More Information on Smoke Forecaster

1 Introduction

Smoke Forecaster is an online wildfire smoke health impact assessment tool designed to provide environmental and public health officials at the state and local level with information about the potential health risks of wildfire smoke exposure in their communities.

2 Acronyms

ED: Emergency department

HIA: Health impact assessment

HIF: Health impact function

HMS: Hazard mapping system

HYS-PLIT: Hybrid Single-Particle Lagrangian Integrated Trajectories

NOAA: National Oceanic and Atmospheric Administration

PM_2.5: Particulate matter with an aerodynamic diameter less than 2.5 \(\mu\)m

RR: Relative risk

RX: Prescribed burn

SEDAC: Socioeconomic Data and Applications Center

US EPA: United States Environmental Protection Agency

WF: Wildfire

WRF-Chem: Weather Research and Forecasting (WRF) model coupled with Chemistry (Chem)

3 Explanation of Key Terms

Particulate matter (PM_2.5): is particulate matter with an aerodynamic diameter less than 2.5 \(\mu\)m. PM_2.5 is also sometimes called “fine particulate matter.” Particulate matter consists of all the solid particles or liquid droplets suspended in air. Particulate matter can come from a number of natural and human-made sources, including emissions from vehicles and industrial facilities, windblown dust and sea spray, and wildfires. PM_2.5 is a subset of these naturally-occurring anthropogenic particles with a diameter smaller than \(\frac{1}{30}\)th the width of a human hair or smaller (Figure 1). We are concerned about PM_2.5 because it has the ability to reach deep into the lungs when inhaled and is known to cause serious health effects, including respiratory and cardiovascular disease among children and older adults. More information on PM_2.5 is available from the US EPA.

Figure 1: Size comparisons for particulate matter. Image Credit: US EPA

Relative Risk (RR): The relative risk (RR) is a type of effect estimate, which is a measure of the relationship between an exposure and an outcome. Relative risk is a ratio defined as the risk for an individual who is exposed to a hazard to experience a negative health outcome relative to a person who is not exposed to that hazard (Figure 2). For Smoke Forecaster, we use a relative risk for emergency department visits due to wildfire smoke exposure. For this tool, we consulted epidemiological studies that compared the likelihood that a person who was exposed to wildfire smoke would seek care in an emergency department to the likelihood of an ED visit for an person who was not exposed to wildfire smoke. The ratio of these risks is called the relative risk of an ED visit due to wildfire smoke exposure. A relative risk above 1 means that exposed populations are more likely to need to visit the ED when exposed to wildfire smoke. The larger the relative risk, the stronger the link is between the hazard (in our case: wildfire smoke) and the outcome (ED visits). The Centers for Disease Control and Prevention has more information on how relative risks are calculated here.

Figure 2: Relative Risk Image Credit: Coriell Institute

Respiratory Relative Risk: This is the relative risk for going to the ED for any respiratory health condition (for example, asthma, chronic obstructive pulmonary disease [COPD], or pneumonia) for people who are exposed to wildfire smoke compared to people who are not exposed to wildfire smoke. Smoke Forecaster provides estimates of the relative risk of going to the ED for a respiratory health condition for the modeled wildfire smoke concentration at a specific location compared to a hypothetical scenario where there is no wildfire smoke at that location. Numbers greater than 1 indicate there is an increased risk of an ED visit for respiratory diseases for people living at that location. The designation of a respiratory health condition is based on the diagnosis listed in a patient’s medical record. These diagnosis codes are usually based on the International Classification of Disease (ICD) system. Examples of which diagnoses are included in the respiratory relative risk are outlined in Table 1 (see the section titled “What information do we need for the Smoke Forecaster HIA”).

Asthma Relative Risk: This is the relative risk for going to the ED for an asthma exacerbation or asthma symptoms (e.g., wheeze, cough, or shortness of breath) for people who are exposed to wildfire smoke compared to people who are not exposed. Smoke Forecaster provides estimates of the relative risk of going to the ED for asthma for the modeled wildfire smoke concentration at a specific location compared to a hypothetical scenario where there is no wildfire smoke at that location. Numbers greater than 1 indicate there is an increased risk of an ED visit for asthma for people living at that location.The asthma relative risk is more specific than the respiratory relative risk, which includes all respiratory health diagnoses. The asthma relative risk only includes ED visits that fall under the asthma category (ICD code 493).

4 Health Impact Assessment

4.1 What is health impact assessment?

Health impact assessment (HIA) is a policy- and decision-making tool that provides information on the potential health impacts of an environmental exposure scenario. For Smoke Forecaster, we use HIA to estimate the number of emergency department (ED) visits we might see due to wildfire smoke. This information helps environmental managers, city and county governments, and individuals plan for and make healthier choices during wildfire events.

