An approach to evaluate the absolute accuracy of WVR water vapor measurements inferred from multiple water vapor techniques

Abstract

This paper uses three different types of water vapor observation instruments, radiosonde, AERONET sunphotometer and GPS, to infer the regression coefficients of one WVR (model: WVR-1100) in Hong Kong – a coastal city with high humidity. The regression using the three types of reference water vapor data is performed on a monthly basis for 6 months from January to June 2012. In order to evaluate the WVR regression accuracies, a water vapor-assisted (WV-assisted) GPS Precise Point Positioning (PPP) method is proposed. The inferred water vapor data are directly injected into PPP computation to correct the water vapor wet tropospheric delay in GPS signals. In principle, water vapor of better accuracy will produce GPS PPP solutions of higher accuracy. Our analysis results show that the radiosonde, AERONET and GPS data all can be used to regress WVR and produce accurate WVR water vapor if the regressed instruments have good data quality. We find that the WVR water vapor inferred from GPS water vapor regression has the most reliable regression results. The vertical component of PPP solutions is very stable, with consistent biases (bias varying by 0.38cm) and standard deviations (bias variation by 0.59cm) over a 6-month period in 2012. When sufficient AEROENT water vapor data are available for WVR regression, the WVR water vapor accuracy will become compatible with that inferred from GPS water vapor regression. However AERONET water vapor measurements are seriously affected by weather condition and can be obtained only in sunny and clear conditions. Compared with the bias variation of 0.38cm using GPS water vapor to regress WVR, the WVR water vapor data regressed by radiosonde result in a bias variation of 3.95cm in the PPP vertical component during the 6-month period. All of the regressed WVR contain a bias, which possibly results from the fact that the WVR, GPS, AERONET and radiosonde stations are all horizontally and vertically separated. Overall, the WVR water vapor data inferred from GPS water vapor regression are regarded to be most reliable and it is the most suitable data source for WVR regression.