<abstract xml:lang="en" xmlns="http://www.w3.org/1999/xhtml"><p>This paper presents the application of a neural network methodology to historical time series of GPS data from the IGS (International GPS Service) network, based on terrestrial water storage information. Hydrology signals at the GPS sites are important for including water loading corrections in GPS data processing. However, it is quite common that a correct global water storage model may not be available for this purpose, due to lack of science data. It is therefore mostly assumed that water mass redistribution is one of the potential contributors to the seasonal variations in GPS station position results, particularly, in the vertical direction. Presently, the IERS Special Bureau for Hydrology (SBH) has archived continental water storage data from some of the latest model developments. Examples include the monthly (GRACE, NOAA CPC, NCEP/NCAR CDAS-1) and daily (NCEP/NCAR and ECMWF reanalyses) solutions. It is valuable to study the relationship between these solutions and long-term geodetic results, especially as the water storage models continue to be refined. Using neural networks offers an effective approach to correlate the non-linear input of hydrology signals and output of geodetic results by recognizing the historic patterns between them. In this study, a neural network model is developed to enable the prediction of GPS height residuals based on the input of NOAA CPC hydrology data. The model is applied to eight GPS sites with satisfactory results.</p></abstract>

This paper presents the application of a neural network methodology to historical time series of GPS data from the IGS (International GPS Service) network, based on terrestrial water storage information. Hydrology signals at the GPS sites are important for including water loading corrections in GPS data processing. However, it is quite common that a correct global water storage model may not be available for this purpose, due to lack of science data. It is therefore mostly assumed that water mass redistribution is one of the potential contributors to the seasonal variations in GPS station position results, particularly, in the vertical direction. Presently, the IERS Special Bureau for Hydrology (SBH) has archived continental water storage data from some of the latest model developments. Examples include the monthly (GRACE, NOAA CPC, NCEP/NCAR CDAS-1) and daily (NCEP/NCAR and ECMWF reanalyses) solutions. It is valuable to study the relationship between these solutions and long-term geodetic results, especially as the water storage models continue to be refined. Using neural networks offers an effective approach to correlate the non-linear input of hydrology signals and output of geodetic results by recognizing the historic patterns between them. In this study, a neural network model is developed to enable the prediction of GPS height residuals based on the input of NOAA CPC hydrology data. The model is applied to eight GPS sites with satisfactory results.

A Neural Network Relation Of Gps Results With Continental Hydrology

Geospatial & Earth Monitoring Division, Geoscience Australia, GPO Box 378, Canberra, ACT 2601, AUSTRALIA

Artificial Satellites

{"article-title":"A Neural Network Relation Of Gps Results With Continental Hydrology"}

This paper presents the application of a neural network methodology to historical time series of GPS data from the IGS (International GPS Service) network,...

A Neural Network Relation Of Gps Results With Continental Hydrology

Publié en ligne: 03 mai 2007

Pages: 23 - 32

DOI: https://doi.org/10.2478/v10018-007-0003-7

Mots clés
continental water storage, geodetic time series, hydrological models, neural networks

This content is open access.

A Neural Network Relation Of Gps Results With Continental Hydrology

Publié en ligne: 03 mai 2007

Pages: 23 - 32

DOI: https://doi.org/10.2478/v10018-007-0003-7

Mots cléscontinental water storage, geodetic time series, hydrological models, neural networks

This content is open access.

Mots clés
continental water storage, geodetic time series, hydrological models, neural networks