Transcript Predicting “near-term” climate change

The Decade After Tomorrow:
“Near-term” climate change
Arthur M. Greene, Lisa Goddard
International Research Institute
for Climate and Society
Palisades, NY USA
The International
Research Institute
for Climate and Society
Outline
“Near-term” climate change: What is it?
Large-scale modes of variability: What are they?
Near- and far-field effects: Maybe in my backyard
Value in paleorecords: Characterize, if not predict…
Prediction: The wave of the future
The International
Research Institute
for Climate and Society
Why “Near-term?”
Interest in regional climate change is increasing, but
the 100-year time scales considered by IPCC are
often deemed too long to be “actionable.” Thus is
born the concept of “near-term climate change” – out
to about three decades.
But there’s a catch: Decadal variability is much less
well-understood than, say, ENSO. It may even
originate from completely random interactions
between atmosphere and ocean.
The International
Research Institute
for Climate and Society
ENSO: El Niño-Southern Oscillation
“Near-term” prediction in context
SI forecast / verification
“Near-term CC”
IPCC “climate change” time scale
The International
Research Institute
for Climate and Society
SI: Seasonal-to-Interannual (ENSO-based)
IPCC: Intergovernmental Panel on Climate Change
Aside: Observational datasets are not all identical
Atmosphere-ocean randomness: Brownian motion
Animation: http://www.phy.ntnu.edu.tw/ntnujava/index.php?topic=24
The International
Research Institute
for Climate and Society
Path of the (massive) red object is smoother, and evolves more slowly,
than those of the rapidly fluctuating grey “molecules.”
Decadal complexity
Unlike in the case of ENSO, where a single, fairly wellunderstood mechanism controls the evolution of events, the
physical processes that produce decadal climate fluctuations
differ from ocean basin to basin; there is no single dominant
process. The processes themselves are generally not wellunderstood, nor is the degree to which they may interact.
The International
Research Institute
for Climate and Society
Principal mode in the Atlantic: AMV (Atlantic Multidecadal Variability)
Gray et al, 2004
“Oscillation” vs “Variability”: The
reconstruction suggests less
regularity than might be inferred
from the instrumental record alone.
Colored line (middle plot) can be
thought of as the slowly varying path
of the red “Brownian” object.
Density-driven “Overturning” circulation
T
D
-- Circulation slows, temperature falls.
T
D
-- Circulation increases, temperature rises.
Result: A natural oscillation
Complication: Salinity
AMV far-field effects on rainfall, in observations and models
Decadal correlation maps for AMV and mean annual precipitation for (a) observations and
picntrl runs for (b) GFDL-CM2.0, (c) GISS-ER and (d) NCAR-CCSM3.0
AMV effects on Atlantic hurricanes
“During warm phases of the AMO,
the numbers of tropical storms that
mature into severe hurricanes is much
greater than during cool phases, at least
twice as many.”
Two 24-yr periods compared
Green lines: Tropical storms
Red lines: Hurricanes
Worth keeping in mind: Other possible
influences on Atlantic hurricanes are not
taken into account here. But the research
argument is in fact supported on basic
physical grounds as well.
Goldenberg et al., 2004
Pacific decadal variability: the PDO
First identified in a research
paper about Salmon production.
Expressed as a pattern, where
SST in the tropical Pacific
varies out of phase with higher
latitudes. “Warm” phase is
shown at left.
Accompanied by changes in
winds (arrows), SLP.
Teleconnections affect
precipitation in many regions
around the globe.
Cause is presently not
understood; decadal variations
in ENSO may play a role
(pattern similarity).
(Source: http://jisao.washington.edu/pdo)
PDO far-field effects on rainfall
Scale: Amount (mm/day) by which rainfall changes
for a 1-unit increase in the PDO index.
All together, now…
AMO/PDO combined effects on drought
- PDO
The International
Research Institute
for Climate and Society
Periods when these conditions applied
+AMO
McCabe et al, 2004
 Recent states
Value of paleodata: Tree-ring reconstruction of PDSI in the Western U.S.
