporal gait parameters.
Gait & Posture 27(1): 160-163.
Mahar, M. T. and D. A. Rowe (2002). Construct validity in physical activity research. Physical
Activity Assessments for Health-related Research. G. J. Welk. Champaign, IL, Human
Kinetics Publishers: 51-72.
Malavolti, M., A. Pietrobelli, M. Dugoni, M. Poli, E. Romagnoli, P. De Cristofaro and N. C.
Battistini (2007). A new device for measuring resting energy expenditure (REE) in
healthy subjects. Nutr Metab Cardiovasc Dis 17(5): 338-43.
Marca, J. (2002). The Design and Implementation of An On-Line Travel and Activity Survey.
The American Society of Civil Engineers, 7th International Conference on Applications
of Advanced Technology in Transportation, Cambridge, Massachusetts.
Marrow, J. R. (2002). Measurement issues for the assessment of Physical activity. Physical
Activity Assessments for Health-related Research. G. J. Welk. Champaign, IL, Human
Kinetics Publishers: 37-50.
Marsh, A. P., R. M. Vance and T. L. Frederick (2007). Objective Assessment of Activity in
Older Adults at Risk for Mobility Disability. Medicine and Science in Sports and
Exercise 39(6): 1020-6.
Mathie, M. J., A. C. F. Coster, N. H. Lovell and B. G. Celler (2004). Accelerometry: providing
an integrated, practical method for long-term, ambulatory monitoring of human
movement. Physiological Measurement 25(2): R1-R20.
Matthews, C. E. (2002). Use of self-report instuments to assess physical activity. Physical
Activity Assessments for Health-related Research. G. J. Welk. Champaign, IL, Human
Kinetics Publishers: 107-123.
McCormack, G., B. Giles-Corti, A. Lange, T. Smith, K. Martin and T. J. Pikora (2004). An
update of recent evidence of the relationship between objective and self-report measures
of the physical environment and physical activity behaviours. Journal of Science and
Medicine in Sport 7(1): 81-92.
McDoniel, S. O. (2007). A Systematic Review on Use of a Handheld Indirect Calorimeter to
Assess Energy Needs in Adults and Children. International Journal of Sport Nutrition
and Exercise Metabolism 17: 491-500.
McKenzie, T. L. (2002). Use of direct observation to assess physical activity. Physical Activity
Assessments for Health-related Research. G. J. Welk. Champaign, IL, Human Kinetics
Publishers: 179-195.
McKenzie, T. L., D. A. Cohen, A. Sehgal, S. Williamson and D. Golinelli (2006). System for
Observing Play and Recreation in Communities (SOPARC): Reliability and Feasibility
Measures. Journal of Physical Activity and Health 3(Supp 1): S208-S222.
McKenzie, T. L., S. J. Marshall, J. F. Sallis and T. L. Conway (2000). Leisure-time physical
activity in school environments: an observational study using SOPLAY. Prev Med 30(1):
70-7.
Mokhtarian, P. L. and X. Cao (2008). Examining the impacts of residential self-selection on
travel behavior: A focus on methodologies. Transportation Research Part B:
Methodological 42(3): 204-228
Montoye, J., H. Kemper, W. Saris and R. Washburn (1996). Measuring physical activity and
energy expenditure. Champaign, IL, Human Kinetics Publishers.

