Quantifying the impacts of ENSO and IOD on rain gauge and remotely sensed precipitation products over Australia

Abstract

Large-scale ocean-atmospheric phenomena like the El Niño Southern Oscillation (ENSO) and Indian Ocean Dipole (IOD) have significant influence on Australia's precipitation variability. In this study, multi-linear regression (MLR) and complex empirical orthogonal function (CEOF) analyses were applied to isolate (i) the continental precipitation variations likely associated with ENSO and IOD, here referred to as 'ENSO/IOD mode', and (ii) the variability not associated with ENSO/IOD (the 'non-ENSO/IOD mode'). The first is of interest due to its dominant influence on inter-annual variability, while the second may reveal lower frequency variability or trends. Precipitation products used for this study included gridded rainfall estimates derived by interpolation of rain gauge data from the Australian Bureau of Meteorology (BoM), two satellite remote sensing products (CHIRP and TRMM TMPA version 7), and two weather forecast model re-analysis products (ERA-Interim and MERRA). The products covered the period 1981-2014 except TMPA (1998-2014). Statistical and frequency-based inter-comparisons were performed to evaluate the seasonal and long-term skills of various rainfall products against the BoM product. The results indicate that linear trends in rainfall during 1981-2014 were largely attributable to ENSO and IOD. Both intra-annual and seasonal rainfall changes associated with ENSO and IOD increased from 1991 to 2014. Among the continent's 13 major river basins, the greatest precipitation variations associated with ENSO/IOD were found over the Northern and North East Coast, while the smallest contributions were for Tasmania and the South West Coast basins. We also found that although the assessed products show comparable spatial variability of rainfall over Australia, systematic seasonal differences exist that were more pronounced during the ENSO and IOD events.