Available via license: CC BY 4.0
Content may be subject to copyright.
1
Title: Potentials and Limitations of Large-scale, Individual-level Mobile Location
Data for Food Acquisition Analysis
Duanya Lyu a, Luyu Liu b, *, Catherine Campbell c, Yuxuan Zhang d, Xiang Yan a
a Department of Civil & Coastal Engineering, University of Florida, Gainesville FL, United States
b Department of Geosciences, Auburn University, Auburn AL, United States
c Department of Family, Youth and Community Sciences, University of Florida, Gainesville FL,
United States
d Department of Computer & Information Sciences & Engineering, University of Florida,
Gainesville FL, United States
* Corresponding author: Luyu Liu, luyuliu@auburn.edu
2046J Haley Center, Auburn University, Auburn, AL 36849, United States
Abstract
Understanding food acquisition is crucial for developing strategies to combat food
insecurity, a major public health concern. The emergence of large-scale mobile location
data (typically exemplified by GPS data), which captures people’s movement over time
at high spatiotemporal resolutions, offer a new approach to study this topic. This paper
evaluates the potential and limitations of large-scale GPS data for food acquisition
analysis through a case study. Using a high-resolution dataset of 286 million GPS records
from individuals in Jacksonville, Florida, we conduct a case study to assess the strengths
of GPS data in capturing spatiotemporal patterns of food outlet visits while also
discussing key limitations, such as potential data biases and algorithmic uncertainties.
Our findings confirm that GPS data can generate valuable insights about food acquisition
behavior but may significantly underestimate visitation frequency to food outlets.
Robustness checks highlight how algorithmic choices-especially regarding food outlet
classification and visit identification-can influence research results. Our research
underscores the value of GPS data in place-based health studies while emphasizing the
need for careful consideration of data coverage, representativeness, algorithmic choices,
and the broader implications of study findings.
Keywords: Food security; Food acquisition; Human mobility data; GPS data;
Spatiotemporal analysis
2
1. Introduction
Food insecurity, which refers to the lack of stable access to sufficient, safe, and nutritious
food for a healthy, active life (Simelane & Worth, 2020), is a major public health concern
(Caspi et al., 2012). In 2023, 13.5% of U.S. households experienced food insecurity
(Matthew P. Rabbitt et al., 2025). A critical aspect of addressing this issue is
understanding food acquisition, i.e., the processes and actions through which people
obtain food. Examining how households acquire food from food-selling places to their
home provides essential insights for developing effective interventions and policies to
combat food insecurity (Rabbitt et al., 2023).
Prior studies have commonly used surveys to gather detailed data on household
food acquisition behavior, including shopping habits, spending, and food consumption
(Coleman-Jensen et al., 2019). Well-structured surveys enable comparisons of food
acquisition behavior across groups and longitudinal analyses (Anekwe & Zeballos, 2019).
Also, interviews and focus groups provide qualitative insights such as how cultural and
socio-economic factors influence food acquisition (Shier et al., 2022). However, these
methods usually rely on personal memories to report the spatiotemporal information
associated with food store visits, which may result in bias, inaccuracies, and limited
spatial and temporal insights (Hillier et al., 2017). In recent years, researchers have used
Global Positioning System (GPS) devices to augment traditional approaches, such as
conducting geo-tagged surveys (Elliston et al., 2020) as well as distributing tracking
devices to record geo-fenced visits (Wray et al., 2023) or mobility trajectories (Zenk et
al., 2011). By tracking people’s movement with high spatial and temporal resolution
(Chen et al., 2016), GPS data allows one to reconstruct their activity-travel pattern and
support more detailed spatiotemporal analyses. For instance, researchers can examine
food exposure based on activity spaces, beyond one’s home or workplace (Elliston et al.,
2020). However, collecting GPS-augmented survey data is resource-intensive and so only
small sample sizes are typically achieved, making it challenging to engage representative
participant groups. Cetateanu and Jones (2016) and Siddiqui et al. (2024) reviewed the
papers on GPS and food environment exposure and reported sample sizes ranging from
12 to 654 individuals. Moreover, the awareness of carrying monitoring devices can also
lead to increased consciousness of actions and potential behavioral changes over the study
period (Zhang et al., 2021).
With the widespread use of location-enabled mobile devices such as smartphones
and smart watches, large-scale GPS data collected by location-tracking apps are
increasingly common (Smith et al., 2023; Xie et al., 2023; Zhou et al., 2022). These
datasets, made available to the research community by data vendors such as SafeGraph,
Cuebiq, and Gravy Analytics, capture the movements of millions of individuals over
extended periods (e.g., weeks, months, and years), with spatial and temporal resolutions
in meters and seconds. Such data offer a new approach to study human mobility (Kwan,
2016; K. Zhao et al., 2016). For example, they have been used to analyze neighborhood-,
region- or even national-level associations between food outlet visits and related health
and social outcomes (Chang et al., 2022; Hu et al., 2021; Xie et al., 2023; R. Xu et al.,
2023). So far, most published works analyze these data at the aggregate levels such as
census tracts or points-of-interest (POIs), partly because SafeGraph made its aggregate-
level mobile location data freely available to researchers around the world from 2020 to
2023 (Xie et al., 2023; K. Zhao et al., 2016). While such data provide valuable insights
on human mobility at the aggregate level, they do not allow one to examine behavioral
differences at the individual level. Also, since these data are pre-processed by vendors
using proprietary algorithms, the research community has limited insight into potential
3
biases in both the derived mobility metrics (e.g., food outlet visits) and the underlying
algorithms. Despite their large sample sizes and high resolution, GPS datasets often have
uneven spatial and temporal coverage and can exhibit sample bias, such as the
underrepresentation of disadvantaged groups (Li et al., 2023). Moreover, other
researchers have noted that analytical choices in processing GPS data—such as algorithm
design and parameter setting—can significantly impact the reliability and accuracy of the
derived mobility measures (Kwan, 2016).
In light of the above discussions, this paper analyzes a unique dataset—GPS
mobility traces from millions of individuals in Florida—to explore the potentials and
limitations of using large-scale, longitudinal, individual-level GPS data for food
acquisition analysis. We first present the analytical steps involved to process raw GPS
data collected from smart mobile devices into food-acquisition-related metrics for further
analysis. By breaking down the algorithmic considerations and parameter choices made
in each step, we reveal how algorithmic uncertainties in GPS data processing can
influence the results. Subsequently, using a case study of Jacksonville, Florida, we discuss
the advantages of leveraging large-scale GPS data for food acquisition analysis compared
to traditional methods, such as surveys, as well as the limitations posed by potential data
biases and algorithmic uncertainties. Specifically, we identify the novel insights GPS data
can provide by examining spatiotemporal patterns of food acquisition behavior at the
individual level. We also explore the potential limitations by examining data
representativeness and evaluating the robustness of study findings to several key
parameter choices.
