A Fractional AOD Approach to Derive PM2.5 Information Using MISR Data Coupled with GEOS-CHEM Aerosol Simulation Results

Yang Liu, Ralph Kahn, Solene Turquety, Robert M. Yantosca, and Petros Koutrakis

with thanks to Lyatt Jaegle and Rynda Hudman

​

April 11, 2007

Can we get more aerosol information from satellites in addition to AOD?

2

Satellite retrieved AOD can improve PM2.5 concentration estimates Valuable in pollution health effect studies (spatial and temporal coverage)

MISR reports aerosol microphysical properties (e.g., particle size, shape and darkness) May provide much needed PM2.5 speciation and size information in health studies

However, total mass is unlikely the only cause of PM2.5 toxicity

The Plan

1. Develop MISR fractional AODs that utilize MISR AOD and aerosol mixture information

​

• Build models using them as predictors to estimate the concentrations of PM_2.5 constituents

​

• Compare model performance with total AOD models

​

• Estimating size distributions of PM_2.5 constituents using model results

3

Main Take-Home Messages

• Regression models developed with MISR fractional AODs as major predictors are more flexible, and have significantly higher predicting powers than the total-AOD models

​

• Much more aerosol information in additional to column AOD is hidden in MISR data. Our approach can be used to extract it

4

MISR fractional AODs break the total AOD into contributions of individual components

5

Total column AOD = sum of all fractional AODs

GC aerosol simulations scale column AODs to surface AOD values

6

Note: currently difficult to match more precisely between MISR and GC due to MISR component definitions

Regression models link fractional AODs with particle concentrations

Compared with total AOD model

​

• Individual components can have different regression coefficients, or even be insignificant
• Each component may assume different growth pattern with increasing RH
• Have the potential to estimate major PM2.5 constituents

7

MISR data contains particle size distributions

8

A Case Study

1. MISR 2005 aerosol data (version 17)
2. EPA STN database (~200 sites, 24-hr concentrations of PM2.5, SO4, NO3, OC, EC)
3. GEOS-CHEM simulated aerosol profiles (V7-02-04)

9

Model Performance – Fractional vs. Total AOD (Eastern US)

10

Model Performance – Fractional vs. Total AOD (Western US)

11

Note: sample size too small, changes in adj. R² are only qualitative indication of improvement

PM2.5 Size Distribution can be estimated using regression coefficients

12

MISR

AERONET

East:

Model PM2.5 Mode Diameter = 0.19 μm

AERONET Mode Diameter = 0.29 μm

West:

Model PM2.5 Mode Diameter = 0.22 μm

AERONET Mode Diameter = 0.25 μm

​

Difference: MISR Sampling bias?

Conclusions

• Fractional AOD values can be calculated using MISR retrieved aerosol microphysical properties – unique to MISR

​

• Regression models using fractional AODs as predictors perform much better than the total AOD models

​

• Additional PM2.5 information such as composition and size distribution can be obtained using this method

​

• Longer MISR data time series are needed to get robust parameter estimates

13