Go to ArcGIS Online to complete exercise 1.3: “Location and distance.”
Spatial patterns
Features on the earth’s surface arrange themselves in spatial patterns. Analyzing these patterns allows geographers to elucidate not only how human and physical features are arranged but also the processes behind their formation.
A commonly used description of spatial patterns is density. Density is the number of features per unit area, as in the number of people per square mile or number of trees per square kilometer. Density is useful for illustrating spatial patterns that would not be seen using raw numbers alone. For example, the population of California is about 39 million people, while the population of Singapore is only 5.5 million. With no additional information, one may get the impression that California is more crowded than Singapore. But when information on area is added, that impression quickly changes. California consists of 163,696 square miles, while Singapore is made up of just 278 square miles. So, in reality, Singapore has a much higher population density than California (figure 1.23).
Figure 1.23.Population density: Singapore. Singapore has one of the highest population densities in the world, with 5.5 million people living in just 278 square miles. Photo by joyfull. Stock photo ID: 138766448. Shutterstock.
Spatial patterns can also be viewed in terms of clustering, randomness, and dispersion (figure 1.24). As the name implies, clustered features are found grouped near each other. Clusters are often identified with hot spot analysis or with a heat map (figure 1.25). Randomly distributed features have no particular spatial pattern. Dispersed features are those that repel each other. They are certainly not clustered and are even farther from each other than if the distribution were random.
Figure 1.24.Spatial patterns can be seen as dispersed, random, or clustered. Image by author.
Analysis of these types of spatial patterns has many applications. For example, if home burglaries are found to be clustered in a specific neighborhood, police can increase patrols in that area, while detectives and community groups can focus on what the underlying causes of the crime cluster are. It may turn out that a prolific burglar lives nearby, or youth from a local high school may be committing crimes after school. If home burglaries are not clustered, but have a more random pattern, then other causes may be at play, such as burglaries being crimes of opportunity, where criminals take advantage of homes with open windows.
Diseases often cluster as well. If cancer rates are found to cluster in a neighborhood, then health researchers may search for environmental causes of the disease, such as a nearby toxic waste site. If cancer cases are randomly distributed around a city, then environmental factors are less likely to be the cause.
Dispersed features can include shopping malls or chain restaurants in an urban region. Mall owners may intentionally maintain a distance from competing malls to avoid competition, while owners of a restaurant chain may space their stores so that they do not cannibalize sales from each other.
Spatial patterns can also be analyzed by measuring the center of features. With a map of consumer purchasing behavior, a business may want to find a new store location that lies at the center of its specific market segment. Likewise, geographers can study shifts in population by mapping the center of US population over time.
Spatial relationships
Mapping the spatial relationships of two or more features can offer insight into why particular patterns exist. Whereas spatial distributions describe how features are clustered or dispersed, spatial relationships depict where features are located in relationship to other types of features. For instance, geographers study the distance between different types of features or whether different feature types overlap (figure 1.26). If there is a disease cluster, geographers can examine the distance between the cluster and factories that emit toxic effluent. If the cluster is nearby, then the effluent may be the cause of the disease. They can also study whether the disease cluster overlaps with the residences of workers in a specific type of occupation. It may turn out that the cluster is not due to nearby toxic effluent but rather that many residents in the disease cluster work in a mine that uses toxic chemicals.
Figure 1.25.Mapping clusters as hot spots. Hot spot analysis can uncover clusters of crime, different demographic groups, disease, natural hazard events, and much more. Map by author. Data source: Long Beach Police Department.
Figure 1.26.Spatial relationships. Geographers study the spatial relationship between features, such as how far apart they are (left) or whether they overlap (right). Image by author.
Statistical tools are often used to study spatial relationships. With spatial correlation, it is possible to analyze the strength and direction of spatial relationships (figure 1.27), be they positive, negative, or unrelated. A positive relationship is when both variables change in the same direction, as when places with high unemployment also have high rates of alcohol consumption. A negative relationship is when an increase in one variable leads to a decrease in another, as when areas with high unemployment have lower traffic fatalities due to people driving less. When there is no pattern of increase or decrease between two variables, they are unrelated, as when places with high unemployment have no correlation with the number of earthquakes.
Figure 1.27.Spatial correlation. Variables in the same place can be plotted to see if they have positive, negative, or no relationship. Image by author.
With quantitative analysis, the phrase “correlation does not imply causation” is commonly used to describe the case where variables can be correlated, but one variable does not cause the other to change. To build on an earlier example, a cancer cluster may be located near a toxic effluent site, leading some people to infer that cancer risk increases because of proximity to the site. But, in reality, there may be a third variable that is not being considered. Even though the cancer cluster correlates with distance to toxic effluent, it may turn out that the cancer is due to where residents of the cluster work. It may be that many residents of the cluster work in a mine that uses toxic chemicals and that exposure to those chemicals is causing the disease. Cancer may have a strong correlation with proximity to toxic effluent, but the proximity is not the cause.
It is therefore important to consider multiple explanations when looking at correlations and to use previous research and theory when determining which variables to include in an analysis. When mapping heart disease by county and determining which factors contribute to it, current scientific research says that variables such as smoking, diet, and physical inactivity are contributing factors. Rates of smoking, rates of high cholesterol from poor diet, and average hours of exercise per capita can be mapped on top of a heart disease map. With spatial statistical analysis, the strength of each variable can be analyzed in relation to rates of heart disease. It may become clear that some counties have high rates of heart disease primarily due to high rates of smoking, while others may have high rates due to a lack of physical activity.
Go to ArcGIS Online to complete exercise 1.4: “Spatial patterns and spatial relationships. An analysis of homicide patterns in Chicago.”
Places and regions
Many people are drawn to geography because they love to explore and learn about the great diversity of the world. From quaint towns along the coast of Italy