2. Processing GPS Data for Food Acquisition Analysis
To obtain food acquisition metrics from raw mobile location data (i.e., GPS data) for
further analysis requires extensive data processing. Figure 1 shows the key analytical
steps we have employed here, where many algorithmic considerations are informed by
the existing literature. Specifically, the process begins with processing the mobile phone
GPS data to infer each device user’s home location and activities (or stays) so as to obtain
their activity-travel patterns. Next, we integrate GPS data and food outlet location data to
infer individual food acquisition behavior. Two key considerations here are regarding 1)
which food outlets to be included for food acquisition analysis and 2) the proximity
criteria used to assign an observed activity (or stay) close to a food outlet as a “food
acquisition visit.” Based on the inferred food acquisition visits, we can then calculate food
acquisition metrics and analyze their spatiotemporal patterns. We now discuss each key
step in detail.
4
Figure 1. Analytical steps to obtain food acquisition metrics from GPS data
2.1. Home Location and Activity Stay Inference
Considering that home location is crucial for analyzing food acquisition behavior
(Coleman-Jensen et al., 2019; Zenk et al., 2018), we inferred home location for device
users. In this study, we use the proxy-home-location algorithm developed in (X. Zhao et
al., 2022). This approach segments the study area into 20-meter grids, counts GPS points
recorded between 10:00 PM and 6:00 AM within each grid, and designates the grid with
the highest GPS point density as the home location. For users whose home locations could
not be inferred due to a lack of nighttime GPS points (approximately 16.2% of device
users), we leveraged weekend data. Assuming users typically spend most of their daytime
at home on weekends, we used GPS points recorded between 6:00 AM and 10:00 PM to
estimate their home locations.
Moreover, we extracted activity stays from the GPS data using the Trackintel
package, which employs a time-space heuristic method. This method identifies activity
stays (referred to as stops in the package) as periods of minimal movement and detects
them using a sliding window algorithm that clusters points (Martin et al., 2023). In this
study, we set thresholds for stop detection to a 100-meter radius and a duration between
5 and 720 minutes. To reduce outliers, we excluded activity stays exceeding two hours,
aligning with the American Time Use Survey (ATUS), which reports a median grocery
shopping duration of 30 minutes with a standard deviation of 30.6 minutes (Brown &
Borisova, 2007). Additionally, we extracted the origins of those activity stays using the
package’s backward searching method, which clusters spatiotemporally linked GPS
points.
5
2.2. Identifying Food Outlet Visits
The activity stays derived from GPS data do not contain information about activity
purposes, which is typically inferred based on the proximity of an activity location to
adjacent places of interest (SafeGraph, n.d.). When identifying food outlet visits from
GPS data, a key consideration is determining which food-selling stores to include. One
should also classify food outlets, which can vary significantly in the type, price, quantity,
and variety of food they offer. Prior research on food acquisition has shown that regional
food environments consist of a wide variety of food outlets, each influencing community
health in different ways and exhibiting distinct visitation patterns (Balagtas et al., 2023;
Shier et al., 2022; Todd & Scharadin, 2016). Another complicating factor involves
individuals also visiting different food outlets for multiple purposes; for example, some
visit big box stores for non-food items and gas stations for food. USDA’s FoodAPS has
indicated that Supplemental Nutrition Assistance Program (SNAP) households allocate
13% of their food spending to convenience stores, dollar stores, and pharmacies (Todd &
Scharadin, 2016).
Here we developed a two-dimensional classification approach to categorize food
outlets (Todd & Scharadin, 2016; R. Xu et al., 2023). First, food locations were grouped
into four groups based on by size and the quality: large groceries/supermarkets that
primarily sell food (e.g., Trader Joe’s and Aldi); big-box stores that offer a full range of
food products along with other goods (e.g., Walmart Supercenter); small healthy food
outlets that carry healthy items include supplements and medicines (i.e. drug stores) and
grocery items like fresh produce, diary and eggs (e.g., food marts, dollar stores)
1
; and
processed food outlets that sell only processed and low-nutrition food (e.g., gas station
stores and food store with processed food only). The second dimension classified
locations by the confidence of food acquisition when visiting them: outlets primarily
selling food where individuals predominantly visit for food acquisition (e.g., groceries,
food marts) versus those visited for various other purposes (e.g., big-box stores, gas
stations).
2.3. Food Acquisition Metrics
We then integrated the food outlet dataset and activity patterns derived from GPS data to
infer food outlet visits. Specifically, we identified food outlet visits from activity stays
using a buffer-based approach, assigning predefined radii based on outlet types: 50 meters
for Small Healthy Outlets and Processed Food Outlets, 200 meters for Big-box Stores,
and 150 meters for Large Groceries.
2
Based on the inferred food outlet visits, we further
calculated four metrics widely discussed in the food access literature to assess food
1
In this study, we classify dollar stores as small food outlets that carry healthy items, despite their typical
association with unhealthy food options and their potential contribution to food deserts by limiting the entry
of supermarkets (Chenarides et al., 2021a), for the following reasons: First, increasing research highlights
the role of dollar stores in enhancing food accessibility and security by providing affordable, widely
available produce and other healthy options (John et al., 2023). These stores may help fill gaps in food
access, particularly in areas where other retailers are unwilling to operate (Chenarides et al., 2021a).
Furthermore, an increasing number of dollar stores now carry fresh produce and dairy products. Ongoing
initiatives and implementations by dollar store companies have highlighted their commitment to expanding
fresh produce offerings. For example, as of Q1 2023, Dollar General has introduced fruits and (continue)
vegetables in nearly 3,900 of their locations (Dollar General Corporation, n.d.). Notably, we differentiate
gas station stores from dollar stores, as the former primarily offer processed or low-quality food.
2
Section 3.4 provides additional details to explain these radius choices.
6
acquisition patterns (Leroy et al., 2015; Todd & Scharadin, 2016).
First, we calculated the number of food outlet visits made by each individual
during the study period, mirroring survey-based approaches that measure food acquisition
frequency. To capture diversity in food acquisition, we counted the number of unique
stores visited by each user. This metric resembles traditional surveys that assess the
diversity of people’s food sources. Regarding spatial metrics used in traditional methods
to assess food accessibility, we computed the home-to-store distance, defined as the
network distance (based on Open Street Map data) between the user's inferred home
location and visited food outlets. We calculated the average distance from a user's home
to all food outlets they visited, as well as the distance to the nearest food store from their
home. Lastly, we calculate the proportion of home-based food outlet visits
3
, a metric
widely used in prior work to examine food purchases in relation to home location so as
to understand the connection between residential food environments and food acquisition
behaviors (Thornton et al., 2017; Ver Ploeg et al., 2015).
