Classification of the wildland–urban interface: A comparison of pixel- and object-based classifications using high-resolution aerial photography Casey Cleve a, * , Maggi Kelly b , Faith R. Kearns a , Max Moritz a a Center for Fire Research and Outreach, University of California, Berkeley, 137 Mulford Hall #3114, Berkeley, CA 94720, United States b Geospatial Imaging and Informatics Facility, University of California, Berkeley, 137 Mulford Hall #3114, Berkeley, CA 94720, United States Received 21 June 2007; received in revised form 5 October 2007; accepted 8 October 2007 Abstract The expansion of urban development into wildland areas can have significant consequences, including an increase in the risk of struc- tural damage from wildfire. Land-use and land-cover maps can assist decision-makers in targeting and prioritizing risk mitigation activ- ities, and remote sensing techniques provide effective and efficient methods to create such maps. However, some image processing approaches may be more appropriate than others in distinguishing land-use and land-cover categories, particularly when classifying high spatial resolution imagery for urbanizing environments. Here we explore the accuracy of pixel-based and object-based classification methods used for mapping in the wildland–urban interface (WUI) with free, readily available, high spatial resolution urban imagery, which is available in many places to municipal and local fire management agencies. Results indicate that an object-based classification approach provides a higher accuracy than a pixel-based classification approach when distinguishing between the selected land-use and land-cover categories. For example, an object-based approach resulted in a 41.73% greater accuracy for the built area category, which is of particular importance to WUI wildfire mitigation. Ó 2007 Elsevier Ltd. All rights reserved. Keywords: Wildland–urban interface; Classification; Image analysis; Object-based methods; Remote sensing 1. Introduction The increased proximity between developed and wild- land areas that often accompanies new development places people and homes at risk from wildfire. In California alone, over five million homes are located in wildland–urban interface (WUI) areas (Radeloff et al., 2005) and that num- ber is likely to grow as the state population continues to increase. Accurate and timely land-use and land-cover (LULC) maps are needed for wildfire management; maps are needed before fires happen to model potential risk, dur- ing fires to assist fire safety personnel in fire-fighting and evacuation, and after fires to better mitigate the ecological consequences of fire. In all cases, it is important to accu- rately distinguish homes and other features of the built environment from vegetation across a large geographic extent. The location of a structure, and its arrangement rel- ative to other structures or flammable materials, is of key interest in preventing wildfire-related losses in the WUI (Cohen 2000; Frontiera & Kearns 2007; Murname 2006). Over regional scales, LULC maps are typically pro- duced from remotely sensed image analysis using moderate resolution satellite imagery such as Landsat TM (Alberti et al., 2004; Cihlar, 2000; Hollister et al., 2004; Vogelmann et al., 1998, 2001; Walsh et al., 2001). While these products are useful for producing coarse-scale classifications, they are inadequate for detailed mapping (e.g., species-level veg- etation or buildings) (Harvey & Hill, 2001; Kalliola & Syrja ¨nen, 1991). LULC maps of urban environments require finer detail, and utilize either photointerpretation 0198-9715/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.compenvurbsys.2007.10.001 * Corresponding author. Tel.: +1 510 643 0409; fax: +1 510 643 3490. E-mail address: caseycleve@gmail.com (C. Cleve). www.elsevier.com/locate/compenvurbsys Available online at www.sciencedirect.com Computers, Environment and Urban Systems 32 (2008) 317–326