Fine-grained Analysis of WUI and its Implications on Wildfire Occurrence in the state of Colorado from 1990s to 2010s

Abstract:

The Wildland-Urban Interface (WUI) is defined as the area where houses are in or near wildland or vegetated areas, which usually indicates elevated risk of wildfire damage to built infrastructure and people. The WUI represents an important concept that researchers, risk managers and planners have been using to analyze potential wildfire risks for many years. Colorado is one of the states in the western United States that experienced devastating wildfires in history. Due to its diverse landscapes ranging from mountains to deserts, wildfire management is difficult and requires different approaches across different parts of the state. This study focuses on two aspects: (1) to explore how the high resolution human settlement data provided by Zillow’s Assessor and Real Estate Dataset (ZTRAX) can be used to revise and refine the WUI; and (2) to understand the importance of different WUI classes for wildfire risk analysis by incorporating FPA-FOD historical wildfire dataset.

By using the high-resolution human settlement data for the creation of gridded spatial layers, WUI classes are mapped at 30m spatial resolution with 2-3 years temporal resolution. This approach mitigates effects of the modifiable areal unit problem, which is typically caused by aggregating structures’ locations to administrative units of large areal extents in rural settings such as census blocks. To better understand how WUI areas and potential high-risk areas develop over time, we assess classification agreement between WUI and non-WUI classes from the 1990s to 2010s. The resulting confusion matrices allow us to identify class transitions from non-WUI classes to WUI classes, which implies effects of increasing human settlement expansion into wildland areas during the time period. The fine-grained WUI layers are overlain with FPA-FOD fire data to explore the relationships between wildfire occurrence and local WUI classes in Colorado. We examine spatial and temporal lag effects to better understand the sensitivity in this relationship. We differentiate wildfire ignition or causes to better understand how wildfires spread differently in WUI and non-WUI land cover classes after filtering the data. Our time series of revised fine-grained WUI layers will be useful for wildfire managers to conduct local risk analysis that can be applied to other regions.