Climate Sense

[

] 146

Seasonal to decadal prediction: improving

accessibility and widening applications

Richard Graham, Adam Scaife, Anca Brookshaw and Mike Davey, Met Office, UK;

Rowan Sutton, National Centre for Atmospheric Science, UK

he potential benefits of reliable seasonal to decadal

climate prediction are enormous, and evident in both

the impacts that could be avoided as a result, and the

forecast applications developed by various centres. Users of

these predictions are often most interested in extremes such

as drought, flooding or heat waves. Studies show it is when

natural and anthropogenic climate change combine that

extremes become more frequent and more intense, so predic-

tion systems need to include both sources of variability.

This understanding has led to a growing need for initial-

ized predictions to provide advance warning of impending

extremes, and to respond and adapt to climate change. State-

of-the-art seasonal to decadal predictions are best placed to

predict these events because they are initialized with the

current state of the climate system while also including

expected changes in anthropogenic forcing. This has accel-

erated the requirement for seasonal to decadal predictions

to be incorporated into the operational output of National

Meteorological and Hydrological Services (NMHSs).

The Met Office makes routine predictions across seasonal to multi-

decadal timescales. Relatively detailed predictions are made in

some regions, mainly in the tropics, where skill levels are highest.

Decadal predictions offer the possibility of additional extratropical

predictability, due in part to the growing anthropogenic climate

change signal and in part to low-frequency Atlantic variability. The

main forecasting tools are dynamical prediction systems which use

coupled ocean-atmosphere general circulation models to develop

ensemble predictions, allowing assessment of uncertainties in the

forecast. Below are some examples of long-range forecast applica-

tions, covering a range of (tropical and extratropical) locations,

timescales, customers and impacts, and illustrating international

collaborations.

Regional Climate Outlook Forums (RCOFs) are currently

convened for regions with strong rainfall seasonality (such as

the West African monsoon, and the East African wet seasons).

Their main objective is to develop a consensus broad-scale

outlook, which may be elaborated to national scales by the

respective NMHSs, and disseminated to sectors including agri-

culture, water resources, energy, health and media. The forums

also provide opportunities to form alliances between forecast

users and producers, for users to feed back requirements to

producers and for institutional capacity building. The consen-

sus outlooks are developed using statistical and dynamical

forecasts, blended where discrepancies occur using interpreta-

tion from regional experts. An example of dynamical

model input to RCOFs is the rainfall prediction over

West Africa for July-September 2008 from the Met

Office coupled ocean-atmosphere seasonal predic-

tion system (GloSea). The prediction formed one

input to an update forecast for the region, devel-

oped following the West Africa RCOF (PRESAO 11)

and issued by the African Centre of Meteorological

Applications for Development on 27 June 2008.

Forecast statements highlighted enhanced risk

of wet conditions and advised a strengthening of

local weather watches. In response to the forecast a

number of flood mitigation procedures were imple-

mented by the International Federation of Red Cross

and Red Crescent Societies, including pre-position-

ing of emergency stocks. In the event, heavy rain

and flooding affected many West African countries

in July 2008.

The relatively good seasonal prediction skill for

West Africa has also led to the development of a

number of applications for river basin and reservoir

management in the region. The Met Office has devel-

oped a system to predict water volume inflow into

Lake Volta, to assist management of the 1,000-mega-

watt hydro-electric power plant near Akosombo, at

the lake’s southern end. Inflow peaks in September,

with around 95 per cent occurring between June

and November. Predictions for total inflow over this

period are made from May, with updates issued each

month. The predictions are derived using observed

preceding rainfall and river flow in the catchment,

global sea surface temperature patterns and GloSea-

predicted rainfall spanning the target period. On

average, over a 20-year period, these predictions are

more accurate than predictions using only catchment

observations.

Hindcast skill of dynamical models over much of

Europe is currently marginal, especially for precipi-

tation. However, there is the potential for predictions

in individual years to be informative. The Met Office

has adopted a process for generating seasonal predic-

tions for Europe by combining dynamical-model

output with data from several other sources. This

approach was first used for the winter 2005-2006,

exploiting connections between sea-surface temper-

bserving

, P

redicting

and

rojecting

limate

onditions