Ocean/Atmosphere Memory Makes Extreme Events More Likely
By Dr. Gordon J. MacDonald, Director,
International Institute for Applied Systems Analysis,
Laxenburg, Austria
Discussions of global change often emphasize changes in the mean
global temperature, where the average is taken over the globe
and over the year. Depending on the particular climate model used,
the expected changes are on the order of 1 to 4 degrees Celsius
for a doubling of the concentrations of carbon dioxide. As has
often been pointed out, changes of this magnitude can be achieved
by moving a relatively small distance north or south, or even
moving 1000 m down a mountainside.
Statistics such as mean temperature, therefore, do not capture
the true implications of climate change. Instead, economic costs
and social disruption arise from extreme atmospheric events, such
as hurricanes, typhoons, floods, droughts, windstorms, or prolonged
periods of cold. The most significant issue in future climate
change is whether these extreme events will become more frequent
and/or more intense. In statistical terms, the question is how
the probability of extreme events that are characterized by the
thickness of the tails of the distribution will change with alteration
of climate. Most current climate models do not capture the climatic
extremes for a variety of reasons, including that the spatial
resolution of these models is inadequate to represent many extreme
events.
Spatial Correlation
Means and standards deviations can be exceedingly useful statistics
for describing independent events, such as coin-tosses. But weather
events are not independent, either in time or in space. For example,
a typical high-pressure system has a spatial scale on the order
of 1000 km, and, in the Northern Hemisphere summer, an air mass
circulating clockwise about a high-pressure center generally remains
over a given geographical area for a few days. During that interval,
weather at various points in the region is correlated both spatially
and temporally. Thus, a prediction that the next day's weather
at a particular site will be similar to today's stands a good
chance of being correct. Stated another way, the spatial correlation
implies that the atmosphere has a memory. What happens today depends
on what happened in previous days.
Role of Memory
Independent events commonly exhibit the normal or bell-shaped
statistical distribution. In such cases, the tails of the distribution
are very thin and the probability of extreme events is low. By
contrast, when there is memory in the system, as is the case in
the atmosphere, the tails of the distribution become thicker and
the probability of extreme events rises as correlated events combine
to yield larger deviation from the mean.
In considering climate, the ocean interactions with the atmosphere
are critical. The large thermal capacity of the ocean makes it
able to influence the thermal character of the overlying atmosphere.
However, unlike the atmosphere, with correlation times on the
order of days, motions within the ocean have time scales of years
and decades, perhaps even longer. Thus, atmospheric events can
also show correlations over periods of years and decades. The
ENSO (El Nino, Southern Oscillation) phenomenon clearly illustrates
the role of the oceans, as every few years a warm pool of water
builds up off the west coast of South America, bringing torrential
rains to desert regions and influencing climate over much of North
America. A North Atlantic oscillation demonstrates a somewhat
analogous phenomenon in the Atlantic circulation. Satellite observations
show that anomalies in the Atlantic sea surface temperature migrate
northward with a time scale of a decade. The northward travelling
sea surface temperature anomalies give the ocean a memory of past
atmospheric conditions.
More Extremes Likely
These correlated events lead to thicker tail distribution and
a greater frequency of extreme events than would be expected if
weather were made up of a sequence of independent events. For
example, the pool of warm water found off the coast of Florida
&emdash; probably associated with 1992's Hurricane Andrew or possibly
with Hurricane Hugo in 1989 &emdash; may in part be responsible
for the flooding of Central Europe during the summer of 1997.
In fact, flooding provides a good example of extremes, as illustrated
by El Nino events. Because evaporation depends exponentially on
temperature, warm surface waters pump into the atmosphere at a
greater rate than cooler waters. As long as the pool of warm water
is exposed to the atmosphere, the atmosphere will carry the water
over continental areas, leading to increased rainfall and increased
probability of flooding.
The long-term memory of oceans implies that heavy rainfalls may
persist over a given region for periods of years. Therefore, accumulated
precipitation in river basins can give rise to disastrous floods
that carry heavy costs. Flooding in Central Europe in the summer
of 1997 was greater than any recorded in the last 200 years. To
illustrate the impact of this catastrophe, the cost to Poland
of the Oder floods of 1997 exceeded US$4 billion, or 4.4 percent
of Poland's GDP.
Methane Hydrates
The vast quantity of methane contained in hydrates of methane
stored in continental shelf sediments poses another potential
long-term threat to climate stability. Exposing the sediments
to warmer ocean currents, and the consequent diffusion downwards
of the thermal load, could release this methane, which would then
add significantly to the greenhouse gas burden of the atmosphere.
Such releases cannot be confidently predicted because of the continuing
uncertainty as to the long-term nature of ocean circulation.
The climate of the future is certain to bring surprises. The
recently discovered rapid oscillations (100 years) in climate,
shown in the ice core from Greenland, may have been associated
with sharp shifts in ocean currents or with totally new processes
not yet discovered.
Article List | back:
Climate Surprises | proceed:
Heavy Rainfall Threat
