Independent patterns of water mass anomalies over Australia from satellite data and models

The Gravity Recovery and Climate Experiment (GRACE) products allow the quantification of total water storage (TWS) changes at global to regional scales. However, the quantity measured by GRACE represents mass signals integrated over vertical columns, requiring their separation into their original sources. Such a separation is vital for Australia, for which GRACE estimates are affected by leakage from the surrounding oceans. The independent component analysis (ICA) method that uses higher-order statistics, is implemented here to separate GRACE-derived water storage signals over the Australian continent from its surrounding oceans, covering from October 2002 to May 2011. The performance of ICA applied to GRACE is then compared to the ICA of WaterGAP Global Hydrology Model (WGHM) and the ICA of the Australian Water Resources Assessment (AWRA) system. To study the influence of rainfall variability on the derived independent patterns. use is made of Tropical Rainfall Measuring Mission (TRMM) data set, from January 2000 to May 2011. Implementing ICA on GRACE-INNS showed a remarkable improvement in separating the continental hydrological signals from the surrounding oceanic anomalies, which was not achievable using a conventional principle component analysis. Reconstructing the continental TWS changes using only those independent components of GRACE that were located over the continent showed a high correlation with WGHM-TWS and AWRA-TWS. Mass concentrations over the oceans and particularly 52 semi-diurnal aliased pattern were separated as independent modes. Correlation analysis between the independent components of GRACE and climate teleconnections showed that the mass anomalies over the northern ocean, Gulf of Carpentaria and north-eastern parts of Australia were significantly correlated with the El Nino-Southern Oscillation, while those over south and south-eastern parts of Australia were mainly linked to the Indian Ocean Dipole. (C) 2012 Elsevier Inc. All rights reserved.