Abstract: Using the monthly mean Met Office Hadley Center sea surface temperature (SST) analyses data, Global Ocean Data Assimilation System ocean temperature data, and NCEP circulation reanalysis datasets for the period 1980-2010, the main ocean?atmosphere coupled processes over the tropical Pacific Ocean were investigated through the multivariate empirical orthogonal functions (MV-EOFs) of two thermal ocean parameters —SST and upper-ocean heat content (HC) — and five meteorological elements — sea level pressure (SLP), 850-hPa u wind, 850-hPa v wind, 200-hPa velocity potential, and outgoing longwave radiation (OLR). Our results show that the temporal and spatial distribution of the first and second MV-EOF modes of the ocean-atmosphere elements capture the ocean?atmosphere couplings of the canonical El Niño and El Niño Modoki events, respectively. Associated with the warming center in the central and eastern tropical Pacific, the canonical El Niño phenomenon is also characterized by the “seesaw” pattern of east?west anti-phase changes in ocean?atmosphere elements such as HC, SLP, and 200-hPa velocity potential. Meanwhile, there are strong westerly anomalies in the central equatorial Pacific and anticyclonic circulation anomalies over the western North Pacific at low atmospheric levels. On the other hand, during the El Niño Modoki events, the centers of the positive anomalies of SST and HC moved west to the central equatorial Pacific, and the anomalies of SLP and 200-hPa velocity potential were characterized by a three-polar pattern. Moreover, the low-level strong westerly anomalies moved west to the east of the warm pool, and cyclonic circulation anomalies occurred over the western North Pacific. The ocean?atmosphere couplings of the two types of El Niño episodes were significantly different, which suggests that more sophisticated indices for El Niño are required to represent and separate the two events more effectively and to better identify their different atmospheric response. Here, in contrast to previously reported El Niño indices based on SST or SLP, the El Niño signal is represented by HC, and new indices based on HC are defined for the two types of events, which not only capture and characterize the two types more distinctly, for example, the 1993 El Niño and 2006 El Niño Modoki event, but better identify their ocean?atmosphere coupled process, providing a new approach to monitoring and short-term climate prediction of El Niño episodes.