Tracking Physical Activity in the Built Environment 28
Morrow, J. R. (2002). Measurement issues for the assessment of physical activity. Physical
Activity Assessments for Health-related Research. G. J. Welk. Champaign, IL, Human
Kinetics Publishers: 35-50.
Murakami, E. and D. P. Wagner (1999). Can using Global Positioning System (GPS) improve
trip reporting? Transportation Research C 7(2/3): 149-165.
Papas, M. A., A. J. Alberg, R. Ewing, K. J. Helzlsouer, T. L. Gary and A. C. Klassen (2007). The
Built Environment and Obesity. Epidemiologic Reviews 29: 129-143.
Papazoglou, D., G. Augello, M. Tagliaferri, G. Savia, P. Marzullo, E. Maltezos and A. Liuzzi
(2006). Evaluation of a multisensor armband in estimating energy expenditure in obese
individuals. Obesity (Silver Spring) 14(12): 2217-23.
Parkka, J., M. Ermes, P. Korpipaa, J. A. Mantyjarvi, J. A. Peltola and I. A. Korhonen (2006).
Activity classification using realistic data from wearable sensors. IEEE Transactions on
Information Technology in Biomedicine 10(1): 119-128.
Reeves, M. M., S. Capra, J. Bauer, P. S. Davies and D. Battistutta (2005). Clinical accuracy of
the MedGem indirect calorimeter for measuring resting energy expenditure in cancer
patients. The European Journal of Clinical Nutrition and Metabolism 59(4): 603-10.
Retscher, G. (2007). Test and Integration of Location Sensors for a Multi-sensor Personal
Navigator. Journal of Navigation 60(1): 107-117.
Richardson, A., E. Ampt and A. Meyburg (1995). Survey Methods for Transport Planning.
Melbourne, Eucalyptus Press.
Rodriguez, D. A., A. L. Brown and P. J. Troped (2005). Portable Global Positioning Units to
Complement Accelerometry-Based Physical Activity Monitors. Medicine & Science in
Sports & Exercise 37(11): 572-581.
Rubenbauer, J. R., D. L. Johannsen, S. M. Baier, R. Litchfield and P. J. Flakoll (2006). The Use
of a Handheld Calorimetry Unit to Estimate Energy Expenditure During Different
Physiological Conditions. Journal of Parenteral and Enteral Nutrition 30(3): 246-50.
Ruiz, T. and H. Timmermans (2006). Changing the timing of activities in resolving Scheduling
Conflicts. Transportation 33: 429 - 445.
Sallis, J. and B. Saelens (2000). Assessment of physical activity by self-report: Limitations and
future directions. Research Quarterly for Exercise and Sport 71: S1-S14.
Sayed, A. H., A. Tarighat and N. Khajehnouri (2005). Network-based wireless location finding.
Signal Processing Magazine, IEEE 22(4): 24 - 40.
Schutz, Y. and A. Chambaz (1997). Could a satellite-based navigation system (GPS) be used to
assess the physical activity of individuals on earth? Eur J Clin Nutr 51(5): 338-9.
Schutz, Y. and R. Herren (2000). Assessment of speed of human locomotion using a differential
satellite global positioning system. Med Sci Sports Exerc 32(3): 642-6.
Sener, I. N., R. B. Copperman, R. M. Pendyala and C. R. Bhat (2008). An Analysis of Children’s
Leisure Activity Engagement: Examining the Day of Week, Location, Physical Activity
Level, and Fixity Dimensions. 2008 Annual Meeting of the Transportation Research
Board (note: under review for Transportation), Washington, DC.
Shoval, N. and M. Isaacson (2006). Application of Tracking Technologies to the Study of
Pedestrian Spatial Behavior. Professional Geographer 58(2): 172-183.
Starling, R. D. (2002). Use of doubly labeled water and indirect calorimetry to assess physical
activity. Physical Activity Assessments for Health-related Research. G. J. Welk.
Champaign, IL, Human Kinetics Publishers: 197-209.

