Transcript Change - hvonstorch.de

20. September 2010 - Universitet Göteborg
The regional issue
of detection and
attribution
Hans von Storch
Institute of Coastal Research
Helmholtz-Zentrum Geesthacht
Germany
with help of Jonas Bhend,
Armineh Barkhordarian
and Michael Richter
Observed temperature anomalies
2
Change !
Change is all over the place,
Change is ubiquitous.
What does it mean?
Anxiety; things become more extreme, more dangerous; our
environment is no longer predictable, no longer reliable.
Change is bad; change is a response to evil doings by egoistic social
forces. In these days, in particular: climate change caused by
people and greedy companies.
Change !
Change is all over the place,
Change is ubiquitous.
What does it mean?
There are other perceptions of change: it provides opportunities; it
is natural and integral part of the environmental system we live
in.
The environmental system is a system with enormous many
degrees of freedom, many non-linearities – is short: is a
stochastic system, which exhibits variations on all time scales
without an external and identifiable “cause”. (Hasselmann’s
“Stochastic Climate Model”)
Assessing change
First task: Describing change
Second task: “Detection” - Assessing change if consistent with
natural variability (does the explanation need invoking external
causes?)
Third task: “Attribution” – If the presence of a cause is “detected”,
determining which mix of causes describes the present change best
First task: Example of inhomogeneous data
Wind speed measurements
 SYNOP Measuring net (DWD)
 Coastal stations at the German
Bight
 Observation period: 1953-2005
This and the next 3 transparencies:
Janna Lindenberg, HZG
First task: Inhomogeneity of wind data
1.25
m/s
First task: Inhomogeneity of wind data
The issue is
deconstructing a given record
with the intention to identify „predictable“ components.
„Predictable“
-- either natural processes, which are known of having
limited life times,
-- or man-made processes, which are subject to decisions
(e.g., GHG, urban effect)
„Significant“ trends
Often, an anthropogenic influence is assumed to be in operation when
trends are found to be „significant“.
• If the null-hypothesis is correctly rejected, then the conclusion to be drawn
is – if the data collection exercise would be repeated, then we may expect
to see again a similar trend.
• Example: N European warming trend “April to July” as part of the seasonal
cycle.
• It does not imply that the trend will continue into the future (beyond the
time scale of serial correlation).
• Example: Usually September is cooler than July.
„Significant“ trends
Establishing the statistical significance of a trend may be a necessary
condition for claiming that the trend would represent evidence of
anthropogenic influence.
Claims of a continuing trend require that the dynamical cause for the
present trend is identified, and that the driver causing the trend itself is
continuing to operate.
Thus, claims for extension of present trends into the future require
- empirical evidence for an ongoing trend, and
- theoretical reasoning for driver-response dynamics, and
- forecasts of future driver behavior.
Detection and attribution
of non-natural ongoing change
• Detection of the presence of non-natural signals: rejection
of null hypothesis that recent trends are drawn from the
distribution of trends given by the historical record.
Statistical proof.
• Attribution of cause(s): Non-rejection of the null hypothesis
that the observed change is made up of a sum of given
signals. Plausibility argument.
Detection and attribution
Detection
Climate system
Internal
variability
External forcings
Observations
Attribution
Anthropogenic
Natural
13
Dimension of D&A
• Purely scientific
• Stakeholder utility
• Attribution – the competitors
• Falsification
Dimension of D&A
Purely scientific
• Statistical rigor (D) and plausibility (A).
• D depends on assumptions about “internal variability”
• A depends on model-based concepts.
• Thus, remaining doubts exist beyond the specified.
Dimension of D&A
Stakeholder utility
Evidence that anthropogenic warming is related to human drivers –
serving as an arguments to implement broad global mitigation
measures.
Evidence that recent change is part of an ongoing (predictable) pattern,
or not. Serving as information to guide regional and local adaptation
measures.
This is what we
need to know
more about
„Global clients“ want to have proof that the basic concept of man-made
global climate change is real. The best answer for this client is an answer
which is very robust and not critically dependent on models. – Mostly
done.
„Regional clients“ want to have best guesses of the foreseeable future, in
order to institute adaptive measures – on the scale of medium-size
catchment basins not many clear results.
„Local clients“ want know how global and local drivers shape the future of
the local environment, and which measures for mitigation are available,
and which levels of adaptation are required. – very little done.
Dimension of D&A
Attribution – the competitors
Climate drivers – relatively easy; not too many drivers, such as
urbanization, aerosol, land-use.
Impact drivers – hardly dealt with.
Many drivers: eutrophication, pollution, overuse,
regulation, globalization, urbanization
Example: Baltic Sea ecosystems
“Mini-IPCC” assessment
on knowledge about
climate change in the
Baltic Sea Basin
Storm surges in Hamburg
Storm surges in the Elbe estuary
Difference in storm surge height
between Cuxhaven and Hamburg
Height massively increased
since 1962 – after the 1962
event, the shipping channel was
deepened and retention areas
reduced.
