human health applications of atmospheric remote sensing simon hales, housing and health research...
Post on 21-Dec-2015
217 views
TRANSCRIPT
Human health applications of atmospheric remote sensing
Simon Hales, Housing and Health Research Programme, University of Otago,
Wellington, New Zealand [email protected]
OutlineAir pollution as a global public health issue3 examples:
– Global burden of disease from ambient PM estimated using MODIS aerosol
– Calibrating SCIAMACHY NO2 with surface monitoring: USA and Europe
– Assessing implications of climate/energy/transport policy on air pollution exposure and health impacts in Australia
Next steps– Estimating spatial distribution of exposure in New Zealand
using OMI NOx– UVB Vitamin D
Conclusion
Air pollution – a global health issue• Estimated 1.2% of deaths 0.5% PYLL (measures of
global burden of disease, BoD)• Data inputs for BoD from ambient air pollution– Population exposed– Long term average exposure (PM10 preferred)– Dose-response from cohort studies: deaths, hospital
admissions• Exposure uncertain, derived from sparse network of
fixed (usually urban) monitoring sites, plus empirical modelling: – economic, weather, population data and available PM
measurements in 304 cities used to estimate PM10 levels in 3000 cities with populations greater than 100,000.
Remote sensing?
• More detailed exposure data would be preferable
• MODIS AOT calibrated using urban station data• Result extrapolated to all land areas, population
weighted and then aggregated at country scale• Estimate of 20% global mortality; (which is
unfeasibly large)• Probable over estimation of exposure, due to
predominance of monitors in regions that are more polluted at the surface.
SCIAMACHY 1:USA
• Annual (2003) average NO2 and PM2.5 data for monitoring stations in the USA
• The annual average NO2 data were derived from hourly averages (up to 24 measurements per day, or 8760 measurements per year).
r N
NO2satellite vs NO2station 0.70 457
NO2satellite vs PM2.5 0.64 130
NO2station vs PM2.5 0.57 130
SCIAMCHY 2:Australia
• Modelling relations between emissions and surface concentration
• Prediction of public health (mortality) implications of hypothetICAL transport policy
Two step modelling approach
• Model A: relationships between surface monitoring and average SCIAMACHY tropospheric retrievals for Sydney and Melbourne, Australia:
• Found similar (linear) relationship for each city
• Spatial averaging of model predictions by small area (statistical local area, SLA):
#*
#*
#*
#*
#*
#*
#*
#* #*#*
#*
#*
#*
#*
#*#*
#*
#*#*
#*
#*
#*
#*
#*
#*
#*
#*
#*
#*
#*
#*
#*
#*
#*
#*#*
#*
#*
#*
#*
#*
#*
#*
#*#*
#*
#*
#*
#*#* #* #*
#*
#*
#*
#*
#* #*#*
#*#*
#*
#* #*#*
#*#*
#*#*
#*
#*
#*#*
#*
#*
#*
#*#*
#*
#*
#*
#*
#*
#*
#*
#*
#*
#* #*
#*
#*
#*
#*
#*
#*#*
#*
#*#*#*#*
#*
#*
#*
#*#*
#*
#*
#*
#*
#*
#*
#*
#*
#*
#*#*
#*
#*
#*
#*#*
#*
#*
#*#* #*
#*
#*
#*
#*
#*#*
#*
#* #*
#*
#*
#*
#*
#* #*#*
#*
#* #*#*
#*
#* #*
#*
#*
#*
#*
#*
#*
#*#*
#*
#* #*#*
#*
#*
#*
#*
#*#*
#*
#*
#*
#*
#*#*
#*
#*
#*
#*
#*
#*#*
#*
#*
#*
#*
#*
#*
#*
#*
#*
#* #*
#*
#*
#*
#*
#*
#*
#*
#*
#*
#*
#*
#*#*
#*
#*
#*
#*
#*
#*
#*
#*
#*#*
#*#*
#*
#*#*
#*
#*
#*
#*#*
#*#*#*
#*
#*
#*
#*
#*
#*#*
#*#*
#*
#*
#*
#*
#*
#*
#*
#*#*
#*
#*#*
#*
#*#*
#*#*
#*
#*#*
#*#*#*
#*
#*
#*#*
#* #*#*
#*
#*
#*
#*
#*
Model B
• Use external data on point source and diffuse (vehicle) emissions
• Model relationship between average NOx and natural log of total emissions, by SLA
• Predict effect of changing emissions on exposure within SLAs
• Potentially important input to climate/energy policy: could help validate emission reductions?
• Can also estimate likely effects of energy/transport policy changes on human health
• In this example, the effect of 50% reduction in vehicle emissions is substantial (several hundred early deaths per year in each city)
Next steps: RS data and public health
• OMI data for NZ: will be used as input to several epidemiological studies
• Estimates of spatial patterns of NOx for study of seasonal patterns of heart disease currently underway
• Applications of surface UVB estimates – effect on Vitamin D synthesis in skin: – many public health implications emerging:– need to understand how much UVB exposure is
optimal for different populations
Conclusions• Simple regression method using SCIAMACHY NOx data works
quite well for USA, Australia but not Europe or New Zealand• Possibly relates to:
– Scale of satellite observations vs scale of spatial variation of NOx in different regions?
– Time of observations not representative? – Regional differences in vertical profile (tropospheric column not
representative of surface levels)?– Cloud effects??– Could be resolved by meteorological/transport modelling (for
discussion)• Thanks to Folkert Boersma, Ronald van der A for the
invitation and travel funding• SH is funded by the National Heart Foundation of NZ