Comparisons of GPS‐derived precipitable water vapors with radiosonde observations in Japan

Abstract

We investigated the accuracy and features of precipitable water vapor (PWV) obtained from the Japanese Global Positioning System (GPS) network. We compared PWVs estimated from the Geographical Survey Institute (GSI) routine geodetic analysis using the Bernese software with those derived from the nearest 10 radiosonde stations over the Japanese islands. The comparisons for a half year showed that the agreement of the GPS‐derived PWV was 3.7 mm in terms of rms difference, with a standard deviation of 2.6 mm and a mean bias of −2.7 mm, which is less precise than those obtained in central North America. This may reflect the high temporal and spatial variability of water vapor over the Japanese islands, causing differences between PWVs if GPS and radiosonde launch stations are not colocated. We also found systematic biases in the GPS‐derived PWV. The absolute value of the mean bias at 1200 UT was systematically larger than that at 0000 UT. Their difference reached as much as 2 mm. Moreover, the mean biases tended to increase as PWV increased. We also confirmed that there were similar systematic biases in the PWV obtained from our GPS analysis using the GIPSY software with high temporal resolution. This fact indicates that the systematic biases depend neither on analysis software nor on temporal resolution of PWV estimation. We analyzed GPS data collected at the Tsukuba station using GIPSY for 2 years and showed a possibility that the systematic biases also found at the Tsukuba station were attributed to ocean tidal loading and seasonal variation of the mapping function both of which were not taken into account in the GPS analysis with the routine analysis using Bernese and our analysis using GIPSY. These results suggest the importance of implementing them in the analysis for accurate estimates of absolute value of PWV from the Japanese GPS network.