Lag correlations of area-averaged Queensland rainfall (land points in 138° –154°E, 9°–29°S) and Niño 4 SSTs
reveal that while winter, spring and early summer HiGEM rainfall shows similar predictability from Niño 4 (Fig.
6b) as observations (Fig. 6a), late summer and autumn rains in HiGEM are only weakly predictable from the
ENSO. In both HiGEM and SILO/HadISST, June Niño 4 SSTs are strongly correlated with rainfall from June
through December, as demonstrated by the downward (i.e. longer lead time) slope of the correlation contours in
June–December. In HiGEM, however, January rainfall displays an erroneously strong correlation with Niño 4 at
long lead times, while the February–April lag correlations are much weaker than observed. The positive
correlations in HiGEM for positive lags (i.e. rainfall leading Niño 4) in February and March are indicative of overly
bi-annual ENSO variability; in other words, the ENSO in HiGEM too often switches sign in austral El Niño years.
Further, the correlation for La Niña years is not affected by the outlying years, as when the four strongest La Niña
events—those with May–April mean Niño 4 < -1.0°C—are removed, the correlation coefficient weakens only
slightly (the dashed blue line in Fig. 7a).