## Abstract

The sunphotometer is broadly adopted in global or local aerosol measurement networks (e.g., CE318 in AERONET) or field observing campaigns. The water absorption 940nm combination-vibrational band of sunphotometers is always used to retrieve vertical columnar water content (VCW). The retrieval principle is based on the ‘Reagan’s model’, like Tw=exp(-a* (m*PW)^b), where Tw is band transmittance, PW is VCW and m is the relative air mass. The process includes two steps: 1) calculating the water absorption transmittance from the measurement by applying the calibration coefficient, which was always determined by the ‘modified Longley method’ from measurements under stable atmospheric conditions; 2) deriving the VCW by the ‘Reagan’s model’, the constants parameters a and b were easily determined by fitting the transmittances calculated by MODTRAN or other radiative transfer models.

Actually, there has been difference in transmittance if the observation angle is different, even though the total path water column abundance is the same. The difference reaches ~10% under certain conditions. ‘Regan’s model’ did not consider these factors. In this paper, the band transmittances of one CE318’s 940nm channel are calculated under different VCW, different zenith angles by Line-By-Line Radiative Transfer Model (LBLRTM). Then, by best fitting all the calculation data, we propose a new model of the dependence of the band transmittance on vertical water column abundance PW and relative air mass m. The model is taken the form as Tw=exp(-a*m^(c+d*lnPW)*PW^b). We determine these constants a, b, c, d by changing observation angle equals and VCW. As expected, the new model achieves far superior retrieval results and the relative error is less than 0.4% within the 63° zenith angle, but the error may reach ~10% for the Reagan’s model. It should be note that the retrieval error for new model is also less than for Reagan’s model, although the error becomes large when the zenith angle and VCW are large for either model.

An iterative calibration procedure is proposed based on the revised model. The calibration algorithm is validated by the CE318’s measurements over Dunhuang and Qing-Tibetan Plateau, where the atmosphere is stable and clean. The relative error of the calibration coefficients is within 2%. It is far less than the general calibration error of V0 ~10% published in formal publications. It also indicates that the new model has the superior advantage than the Reagan’s model.

In conclusion, compared with Reagan’s model, the new model we proposed shows advantages for compensating the differences in transmittance under the same total path water abundance with different observation angles. From simulation results, the retrieval error of VWC will reach ~10% by Reagan’s model when the observation angles as large as 63°, but 0.4% for our new model. So, our new retrieval model is better than Reagan’s model in retrieving water abundance, especially in large observation angles.

Improved Algorithm for Water Vapor Retrieval and Field Calibration in the Channell 940-nm of Sun-photometer

The sunphotometer is broadly adopted in global or local aerosol measurement networks (e.g., CE318 in AERONET) or field observing campaigns. The water absorption 940nm combination-vibrational band of sunphotometers is always used to retrieve vertical columnar water content (VCW). The retrieval principle is based on the ‘Reagan’s model’, like Tw=exp(-a* (m*PW)^b), where Tw is band transmittance, PW is VCW and m is the relative air mass. The process includes two steps: 1) calculating the water absorption transmittance from the measurement by applying the calibration coefficient, which was always determined by the ‘modified Longley method’ from measurements under stable atmospheric conditions; 2) deriving the VCW by the ‘Reagan’s model’, the constants parameters a and b were easily determined by fitting the transmittances calculated by MODTRAN or other radiative transfer models.

Actually, there has been difference in transmittance if the observation angle is different, even though the total path water column abundance is the same. The difference reaches ~10% under certain conditions. ‘Regan’s model’ did not consider these factors. In this paper, the band transmittances of one CE318’s 940nm channel are calculated under different VCW, different zenith angles by Line-By-Line Radiative Transfer Model (LBLRTM). Then, by best fitting all the calculation data, we propose a new model of the dependence of the band transmittance on vertical water column abundance PW and relative air mass m. The model is taken the form as Tw=exp(-a*m^(c+d*lnPW)*PW^b). We determine these constants a, b, c, d by changing observation angle equals and VCW. As expected, the new model achieves far superior retrieval results and the relative error is less than 0.4% within the 63° zenith angle, but the error may reach ~10% for the Reagan’s model. It should be note that the retrieval error for new model is also less than for Reagan’s model, although the error becomes large when the zenith angle and VCW are large for either model.

An iterative calibration procedure is proposed based on the revised model. The calibration algorithm is validated by the CE318’s measurements over Dunhuang and Qing-Tibetan Plateau, where the atmosphere is stable and clean. The relative error of the calibration coefficients is within 2%. It is far less than the general calibration error of V0 ~10% published in formal publications. It also indicates that the new model has the superior advantage than the Reagan’s model.

In conclusion, compared with Reagan’s model, the new model we proposed shows advantages for compensating the differences in transmittance under the same total path water abundance with different observation angles. From simulation results, the retrieval error of VWC will reach ~10% by Reagan’s model when the observation angles as large as 63°, but 0.4% for our new model. So, our new retrieval model is better than Reagan’s model in retrieving water abundance, especially in large observation angles.