These metrics would allow us to assess how well GPS data can replicate and
enhance traditional survey approaches for food acquisition analysis. Specifically, we will
analyze visitation patterns of various food outlets using GPS data to assess their alignment
with traditional survey findings and reliability. Further exploring the temporal trends (e.g.,
time-of-day and day-of-week patterns) and spatial aspects (e.g., home-to-store distance
and its distribution) allows us to explore novel insights generated by GPS data not easily
captured in traditional survey studies.
2.4. Algorithmic Uncertainties
The above discussion suggests that, when using GPS data for food acquisition analysis,
one must make many algorithmic choices in data processing. These choices—often made
based on the literature findings or the analyst’s subjective judgment—may introduce
significant uncertainties to study results. Notably, prior work on leveraging large-scale
GPS data for human mobility analysis highlights two key challenges affecting activity
detection and classification (Kwan, 2016): (1) determining the location relevant to the
activity and (2) inferring the activities performed at those locations. Here we discuss the
algorithmic uncertainties associated with these two challenges for food acquisition
analysis.
(1) Uncertainty in food outlet visit identification. Radius is a crucial parameter
in visit attribution, as it determines whether an activity inferred from the GPS data is
considered a visit to a food outlet. A smaller radius risks missing visits (false negatives),
while a larger radius may capture unrelated visits (false positives). For example,
supermarkets may have visits missed with a smaller radius (Figure 2, left), while small
grocery stores may capture unrelated visits with a larger one (Figure 2, right). In this study,
we test radii of 50m, 100m, 150m, and 200m to evaluate how this parameter affects the
robustness of our GPS-based food acquisition analysis.
3
Here we defined home-based food outlet visits as those originating within 200 meters of the inferred home
location. The 200-meter radius—as opposed to the residential parcel—is used to mitigate the potential
impacts of GPS location errors (up to 100 meters).
7
Figure 2. Identification radii in food-related trip extraction
(2) Uncertainty in Activity Purpose Determination. Visits to the same location
may serve different purposes. In the context of food acquisition, visits to food outlets may
not always be for food acquisition. To explore this issue, we considered two food outlet
inclusion criteria to calculate food acquisition metrics: (a) including all locations and (b)
limiting analysis to primary food-selling locations. For instance, for visits to small healthy
food outlets, the former include dollar stores and drug stores but latter does not. Excluding
non-primary-food-selling locations offers a more conservative estimate, potentially
omitting some food acquisition visits but increasing the likelihood that the identified visits
accurately represent actual food acquisition.
3. Case Study: GPS-data-based Food Acquisition Analysis in Jacksonville, Florida
We further conduct a case study in Jacksonville, Florida to shed light on the potentials
and limitations of using large-scale, individual-level GPS data for food acquisition
analysis. After describing the case study area and data sources, we investigate the
sampling rate and inferred food acquisition visit to assess data coverage and
representativeness. Next, we examine key food acquisition metrics derived from GPS
data and the associated spatiotemporal patterns to evaluate to what extent GPS data can
replicate traditional survey approaches for food acquisition analysis and generate novel
insights. Finally, we conduct robustness checks to assess how algorithmic uncertainties
in food outlet visit identification and activity purpose determination shape the results of
GPS-data-based food acquisition analysis.
3.1. Case Study Area and Data Sources
Our case study area is the City of Jacksonville, Florida, the largest municipality in the
state. The city’s demographics reveal a complex socioeconomic landscape. According to
the American Community Survey (ACS) 2018-2022 five-year estimates (U.S. Census
Bureau, 2022), the population of over 950,000 includes 53.1% White, 30.4% Black, and
11.3% Hispanic or Latino, and with a median age of 36.3 years, a median household
income of $64,138, and 14.8% living in poverty. Figure 3 illustrates the spatial
distribution of sociodemographic characteristics by census tracts. As shown in the figure,
urban tracts generally exhibit higher population density, greater percentage of individuals
aged 18-39, lower household vehicle ownership, and higher poverty rates. The northwest
part of the study area demonstrates a lower percentage of the White population.
8
Figure 3. Tract-level socio-demographic characteristics in Jacksonville, Florida
This study uses a large-scale mobile device location dataset from Gravy Analytics,
which aggregates data from over 150 million U.S. mobile devices through various apps
(Gravy Analytics, 2023b). The data vendor suggests that it complies with privacy laws,
sourcing data only from users who opt in, with a 48- to 72-hour processing delay (Gravy
Analytics, 2023a). The dataset is also pre-processed for positioning errors, with accuracy
indicated by a forensic identifier field (Gravy Analytics, 2023c; Y. Xu et al., 2022). For
this study, we included only records classified as “high accuracy,” where GPS positioning
errors do not exceed 35 meters. This pre-analysis filtering minimizes positioning errors,
ensuring more reliable results. After pre-processing, we retained 286.4 million
disaggregated records from September 1 to October 15, 2022 (45 days). The data fields
included device identifiers, latitude, longitude, geohash, and timestamp.
Another key data source is a comprehensive food outlet database for North Florida
developed by the University of Florida GeoPlan Center, covering various components of
the local food system, including food production, retail, and distribution sites (Alachua
County, 2022). Our study focuses on food retail outlets that provide food-at-home (FAH)
access, including grocery stores, supermarkets, drug stores, corner stores, gas station
convenience stores. Figure 4 shows the distribution of the four food outlet types in
Jacksonville, based on the classification in Section 2.2.
Figure 4. Distribution of various types of food outlets in the study area
9
3.2. Data Coverage and Representativeness
3.2.1. Sampling Rate
Figure 5 shows the spatial distribution of the 93,854 individual device home locations
and the histogram of the sampling rate across census tracts. On average, 10.4% of each
tract’s population, based on census estimates, is represented in our sample, highlighting
that GPS data achieves a better coverage than survey. Regarding the spatial distribution
of sampling rates across census tracts, the map suggests a reasonable spatial coverage and
the histogram shows a normal-like distribution centered around 8%. However, the two
figures also show that sampling rate varies significantly across census tracts, which
indicate potential spatial bias.
Figure 5. Distribution of sampling rate across census tracts
3.2.2. Inferred Food Outlet Visits
We extracted a total of 852,224 food outlet visits, with the following distribution: 250,916
for Large Groceries, 76,979 for Big Box Stores, 191,796 for Small Healthy Food Outlets,
and 332,533 for Processed Food Outlets. The number of visits captured is significantly
larger compared to traditional surveys
4
.
3.3. Understanding Food Acquisition Patterns with GPS Data
3.3.1. Food Acquisition Metrics
In this subsection, we analyze various food acquisition metrics to examine whether the
results derived from GPS data align with or differ from those obtained from surveys.
Based on the inferred food outlet visits, we analyzed visitation frequency, unique stores
visited, home-to-store distances, and percentage of home-based visits. Table 1 presents
the population-averaged metrics derived from our GPS data.
