An optimal calibration method for MODIS precipitable water vapor using GNSS observations

Abstract

Precipitable water vapor (PWV) plays an important role in the global water and energy cycle. Compared with GNSS and radiosonde which are distributed in the form of scatters, near-infrared -derived PWV has a higher spatial resolution, meeting more comprehensive investigation of regional climate change. However, PWV derived from the near-infrared water vapor channels of the Moderate Resolution Imaging Spectroradiometer (MODIS), onboard Aqua and Terra satellites, exhibits limitations of missing data and poor accuracy especially when data are collected under cloudy conditions. Many researches have been made to evaluate and calibrate MODIS-PWV with a cloud-free probability >95%, there is still very little research on improving accuracy of PWV under all-weather conditions. Therefore, an optimal calibration method was proposed, in which a filling algorithm was applied to enhance availability of the PWV data, and a linear and periodic calibration scheme based on the analysis of residuals was utilized to improve its accuracy. The experiment was conducted in Hong Kong with 11 uniformly distributed GNSS stations and the station cross-validation was employed using the GNSS-PWV as the references. The results show that the filling algorithms can first effectively fill the data vacancy and generate complete MODIS-PWV with data coverage reaching up to 100%, further make MODIS-PWV more conducive to construct following calibration model. The R-Square(R²), root mean square error (RMSE), mean absolute error (MAE) and relative error of the MODIS-PWV obtained by the proposed method are 0.67, 9 mm,7 mm and 22.5% compared with GNSS-PWV. In comparison to original MODIS-PWV, the RMSE and MAE are reduced by 62% and 59%. For the four seasons, the average value of RMSE is improved from 25 to 9 mm, 32 to 10 mm, 22 to 9 mm, and 12 to 8 mm, respectively. Moreover, the RMSEs and MAEs are about 6–8 mm and 5–6 mm for station cross-validation in every month. These results confirm that the proposed method can provide a complete and continuous PWV data and improve the accuracy of original MODIS-PWV, which is benefit for the hydrological and ecological research.