Tracking Physical Activity in the Built Environment 29
Stel, V. S., J. H. Smit, S. M. F. Pluijm, M. Visser, D. J. H. Deeg and P. Lips (2004). Comparison
of the LASA Physical Activity Questionnaire with a 7-day diary and pedometer. Journal
of Clinical Epidemiology 57(3): 252-258.
Stopher, P. and A. Collins (2005). Conducting a GPS Prompted Recall Survey over the Internet.
The 84th Annual Meeting of the Transportation Research Board, January 9-13,
Washington, D.C.
Stopher, P., C. FitzGerald and M. Xu (2007). Assessing the accuracy of the Sydney Household
Travel Survey with GPS. Transportation 34(6): 723 - 741.
Stopher, P. and S. P. Greaves (2007). Household travel surveys: Where are we going?
Transportation Research Part A 41(5): 367-381.
Stopher, P., Q. Jiang and C. FitzGerald (2005). Processing GPS Data from Travel Surveys. 28th
Australasian Transport Research Forum, September, Sydney, Australia.
Stopher, P., Q. Jiang and C. FitzGerald (2008). Deducing mode and purpose from GPS data.
Sydney, Australia, The Institute of Transport and Logistics Studies, The University of
Sydney.
Stopher, P. and P. Jones, Eds. (2003). Transport Survey Quality and Innovation. New York,
Pergamon.
Strath, S. J., A. M. Swartz, W. L. Bassett, D. R. O’brien, G. A. King and B. E. Ainsworth (2000).
Evaluation of heart rate as a method for assessing moderate intensity physical activity.
Medicine & Science in Sports & Exercise 32(supplement): 465-470.
Szeto, A. Y. and S. K. Sharma (2007). RFID based indoor navigational aid for persons with
severe visual impairments. Conf Proc IEEE Eng Med Biol Soc 2007: 6361-4.
Terrier, P., Q. Ladetto, B. Merminod and Y. Schutz (2001). Measurement of the mechanical
power of walking by satellite positioning system (GPS). Med Sci Sports Exerc 33(11):
1912-8.
Terrier, P., V. Turner and Y. Schutz (2005). GPS analysis of human locomotion: further
evidence for long-range correlations in stride-to-stride fluctuations of gait parameters.
Hum Mov Sci 24(1): 97-115.
Townshend, A. D., C. J. Worringham and I. B. Stewart (2008). Assessment of speed and position
during human locomotion using nondifferential GPS. Med Sci Sports Exerc 40(1): 124-
32.
Transportation Research Board and Institute of Medicine of the National Academies (2005).
Does the Built Environment Influence Physical Activity? Examining the Evidence
Washington, DC, Transportation Research Board. Special Report 282.
Treuth, M. S. (2002). Applying multiple Methods to improve the accuracy of activity
assessments. Physical Activity Assessments for Health-related Research. G. J. Welk.
Champaign, IL, Human Kinetics Publishers: 213-225.
Tsui, S. Y. A. and A. Shalaby (2007). An Enhanced System for Link and Mode Identification for
GPS-based Personal Travel Surveys. Journal of the Transportation Research Board
1972: 38-45.
Van Loan, M. D. (2007). Do Hand-Held Calorimeters Provide Reliable and Accurate Estimates
of Resting Metabolic Rate? J Am Coll Nutr 26(6): 625-629.
Welk, G. J. (2002). Introduction to physical activity research. Physical Activity Assessments for
Health-related Research. G. J. Welk. Champaign, IL, Human Kinetics Publishers: 1-17.

Tracking Physical Activity in the Built Environment 30
Welk, G. J. (2002). Use of accelerometry-based activity monitors to assess physical activity.
Physical Activity Assessments for Health-related Research. G. J. Welk. Champaign, IL,
Human Kinetics Publishers: 125-141.
Welk, G. J., J. J. McClain, J. C. Eisenmann and E. E. Wickel (2007). Field Validation of the MTI
Actigraph and BodyMedia Armband Monitor Using the IDEEA Monitor. Obesity 15(4):
918-928.
Weston, A. T., R. Petosa and R. R. Pate (1997). Validation of an instrument for measurement of
physical activity in youth. Medicine & Science in Sports & Exercise 29(1): 138-143.
Witte, T. H. and A. M. Wilson (2005). Accuracy of WAAS-enabled GPS for the determination
of position and speed over ground. J Biomech 38(8): 1717-22.
Wolf, J., M. Loechl, M. Thompson and C. Arce (2003). Trip Rate Analysis in GPS-Enhanced
Personal Travel Surveys. Transport Survey Quality and Innovation. P. R. Stopher and P.
M. Jones. New York, Pergamon: 483-498.
Wormald, H., M. Sleap, J. Brunton, L. Hayes, P. Warburton, M. Waring and M. White (2003).
Methodological issues in piloting a physical activity diary with young people. Health
Education Journal 62(3): 220-233.
Zhang, K., F. X. Pi-Sunyer and C. N. Boozer (2004). Improving Energy Expenditure Estimation
for Physical Activity. Medicine & Science in Sports & Exercise 36(5): 883-889.
Zhang, K., P. Werner, M. Sun, F. X. Pi-Sunyer and C. N. Boozer (2003). Measurement of
Human Daily Physical Activity. Obesity Res 11(1): 33-40.
Zhou, J. and R. Golledge (2007). Real-time tracking of activity scheduling/schedule execution
within a unified data collection framework. Transportation Research Part A 41: 444-463.

