Climate Sense

[

] 147

bserving

, P

redicting

and

rOjecting

limate

OnditiOns

In hindcasts, the method predicts well the interan-

nual variability of storm numbers and ACE index.

Real-time forecasts since 2007 have also been

successful: 12 storms occurred in 2007 when the best

estimate prediction was ten, and 15 storms in 2008

when the best estimate prediction was 15.

Currently applications in the health sector are prob-

ably best developed for malaria prediction in Africa. In

malaria-prone regions incidence of the disease is posi-

tively correlated with seasonal rainfall totals. Research has

demonstrated the potential of using dynamical seasonal

predictions to predict malaria incidence in southern Africa.

Ensemble-mean November-February rainfall anomalies

from dynamical models of three centres participating

in the EU project DEMETER (Met Office, ECMWF and

Météo-France) show spatial distributions that are similar

to observed distributions when composited for years with

atures in the North Atlantic and northern European winter

climate, and later for the summers of 2007 and 2008, when tele-

connections between Europe and the tropical Pacific (in La Niña

phase) were invoked. In all these cases the ‘consolidated’ forecast

provided generally accurate guidance.

The cost of tropical storm damage represents one of the

largest risks covered by the insurance sector, and consequently

there is considerable interest in its long-lead prediction. Recent

research has shown that the skill of dynamical systems, such as

GloSea, for predicting the numbers of North Atlantic tropical

storms is comparable to that of statistical prediction methods,

which have traditionally formed the basis of most published

predictions. Based on research initiated at ECMWF and using

GloSea, the Met Office issues predictions of tropical storm

numbers and the accumulated cyclone energy (ACE) index

(a measure of the collective intensity and duration of storms)

during July-November.

Predictions and observations for the 2008 wet season in West Africa

a) GloSea-predicted probability for rainfall above the 1987-2001 average, for JAS08 b) Prediction skill based on hindcasts for 1987-2001, scores greater

than 0.5 indicate forecasts are better than guesswork and perfect forecasts would attain a score of 1.0 c) PRESAO 11 consensus update forecast showing

predicted probabilities of (top to bottom) above-average, average, below-average rainfall for each geographical zone indicated by the colour shading. d)

observed JAS08 rainfall anomalies relative to 1979-2000

Source: a) and b) Met Office Hadley Centre, c) African Centre for Meteorological Applications for Development, d) International Research Institute for Climate and Society