S2 Time-filtered heat transport

S2 Time-filtered heat transport
The present paper analyses heat transport by transient motions primarily by using
unfiltered data. Applying a high-pass filter to the data, with an upper cut-off around 68 days, will remove the large-scale component. Since this component is found to be
an important driver of zonally integrated heat transport, it is expected that the
distributions discussed in the paper will also be affected by the filtering procedure,
both in terms of magnitude of the extremes and shape of the distributions. A test was
performed by using a finite-impulse recursive high-pass filter which retains periods
below 8 days, as discussed in section 2.2 in the paper.
Figure 8b in the paper displays the zonally integrated heat transport distribution of the
filtered data for NH DJFs. This can be compared to Figure 7a, which displays the
same distribution for the unfiltered data, plotted using the same bins. As hypothesised,
the magnitude of the extremes is greatly reduced, and the shape of the distribution is
altered. However, the filtered transport data still displays a positive skewness (1.34)
and a pronounced, near-zero most likely value (0 W/1000hPa). The high-frequency
end of the heat transport spectrum therefore retains the salient features of the full
transport distribution.
Because of the significant differences in the shape and magnitude of the distributions,
we do not necessarily expect the zonal events exceeding the 95th percentile of the
unfiltered distribution to coincide with the events exceeding the 95th percentile of the
filtered distribution. Indeed, the results in the paper suggest that the extremes we
select are due to both a local and a zonal contribution, while the filtered data will only
retain the former. One would therefore expect the largest extremes, where both the
baroclinic and longer timescales must provide a large contribution, to match; the
extremes which are closer to the 95th percentile threshold will display a looser
Figure S1: Composite pressure versus longitude v’ and H’ sign combination
colourmap for extreme events. The colour bar shows the sign combinations
corresponding to the different colours. The data cover NH DJFs from December 1989
to February 2011. All latitude circles between 30° and 89°N are taken into account.
The continuous black contours mark regions where 30%, 40%, 50% and 60% of
extreme events share the same sign combination. The diagonal striping marks regions
of equatorward transport.
Figure S2: Composite pressure versus longitude composite covariance map for
meridional heat transport extreme events. The data cover the same range as in Figure
S1. Darker (redder) colours correspond to larger (more positive) covariance values.
The covariance is normalised relative to the maximum. The diagonal striping marks
regions of equatorward transport. Note that the colour bar is not symmetric about