Table 1. Food acquisition metrics for each type of store
Metrics
Large
Groceries
Big
Box
Stores
Small
Healthy Food
Outlets
Processed
Food
Outlets
All
Food
Outlets
4
For the “proportion of home-based visits” metric, we considered only visits with inferred origins, resulting
in 1,336 visits for Large Groceries, 646 for Big Box Stores, 801 for Small Healthy Food Outlets, and 1,808
for Processed Food Outlets. The significant reduction in number of visits is because inferring the origin of
these visits requires continuous GPS tracking of the trip trajectory, which is not available for most visits.
10
Number of visits per individual in
1.5 months (visits)
4.74
3.12
4.1
5.02
9.08
Number of unique stores visited per
individual in 1.5 months (stores)
1.86
1.31
1.93
2.38
3.85
Distance of visited store
to home (km)
Euclidean
5.29
6.54
5.63
5.8
5.62
Network
7.43
8.69
7.21
7.47
7.61
Proportion of home-based visits (%)
18.65
14.44
18.33
16.02
17.84
The average number of food outlet visits per individual over 1.5 months is 9,
equating to 1.4 visits per week. In contrast, USDA FoodAPS reports higher rates for the
South US: 2.86, 1.33, and 2.37 food acquisition events per week at Large grocery stores,
Small and specialty stores and All other food-at-home stores, respectively—equating to
18.4, 8.5, and 15.2 over 1.5 months (Todd & Scharadin, 2016). A Florida survey with
1,594 respondents reports more comparable values, finding that the most common
grocery store shopping frequency was weekly (Hodges & Stevens, 2013). We shall
should note that the referenced studies are household-based, whereas our analysis relies
on individual devices, which may contribute to underestimation (Ver Ploeg et al., 2015).
Moreover, we observed a significant underestimation in the proportion of home-based
visits, as ATUS reports 64% of grocery shopping trips begin and end at home (Ver Ploeg
et al., 2009). A main source for these underestimations is the gaps in GPS location
tracking, which leaves out some activities and trips taken by the device owner. On the
other hand, the number of unique stores visited is closer to the 3.5 retail banners visited
per month reported in a national sample of 2,091 grocery shoppers (FMI and the Hartman
Group, 2022). Moreover, the GPS data’s home-to-store distance aligns closely with the
literature, showing an average of 3-4 miles (4.8-6.4 km) (Euclidean distance, to primary
store) (Todd & Scharadin, 2016).
The findings echo previous literature, emphasizing that reliable and valid
measurement is essential when using mobility data to study food environments (Zenk et
al., 2018). Inconsistent tracking may lead to missing records, resulting in the
underestimation of food acquisition activity and a low proportion of home-based visits,
which are familiar environments to individuals. Despite these limitations, GPS data show
potential into capturing actual locations visited that were tracked and distances travelled,
which are not always well-captured in self-reported surveys. While not fully replicating
survey methods, GPS data complements food access research by offering more accurate
insights into individuals' actual movements.
3.3.2. Spatial and Temporal Patterns
Mobile GPS location data continuously tracks millions of people’s movements with high
spatial resolution, enabling analysts to derive novel spatial and temporal insights that
enhance the understanding of food acquisition behavior. This is a key advantage that GPS
data has compared to traditional surveys, which usually lack the sample size, spatial and
temporal coverage, or granularity for detailed spatiotemporal analyses. To empirically
test and validate this point, we analyze the spatial patterns of home-to-store distances and
the temporal variation in food outlet visitation. These aspects of food acquisition are
important yet remain underexplored in survey-based studies due to smaller sample sizes
and lower spatial or temporal resolution (Todd & Scharadin, 2016; Ver Ploeg et al., 2015).
(1) Spatial Patterns
11
Figure 6 shows the distribution curves of the distance from individual home to store. The
left-side figure, depicting the home-to-nearest-store distance, reveals distinct
distributions by food outlet type. These distance distributions reflect the densities of
various types of food outlets, where a longer distance indicates a smaller density.
Interestingly, the right-side figure, showing home-to-visited-store distances, reveals
similar distance distributions across outlet types, suggesting that individuals often bypass
the nearest store to shop at stores further from home (J. L. Liu et al., 2015).
Figure 6. Distribution of home-to-store distances within the population
Figure 7 presents scatter density plots, with nearest- and visited-store distances on
the x and y axes, respectively. Darker colors indicate higher density. The slopes in the
figure indicate the degree to which people prefer closer stores: a slope of 1 means
exclusive visits to the nearest store, while a slope of 0 means distance is not a factor. Big-
box Stores show the steepest contours, indicating a strong preference for nearby locations,
followed by Large Groceries. In contrast, Small Healthy Outlets and Processed Food
Outlets have flatter contours, suggesting distance is less influential in choosing these
stores. This pattern likely results from business strategies that standardize Big-box Stores
and Large Groceries, making location differences less significant, while the diversity in
quality and price at Small Healthy Outlets and Processed Food Outlets drives people to
visit specific locations.
Figure 7. Density contours of home-to-store distances
We aggregated individual-level measurements to the tract level in Figure 8,
mapping home-to-nearest store distance, home-to-visited store distance, and their
12
difference. Darker colors indicate larger distances. The first row shows greater nearest-
store distances in rural areas, reflecting limited food access. However, the second row
reveals large visited-store distances in some urban areas, aligning with lower-income,
high-density, and predominantly non-white populations, consistent with Jacksonville’s
urban food deserts phenomenon (Lewis et al., 2018). The last row highlights where
individuals bypass closer stores. Big-box stores and large groceries generally show
smaller differences, consistent with Figure 7. However, in the central-west area, Big-box
Store deviations are greater, corresponding to higher-income, predominantly white
populations, possibly reflecting limited nearby options or store preferences. In contrast,
larger deviations for Large Groceries visits appear in the northwest, where more
underprivileged populations reside, likely due to affordability or cultural preferences.
Small Healthy Food and Processed Food outlets exhibit larger differences but with less
clear spatial patterns.
Figure 8. Spatial distribution of home-to-store distances
(2) Temporal Pattern
Prior research on food acquisition has shown distinct temporal visitation patterns across
food outlets. For example, fast food visits peak on weekdays, while supermarkets see
more traffic on weekends (East et al., 1994; García Bulle Bueno et al., 2024). Building
on these findings, we leveraged GPS data to analyze temporal patterns in food outlet
visitation across different temporal dimensions, including day of the week, time of day,
and daily variations.
Figure 9 presents the patterns. The time-of-day curves (first row) reveal generally
similar weekday and weekend patterns. Though, weekday visits vary more by outlet type
13
with the two large outlet types showing clearer daytime peaks; weekend patterns are more
uniform (except for Big-box Stores), with a weaker evening peak but a stronger midday
peak, consistent with literature on weekend activity shifts (East et al., 1994). The day-of-
week trends (second row) highlight a Friday peak in general and a higher weekend share
for Big-box Stores, trends also noted in prior studies (Cai, 2006). The last row presents
the longitudinal trends in number of visits extracted each day. It shows overall stability
with a notable drop around Labor Day (September 5, 2022), suggesting a temporary shift
in food shopping behavior due to the holiday.