Stahle et al., 2007
The International
Research Institute
for Climate and Society
Red arrow: Date of the Colorado Water Compact (1922)
Blue line: Portion of record used to estimate flows.
If they had known then what we know now…what?
“In short, the river is allocated to the tune of 17.5 MAF (million acre-feet) …
whereas the observed average yield is 14.8 MAF (1896-2004), and tree-ring
reconstructions suggest a longer-term average yield of 13.5 MAF (1520-1961).”
– Hard Times on the Colorado River: Growth, Drought and the Future of the Compact, 2005. Natural
Resources Law Center, University of Colorado School of Law
Level in 2002 (Lake Mead)
The International
Research Institute
for Climate and Society
Ken Dewey, HPRCC (U. Nebraska)
Important to keep in mind: Natural variability is only part of the picture!
“If these models are correct…
levels of aridity of the recent
multiyear drought… will become the new climatology of the
American Southwest within a
time frame of years to decades.”
– Seager et al., 2007
So what about prediction?
Since the long-term “memory” of the climate system must
reside in the ocean, it is initialization of the model ocean
that holds the key to decadal prediction. However:
Subsurface ocean observations are sparse, except for the past
few years. This limits the accuracy with which models can be
initialized and “hindcasts” can be verified.
Dynamical models currently do not do a very good job of
simulating the observed space-time variability of the large-scale
decadal modes.
Statistical prediction, which depends neither on
subsurface ocean data nor dynamical model simulations,
may be possible: This is an active area of research.
The International
Research Institute
for Climate and Society
Model ensemble members limn the range of natural variability…
The International
Research Institute
for Climate and Society
Mean annual Atlantic SST, ensemble of simulations using GFDL-CM2.1,
for SRES A1B. Ensemble members differ only in the initial conditions
utilized, indicating range of internal variability. Heavy black line is
ensemble mean; values are relative to 1991-2000.
Sampling the subsurface ocean: The ARGO program
Positions of ARGO floats that have delivered
data within the past 30 days. (http://www.argo.ucsd.edu/ )
The International
Research Institute
for Climate and Society
Dynamical prediction: State of the art
Ten-year forecast beginning June 2005 (white line, red
uncertainty bounds) and two 10-year hindcasts,
beginning 1985, 1995. Blue line represents uninitialized
forecasts; black line, observations (HADCRUT2v).
(Smith et al., Science, 2007).
Improvements in hindcast RMSE (9-year
means) due to ocean initialization:
A, B, C, surface temperature anomaly;
D upper ocean heat content anomaly.
The International
Research Institute
for Climate and Society
Statistical prediction: SST in the Southeast Asia domain
Forced and natural signals are separated; each is then projected forward in time…
Forced and natural projections combined
Not shown here, but important: Uncertainty envelope
Summary
Decadal climate variations have received increasing attention of late:
Possibly more relevant for adaptation than 100-year projections.
Less is known about underlying physical processes at this time scale.
Centennial, “IPCC-type” trends are not perforce irrelevant!
At least some decadal variability is associated with large-scale intrinsic modes.
(Random atmosphere-ocean interaction may also play a role.)
Atlantic Multidecadal Variability (AMV or AMO)
Pacific Decadal Oscillation (PDO)
NAO, AO, NAM, SAM, PNA…
Teleconnections to many parts of globe
Physics relatively poorly understood; so far, not well modeled (by GCMs)
Paleorecords may aid in characterizing regional decadal variations of the past.
Tree-ring reconstructions
Corals, other paleoproxies
Potential utility for decision-making
Both dynamical and statistical forecasts may be possible:
Potential predictability, forecast skill, will vary from region to region.
Dynamical prediction depends on ocean initialization, since climate
“memory” resides in the oceans.
Statistical forecasts (of the type discussed) decompose the signal, project
components separately, recombine: Voilà, the “wave” of the future!
Thank you!
[email protected]