Wildfire episodes pose a significant public health threat in the United States. The potential impact on health, arising from the flame activity near active fires as well as exposures to surface smoke concentrations downwind of the burn area, represent several far-reaching consequences of wildfires (Williamson et al., 2016). Smoke plumes from wildfires represent a complex mixture of pollutants and its composition depends on several factors, including the fuel type and prevailing meteorological conditions (Urbanski et al., 2008). As a result, plume composition can vary over time and space (Lassman et al., 2017) and can consist of several air pollutants, including high concentrations of fine particulate matter (PM_2.5) and ozone (Jaffe et al., 2008; Pfister et al., 2008). Health effects associated with wildfire smoke can range from eye, nose, and throat irritations to more serious disorders, such as asthma exacerbation, bronchitis, decreased lung function, and premature death (Reid et al., 2016).

A comprehensive strategy to mitigate adverse health impacts associated wildfire disasters necessitates a thorough understanding of population-level exposures to surface smoke PM_2.5 concentrations. Additionally, identifying vulnerable populations and places, and quantifying the disease burden associated with surface smoke exposures are critical to strengthening public health preparedness capabilities for wildfires. Accordingly, to support preparedness efforts, a need emphasized by state, local, territorial, and tribal (SLTT) health departments (HDs), is access to smoke forecast data, visualization tools, and communication documents, as well as surveillance data on population health- and on vulnerable populations, — all of which can be made available on one information system and on a real-time basis. Unfortunately, such a health information system does not currently exist. However, CDC’s Environmental Public Health Tracking Network (Tracking Network), which is a surveillance platform with a well-designed repository of environmental health data and user-friendly visualization capabilities, has the necessary information technology infrastructure and expertise to undertake such tool development efforts (CDC 2010).

In this manuscript, we describe CDC’s efforts to develop an online tool, which can assist SLTT HDs with conducting a real-time vulnerability assessment and identify at-risk populations to wildfire smoke impacts.

CDC’s National Environmental Public Health Tracking Program (Tracking Program) has been collaborating with a multi-disciplinary team of experts from several academic, state, and federal agencies to support this effort. In this section, we describe the different phases involved in the tool-building effort. Fig. 1 describes our tool-building framework with specific tasks, which can be classified under the substantive areas of environmental epidemiology, health informatics, and risk communication.

Databases that are necessary to identify at-risk populations and advance our understanding of the health impacts associated with wildfire smoke are highlighted in Table 1.

We selected 11 major wildfires in multiple western states during 2008–2013 to conduct a pilot study. The selection criteria for these fires was based on the estimated potential adverse health impacts as well as availability of data to carry out tasks related to social vulnerability and health burden assessments. The data availability for these fires are explained in Table 2.

Our assessment of the 11 wildfires (Fig. 2) indicates that, areas directly impacted by fires (denoted by the area highlighted in red in Fig. 2), have communities that rank poorly for social vulnerability. Surrounding areas to those directly impacted by fires, which are denoted by a cross-hatched pattern, do not directly bear the brunt of wildfire damage but may still be impacted by wildfire smoke. While, using a proximity relationship to identify vulnerable populations to wildfire smoke may overlook transport issues due to meteorology, terrain, and chemistry, it provides an overview of areas likely to be impacted by wildfire smoke.

Predictions from air quality models are used to understand transport of smoke during major wildfire episodes, but information from these historical model runs are not systematically archived for public access. Toward this end, we relied on historical model- and monitor-based air quality data from academic and federal partners. The model building strategy, including an ensemble approach using satellite-based aerosol optical depth and an evaluation of various smoke estimation methods was conducted by Colorado State University (Lassman et al., 2017). The health risk assessment that was conducted for Washington 2012 fires evaluated the health risk estimates generated from various PM_2.5 estimation methods (Gan et al., 2017). After comparing results, we selected C-R relationships generated using a geographically weighted ridge regression — an ensemble approach using multiple air quality inputs.

Research is being conducted evaluating the impact of wildfire smoke on human health in the context of varying population cohorts and geographical locations (Liu et al., 2015). However, health risk assessments exploring adverse health effects associated with wildfire smoke in the U.S. are limited to a few wildfires. A critical review (Liu et al., 2015; Reid et al., 2016) of available evidence related to health effects associated with wildfire smoke demonstrated that there is a robust association between wildfire smoke and respiratory morbidity, including exacerbations of asthma and chronic obstructive pulmonary disease (COPD). Reid et al., 2016 noted that the evidence for all-cause mortality is growing while associations for other outcomes, particularly cardiovascular morbidity, remain unclear. We were able to compile a database of estimated effect sizes for various health end points, using C-R relationships based on 2012 Washington fires and those abstracted from literature. In Table 3, we provide a summary of effect estimates from U.S. based studies for asthma exacerbations and COPD associated with smoke PM_2.5.

