User:Jf umbc/sandbox

9/25/11

GES 679, Geospatial Technologies Seminar

Recognizing MAUP in Research
Increased computing capabilities have improved the way the world can be visualized, analyzed and understood. Of these capabilities, Geographic Information Systems (GIS) has emerged as a powerful tool with which to research and analyze data geographically. Previously incomparable or incompatible data can now easily be analyzed against another by associating them to a common platform (e.g., coordinate systems in the case with GIS). However, with the increased ability to analyze data via GIS comes another problem: how does one represent the data to accurately (or inaccurately) represent its geographic distribution? This, in essence, defines the Modifiable Areal Unit Problem (MAUP) and persists as potentially an unsolvable problem in geographic research.

GIS has the capability to visualize the world in a computer system, able to define objects down to a point location. Each person, fire hydrant, or house has a precise position on the earth that can be described as a specific latitude and longitude. However, in some cases, point data is neither required nor preferred. Each of these points are aggregated to some larger area (areal unit), whether it be the sum of houses in a neighborhood, people in a county, or fire hydrants in a city. Sometimes there are simply too many data points for time sensitive calculations. Whatever the reason for aggregation may be, there is always larger, polygonal, areal unit to which these points are assigned.

Areal units can simply be described as polygons that define an area. Examples of areal units are census tracts, police districts, zip codes, counties, and states. There is no prescribed base ‘areal unit.’ Areal units range from large to small, thin and short to wide and long. Because there is no set of standards by which areal units are defined, areal units can be whichever shape and size the researcher desires. These two items – shape and size – are a double-edged sword. They are what give GIS its unprecedented ability to derive spatial information, but it is the fundamental base of the Modifiable Areal Unit Problem.

MAUP is split into two main subproblems: scale problems and zonal problems. These problems were identified by Openshaw as early as 1977, before the application of GIS in research. The scale problem is exhibited when census block information is aggregated to the tract level, and then to the county level. Zonal problems occur when the number of units remains the same, but the shape of the units changes (Wilson, 2011). In some cases, the zonal problem is referred to as an aggregation problem (Openshaw, 1985).

Census blocks and tracts are a perfect example of both the scale and zonal problems. Census blocks are smaller units than the tracts. It is therefore assumed that since you have defined a smaller areal extent, you know more precise information about the people that reside within the boundaries. Differences that distinguish nearby blocks are eliminated when information is aggregated to the tract level. The information that was defined to a smaller geographic region has been normalized to a larger area.

Since these areal units (blocks and tracts) generally contain an equal number of people between like units, their shape and size may vary. For example, densely populated urban areas will have smaller areal units compared to rural areas, where the population is distributed over a larger area. A block for an urban area may measure to a few square miles whereas rural blocks may range into the 10s and 100s of square miles and still contain the same number of residents as the urban block.

With these two subproblems also comes the ecological fallacy problem. This occurs when data, resultant of areal unit analysis, is applied to the individuals within the areal unit (Openshaw, 1985). This problem may occur when analyzing counties by political vote. For example, in landslide election counties, where votes for a political candidate were separated by more than 20%, frequently the residents can be thought of as ‘all blue’ or ‘all red.’ However, in the landslide counties, the individuals who are against the norm are ignored (or forgotten) when analyzing these regions. Interestingly enough, they mention cases where neighboring counties are of opposite political leanings. If these counties were aggregated to a larger areal unit, or if the cities also included the exurbs, these ‘landslide’ areas may no longer be defined as landslides, but more of the national average. (Bishop, 2008)

In summary, the Modifiable Areal Unit Problem is amplified by the capability to quickly change shape and size of geographical units. It has effects in both academic research and day to day lives, impacting the way the world defines and sorts itself.