Sensitivity study on polarized aerosol retrievals of PARASOL in Beijing and Kanpur

Abstract. Sensitivity study on the PARASOL aerosol retrieval algorithm over land is presented for aerosol mixtures composed of fine mode pollution particles combined with coarse mode desert dust. First an assessment of the PARASOL aerosol products during the period of 2005–2009 is investigated by comparison with AOD derived by AERONET (Aerosol Robotic Network) at Beijing and Kanpur.

Validation against AERONET fine mode AOD shows an overall high correlation of R² = 0.79 for Beijing and R² = 0.83 for Kanpur. However, the PARASOL retrievals are found to underestimate aerosol optical depth by about 27% and 34% for Beijing and Kanpur, respectively. The AOD agreement is obviously poorer as compared to AERONET total AOD, showing underestimation by 60% and 67%. At both sites, the PARASOL retrieval algorithm performs better in autumn and winter seasons with the best appearing in autumn.

As PARASOL aerosol algorithm is sensitive to the accumulation mode of the aerosol size distribution, we conduct study on the threshold radius of this fraction of size distribution, named as sensitive radius, for different seasons at both Beijing and Kanpur. The results show that the sensitive radius for polarized aerosol retrieval is 0.35 μm for all seasons. And the agreement is significantly improved by employing comparison against the AERONET AOD recomputed for radius <0.35 μm, showing a correlation coefficient (R²) of 0.82 with relative difference being 12% for Beijing and 0.87 with relative difference being 19% for Kanpur.

The sensitivity study on uncertainty of PARASOL aerosol retrieval demonstrates that uncertainties caused by the algorithm-assumed refractive index and size distribution are significantly higher in spring than those of autumn and winter seasons. The aerosol retrieval errors caused by aerosol polarized phase function q_a(Θ) for spring are found to be higher at Kanpur, due to the obviously higher content of coarse dust particles.

For all seasons the aerosol retrieval errors contributed by uncertainty in q_a(Θ) are much close to the total retrieval errors (accounts for about 65% to 94% in different seasons), indicating that the overestimate of q_a(Θ) in PARASOL algorithm accounts for most of the underestimate of retrieved AOD at both sites.

Investigation on the uncertainty of surface contribution shows that the surface model overestimates surface polarization from about 20% to 50% with the maximum uncertainties occurring in winter.