Challenges in Studying Wildfire Health Impacts - Exposure Assessment - The National Academies of Sciences, Engineering ...

Page created by Sherry Lindsey
 
CONTINUE READING
Challenges in Studying Wildfire Health Impacts - Exposure Assessment - The National Academies of Sciences, Engineering ...
Challenges in Studying Wildfire
                       Health Impacts – Exposure
                              Assessment
                National Academies of Sciences – Implications of the California Wildfires for
                                Health, Communities, and Preparedness

                       Colleen Reid, PhD MPH
           Assistant Professor, Department of Geography
                   University of Colorado Boulder
                 Email: Colleen.Reid@Colorado.edu

Image Credit: NASA                           Reid CE | NAS 2019
Challenges in Studying Wildfire Health Impacts - Exposure Assessment - The National Academies of Sciences, Engineering ...
When investigating a wildfire
retrospectively, how do you
assess which people were
exposed?

               Reid CE | NAS 2019
Challenges in Studying Wildfire Health Impacts - Exposure Assessment - The National Academies of Sciences, Engineering ...
Exposure Assessment Methods Used in
Wildfire Epidemiological Studies
• Temporal Comparisons
  • Relatively easy
  • Can’t estimate dose-response
  • Associations may not be solely
    due to smoke exposure

                                                          Holstius et al. 2012, EHP

                                     Reid CE | NAS 2019
Challenges in Studying Wildfire Health Impacts - Exposure Assessment - The National Academies of Sciences, Engineering ...
When investigating a wildfire
retrospectively, how do you
assess how much exposure those
people experienced?

             Reid CE | NAS 2019
Challenges in Studying Wildfire Health Impacts - Exposure Assessment - The National Academies of Sciences, Engineering ...
Exposure Assessment Difficulties with
Wildfires
 •Monitoring data
    • Lots of data – relatively easy to access
 • The monitors are not always in all of the locations
   that you want
 • Many EPA PM2.5 monitors only measure every
   sixth or third day
 • Leads to spatial and temporal averaging of
   exposure measurements
     • But, smoke plumes migrate quickly, changing
       exposures over smaller spatial and temporal
       scales

                                           Reid CE | NAS 2019   5
Challenges in Studying Wildfire Health Impacts - Exposure Assessment - The National Academies of Sciences, Engineering ...
Exposure Assessment Methods Used in
Wildfire Epidemiological Studies
• Nearest air pollution
  monitor or average of
  regional monitors

https://www.epa.gov/out
door-air-quality-
data/interactive-map-air-
quality-monitors
                            Reid CE | NAS 2019
                              Reid CE | NAS 2019
Challenges in Studying Wildfire Health Impacts - Exposure Assessment - The National Academies of Sciences, Engineering ...
Exposure Assessment Methods Used in
 Wildfire Epidemiological Studies
                                                               MODIS AOD August 2018
• Satellite Data – Aerosol
  Optical Depth
   • Benefits
      • Full spatial coverage (except
        for cloud masking)
      • Actual measurement
   • Drawbacks
      • Full column measurement
        and not just at ground level
      • Missing data

                                        https://earthobservatory.nasa.gov/global-maps/MODAL2_M_AER_OD

                                          Reid CE | NAS 2019
Challenges in Studying Wildfire Health Impacts - Exposure Assessment - The National Academies of Sciences, Engineering ...
Exposure Assessment Methods Used in
Wildfire Epidemiological Studies
• Air pollution models                               • Advantages
   • Dispersion models                                         • Complete spatial coverage
      • HYSPLIT (Thelen et al., 2013)                          • Can get just ground-level
      • CalPuff (Henderson et al. 2011)                          concentration estimates
   • Chemical Transport Models                                 • Allow estimation of counterfactual
     (CTMs)
      • GEOS Chem (Liu et al. 2016; Liu et al.
                                                     • Disadvantages
        2017)                                                  • Dependent on the inputs
      • WRF-Chem (Gan et al. 2017)                             • uncertainties in emissions
      • CMAQ (DeFlorio-Barker et al., 2019)                      estimates from fires