Figure 9. Temporal patterns of food acquisition activities
3.4. Robustness Checks
In Section 2.4, we have discussed that algorithmic uncertainties can pose a major threat
to the validity and reliability of study findings on food acquisition behavior derived from
GPS data. This section presents results from the robustness checks we have performed to
address the two key challenges associated with leveraging GPS data for food acquisition
analysis: uncertainty in food outlet visit identification and uncertainty in activity purpose
14
determination.
3.4.1. Uncertainty in Food Outlet Visit Identification
Figure 10 shows the metrics calculated with food outlet visits identified under different
radii. The general trend remains consistent across radius variations, with no drastic
changes in patterns. However, the number of visits shows a notable increase for Small
Healthy Outlets and Processed Food Outlets as the radius expands from 50m to 100m,
while the increase for Big-box Stores is more gradual. This supports the expectation that
food outlet visit identification is sensitive to radius selection, especially for smaller outlets,
which have a limited venue area and are often clustered with other points of interest
(Figure 4). These findings informed our parameter choices for the case study, that is, we
selected a 50m for the two smaller outlet types, 200m for Big-box Stores, and 150m for
Large Groceries.
Figure 10. Food access metrics calculated under different radii
3.4.2. Uncertainty in Activity Purpose Determination.
Table 2 and Figure 11 present the food acquisition metrics derived from GPS data when
we focus solely on primary food-selling locations. The results differ from those show in
Table 2 and Figure 6 in that visits to Big-box Stores are now excluded.
As expected, we observed a decrease in total visits and the number of unique
stores visited compared to the results shown in Table 2. Focusing on primary food-selling
locations also led to smaller home-to-store distances and more home-based visits,
aligning with findings from activity-space studies which shown maintenance activities to
have smaller radii (Gong et al., 2020). Differentiating store types revealed distinct
patterns. Visits to Processed Food Outlets showed similar trends, while visits to Large
Groceries saw increased home-to-store distances, reduced home-based visits, and more
frequent trips to more unique stores. These differences likely arise from the behaviors of
individuals who visit exclusively food-selling stores versus those who shop at outlets with
broader inventories. Shoppers at specialized grocery stores visit more often and travel
farther, while those at gas stations or dollar stores tend to visit more frequently than fast-
15
food or specialized processed outlets, consistent with findings in the literature (Todd &
Scharadin, 2016; Ver Ploeg et al., 2015).
Table 2 Food acquisition metrics by food outlet type (primary food-selling locations
only)
Metrics
Large
Groceries
Small
Healthy
Food
Outlets
Processed
Food
Outlets
All Food
Outlets
Number of visits per individual in
1.5 months (visits)
5.13
3.47
6.04
5.13
Number of unique stores visited per
individual in 1.5 months (stores)
1.87
1.46
2.41
1.87
Distance of visited store to home
(km)
8.15
7.82
7.03
7.52
Proportion of home-based visits
(%)
17.95
21.62
18.9
17.95
Figure 11. Distribution of home-to-store distances (primary food-selling locations only)
Overall, the robustness checks suggest that the food outlet visit identification and activity
purpose determination methods are sensitive to algorithmic uncertainties. Radius
selection and outlet classification can influence the number of visits, store type
differentiation, and distances traveled. Despite these uncertainties, the methods still
provide valuable insights into food access patterns. Adjustments to algorithmic
parameters can allow for more accurate representation of food acquisition behavior,
enhancing the reliability of study findings.
4. Discussion
4.1. The Potential of Using GPS Data for Food Acquisition Analysis
Our case study highlights the potential of using GPS data to significantly enhance food
acquisition analysis. GPS data can be an effective tool for studying food acquisition,
offering advantages over traditional methods.
16
First, our study demonstrates the capability of GPS data in capturing food
acquisition patterns at the individual level with a large sample size, which is often
challenging with traditional methods. GPS enables the calculation of individual food
acquisition metrics as well as their spatial and temporal distribution across large areas and
extended periods, offering a more comprehensive understanding than conventional
surveys. This broad yet high-resolution coverage is particularly valuable for informing
policy design and evaluation. For example, policymakers can monitor food security
programs’ effectiveness spatially across areas and assess their impact temporally by
tracking changes in visitation patterns to food outlets before and after program
implementation. This allows for more efficient adjustments to policies based on long-
term trends.
Additionally, GPS data offer flexibility in defining and refining analytical
approaches to suit different research questions with the same dataset. For instance, by
adjusting classification criteria, we examined food acquisition patterns both with and
without non-traditional food outlets. The results showed that including gas stations and
dollar stores increased visits per individual from 3.47 to 5.02 and unique stores visited
from 1.46 to 2.41. This flexibility can provide new angles for assessing interventions,
such as the impact of introducing healthy produce into dollar stores on food accessibility
and security—an important topic in food security research (Chenarides et al., 2021b; John
et al., 2023; Lucan et al., 2018).
4.2. GPS Data’s Limitations for Studying Food Acquisition
Our analysis also reveals some key limitations that influence the validity and reliability
of study findings derived from the use of GPS data in food acquisition research.
First, despite large sample sizes, our GPS data exhibit spatial sampling biases,
with varying rates across areas. This imbalance can contribute to ecological fallacy, where
aggregate trends misrepresent individual behaviors (Chen et al., 2016). In food
acquisition analysis, this is particularly problematic, as it affects the equity of food access
interventions. Marginalized communities face disproportionate barriers to healthy food
(Jin et al., 2023), prior studies show GPS data tend to underrepresent them (Coston et al.,
2021; Li et al., 2023; Squire, 2019). Future research should address these biases and
exploring the role of sociodemographic factors to improve understanding (Singleton et
al., 2023).
Another issue is inconsistent tracking. From the case study, we found that food
outlet visitation frequencies derived from GPS data are lower than those reported in
surveys. These discrepancies could be affected by whether and when a user activates
location-tracking for their mobile devices and other factors such as signal coverage and
strength across space. Regarding food acquisition, many individuals may not use
navigation apps for routine grocery trips, especially those originated from their home,
which may result in underestimations of food outlet visits. This introduces another source
of bias in GPS-based food acquisition analysis, potentially compromising both the
internal and external validity of the study findings.
Additionally, the robustness checks we have performed in this study affirm that
assumptions, algorithmic choices and parameter settings used in the process of inferring
food acquisition metrics from GPS data can significantly influence study results (Kwan,
2016). Although we found that the general patterns and trends were consistent across the
decision space, algorithmic uncertainties clearly had a major influence on both the content
17
and reliability of food acquisition result results. As big data such as mobile location data
are increasingly used for human mobility analysis, our work stresses the importance of
paying attention to the algorithms used throughout the analytical process.
