Sources and Implications of Bias and Uncertainty in a Century of US Wildfire Activity Data

The statistical analysis of wildfire activity is a critical component of national wildfire planning, operations, and research in the United States (US). Wildfire activity data have been collected in the US for over a century. Yet, to this day, no single unified system of wildfire record-keeping exists. Data for analysis are generally harvested from archival summary reports from federal or interagency fire organizations; incident-level wildfire reporting systems of the federal, state, and local fire services; and, increasingly, remote-sensing programs. It is typical for research into wildfire activity patterns for all or part of the last century to require data from several of these sources and perhaps others. That work is complicated by the disunity of the various datasets and potentially compromised by inherent reporting biases, discussed here. The availability of wildfire records with the information content and geospatial precision generally sought for increasingly popular climatological analyses and the modeling of contemporary wildfire risk is limited to recent decades. We explain how the disunity and idiosyncrasies of US wildfire reporting have largely precluded true interagency, or all-lands, analyses of even recent wildfire activity and hamstrung some early risk modeling efforts. We then describe our efforts to acquire, standardize, error-check, compile, scrub, and evaluate the completeness of US federal, state, and local wildfire records from 1992-2011 for the national interagency Fire Program Analysis (FPA) application. The resulting FPA Fire-Occurrence Database (FPA FOD) includes nearly 1.6 million records from the 20-year period, with values for at least the following core data elements: location at least as precise as a Public Land Survey System section (2.6-km2 grid), discovery date, and final fire size. The FPA FOD is publicly available from the Research Data Archive of the US Department of Agriculture, Forest Service (http://dx.doi.org/10.2737/RDS-2013-0009). While necessarily incomplete in some aspects, the database is intended to facilitate fairly high-resolution geospatial analysis of wildfire activity over the past two decades, based on available information from the authoritative systems of record. Formal non-federal wildfire reporting has been on the rise over the past several decades, and users of national datasets like the FPA FOD must beware of state and local reporting biases to avoid drawing spurious conclusions when analysing the data. Apparent trends in the numbers and area burned by wildfires, for example, may be the result of multiple factors, including changes in climate, fuels, demographics (e.g. population density), fire-management policies, and - as we underscore here - levels of reporting.