There are generally two types of HIAs: quantitative and qualitative. Quantitative HIAs tend to focus on health impacts for which we can quantify the relationship between exposure and outcome. Smoke Forecaster uses this type of HIA to estimate how ED visits might increase during wildfire events. ED visits are just one of many potential health effects associated with wildfires, many of which we cannot yet make quantitative predictions for. These additional health effects might include more frequent rescue inhaler use among people with asthma or less physical activity due to smoke in the air. Therefore, the estimates generated by Smoke Forecaster should be considered a subset of the total health effects.

4.2 What information do we need for the Smoke Forecaster HIA?

The HIA used by Smoke Forecaster requires four key pieces of information to make estimates of how many additional emergency department visits we might expect during a wildfire event. We’ve outlined these key inputs below.

4.2.1 Wildfire smoke concentrations

First, we need to know the concentration of wildfire smoke in the air at ground level where people breathe. Smoke forecaster uses data from the Blue Sky model developed by the U.S. Forest Service to estimate outdoor levels of wildfire smoke on the day of the analysis (i.e., today) and the next day (i.e., tomorrow). These predictions are made using multiple types of data, including:

• wildfire characteristics, such as how much area has burned;

• emissions data, including how much particulate matter and carbon monoxide are emitted from the wildfire;

• and meteorology, including 72-hour data on temperature, wind speed, and humidity. We Blue Sky uses these data in mathematical models that characterize how smoke disperses in the atmosphere and reaches the breathing zone of people living in the affected area. Figure 3 shows an example of how Blue Sky displays this information. PM_2.5 from wildfire smoke is shown as gray pixels. This model is represented in Smoke Forecaster as the Forecasted Smoke Concentrations layer on the map. These plumes are an estimate of where smoke is currently located and where it will likely be the next day. The Blue Sky model also gives us an estimate of the concentration of wildfire smoke at the ground level. This is important for the Smoke Forecaster HIA because we are primarily concerned with smoke that people breath in while outside during a wildfire event. Blue Sky produces predictions (estimates) of the 24-hour average concentrations on the current day and the next day.

Figure 3: Blue Sky Smoke Plumes

Smoke Forecaster also shows the locations of current wildfires burning in the United States and Canada. These locations are shown on the map in the Fire Locations layer. The fire locations are plotted on the map using data from Blue Sky. Blue Sky models where fires are located using satellite data and other information sources. The models depicts wildfires (in red) and prescribed burns (in yellow), as shown in Figure 4.

Figure 4: Blue Sky Modeled Fires

Data from the Blue Sky model can be accessed here

To help visualize where smoke is located, Smoke Forecaster also includes a layer that shows visible wildfire smoke plumes (called Visible Smoke Plumes). The data showing where these visible smoke plumes are located are generated by the Hazard Mapping System (HMS), which is run by the National Oceanic and Atmospheric Association (NOAA). Each day, NOAA analysts use satellite imagery to identify wildfires and the smoke plumes generated by those fires and create maps that can be downloaded by the public. We incorporate these maps into Smoke Forecaster to help users visualize where wildfires and wildfire smoke are located in North America. The HMS website has more information on how these maps are developed and provides links to download the data.

4.2.2 Exposed population

Second, we need to know how many people are exposed to wildfire smoke. Smoke forecaster uses data from the 2015 American Community Survey (an annual survey conducted by the US Census Bureau) to estimate how many people live in each county in the state. To calculate the population-weighted exposure to wildfire smoke, we use the wildfire smoke concentration data from the Blue Sky model and population density data from the Socioeconomic Data and Applications Center (SEDAC) at Columbia University. Figure 5 shows the population density raster for area surrounding Denver, CO.

Figure 5: SEDAC Population Denisty estimates for Denver, CO

4.2.3 Daily Emergency Department visit rates

Third, we need to know how many ED visits for respiratory health outcomes typically occur each day in each county in the state. To do this, we use data from the Colorado Hospital Association (CHA). CHA collects data from participating hospitals across the state on who is admitted to a hospital’s emergency department and why. This data set also includes information on where each person who is admitted lives, which allows us to aggregate the data to the county level. Using annual counts for ED visits and the population in each county, we can calculate an annual rate (as admissions per county per 10,000 persons per year) for ED visits for respiratory outcomes. To get a daily rate (as admissions per county per 10,000 persons per day), we assume that the number of ED visits for respiratory outcomes is relatively constant day-to-day and divide the annual rate by 365 days per year.

4.2.4 Relationships between wildfire smoke and ED visits for respiratory health outcomes

Fourth, we need to know how health outcomes and exposures are related. In our HIA for Smoke Forecaster we use the relative risk values reported by previous studies to understand the relationship between wildfire smoke and ED visits for For Smoke Forecaster, we our health outcomes of interest based on existing studies of the health effects of wildfire smoke. In the last few years, a few studies have been published that have identified relationships between emergency department visits for respiratory or cardiovascular diseases and wildfire smoke exposures (Table 1). These include studies that were conducted in Colorado, Washington, California, the western United States, and North Carolina. In general, the evidence is strongest for associations between wildfire exposures and ED visits for respiratory health issues. The evidence of a link between wildfire smoke and ED visits for cardiovascular diseases is weaker, so we have elected not to include these types of ED visits in our tool at this time.

