Census data in combination with GIS are increasingly being used to analyze urban expansion and develop models for identifying landscape change in the urban fringe. Census data are aggregated along the large-to-small-unit gradient of county, tract, census block group (CBG), and census block. The multiple scale availability often confounds the selection of an appropriate level of data in research pertinent to using census data. This study addressed the modifiable areal unit problem of census data through comparing spatial pattern and area of wildland-urban interface (WUI) determined at different levels of census aggregation (county, census tract, CBG, and census block). Total WUI area in each single year decreased along the shrinking census unit gradient from county to census block. Area converted from wildland to WUI between 1990 and 2000 decreased along the census gradient of the tract, CBG, census block, county level. The number of WUI patches decreased, and area of WUI patches increased along the decreasing census gradient of county, tract, CBG, block. In contrast to 60% of WUI blocks falling inside WUI CBGs or tracts, more than 80% of WUI tracts fell inside WUI counties, and 76.8% of WUI CBGs fell inside WUI tracts. WUI at the block level showed a different spatial pattern from those at the tract and CBG levels in that it represented more spatial detail. County-level data tended to overestimate WUI area while underestimating area converted to WUI. The study concluded that coarse sale data, such as those at the county level, were suitable for detecing a regional pattern. Fine-scale data, such as those at the census block level, need to be used in addressing issues at a landscape pattern. © 2007 by the Society of American Foresters. Abstract reproduced by permission.