Development of on-site self-calibration and retrieval methods for sky-radiometer observations of precipitable water vapor

The Prede sky radiometer measures direct solar irradiance and the angular distribution of diffuse radiances at the ultraviolet, visible, and near-infrared wavelengths. These data are utilized for the remote sensing of aerosols, water vapor, ozone, and clouds, but the calibration constant, which is the sensor output current of the extraterrestrial solar irradiance at the mean distance between Earth and the Sun, is needed. The aerosol channels, which are the weak gas absorption wavelengths of 340, 380, 400, 500, 675, 870, and 1020 nm, can be calibrated by an on-site self-calibration method, the Improved Langley method. This on-site self-calibration method is useful for the continuous long-term observation of aerosol properties. However, the continuous long-term observation of precipitable water vapor (PWV) by the sky radiometer remains challenging because calibrating the water vapor absorption channel of 940 nm generally relies on the standard Langley (SL) method at limited observation sites (e.g., the Mauna Loa Observatory) and the transfer of the calibration constant by a side-by-side comparison with the reference sky radiometer calibrated by the SL method. In this study, we developed the SKYMAP algorithm, a new on-site method of self-calibrating the water vapor channel of the sky radiometer using diffuse radiances normalized by direct solar irradiance (normalized radiances). Because the sky radiometer measures direct solar irradiance and diffuse radiance using the same sensor, the normalization cancels the calibration constant included in the measurements. The SKYMAP algorithm consists of three steps. First, aerosol optical and microphysical properties are retrieved using direct solar irradiances and normalized radiances at aerosol channels. The aerosol optical properties at the water vapor channel are interpolated from those at aerosol channels. Second, PWV is retrieved using the angular distribution of the normalized radiances at the water vapor channel. Third, the calibration constant at the water vapor channel is estimated from the transmittance of PWV and aerosol optical properties. Intensive sensitivity tests of the SKYMAP algorithm using simulated data of the sky radiometer showed that the calibration constant is retrieved reasonably well for PWV < 2 cm, which indicates that the SKYMAP algorithm can calibrate the water vapor channel on-site in dry conditions. Next, the SKYMAP algorithm was applied to actual measurements under the clear-sky and low-PWV (< 2 cm) conditions at two sites, Tsukuba and Chiba, Japan, and the annual mean calibration constants at the two sites were determined. The SKYMAP-derived calibration constants were 10.1% and 3.2% lower, respectively, than those determined by a side-by-side comparison with the reference sky radiometer. After determining the calibration constant, we obtained PWV from the direct solar irradiances in both the dry and wet seasons. The retrieved PWV values corresponded well to those derived from a global-navigation-satellite-system-global-positioning-system receiver, a microwave radiometer, and an AERONET (Aerosol Robotic Network) sun-sky radiometer at both sites. The correlation coefficients were greater than 0.96. We calculated the bias errors and the root mean square errors by comparing PWV between the DSRAD (direct solar irradiance) algorithm and other instruments. The magnitude of the bias error and the root mean square error were < 0.163 and < 0.251 cm for PWV < 3 cm, respectively. However, our method tended to underestimate PWV in the wet conditions, and the magnitude of the bias error and the root mean square error became large, < 0.594 and < 0.722 cm for PWV > 3 cm, respectively. This problem was mainly due to the overestimation of the aerosol optical thickness before the retrieval of PWV. These results show that the SKYMAP algorithm enables us to observe PWV over the long term, based on its unique on-site self-calibration method.