Table 1. Summary of available studies on the respiratory health effects of wildfire smoke

Authors	Year	Location	Ages	Exposure Method	ED Visit Reasons	Effect Estimate (RR or %Change)	Exposure Contrast (µg/m³)
Alman et al.	2016	Colorado	All	WRF-Chem	Respiratory disease (ICD-9: 460–465, 466.0, 466.1, 466.11, 466.19, 480–486, 487, 488, 490, 491, 492, 496, 493–786.07)	1.02 (1.01, 1.03)	5
Gan et al.	2017	Washington	<65	WRF-Chem, area monitors, and satellite data	Respiratory disease (ICD‐9‐CM: 460–519)	1.052 (1.025, 1.080)	10
Hutchinson et al.	2018	California	<65	HYS-PLIT	Respiratory disease (ICD-9: 277, 460-464, 466, 480-487, 490-496, 506, 508, 786)	1.02 (1.01, 1.04)	10
Rappold et al.	2012	North Carolinal	18+	HYS-PLIT	Asthma (ICD-9: 493)	66% (28%, 117%)	100
Reid et al.	2016	California	All	Satellite data, WRF-Chem, and meterology	Asthma (ICD-9: 493), COPD (496, 491–492), and pneumonia (480–486)	1.015 (1.009, 1.020)	10
Wettstein	2018	California	18+	HMS smoke plume denisty: light (0- 10 µg/m³)	Respiratory (ICD-9: 480–486, 491–493, 496, 786)	1.02 (1.00, 1.04)
				HMS smoke plume denisty: medium (10.5 - 21.5 µg/m³)	Respiratory (ICD-9: 480–486, 491–493, 496, 786)	1.09 (1.05, 1.12)
				HMS smoke plume denisty: dense (22+ µg/m³)	Respiratory (ICD-9: 480–486, 491–493, 496, 786)	1.09 (1.03, 1.15)
Tinling et al.	2016	North Carolina	All	NOAA Smoke Forecasting System (SFS)	Respiratory conditions (ICD-9: 460-466, 480-486, 490-493, 496)	1.04 (0.99, 1.09)	10

For Smoke Forecaster, we elected to use the effect estimates (relative risks) for all respiratory-related ED visits and asthma ED visits from the study by Gan et al. (2017). This study used a similar method as Blue Sky (a combination of air quality modeling, satellite data, and ground-based measurements) to assess exposures to wildfire smoke. This study also included a large segment of the population and included several types of respiratory diseases. Importantly, the relative risk estimate provided by Gan et al. falls within the range of relative risk values reported by the literature (Table 1). For these reasons, we felt that this study provided a reasonable effect estimate to use in our health impact assessment.

4.3 How does a health impact assessment estimate the number of ED visits that might occur during a wildfire?

To estimate the number of additional ED visits for respiratory diseases or asthma we might expect over the typical number of ED visits that occur during periods without wildfires, Smoke Forecaster uses a health impact function (HIF). The HIF calculates the number of additional ED visits (\(Y\), number of ED visits) for each county separately and uses four key pieces of information described above: the concentration of wildfire smoke in the air (\(x\), \(\mu\)g/\(m^3\)), the number of people exposed to wildfire smoke in the county (P, number of persons), the daily rate of ED visits in the county (\(y_{0}\), number of ED visits per person per day) and the effect estimate for wildfire smoke on ED visits (\(\beta\), 1/(\(\mu\)g/\(m^3\))). The HIF takes the form:

The HIF is a log-linear function, which the relationship between smoke and ED visits for respiratory health outcomes is strongest when concentrations are lower and starts to level off with increasing concentrations of wildfire smoke. Importantly, the relationship is always positive, meaning there may be negative health effects at any level of wildfire smoke exposure.

Smoke Forecaster displays the number of additional ED visits predicted for each county on the map using the HIA layer. Figure 6 shows an example of what the HIA layer looks like for Utah on August 21, 2019. In this figure, there are wildfires burning across central Washington state, Oregon, California, and Nevada (shown as red dots and red and grey pixels). This wildfire is associated with increased wildfire smoke exposures and 10 additional ED visit per day in Yakima county, Washington.

Figure 6: Smoke-Related ED Visits in Yakima County (Aug 21, 2019)

4.4 What can we do with HIA estimates?

The HIF produces an estimate of the number of additional ED visits there might be over the typical number of ED visits in a county. This prediction is based on the number of ED visits that typically occur on a single day in a county, the number of people who live in that county, and the estimated (predicted) concentration of wildfire smoke in the county.

The wildfire smoke exposures data and predicted number of ED visits estimated by the Smoke Forecaster tool are intended to provide environmental and public health officials information on where the most vulnerable communities in their jurisdictions might be during a wildfire event. This information can help officials plan their communication efforts and make decisions about where to allocate resources.