4.3. Study Limitations and Future Research
A key limitation of this study is that we have attempted to shed light on the potentials and
limitations of using GPS data for food acquisition analysis based on a single case study.
The specific GPS dataset used here may not be representative of other GPS datasets used
in the research community. Also, as briefly mentioned in the results section, GPS data
tracks individuals rather than households, whereas food shopping is often a household-
level activity, with multiple individuals from the same household visiting food outlets
together (Todd & Scharadin, 2016). Our work can thus be enhanced by performing a
household-level analysis. Moreover, since the case study is based on Jacksonville, FL, its
findings may be influenced by the unique characteristics of the study area, potentially
limiting the external validity of the results. Due to resource constraints, we could not
access multiple popular GPS datasets. Future research can build on our work by analyzing
multiple GPS datasets across diverse study areas to enhance validity and generalizability.
In addition, there can be concerns about the temporal and spatial generalizability
of the empirical findings. Temporally, research on food sales has shown seasonality in
food demand (Balagtas et al., 2023; Hu et al., 2021), which may influence mobility
patterns. Therefore, generalizing our findings from the 45-day study period could
introduce bias and limit the representativeness of the results. Spatially, study findings
from Jacksonville may not be transferable to other contexts. A 2012 study noted
disparities in food acquisition among Health Zones within the city, with Urban Core
residents facing a greater health burden (Healthy Jacksonville Children Obesity
Prevention Coalition, 2012). Additionally, Jacksonville’s poverty rate (14.8%) exceeds
both the national (12.5%) and state averages (12.9%) (U.S. Census Bureau, 2022). These
socio-economic factors should be considered when generalizing the results to other
contexts. Future research can apply the analytical steps outlined in Section 2 to explore
food acquisition patterns across different locations and time periods, which can validate
our empirical findings.
5. Conclusion
Mobile location data provides a novel approach to studying human mobility. This study
presents a systematic analysis of the potentials and limitations of using large-scale,
individual level GPS data for food acquisition analysis. Using a large-scale mobile
location dataset with 286 million GPS records, we conducted a case study in Jacksonville,
Florida. We inferred several food acquisition metrics commonly examined in the
literature and explored their spatial and temporal patterns. The results demonstrate the
capability of GPS data to extract key insights regarding food acquisition patterns that
confirm findings from prior survey-based studies. On the other hand, our analysis
suggests that relying on GPS data would significantly underestimate food outlet visitation
frequency. The robustness checks, focusing on examining how algorithmic uncertainties
in the classifications of food-selling stores and identification radii of food outlet visits
shape research results, affirm general patterns and trends across the decision space but
also reveal some inherent challenges of extracting food acquisition visits from GPS data.
18
Overall, this study confirms the potential of GPS data for analyzing food
acquisition while also underscoring the need for careful interpretation and application.
Our research highlights the need for critical reflexivity with respect to data coverage and
representativeness, algorithmic choices, and the findings generated from them. We
suggest future research to apply a mixed-methods research design and integrate diverse
data sources to gain a more holistic understanding of people’s food acquisition patterns.
Studies integrating small-scale GPS data with surveys have uncovered novel insights that
challenged previously held behavioral assumptions (B. Liu et al., 2020; Sadler et al.,
2016). Similarly conducting surveys on the same population from which large-scale GPS
data are collected can allow for the triangulation of results and enhance the
generalizability of study findings, ultimately contributing to more effective interventions
and policies for combating food insecurity.
Author contributions
Conceptualization: XY, CC, LY
Data Curation: YZ, DL
Formal Analysis: DL
Methodology: XY, CC, LY
Supervision: XY, CC, LY
Visualization: DL
Writing – Original Draft: DL, LL
Writing – Review and Editing: XY, CC, LY, DL
Declarations of Interest statement
The authors report there are no competing interests to declare.
Funding
This work was supported the US Department of Transportation Center for Transit-
Oriented Communities Tier-1 University Transportation Center (Grant No.
69A3552348337) and the National Science Foundation (Award # 2416202).
Acknowledgements
We thank Erik Finlay at the University of Florida GeoPlan Center for sharing the North
Florida food retailer location database with us.
Data statement
The data that support the findings of this study are available from Gravy Analytics.
Restrictions apply to the availability of these data, which were used under license for this
study. Aggregate-level data are available from the authors with the permission of Gravy
Analytics. All the codes are available at https://anonymous.4open.science/r/GPS-Food-
Accessibility-9B32.
19
References
Alachua County. (2022). Alachua County Foodshed Map Sample.
https://ufl.maps.arcgis.com/apps/instant/interactivelegend/index.html?appid=af5
fecad950a46089d23fd63a2dc835b
Anekwe, T. D., & Zeballos, E. (2019). Food-related time use: Changes and
demographic differences.
Balagtas, J. V., Cooper, J., McLaughlin, P., & Qin, F. (2023). Consumer demand for
food at home and food away from home: Understanding economic linkages
during the pandemic. Applied Economic Perspectives and Policy, 45(3), 1604–
1617. https://doi.org/10.1002/aepp.13361
Brown, C., & Borisova, T. (2007). Understanding commuting and grocery shopping
using the American Time Use Survey. Presentation at the International
Association of Time Use Research XXIX Conference, Washington DC.
Cai, J. (2006). Household production or leisure? A study of grocery and other shopping
(Preliminary Draft).
Caspi, C. E., Sorensen, G., Subramanian, S., & Kawachi, I. (2012). The local food
environment and diet: A systematic review. Health & Place, 18(5), 1172–1187.
Cetateanu, A., & Jones, A. (2016). How can GPS technology help us better understand
exposure to the food environment? A systematic review. SSM-Population
Health, 2, 196–205.
Chang, T., Hu, Y., Taylor, D., & Quigley, B. M. (2022). The role of alcohol outlet visits
derived from mobile phone location data in enhancing domestic violence
prediction at the neighborhood level. Health & Place, 73, 102736.
https://doi.org/10.1016/j.healthplace.2021.102736
Chen, C., Ma, J., Susilo, Y., Liu, Y., & Wang, M. (2016). The promises of big data and
small data for travel behavior (aka human mobility) analysis. Transportation
20
Research Part C: Emerging Technologies, 68, 285–299.
https://doi.org/10.1016/j.trc.2016.04.005
Chenarides, L., Cho, C., Nayga Jr, R. M., & Thomsen, M. R. (2021a). Dollar stores and
food deserts. Applied Geography, 134, 102497.
Chenarides, L., Cho, C., Nayga Jr, R. M., & Thomsen, M. R. (2021b). Dollar stores and
food deserts. Applied Geography, 134, 102497.