Tracking Physical Activity in the Built Environment 31
Figure 1 MedGem Handheld indirect calorimeter
From McDoniel (2007, p. 493). See also http://www.metabolismmatters.com/medgem/index.html
Figure 2 Polar Heart Rate Monitor
From the Polar S120 Manual, available at http://www.polarusa.com/manuals/s120.pdf
Figure 3 SenseWear Pro Armband
Armband Skin sensors on flip side
Images from http://www.sensewear.com/index.php

Tracking Physical Activity in the Built Environment 32
Figure 4 Retrospective Physical Activity Questionnaire Example: California Health Interview Study 2005
…
The next questions are about walking for transportation. Please only include walks that involved an errand or to get
some place. I will ask you separately about walking for relaxation or exercise.
During the past seven days, did you walk for at least ten minutes at a time to get some place such as work,
school, a store, or restaurant?
How much time did you [usually] spend walking on [one of those days/that day]?
Sometimes you may walk for fun, relaxation, exercise, or to walk the dog. During the past seven days did you walk
for at least ten minutes at a time for any of these reasons? Please do not include any walking that you already told
me about.
How much time did you (usually) spend walking on (one of those days/on that day)?
The next questions are about physical activities or exercise you may do in your free time for at least 10 minutes,
other than walking. First, think about activities that take moderate physical effort, such as bicycling, swimming,
dancing, and gardening.
During the last 7 days, did you do any moderate physical activities in your free time for at least 10 minutes,
other than walking?
How much time did you [usually] spend on [one of those days/that day] doing moderate physical activities
in your free time?
Now think about vigorous activities you did in your free time that take hard physical effort, such as aerobics,
running, soccer, fast bicycling, or fast swimming. Again, do not include walking.
During the last 7 days, did you do any vigorous physical activities in your free time?
How much time did you [usually] spend on [one of those days/on that day] doing vigorous physical
activities in your free time?
Now think about activities specifically designed to STRENGTHEN your muscles, such as lifting weights or other
strength-building exercises. Include all such activities even if you have mentioned them before.
During the last 7 days, on how many days did you do activities to strengthen your muscles?
…
Interview Script Extracted from California Health Interview Study 2005, Adult Questionnaire Version 6.3, October
10, 2006 “Section C – Health Behaviors”. Does not include coding and other prompting script. For full details, see
http://www.chis.ucla.edu/pdf/CHIS2005_adult_q.pdf
.

Tracking Physical Activity in the Built Environment 33
Figure 5 Stel et al. (2004) Example 7-day Physical Activity Diary
Figure 6 Weston et al.’s (1997) Previous Day Physical Activity Recall Diary

Tracking Physical Activity in the Built Environment 34
Figure 7 Bouchard et al.’s (1983; 1997) 3 Day Physical Activity Diary by Time, Type and Intensity
Figure 8 Wormald et al. (2003) Physical Activity Diary

