Presentation is loading. Please wait.

Presentation is loading. Please wait.

Estimating PM 2.5 from MODIS and MISR AOD Aaron van Donkelaar and Randall Martin March 2009.

Similar presentations


Presentation on theme: "Estimating PM 2.5 from MODIS and MISR AOD Aaron van Donkelaar and Randall Martin March 2009."— Presentation transcript:

1 Estimating PM 2.5 from MODIS and MISR AOD Aaron van Donkelaar and Randall Martin March 2009

2 Objective Get feedback Estimate long-term, high-resolution, fine particulate matter concentrations globally, using satellite-based measurements Satellite-based Sparse networks outside North America Global coverage Multi-year coverage – (2001-2006) Fine resolution –(0.1º x 0.1º) Fine Particulate Matter (PM 2.5 ) Dry weight of aerosol particles < 2.5 µm in diameter Infiltrates natural biological defenses Linked to lung cancer, cardiopulmonary mortality lowers life expectancy Often 24h measurements Annual EPA standard < 15 μg/m 3

3 τ -PM 2.5 Relation Most τ -PM 2.5 studies are –Regional –High temporal –Empirical or Hybrid –Reliant on a single instrument Relationship can be reduced to: Accuracy of PM2.5 will depend upon accuracy of both η and τ Estimated PM 2.5 = η· τ vertical structure aerosol type meteorological effects meteorology η

4 MODIS and MISR τ MODIS τ 1-2 days for global coverage Requires assumptions about surface reflectivity Version 5, best quality MISR τ 6-9 days for global coverage Simultaneous surface reflectance and aerosol retrieval Most recent available version, (F09_0017-F11_0021), best estimate Mean τ 2001-2006 at 0.1º x 0.1º 0 0.1 0.2 0.3 τ [unitless] MISR MODIS r = 0.40 vs. PM 2.5 r = 0.54 vs. PM 2.5

5 How to identify accurate regions? <0.40.4-0.6>0.6 <0.4 0.4-0.6 >0.6 ρ 650 nm / ρ 2.1 µm ρ 470 nm / ρ 650 nm Need way to group regions of similar quality Compare with AERONET - is the nearest station best? Divide world into regions of similar surface albedo ratios using MODIS surface albedo product (MCD43) Use same regions for MISR Similar to MODIS –ρ 650 nm = (0.39 – 0.67) * ρ 2.1 um –ρ 470 nm = 0.49 * ρ 650 nm + 0.005 Evaluate monthly AERONET and Satellite τ agreement at overpass within each region Remove τ from regions with mean error greater than ±(0.1+20%)

6 Global Monthly Filter Number of Included Months

7 Combined τ mixes MODIS and MISR MODIS r = 0.40 (vs. PM 2.5 ) MISR r = 0.54 (vs. PM 2.5 ) Combined MODIS/MISR r = 0.63 (vs. PM 2.5 ) 0.3 0.25 0.2 0.15 0.1 0.05 0 τ [unitless]

8 How do we find η? Estimated PM 2.5,24h = η· τ η= τ fτ f · PM 2.5,overpass · PM 2.5,24h τ τ fτ f PM 2.5,overpass v8-01-04 assimilated meteorology –GEOS-4 (2001-2006) GEOS-Chem Global Chemical Transport Model 2º x 2.5º Aerosols include: –Dust, sea salt, sulfate, organic carbon, black carbon, SOA Explore model and satellite Combine model and satellite model

9 How to best represent typical fine fractions? r GEOS- Chem 0.73 MODIS0.26 MISR ( τ > 0.2) 0.39 Based on non-coincident, monthly mean values τfτf = fine fraction τ

10 2º x 2.5º can contain a lot of variation PM 2.5,overpass → ν ·PM 2.5,overpass τ fτ f ν · τ f - ( ν -1)· τ f,free trop Can use observed variation within a model grid: where PM 2.5,overpass, τ f and τ f,tree trop are simulated Based upon Lamsal et al., 2008 ν = τ 0.1ºx0.1º τ 2ºx2.5º

11 Significantly improved agreement with coincident ground measurements over NA Satellite Derived Measured Satellite-Derived [ μ g/m3] Measured PM 2.5 [μg/m 3 ] PM 2.5 [ μ g/m 3 ] r MODIS τ 0.40 MISR τ 0.54 Combined τ 0.63 Combined PM 2.5 0.76

12 PM 2.5 -AOD relations can be globally estimated where: PM 2.5,24h = η· τ

13 Annual mean measurements –Outside Canada/US –295 sites (105 non-EU) r = 0.70 (0.72) slope = 0.77 (0.81) bias = 0.86 (-0.80) μg/m 3 Significant agreement with global mean measurements

14 Coincident error has two sources Satellite Error limited to 0.1 + 20% by AERONET filter Implication for satellite PM 2.5 determined by η Estimated PM 2.5 = η· τ Model Affected by aerosol optical properties, concentrations, vertical profile, relative humidity Most sensitive to vertical profile [van Donkelaar et al., 2006]

15 τ(z)/τ surface Altitude [km] CALIPSO allows profile evaluation Coincidently sample model and CALIPSO extinction profiles –Jun-Dec 2006 Compare % within boundary layer Model (GC) CALIPSO (CAL) Optical Depth from TOA Optical Depth at surface

16 Coincident measurements within estimated error Combined effect of τ and vertical structure: ±(5 μg/m 3 + 25%) Contains 95.1% of NA data Satellite-Derived PM 2.5 [ μ g/m 3 ]

17 Sampling frequency varies with region Potential loss of representativeness relative to annual mean

18 Sampling error is regional Compare continuous and coincident model results Plot sampling-induced error in excess of ±2 μg/m 3 Sampling-Induced Error = Annual PM 2.5 – Coincident PM 2.5 ± 2 μg/m 3 Annual PM 2.5 Sampling Error [%]

19 Satellite PM 2.5 deviates from model All East West All East West

20 Satellite-model deviation not just resolution All East West All East West

21 Significant global deviations from model Annual mean measurements –295 sites (105 non-EU) r = 0.70 (0.72) slope = 0.77 (0.81) bias = 0.86 (-0.80) μg/m 3 r = 0.39 (0.52) slope = 0.38 (0.37) bias = 7.29 (4.20) μg/m 3

22 Satellite-Derived PM 2.5 [μg/m 3 ]

23

24

25 Pope et al. [2009] –Mean life expectancy decreases 0.61±0.20 year / 10 μg/m 3 satellite-PM 2.5 + population map → –estimate of lost life expectancy ~2 years lost for 30-40% of Asian population ~6 months lost for 70% of E. North American population Loss in Expected Lifetime [years] PM 2.5 Exposure [μg/m 3 ] Population [%] High global impact of PM 2.5

26 Summary Satellite-derived PM 2.5 asset to global air quality monitoring Quantifiable Error –Coincident: ±(5 μg/m 3 + 25%) –Sampling: ±(2 μg/m 3 + 10%) Distinct from model data Potential for health studies


Download ppt "Estimating PM 2.5 from MODIS and MISR AOD Aaron van Donkelaar and Randall Martin March 2009."

Similar presentations


Ads by Google