Coleman-Jensen, A., Rabbitt, M. P., Gregory, C. A., & Singh, A. (2019). Household
food security in the United States in 2018. USDA-ERS Economic Research
Report, 270. https://www.ers.usda.gov/webdocs/publications/94849/err-270.pdf
Coston, A., Guha, N., Ouyang, D., Lu, L., Chouldechova, A., & Ho, D. E. (2021).
Leveraging Administrative Data for Bias Audits: Assessing Disparate Coverage
with Mobility Data for COVID-19 Policy. Proceedings of the 2021 ACM
Conference on Fairness, Accountability, and Transparency, 173–184.
https://doi.org/10.1145/3442188.3445881
Dollar General Corporation. (n.d.). Find a DG With Produce Near You. Retrieved
February 10, 2025, from https://www.dollargeneral.com/fresh-produce
East, R., Lomax, W., Willson, G., & Harris, P. (1994). Decision making and habit in
shopping times. European Journal of Marketing, 28(4), 56–71.
Elliston, K. G., Schüz, B., Albion, T., & Ferguson, S. G. (2020). Comparison of
Geographic Information System and Subjective Assessments of Momentary
Food Environments as Predictors of Food Intake: An Ecological Momentary
Assessment Study. JMIR mHealth and uHealth, 8(7), e15948.
https://doi.org/10.2196/15948
21
FMI and the Hartman Group. (2022). U.S. Grocery Shopper Trends 2022: Shopper
Landscape. https://www.fmi.org/forms/store/ProductFormPublic/u-s-grocery-
shopper-trends-2022-april
García Bulle Bueno, B., Horn, A. L., Bell, B. M., Bahrami, M., Bozkaya, B., Pentland,
A., De La Haye, K., & Moro, E. (2024). Effect of mobile food environments on
fast food visits. Nature Communications, 15(1), 2291.
https://doi.org/10.1038/s41467-024-46425-2
Gong, L., Jin, M., Liu, Q., Gong, Y., & Liu, Y. (2020). Identifying Urban Residents’
Activity Space at Multiple Geographic Scales Using Mobile Phone Data. ISPRS
International Journal of Geo-Information, 9, 241.
https://doi.org/10.3390/ijgi9040241
Gravy Analytics. (2023a). Consumer privacy. Gravy Analytics.
https://gravyanalytics.com/consumer-privacy/
Gravy Analytics. (2023b). Frequently asked questions. Gravy Analytics.
https://gravyanalytics.com/frequently-asked-questions/
Gravy Analytics. (2023c, October). How to Use Gravy Forensic Flags for Retail and
Transportation Analysis. Gravy Analytics. https://gravyanalytics.com/blog/use-
gravy-forensic-flags-retail-transportation-analysis/
Healthy Jacksonville Children Obesity Prevention Coalition. (2012). Duval County food
summit fact sheet.
https://www.bcbsfl.com/DocumentLibrary/AboutOurCompany/HealthyKids_Ac
tionPlanRev.pdf
Hillier, A., Smith, T. E., Whiteman, E. D., & Chrisinger, B. W. (2017). Discrete choice
model of food store trips using National Household Food Acquisition and
22
Purchase Survey (FoodAPS). International Journal of Environmental Research
and Public Health, 14(10), 1133. https://doi.org/DOI: 10.3390/ijerph14101133
Hodges, A. W., & Stevens, T. J. (2013). Local food systems in Florida: Consumer
characteristics and economic impacts. Proceedings of the Florida State
Horticultural Society, 126, 338–345.
Horn, A. L., Bell, B. M., Bulle Bueno, B. G., Bahrami, M., Bozkaya, B., Cui, Y.,
Wilson, J. P., Pentland, A., Moro, E., & De La Haye, K. (2023). Population
mobility data provides meaningful indicators of fast food intake and diet-related
diseases in diverse populations. Npj Digital Medicine, 6(1), 208.
https://doi.org/10.1038/s41746-023-00949-x
Hu, Y., Quigley, B. M., & Taylor, D. (2021). Human mobility data and machine
learning reveal geographic differences in alcohol sales and alcohol outlet visits
across U.S. states during COVID-19. PLOS ONE, 16(12), e0255757.
https://doi.org/10.1371/journal.pone.0255757
Jin, A., Chen, X., Huang, X., Li, Z., Caspi, C. E., & Xu, R. (2023). Selective Daily
Mobility Bias in the Community Food Environment: Case Study of Greater
Hartford, Connecticut. Nutrients, 15(2). https://doi.org/10.3390/nu15020404
John, S., Sundermeir, S. M., & Gardner, K. (2023, October). Stretching the dollar:
Community-informed opportunities to improve healthy food access through
dollar stores. https://www.cspinet.org/sites/default/files/2023-
10/CSPI_dollarStoreReport_2023_4.pdf
Kwan, M.-P. (2016). Algorithmic Geographies: Big Data, Algorithmic Uncertainty, and
the Production of Geographic Knowledge. Annals of the American Association
of Geographers, 106(2), 274–282.
https://doi.org/10.1080/00045608.2015.1117937
23
Leroy, J. L., Ruel, M., Frongillo, E. A., Harris, J., & Ballard, T. J. (2015). Measuring
the food access dimension of food security: A critical review and mapping of
indicators. Food and Nutrition Bulletin, 36(2), 167–195.
Lewis, F. W., Mandal, A., Stevens, A., & Jones, C. (2018). Food deserts and food
insecurity: A case study in Jacksonville, Florida. Journal of Behavioral & Social
Sciences, 5(1).
Li, Z., Ning, H., Jing, F., & Lessani, M. N. (2023). Understanding the bias of mobile
location data across spatial scales and over time: A comprehensive analysis of
SafeGraph data in the United States. Available at SSRN 4383333.
Liu, B., Widener, M., Burgoine, T., & Hammond, D. (2020). Association between time-
weighted activity space-based exposures to fast food outlets and fast food
consumption among young adults in urban Canada. International Journal of
Behavioral Nutrition and Physical Activity, 17(1), 62.
https://doi.org/10.1186/s12966-020-00967-y
Liu, J. L., Han, B., & Cohen, D. A. (2015). Peer reviewed: Beyond neighborhood food
environments: Distance traveled to food establishments in 5 US Cities, 2009–
2011. Preventing Chronic Disease, 12.
Lucan, S. C., Maroko, A. R., Seitchik, J. L., Yoon, D. H., Sperry, L. E., & Schechter, C.
B. (2018). Unexpected neighborhood sources of food and drink: Implications for
research and community health. American Journal of Preventive Medicine,
55(2), e29–e38.
Martin, H., Hong, Y., Wiedemann, N., Bucher, D., & Raubal, M. (2023). Trackintel: An
open-source Python library for human mobility analysis. Computers,
Environment and Urban Systems, 101, 101938.
