Monday 22 August 2005

G3
1600-1700 hours

083
Fuzzy logic based estimation of continental water storage changes from GRACE observations
Akyilmaz, Orhan¹, Kutterer, Hansjörg², Shum, C.K.¹, Han, Shin-Chan¹
1 Ohio State University, Ohio, USA
2 Geodetic Institute, University Of Hannover, Germany
Author email: kutterer@gih.uni-hannover.de
The low-low satellite-to-satellite tracking mission GRACE provides scientists an efficient and cost-effective way to map the Earth's static and monthly temporal gravity fields with unprecedented accuracy and resolution. One of its major objectives is to measure the climate-sensitive signals generated by mass redistributions on Earth at spatial scales larger than several hundred km and temporal scales longer than 30 days. Fuzzy logic has been widely used in various disciplines for control, prediction and modelling issues. The most important benefit of fuzzy logic modelling is to achieve a linguistic description of the process under investigation. Despite fuzzy logic is a quite proper tool for modelling in case of lack of data, available data helps to create more precise models by fuzzy inference systems (FIS). The system parameters can then be inferred from the data itself by using mathematical optimisation techniques. The classical way of hydrological modeling from satellite based observations is collocation with kernel functions. Here, FIS have been used to recover monthly or sub-monthly mean water storage anomalies (MWSA) in South America from the gravity variations observed by GRACE. To this end, regularly gridded MWSA were computed by averaging daily water storage anomalies derived from NCEP (National Centers for Environmental Prediction) daily mean water storage (MWS) data. We tested the fuzzy logic algorithm on the GRACE Level 2 (L2) data products and the data products generated by processing Level 1B (L1B) data based on the energy conservation methods. Results from the obtained fuzzy model were tested with independent data that were not used for estimation of the model parameters. The performance of the resulting model was compared with those of other models previously used for detection of MWSA from GRACE observations.

Return to Oral Presentations