Ground-based Polarization Remote Sensing of Atmospheric Aerosols and the Correlation between Polarization Degree and PM2.5

The ground-based polarization remote sensing adds the polarization dimension information to traditional intensity detection, which provides a new method to detect atmospheric aerosols properties. In this paper, the polarization measurements achieved by a new multi-wavelength sun photometer, CE318-DP, are used for the ground-based remote sensing of atmospheric aerosols. In addition, a polarized vector radiative transfer model is introduced to simulate the DOLP (Degree Of Linear Polarization) under different sky conditions. At last, the correlative analysis between mass density of PM2.5 and multi-wavelength and multi-angular DOLP is carried out. The result shows that DOLP has a high correlation with mass density of PM2.5, R2>0.85. As a consequence, this work provides a new method to estimate the mass density of PM2.5 by using the comprehensive network of ground-based sun photometer.


Introduction
Polarization is one of the inherent characteristics in atmospheric detection. Because any target in atmosphere will have their unique characteristics nature of polarization, when non-polarized natural light interacts with air molecules and aerosol particles. The key advantage of using polarized observation is that polarization features strongly depend on the characteristics of scattering particles, which provides a new method to detect PM 2.5 (Particle with aerodynamic diameter ≤ 2.5 Micron), whereas the classic approach using natural light is difficult to do. Atmospheric aerosol polarization measurement from ground-based remote sensing has been well progressed in recent years. A new polar metric sun photometer, CE318-DP, has been used in AErosol RObotic NETwork (AERONET) sites [1] . CE318-DP, developed by the CIMEL Electronic (Paris, France), adopts a new mechanical design providing the polarization capability in solar principal plane at all measurement wavelengths centred at 340, 380, 440, 500, 675, 870, 936, 1020 and 1640nm, makes the angular and multiwavelength polar metric measurements available. PM 2.5 , the particle with aerodynamic diameter less than 2.5 micro, is considered as a better exposure indicator than PM 10 (Particle with aerodynamic diameter ≤ 10.0 Micron) for health assessments [2] . However, the lack of a dense network to monitor PM 2.5 hinders health risk assessments at regional scale. Because the polarization features are strongly depending on the characteristics of scattering particles, we try to employ the ground-based polar metric sun photometer to detect PM 2.5 .
In this paper, we present the ground-based remote sensing of atmospheric aerosols, achieving the multi-angular and multi-wavelength DOLP (Degree Of Linear Polarization) from CE318-DP. In addition, a polarized vector radiative transfer model is introduced to simulate the DOLP under different sky conditions. Then we provide a preliminary analysis on the DOLP and mass density of PM 2.5 , and get a reasonable agreement. As a result, we provide a new method to estimate the mass density of PM 2.5 by using the comprehensive network of ground-based sun photometer.

Measurements
The polarization measurements are implemented at RADI (Institute of Remote Sensing and Digital Earth Chinese Academy of Sciences), Beijing, China, using multi-wavelength sun photometer, CE318-DP. The instrument at RADI can be found at Figure 1. By making multi-wavelength sky scanning, CE318-DP can obtain a spectral and angular distribution of DOLP of sky light [3] . Light scattered in the atmosphere by air molecules and aerosol particles is partially polarized and can be described by Stocks vector [4] . In order to detect polarization signal, the CE318-DP polarized radiometers place three linear polarizes, which are keeping 60 o between their axis orientation angles.
where 12 R , 13 R and  are calibration coefficients. 12 R and 13 R are calibrated to correct the difference between the intensity responses of the three polarization units.  is calibrated to correct the measured DOLP with the reference DOLP source. By using this methodology, DOLP was measured with an uncertainty of ~0.01 [5] .
The field experiment of CE318-DP instrument (#350) has been performed in north China at Beijing site from 12/2010 to 03/2011. From these measurements, the DOLP of sky light is calculated at five wavelengths (440, 675, 870, 1020 and 1640nm) in SPP (Sun Principal Plane) geometry, using the algorithm introduced above. Examples of measured multi-wavelength and multi-angular DOLP of sky light can be found at Figure 2.

Simulations
The propagation and redistribution of radiation in surface-atmosphere can be fully described by vector radiative transfer equation, which is the basis of quantitative remote sensing. In this paper, we adopted the SOS model, which solves the vector equation of radiative transfer by using the successive orders of scattering approach, to simulate the Stocks parameters ( ) of skylight, then according the Stocks formalism the DOLP can be simulated. The formula to calculate DOLP [4] [6] : We performed simulations for different air condition in order to make clear aerosols optical properties and its impact on the polarization radiance. The atmospheric aerosols optical properties, including aerosol optical depth (AOD), aerosol particle size distribution and complex refractive index which are all retrieved from CE318-DP observations at RADI. These aerosols properties are input to the SOS model to simulate the DOLP under different sky conditions. The detailed steps of simulation used in this study are given in Figure3.

The Correlation between DOLP and PM 2.5
We present the DOLP measurements and simulations in different air conditions in the section 2. Our primary focus is on the analysis of the each band's maximum DOLP in both sunny and hazy days.  In addition, we have tested our measured and simulated data from 28th of December 2010 to 7th of March 2011. These are specifically in both sunny and hazy condition, mass density of PM 2.5 ranging from lower than 10μg/m³ to higher than 400μg/m³. We tend to establish an exponential regression of DOLP and K. Then an exponential regression was performed, the equation of exponential regression is PM 2.5 = 83.62  exp(5.37  K), squares of correlation coefficient (R²) = 0.86, regression points number (n) = 58. Results show that K value of multi-angular and multi-wavelength DOLP has a high correlation with mass density of PM 2.5 . This simple exponential model links daily ground-based polarization remote sensing and PM 2.5 mass concentration.

Conclusion
This work developed a methodology to estimate the mass density of PM 2.5 using the comprehensive network of ground-based polarization sun photometer. To achieve this objective, first, we gathered the DOLP and daily PM 2.5 from 28th of December 2010 to 7th of March 2011. In a second time, we compared the measured and simulated DOLP in different air conditions. Our results show high correlations (R² = 0.86). It should be noted that this result only tests in a few cases, more extensive studies should be done to verify the result. Conclusions from this work contribute to the ground-based polarization remote sensing which is performed routinely today [1] [7] . As a consequence, this work provides a new method to estimate the mass density of PM 2.5 by using the comprehensive network of ground-based sun photometer.