The relationship between up-down (UD) component of the site coordinates and zenith tropospheric delay (ZTD) is studied by using the correlation analysis method and the principal component analysis method for the purpose of developing techniques to separate vertical coordinates and zenith tropospheric delay more strictly in GPS analysis. The data set used in the study is the solutions of the routine analysis for GSI / GEENET (Geographical Survey Institute/GPS Earth Observation NETwork) with the Bernese GPS software. The two periods in the summer and winter of 1996; from August 8 to September 28 and from October 31 toDecember 30, are selected because continuous data set is available in the two periods. The daily anomalies in UD and ZTD (UD anomaly and ZTD anomaly) from their averaged values are used in the analyses. The strong positive or negative correlation coefficients between UDand ZTD- anomalies can be seen in many sites when the length of the data is 3 to 5 days. Also, the strong correlation coefficients are remained in some parts when the length of the data is longer than 17 days. The spatial patterns of the correlation coefficients with the data length of 9 days are almost steady in summer, while they change with 10-day scale in winter. Some of the patterns are similar to a spatial pattern of clustering structure applied for the GEONET routine analysis. The empirical orthogonal functions (EOF) for the first principal components (PC) in UD- and ZTD- anomalies correspond to the temporal change of the averaged values for all the sites in their anomalies. The EOFs for the second and third PC show common northsouth or east-west oscillation patterns over the Japanese Islands, which may correspond to weather variations over the islands. The spatial pattern of EOFs for UD anomalies shows similar spatial pattern to the clustering structure. These results suggest that changes in UD (site coordinate) are strongly affected by ZTD changes, corresponding to water vapor variations. However, some biases affected by cluster organizations in the GEONET routine nalysis areseen in the correlation maps and EOFs. It is thus necessary to re-analyze GEONET data with other methods such as precise point positioning that can avoid the clustering effects.
View full abstract