Urban Heat Island effect in Stockholm
Average diurnal cycle
of UHI (urban heat
island) intensity for
the whole year,
winter months (DJF),
spring months
(MAM), summer
months (JJA) and
autumn months
(SON) for 1996 to
2009
• Mean UHI intensity 1.2 °C
• Maximum measured UHI intensity 12.9 °C
•Maximum temperature differences urban-rural in warm season
Michael Richter
Urban Heat Island effect in Rostock
Average monthly UHI
intensities for 2001 to 2009,
computed from each
difference of monthly
averages between inner-city
station RostockHolbeinplatz (Ho) and
Rostock-Stuthof, RostockWarnemünde and Gülzow
stations
• Mean UHI intensity 0.3-0.6 °C for different stations
• Maximum measured UHI intensity 8.5 °C
• Maximum temperature differences urban-rural in warm season
Michael Richter
Local change – another major driver: urban warming
Gill et al.,2007
2
3
Dimension of D&A
Falsification
Which observations in the coming 5/10 (?) years would lead to reject
present attributions?
Suggestion: Formulate and freeze NOW falsifiable hypotheses, and test
in 5/10 (?) years time – using the independent data of the additional
years.
Suggestion: Have assessment done by scientists independent of those,
who formulated the hypotheses.
Cases of Global Climate Change Detection Studies
In the 1990s … weak, not well documented signals.
Example: Near-globally distributed air temperature
IDAG (2005), Hegerl et al. (1996), Zwiers (1999)
In the 2000s … strong, well documented signals
Examples: Rybski et al. (2006)
Zorita et al. (2009)
Global
detection
The Rybski et al-approach dealing with global mean temperature
Temporal development of Ti(m,L)
= Ti(m) – Ti-L(m) divided by the
standard deviation of the m-year
mean reconstructed temp record
for m=5 and L=20 (top), and
for m=30 and L=100 years.
The thresholds R = 2, 2.5 and 3σ
are given as dashed lines; they are
derived from temperature
variations are modelled as
Gaussian long-memory processes
fitted to various reconstructions of
historical temperature (Moberg,
Mann, McIntyre)
Global detection
Regional:
Intention: Preparation and design
of measures to mitigate expected
adverse effects of climate change.
Problems: high variability, little
knowledge about natural variability;
more human-related drivers (e.g.
industrial aerosols, urban effects)
Zorita, et al., 2009
Log-probability of the event E that the m largest values of 157 values
occupy the last17 places in long-term autocorrelation synthetic series
Derived from Hadley Center/CRU data for „Giorgi bins“.
Baltic Sea: Observations and simulations used
Observations
Interpolated land station data
Temperature: CRUTEM 3v
Precipitation: GPCC v4
Simulations
Global model data from CMIP3
ALL: anthropogenic and natural forcing
ANT: anthropogenic forcing only
Jonas Bhend
29
Regional JJA temperatures
Baltic Sea: Detection using optimal fingerprinting
Model response
is too weak
No detection
Model response
is consistent with
observed change
31
Detection with different models, 1943-1997
Temperature scaling
Model response
is too weak
Consistency
No detection
32
Δ=0.05%
Regional DJF precipitation
Detection with different models, 1943-1997
Precipitation scaling
Model response
is too weak
Consistency
No detection
34
Consistency of observed trend with a B2 scenario
Pattern correlation
Ratio of area mean change
precipitation
temperature
precipitation
temperature
DJF
0.84* (0.74*)
0.95* (0.73)
2.0 (1.5)
1.3 (0.5)
MAM
0.72* (0.69*)
0.83 (0.79)
2.2 (1.9)
0.9 (0.8)
JJA
-0.28
0.95*
-1.8
1.7
SON
-0.59
0.60
-1.2
0.5
>
Consistency not in all seasons
>
Ppecip change too large compared to scenario
>
NAO (*) has significant influence
35
Consistency analysis: Baltic Sea catchment
1. Consistency of the patterns of model “predictions” and recent
trends is found in most seasons.
2. A major exception is precipitation in JJA and SON.
3. The observed trends in precipitation are stronger than the
anthropogenic signal suggested by the models.
4. Possible causes:
- scenarios inappropriate (false)
- drivers other than CO2 at work (industrial aerosols?)
- natural variability much larger than signal (signal-to-noise ratio 
0.2-0.5).
BACC conclusion
• Detection and consistency within reach for Northern
Europe
• But not really for attribution, since signals for changig
aerosol emissions and land-use change are not known.
• Other signals?
• Falsification of detection and attribution an open problem
37
Med Sea region: Precipitation over land
Observed trend 1966-2005
Ensemble mean 22 models (A1B)
Ensemble mean 18 models (A2)
90% uncertainty range, 9000-year
control runs
Spread of trends of 22 GS signals
Spread of trend of 18 GS signal
Spread of trend of CRU3 and GPCC5
observed trends
There is less than 5% probability that observed trends in DJF, JFM, FMA, ASO, SON are due to
natural (internal) variability alone.
Externally forced changes are significantly detectable in winter and autumn intervals (at 5% level)
2m Temperature
Observed seasonal and annual area mean changes
of 2m temperature over the period 1980-2009 in
comparison with GS signals
Observed trends of 2m temperature (1980-2009)
Projected GS signal patterns (time slice experiment)
23 AOGCMs, A1B scenario derived from the CMIP3
90% uncertainty range of observed trends, derived
from 10,000-year control simulations
The spread of trends of 23 climate change projections
Less than 5% probability that observed warming can be attributed to
natural internal variability alone
Externally forced changes are detectable in all seasons except in winter
41
Rest - Armineh
42