The prediction of high-impact climate
phenomena can be substantially improved
by a new mathematical approach that analyses
the connectivity and patterns between
geographical locations, scientists say
in a new publication.
This can potentially save thousands of lives
and avoid billions in economic losses.
Prediction times for events like El Niño,
monsoons, droughts or extreme rainfall could
be increased substantially, to a month or in
some cases even a year in advance, depending
on the type of the event.
The new framework can thus become key for
improving adaptation to the global warming crisis.
"The new forecasting approach has, in several
instances over the past years, proven to be
highly efficient in predicting different climate
phenomena much earlier than before," said Josef
Ludescher from the Potsdam Institute for Climate
Impact Research (PIK), lead author of the
perspective article published in the Proceedings
of the US National Academy of Sciences.
"El Niño for instance could be predicted up to
one full year early, compared to about six
months with the standard prediction methods.
"The onset of the Indian Summer Monsoon in central
India, vital for the economy in this region, was
predicted more than a month in advance, much
earlier than the forecast currently used, thanks
to the new approach."
Dr Niklas Boers, of the University of Exeter,
PIK and TU Munich, said: "The power of network-based
foreasting approaches also lies in their generality,
making them applicable for different kinds of
climate phenomena, from El Niño events to the
prediction of both extreme rainfall or droughts.
"For example, six out of the seven most severe
Amazon droughts can be predicted using a network
encoding the dependencies between the tropical
Atlantic ocean and the Amazon."
Extreme events like floods, heatwaves or droughts
often arrive with little or no warning time at all,
making effective short-term adaption challenging
if not impossible.
The new prediction framework fundamentally improves
this, as Jürgen Kurths from PIK, a pioneer of
network application to climate-phenomena forecasting
and co-author of the paper, underlines: "Currently,
for instance, there is no reliable prediction of
heavy rainfall in the Easter Central Andes leading
to floods and landslides with devastating impacts
for the inhabitants in that part of South America.
"Our network-based approach can predict those
events up to two days in advance – that is crucial
time for the people to prepare, save lives and
Traditional weather and climate forecasting rely
predominantly on numerical models imitating
atmospheric and oceanic processes.
These models, while generally very useful, can’t
perfectly simulate all underlying processes – and
phenomena like monsoon onsets, floods or droughts
might be predicted too late.
This is where network-based forecasting comes into play.
Ludescher explains: "As opposed to looking at a
huge number of local interactions, which represent
physical processes like heat or humidity exchange,
we look directly at the connectivity between
different geographical locations, which can span
continents or oceans.
"This connectivity is detected by measuring the
similarity in the evolution of physical quantities
like air temperatures at these locations.
"For instance, in the case of El Niño, a strong
connectivity in the tropical Pacific tends to
build up in the calendar year before the onset
of the event."
Kurths adds: "That’s a fundamentally different
approach from traditional numerical modelling
used in weather and climate forecasts.
"It does not simulate the entire Earth system,
but analyses large-scale connectivity patterns
in observational data."
Co-author Maria Martin, also at PIK, added:
"These patterns, that is the connectivity between
the locations and their evolution in time, can
provide critical new information for forecasting
– and, so we hope, make the respective regions safer.
Hans-Joachim Schellnhuber, former Director of the
institute, concludes: “With this perspective,
we have brought together several success stories
that demonstrate the scientific power of the
network approach for forecasting – and, in
consequence, for potentially saving thousands
of lives and avoiding billions of economic costs."
The article is entitled: "Network-based
forecasting of climate phenomena."
UNIVERSITY OF EXETER
Provided by the IKCEST Disaster Risk Reduction Knowledge Service System