Water vapor mapping by fusing InSAR and GNSS remote sensing data and atmospheric simulations

Data fusion aims at integrating multiple data sources that can be redundant or complementary to produce complete, accurate information of the parameter of interest. In this work, data fusion of precipitable water vapor (PWV) estimated from remote sensing observations and data from the Weather Resear...

Full description

Bibliographic Details
Main Authors: F. Alshawaf, B. Fersch, S. Hinz, H. Kunstmann, M. Mayer, F. J. Meyer
Format: Article
Language:English
Published: Copernicus Publications 2015-12-01
Series:Hydrology and Earth System Sciences
Online Access:http://www.hydrol-earth-syst-sci.net/19/4747/2015/hess-19-4747-2015.pdf
Description
Summary:Data fusion aims at integrating multiple data sources that can be redundant or complementary to produce complete, accurate information of the parameter of interest. In this work, data fusion of precipitable water vapor (PWV) estimated from remote sensing observations and data from the Weather Research and Forecasting (WRF) modeling system are applied to provide complete grids of PWV with high quality. Our goal is to correctly infer PWV at spatially continuous, highly resolved grids from heterogeneous data sets. This is done by a geostatistical data fusion approach based on the method of fixed-rank kriging. The first data set contains absolute maps of atmospheric PWV produced by combining observations from the Global Navigation Satellite Systems (GNSS) and Interferometric Synthetic Aperture Radar (InSAR). These PWV maps have a high spatial density and a millimeter accuracy; however, the data are missing in regions of low coherence (e.g., forests and vegetated areas). The PWV maps simulated by the WRF model represent the second data set. The model maps are available for wide areas, but they have a coarse spatial resolution and a still limited accuracy. The PWV maps inferred by the data fusion at any spatial resolution show better qualities than those inferred from single data sets. In addition, by using the fixed-rank kriging method, the computational burden is significantly lower than that for ordinary kriging.
ISSN:1027-5606
1607-7938