                                          Reid CE | NAS 2019
Challenges in Studying Wildfire Health Impacts - Exposure Assessment - The National Academies of Sciences, Engineering ...
Exposure Assessment Methods Used in
Wildfire Epidemiological Studies
• Blended Models
   • Statistically combine CTMs, satellite data, and
     monitoring data
   • Sometimes also auxiliary data

• GWR method – Gan et al. 2017 and
  Lassman et al. 2017
• Machine learning method – Reid et al.
  2015 and Yao et al. 2018

                                     Reid CE | NAS 2019
Challenges in Studying Wildfire Health Impacts - Exposure Assessment - The National Academies of Sciences, Engineering ...
Adapt Land Use Regression Modeling with Machine
Learning and Adding Temporal Component

     • Include novel spatiotemporal datasets
     • Apply machine learning methods to
         • Select from a long list of predictor variables
         • Select from a variety of statistical algorithms

                                       Reid CE | NAS 2019    10
      Image courtesy of Mike Jerrett
Source: Reid et al. 2015. Spatiotemporal Prediction of Fine Particulate Matter During the 2008 Northern California Wildfires Using Machine Learning.
                                                         Environmental Science & Technology.

                                                                                                                                                       11
July 11, 2008

        75   µg/m3
                                            Reid CE | NAS 2019
Large circles are observed values at monitors, small circles are predicted values
Watson et al.
                     Under Revisions
                     at
                     Environmental
                     Pollution

Reid CE | NAS 2019           13
PM2.5 and ozone exposure estimates by ZIP code
by day for the 2008 northern California wildfires

                                            Reid et al. 2019 Env Int
                                                                       14
                       Reid CE | NAS 2019
Variables                                   Data Source                                          Temporal Resolution     Spatial Resolution Buffer Size

       Goal #1: Estimating Exposures
Dependent Variable
                                            US EPA, states
                                            Federal Land Manager Environmental Database, Fire
                                            Cache Smoke Monitor Archive, IMPROVE Network,
PM2.5 from monitoring stations              academic research groups                             Daily or hourly         point
Spatiotemporal Variables
GOES Aerosol and Smoke Product (GASP) AOD   NOAA                                                 Hourly                  4 km
Multi-Angle Implementation of Atmospheric
Correction (MAIAC) AOD                      NASA                                                 Daily                   1 km
MODIS Active Fire Detection                 NASA                                                 Daily                   1 km
VIIRS Fire Occurence                        NASA                                                 Daily                   375 m
Hazard Mapping System (HMS) Smoke and Fire                                                                                                  25km, 50, 100km, 500km,
Product                                    NOAA                                                  Daily                   4km                1000km, 2000km
MODIS Normalized Difference Vegetation Index
(NDVI)                                       NASA                                                Monthly                 1 km
14 Meteorological Variables                 NOAA: North American Mesoscale (NAM) Forecast System 6-Hourly                12 km

Spatial Variables
Elevation (m)                               USGS                                                 Nominal 2-month cycle   1 arc-second

Percentages of land cover types             National Land Cover Database 2011                    Every 5 years           30 m               1km
Kilometers of highway within buffer zones   National Highways Planning Network, US DOT                                                      100, 250, 500, and 1000 m
Temporal Variables

Julian Date                                                                                      Daily

Weekend                                                                                          Daily

     Maestas, Reid, Considine, Li et al. Work In Progress
Blended Models…
• Strengths
   • May be able to combine best of satellite and CTM
   • Full spatial coverage
• Weaknesses
   • There are assumptions of each model that should be understood –
     performance may vary spatially and temporally
   • There are errors in these models that should be considered
   • Results of epidemiological investigations may differ because of the exposure
     assessment method….
Different findings from different exposure
models….