Lastly, based on input from tracking funded partners and HDs, we have identified GIS databases from the Active Fire Mapping Program to summarize state- and county-specific information on the number of fires, acreage burned, and duration of major wildfire episodes. We used SAS v9.4 and ArcGIS 10.3, for data processing and for creating map displays.

CDC’s Tracking Network satisfies a wide array of users by providing reliable environmental and health information via an online platform. The IT platform, upon which the Tracking Network is built, is interoperable and supports integration of real-time feeds from reliable external sources. Currently, we use base map layers from external sources to accomplish our visualization needs. Our real-time tool development efforts will tap into web services for real-time wildfire smoke predictions and forecasts available from NOAA/NWS. Specifically, we have successfully piloted protocols to import NWS’ web feature services for smoke and expose smoke concentrations as a map layer on the Tracking Network’s web portal (https://ephtracking.cdc.gov/DataExplorer/). A description of wildfire smoke model that NOAA/NWS uses is provided in the attached supplemental information (Fig. S1). We also have created a process to integrate real-time smoke information with static measures of vulnerability and other indicators of baseline health information. Fig. 3 provides a snapshot on how a user will query smoke predictions and forecasts along with measures of vulnerability; the graphic is based on smoke predictions for the recent CA wildfires (July 2017).

The tool-building effort helped us gain insights on several knowledge gaps related to reducing public health impacts associated with wildfire smoke exposure. Firstly, epidemiologic studies that have explored the relationship between exposure to smoke and adverse health outcomes are limited to a few fires in the U.S. A systematic assessment presenting evidence from multiple fires across multiple jurisdictions is needed to help generate a nationally representative C-R relationship. In addition, population-level exposure assessment methods rely on a combination of monitor-based measurements and model-based predictions. Absence of a central repository with data collection standards for historical and near-real-time smoke predictions affects our ability to conduct routine health investigations. Further, extant literature sheds little light on the differential toxicity of smoke constituents—smoke is a complex mixture and particulate matter is a marker for many toxic components in smoke. These scientific and data access limitations collectively shaped our decision-making, especially with respect to conducting the vulnerability assessment and selection of C-R functions.

The design and development process has been complicated and resource-intensive. Accordingly, we have leveraged existing partnerships and fostered new ones, especially with agencies and institutions whose priorities related to reducing wildfire smoke exposure complemented or aligned with ours. For example, we partnered with USFS on identifying the right databases on location of fires and burn severity, and worked with Colorado State University on comparing various methods to generate smoke exposure and their implications for conducting health risk assessment. We also plan to enhance existing risk communication documents and synthesize key messages that cater to specific audiences. Recently, a multi-agency consortium of experts updated the Wildfire Smoke Guide—a communication tool for public health officials (https://www3.epa.gov/airnow/wildfire_may2016.pdf). This guide consists of information including an introduction to wildfire smoke and its health effects, information on sensitive populations, specific strategies to reduce smoke exposure, and recommendations for public health action during wildfire emergencies. Our plan is to abstract relevant prevention guidelines from this resource guide and disseminate them along with smoke forecasts and measures of vulnerability.

In addition, we received input from various stakeholders to refine the design of this online tool. A majority of these stakeholders are funded by the Tracking Program or through other programs within CDC, providing a pool of potential end users with interests in reducing public health impact associated with wildfires. As soon as a beta-version of this online tool is ready for testing, we plan on gaining feedback from these knowledgeable end users to resolve any potential inconsistencies in visualization and reporting of information and to improve user experience.

After the initial launch, we will add functionalities to this online tool such as projecting health burden on a real-time basis by using smoke predictions and forecasts, or possibly providing additional information for decision making, including the location of the nearest shelter or hospital. Dissemination of smoke forecasts and identifying vulnerable populations can help emergency responders and public health practitioners, but a unified response is incomplete without engaging the public and providing them with effective messaging to reduce smoke exposure. Once the tool is launched, it will be important to design an evaluation plan in partnership with interested stakeholders to examine the impact of this online tool in reducing the health burden associated with wildfire smoke.

A primary motivation for developing an online tool is to assist public health practitioners and emergency responders in making informed decisions before, during, and after wildfire emergencies. Toward that end, CDC is developing an online tool that utilizes short-term predictions and forecasts of smoke concentrations and integrates them with measures of population-level vulnerability to help identify at-risk populations to wildfire smoke hazards. The tool will be operationalized on a national scale, seeking input and assistance from several academic, federal, and SLTT Partners. We foresee that this tool will decrease the time to identify impacted communities, help to identify and enumerate vulnerable populations, better characterize population-level exposure, and inform implementation of appropriate interventions for those areas affected by wildfire smoke hazards.