EXPLORING Food Environments Assessing access to nutritious food By Karen Richardson, Esri Writer "GIS has a lot to contribute to discussions about how people access food, recreation, and other human needs," according to Jim Herries, a product engineer for cartography at Esri who has been working on mapping access to healthy foods. Because diet has a direct effect on health, people who eat nutritious foods have a lower propensity for diabetes, obesity, and other diseases and chronic conditions. If access to nutritious food is limited or made difficult by factors such as cost or the distance traveled to obtain it, a population can be adversely affected. Areas that lack relatively easy access to nutritious, affordable food have been described as "food deserts." "It's a very useful thing to draw a map of where people are and the types of food available around them. You look at that map and wonder, are people getting what they need?" said Herries. "Esri decided to look into ways that GIS could help people look a little deeper to see how things differ from neighborhood to neighborhood. And we'd like to do this in a way that is relatable to the existing studies in literature and provides an opportunity for communities to improve the quality of life for the people who live there." Early attempts at mapping the food environment used geography aggregated at high levels and simple methods for measuring access to food. "The question is, is this really an accurate representation of what is out there?" said Herries. "One thing I know for sure is that we can use GIS technology and data to define distance a bit better than what we've seen historically." Instead of measuring distances as simple rings or "as the crow flies," GIS can look at distance over a road network. It determines how many people live within the prescribed distance of the grocery store based on travel over streets. Using actual street network topology more accurately models time and distance and produces more reliable results. Even the best optimization algorithms will generate grossly inaccurate results if the underlying time and distance measurements are not correct. How Would a Crow Walk (or Drive)? Herries' analysis used a logical network that incorporated information about the topology of the network so that overpasses, underpasses, 50 ArcUser Fall 2010 This map answers the question, How many residents in this neighborhood live within a walkable distance of a supermarket? for several areas of Cincinnati, Ohio. Green dots represent populations living within one mile of a supermarket. Red dots indicate populations living in areas that require more than a one-mile walk to a supermarket. one-way streets, speed limits, and turn restriction conditions could be accurately modeled. To assess the food environment of an area, Herries generated an origin-destination (O-D) matrix that provided the cost (in time or distance) associated with travel from one location to another. In this case, the cost was the time required to drive or walk from the shopper's location to the nearest grocery store. Street data that included attributes for speed limits, one-way streets, and barriers was combined with starting and ending locations to generate an accurate O-D matrix. The travel time computed by applying a shortest-path algorithm to a street network dataset realistically models the characteristics of the route traveled and avoids unrealistic scenarios like driving across water bodies or through areas with no roads. Moving Analysis to a Useful Scale Historically, many studies of food access have used census tract-level population data because this is a convenient geographic unit. However, census tracts have an average population of 43,000—hardly the size of a typical neighborhood. Smaller geographic units are available: census block groups and census blocks. On average, block groups contain 1,500 people, and blocks contain 40 people. A block group is the smallest geographic unit for which the U.S. Census Bureau publishes sample data. Major retailers use block group and block data in GIS analyses of markets, sites, and competitors. Instead of using census tracts, Herries looked at block data. He looked at how many grocery stores existed within a 10-minute walk of each one of the eight million census blocks in the United States. The results were summarized www.esri.com