  Henderson et al. (2011) EHP                        Gan et al. (2017) GeoHealth
                                Reid CE | NAS 2019
Moving Forward….
• More work to improve and compare exposure models
• More work on integrating exposure measurement error into
  epidemiological models
• Possibly increased use of social media on its own or in blended
  models to estimate exposure (i.e., Ford et al. 2017 Atmos. Chem.
  Phys.)
• Cheaper sensors or personal monitoring

                                Reid CE | NAS 2019
References
• DeFlorio-Barker S, Crooks J, Reyes J, Rappold AG. 2019. Cardiopulmonary Effects of Fine Particulate Matter Exposure among Older Adults,
  during Wildfire and Non-Wildfire Periods, in the United States 2008-2010. Environ Health Perspect 127:37006; doi:10.1289/EHP3860.
• Ford B, Burke M, Lassman W, Pfister G, Pierce JR. 2017. Status update: is smoke on your mind? Using social media to assess smoke
  exposure. Atmos Chem Phys 17:7541–7554; doi:10.5194/acp-17-7541-2017.
• Gan RW, Ford B, Lassman W, Pfister G, Vaidyanathan A, Fischer E, et al. 2017. Comparison of wildfire smoke estimation methods and
  associations with cardiopulmonary-related hospital admissions. Geohealth 1:122–136; doi:10.1002/2017GH000073.
• Henderson SB, Brauer M, Macnab YC, Kennedy SM. 2011. Three measures of forest fire smoke exposure and their associations with
  respiratory and cardiovascular health outcomes in a population-based cohort. Environmental health perspectives 119:1266–71;
  doi:10.1289/ehp.1002288.
• Holstius DM, Reid CE, Jesdale BM, Morello-Frosch R. 2012. Birth weight following pregnancy during the 2003 Southern California wildfires.
  Environmental health perspectives 120:1340–5; doi:10.1289/ehp.1104515.
• Lassman W, Ford B, Gan RW, Pfister G, Magzamen S, Fischer EV, et al. 2017. Spatial and temporal estimates of population exposure to
  wildfire smoke during the Washington state 2012 wildfire season using blended model, satellite, and in situ data. GeoHealth 1:106–121;
  doi:10.1002/2017GH000049.
• Liu JC, Wilson A, Mickley LJ, Dominici F, Ebisu K, Wang Y, et al. 2016. Wildfire-specific Fine Particulate Matter and Risk of Hospital Admissions
  in Urban and Rural Counties. Epidemiology (Cambridge, Mass) 28:77–85; doi:10.1097/ede.0000000000000556.
• Liu JC, Wilson A, Mickley LJ, Ebisu K, Sulprizio MP, Wang Y, et al. 2017. Who Among the Elderly Is Most Vulnerable to Exposure to and Health
  Risks of Fine Particulate Matter From Wildfire Smoke? Am J Epidemiol 186:730–735; doi:10.1093/aje/kwx141.
• Reid CE, Considine EM, Watson GL, Telesca D, Pfister GG, Jerrett M. 2019. Associations between respiratory health and ozone and fine
  particulate matter during a wildfire event. Environment International 129:291–298; doi:10.1016/j.envint.2019.04.033.
• Reid CE, Jerrett M, Petersen ML, Pfister GG, Morefield PE, Tager IB, et al. 2015. Spatiotemporal prediction of fine particulate matter during
  the 2008 northern California wildfires using machine learning. Environmental science & technology 49:3887–96; doi:10.1021/es505846r.
• Thelen B, French NH, Koziol BW, Billmire M, Owen RC, Johnson J, et al. 2013. Modeling acute respiratory illness during the 2007 San Diego
  wildland fires using a coupled emissions-transport system and generalized additive modeling. Environmental health : a global access science
  source 12:94; doi:10.1186/1476-069x-12-94.
• Wu J, Winer A, Delfino R. 2006. Exposure assessment of particulate matter air pollution before, during, and after the 2003 Southern
  California wildfires. Atmospheric Environment 40: 3333–3348.
• Yao J, Brauer M, Raffuse SM, Henderson S. 2018. A machine learning approach to estimate hourly exposure to fine particulate matter for
  urban, rural, and remote populations during wildfire seasons. Environ
                                                                      CE ReidSci Technol;
                                                                              | NAS 2019 doi:10.1021/acs.est.8b01921.                               19
Thank You!!
    Questions?

Colleen.Reid@Colorado.edu
You can also read