24
Matthew P. Rabbitt, Laura J. Hales, & Madeline Reed-Jones. (2025). Food Security in
the U.S. - Key Statistics & Graphics. https://www.ers.usda.gov/topics/food-
nutrition-assistance/food-security-in-the-us/key-statistics-graphics
Rabbitt, M. P., Hales, L. J., Burke, M. P., & Coleman-Jensen, A. (2023). Household
food security in the United States in 2022.
Sadler, R. C., Clark, A. F., Wilk, P., O’Connor, C., & Gilliland, J. A. (2016). Using
GPS and activity tracking to reveal the influence of adolescents’ food
environment exposure on junk food purchasing. Canadian Journal of Public
Health, 107(Suppl 1), eS14–eS20.
SafeGraph. (n.d.). Determining Points of Interest Visits From Location Data: A
Technical Guide To Visit Attribution. Retrieved March 6, 2025, from
https://www.safegraph.com/guides/visit-attribution-white-paper
Shier, V., Miller, S., & Datar, A. (2022). Heterogeneity in grocery shopping patterns
among low-income minority women in public housing. BMC Public Health,
22(1), 1612. https://doi.org/10.1186/s12889-022-14003-0
Siddiqui, N. Z., Wei, L., Mackenbach, J. D., Pinho, M. G. M., Helbich, M.,
Schoonmade, L. J., & Beulens, J. W. J. (2024). Global positioning system-based
food environment exposures, diet-related, and cardiometabolic health outcomes:
A systematic review and research agenda. International Journal of Health
Geographics, 23(1), 3. https://doi.org/10.1186/s12942-024-00362-x
Simelane, K. S., & Worth, S. (2020). Food and Nutrition Security Theory. Food and
Nutrition Bulletin, 41(3), 367–379. https://doi.org/10.1177/0379572120925341
Singleton, C. R., Wright, L. A., McDonald, M., Archer, I. G., Bell, C. N., McLoughlin,
G. M., Houghtaling, B., Cooksey Stowers, K., & Anderson Steeves, E. (2023).
Structural racism and geographic access to food retailers in the United States: A
25
scoping review. Health & Place, 83, 103089.
https://doi.org/10.1016/j.healthplace.2023.103089
Smith, L. G., Ma, M. Y., Widener, M. J., & Farber, S. (2023). Geographies of grocery
shopping in major Canadian cities: Evidence from large-scale mobile app data.
Environment and Planning B: Urban Analytics and City Science, 50(3), 723–
739. https://doi.org/10.1177/23998083221129272
Squire, R. F. (2019). An Interactive Guide to Analyze Demographic Profiles from
SafeGraph Patterns Data.
https://colab.research.google.com/drive/1qqLRxehVZr1OBpnbHRRyXPWo1Q9
8dnxA?authuser=1#scrollTo=fEFiU4ny9LYx
Thornton, L. E., Crawford, D. A., Lamb, K. E., & Ball, K. (2017). Where do people
purchase food? A novel approach to investigating food purchasing locations.
International Journal of Health Geographics, 16, 1–13.
https://doi.org/10.1186/s12942-017-0082-z
Todd, J. E., & Scharadin, B. (2016). Where households get food in a typical week:
Findings from USDA’s FoodAPS.
U.S. Census Bureau. (2022, December). American Community Survey 5-Year Estimates
(2017–2021), Jacksonville city, Florida.
https://data.census.gov/table?q=Jacksonville%20city,%20Florida&y=2021&d=
ACS%205-Year%20Estimates%20Data%20Profiles
Ver Ploeg, M., Breneman, V., Farrigan, T., Hamrick, K., Hopkins, D., Kaufman, P.,
Lin, B.-H., Nord, M., Smith, T. A., Williams, R., & others. (2009). Access to
affordable and nutritious food: Measuring and understanding food deserts and
their consequences: Report to congress.
26
Ver Ploeg, M., Mancino, L., Todd, J. E., Clay, D. M., & Scharadin, B. (2015). Where do
Americans usually shop for food and how do they travel to get there? Initial
findings from the National Household Food Acquisition and Purchase Survey.
Wray, A., Martin, G., Doherty, S., & Gilliland, J. (2023). Analyzing differences
between spatial exposure estimation methods: A case study of outdoor food and
beverage advertising in London, Canada. Health & Place, 79, 102641.
https://doi.org/10.1016/j.healthplace.2021.102641
Xie, H., Li, D., Wang, Y., & Kawai, Y. (2023). An early warning model of type 2
diabetes risk based on POI visit history and food access management. PLOS
ONE, 18(7), e0288231. https://doi.org/10.1371/journal.pone.0288231
Xu, R., Huang, X., Zhang, K., Lyu, W., Ghosh, D., Li, Z., & Chen, X. (2023).
Integrating human activity into food environments can better predict
cardiometabolic diseases in the United States. Nature Communications, 14(1),
7326. https://doi.org/10.1038/s41467-023-42667-8
Xu, Y., Zhao, X., Lovreglio, R., Kuligowski, E., Nilsson, D., Cova, T. J., & Yan, X.
(2022). A highway vehicle routing dataset during the 2019 Kincade Fire
evacuation. Scientific Data, 9(1), 608.
Zenk, S. N., Matthews, S. A., Kraft, A. N., & Jones, K. K. (2018). How many days of
global positioning system (GPS) monitoring do you need to measure activity
space environments in health research? Health & Place, 51, 52–60.
Zenk, S. N., Schulz, A. J., Matthews, S. A., Odoms-Young, A., Wilbur, J., Wegrzyn, L.,
Gibbs, K., Braunschweig, C., & Stokes, C. (2011). Activity space environment
and dietary and physical activity behaviors: A pilot study. Health & Place,
17(5), 1150–1161.
27
Zhang, L., Misir, A., Boshuizen, H., & Ocké, M. (2021). A Systematic Review and
Meta-Analysis of Validation Studies Performed on Dietary Record Apps.
Advances in Nutrition, 12(6), 2321–2332.
https://doi.org/10.1093/advances/nmab058
Zhao, K., Tarkoma, S., Liu, S., & Vo, H. (2016). Urban human mobility data mining:
An overview. 2016 IEEE International Conference on Big Data (Big Data),
1911–1920. https://doi.org/10.1109/BigData.2016.7840811
Zhao, X., Xu, Y., Lovreglio, R., Kuligowski, E., Nilsson, D., Cova, T. J., Wu, A., &
Yan, X. (2022). Estimating wildfire evacuation decision and departure timing
using large-scale GPS data. Transportation Research Part D: Transport and
Environment, 107, 103277.
Zhou, R. Z., Hu, Y., Tirabassi, J. N., Ma, Y., & Xu, Z. (2022). Deriving neighborhood-
level diet and physical activity measurements from anonymized mobile phone
location data for enhancing obesity estimation. International Journal of Health
Geographics, 21(1), 22. https://doi.org/10.1186/s12942-022-00321-4
