Hits:
Impact Factor:4.9
Journal:GPS Solutions
Key Words:Weighted mean temperature; Sparse kernel; Spectral analysis; L1-norm
Abstract:As a crucial parameter in the process of converting the zenith wet delay into precipitable water vapor at Global Navigation Satellite System (GNSS) stations, the weighted mean temperature (Tm) influences the accuracy of GNSS-based water vapor retrievals. We propose an improved atmospheric Tm modeling method by introducing sparse kernel learning to obtain high-accuracy and high-temporal-resolution Tm data. To establish the model, the temporal variation characteristics of Tm time series are first analyzed by spectral analysis using the Lomb–Scargle periodogram. Second, the sparse kernel learning method is introduced to model the residuals from spectral analysis, for which the Gauss radial basis function kernel, L1-norm regularization, and the highly efficient fast iterative shrinkage thresholding algorithm are employed. To verify the performance of the proposed method, we used ERA5 hourly data from China’s 9 International GNSS Service stations produced by the European Centre for Medium-Range Weather Forecasts. ERA5 Tm data with a temporal resolution of 12 h are used as the training data, and the ERA5 hourly Tm data (excluding the data employed for modeling) are used for testing. The results show that compared with the spectral analysis accuracy, the accuracy of the calculated Tm can be improved by approximately 48.3% when the residual sparse kernel learning method is used. Thus, this proposed method for GNSS-based water vapor retrievals can provide high-accuracy and high-temporal-resolution Tm data.
Indexed by:Journal paper
Document Code:28
Discipline:Engineering
First-Level Discipline:Surveying and Mapping
Document Type:J
Volume:25
Issue:1
Translation or Not:no
Date of Publication:2021-01-04
Included Journals:SCI
Links to published journals:https://doi.org/10.1007/s10291-020-01061-3