Tracking Physical Activity in the Built Environment 35
Figure 9 Heart and Stroke Foundation “My Physical Activity Diary”
From http://www.heartandstroke.com/site/c.ikIQLcMWJtE/b.3754199/k.A02C/ My_physical_activity_diary.htm

Tracking Physical Activity in the Built Environment 36
Figure 10 Major Types and Example Specific Activities from Ainsworth et al.’s (2000) Compendium of
Physical Activities (full list has 650 specific activities)
…
…
Figure 11 Example Trip Diary: St. Louis Region Small sample Travel Survey

Tracking Physical Activity in the Built Environment 37
Figure 12 Example Activity and Travel Diaries
a) From Barnard (1986)
b) As used by children in the Mackett et al. (2007) CAPABLE project

Tracking Physical Activity in the Built Environment 38
Figure 13 Time Diary Questionnaires from the Child Development Supplement (CDS) to the Panel Study of
Income Dynamics (Institute for Social Research, 2002)
Full details available at http://psidonline.isr.umich.edu/CDS/TDqnaires.html
Figure 14 Ermes et al. (2008) PDA Software for Observer or Self-Reported Physical Activity Events

Tracking Physical Activity in the Built Environment 39
Figure 15 Direct Observation Form Used in the SOPLAY method to Record Physical Activity of Groups of
Individuals
From McKenzie et al. (2002, p. 194)

Tracking Physical Activity in the Built Environment 40
Figure 16 Example Pedometer (Yamax SW-200), showing spring-suspended horizontal lever arm that moves
up and down in response to vertical accelerations.
Image from Inside look, from Bassett and Strath (2002, 165)
http://www.yamaxx.com/digi/sw-200-e.html

Tracking Physical Activity in the Built Environment 41
Figure 17 IDEEA Multi-sensor Accelerometer
Image from Maffiuletti et al. (2008, p. 161).
Image from Huddleston et al. (2006)

Tracking Physical Activity in the Built Environment 42
Figure 18 Example GPS-tracked physical activity bouts overlaid in a GIS, from Rodriguez et al. (2005)
Note: Physical activity bouts identified via accelerometer in 1-minute epochs.
Figure 19 Example Person-based GPS tracked data before and after processing to detect movements by
Travel mode (from Stopher et al. 2008)
Æ
a) RAW GPS Data after identifying trip ends b) After automated processing to detect travel mode

Tracking Physical Activity in the Built Environment 43
Figure 20 Example Spatial Depiction of GPS and Accelerometer Data Highlighting Physical Activity
a) Overlaid on an aerial photo in an interactive Web display
From http://www.casa.ucl.ac.uk/capableproject/maps/home.asp. See also, Mackett et al. (2007)
b) Highlighting physical activity types and locations
From Mackett et al. (2007, p. 12)

Tracking Physical Activity in the Built Environment 44
Figure 21 Time line showing Physical Activity Type alongside Accelerometer and GPS derived data, from
Ermes et al. (2008)
Figure 22 Example Temporal Prompted Recall Diary
a) Initial Diary Automatically Detected via GPS b) After Prompted Recall interaction with Subject
Æ
Peak frequency of
up-down acceleration
Range of up-down
acceleration
Spectral entropy of
up-down acceleration
Speed from GPS
Activity from direct observation
A) cycling; B) walking; C) football;
D) Nordic walking; E) running.

Tracking Physical Activity in the Built Environment 45
Figure 23 Example Spatial/Temporal Prompted Recall Diaries
a) Example hard copies of map and table mailed to subjects, from Stopher et al. (2007)
b) Web-based Prompted Recall diary from Li and Shalaby (2008)

Tracking Physical Activity in the Built Environment 46
Figure 24 De la Torre and Agell’s (2007) “Multimodal Diary”, combining the Sensewear Biometric monitor,
GPS, and Video-audio recording of work-related tasks.
Figure 25 Example Video Screenshot of Multi Sensor Data from De la Torre and Agell’s (2007) “Multimodal
Diary”
From: http://www.cs.cmu.edu/~ftorre/IcmeVideo.mpg