| Annual series of three
stochastic rainfall model parameters — the seasonal wet day amount (or
intensity), the conditional dry–day probability (or dry–spell
persistence), and the conditional wet-day probability (or wet-spell persistence) — were
examined using daily rainfall records for ten UK
stations for the period 1901–1995. The purpose was first, to determine the
extent to which these indices of summer (June–August) rainfall
were correlated with empirical orthogonal functions (EOFs) of summer North
Atlantic sea surface temperature (SST) anomalies: second,
to evaluate the skill of EOFs of preceding winter (December–February) SSTs for
summer rainfall forecasting and downscaling.Correlation analyses suggest that
observed increases in summer dry-spell persistence since the 1970s coincided
with positive SST anomalies in the North Atlantic. In
contrast, wet-spell persistence and intensities were relatively weakly
correlated with the same patterns, implying that the use
of SSTs is justifiable for conditioning occurrence but not intensity parameters.
Furthermore, the correlation strengths were greater for EOFs
of SSTs than those reported for area-average SST anomalies, indicating that the pattern
of SST anomalies conveys important information about
seasonal rainfall anomalies across the UK. When EOFs of winter SSTs were used to
forecast summer rainfall in Cambridge, the skill was once
again greater for dry-spells than either wet-spells or intensities. However,
even for dry–spells, the correlation with observations — whilst
statistically significant — was still rather modest (r<0.4).
Nonetheless, the results are comparable to previous investigations of summer
rainfall across Europe, and suggest that forecasting skill
(across the UK) originates from the predictability of the rainfall occurrence
process.
Keywords: North Atlantic, ocean temperatures, downscaling, rainfall, forecasting, UK |