UNFCCC Article 2, Stabilisation and Uncertainty

About This Presentation
Title:

UNFCCC Article 2, Stabilisation and Uncertainty

Description:

UNFCCC Article 2, Stabilisation and Uncertainty – PowerPoint PPT presentation

Number of Views:14
Avg rating:3.0/5.0
Slides: 32
Provided by: choosec

less

Transcript and Presenter's Notes

Title: UNFCCC Article 2, Stabilisation and Uncertainty


1
  • UNFCCC Article 2, Stabilisation and Uncertainty
  • probabalistic interactive exploration of
    scenarios using Java Climate Model
  • ASTR seminar 17th Feb 2004Ben Matthews
    matthews_at_climate.be
  • (with Jean-Pascal van Ypersele
    vanyp_at_climate.be)
  • Institut dastronomie et de géophysique G.
    Lemaître,Université catholique de Louvain,
    Louvain-la-Neuve, Belgium
  • www.climate.be (UCL-ASTR)jcm.chooseclimate.org
    (interactive model)

2
(No Transcript)
3
UN Framework Convention on Climate Change
Ultimate objective (Article 2)
  • '...stabilization of greenhouse gas
    concentrations in the atmosphere at a level that
    would prevent dangerous anthropogenic
    interference with the climate system.
  • Such a level should be achieved within a time
    frame sufficient
  • - to allow ecosystems to adapt naturally to
    climate change,
  • - to ensure that food production is not
    threatened and
  • - to enable economic development to proceed in a
    sustainable manner.'

(technologies, lifestyles, policy
instruments) Emissions pathways(biogeochemical
cycles) Critical Levels (global temperature /
radiative forcing) Critical Limits (regional
climate changes) Key Vulnerabilities
(socioeconomic factors)
inverse calculation
4
European Union 2 C limit
  • EU Council Of Ministers 1996
  • "...the Council believes that global average
    temperatures should not exceed 2 degrees Celsius
    above pre-industrial level and that therefore
    concentration levels lower than 550 ppm CO2
    should guide global limitation and reduction
    efforts."
  • "This means that the concentrations of all GHGs
    should also be stabilised. This is likely to
    require a reduction of emissions of GHGs other
    than CO2, in particular CH4 and N2O"
  • However, widely varying interpretations of
    implications for emissions!
  • Why? Java Climate Model may help to investigate...

5
  • Stabilisation scenarios in Java Climate Model
  • (Article 2 critical limits gt critical levels
    gt emissions pathways)
  • Inverse calculation to stabilise
  • CO2 concentration (as IPCC "S"/ WRE scenarios)
  • Radiative Forcing (all-gases, "CO2 equivalent")
  • Global Temperature (e.g. to stay below 2C limit)
  • (Sea-level -difficult due to inertia in ocean /
    ice)
  • JCM core science very similar to IPCC-TAR models,
    but (unlike TAR SYR) JCM stabilisation scenarios
    include mitigation of all (21) greenhouse gases
    and aerosols, scaled w.r.t. SRES baseline.

6
  • Stabilisation scenarios in Java Climate Model
  • CO2 concentration scenario is a Padé
    polynomial(similar to formula of Enting et al
    1994 for IPCC S/WRE)defined by
  • 2000 concentration c
  • 2000 gradient dc/dt,
  • 2000 second derivative d2c/dt2 (ensures smooth
    emissions trajectory),
  • stabn year concentration
  • stabn year gradient (zero if stabilising
    concentration)Also define quadratic curve to
    continue from stabn year until 2300.
  • If stabilising radiative forcing or temperature
    (or...) iterate to find best concentration and
    gradient in stabilisation year.
  • Iterates 1-10 times, depending on magnitude of
    change (reuse of correction factors so efficient
    for dragging control).
  • Explore interactively by dragging target curve
    with mouse
  • Or systematically calculate probabilistic
    analysis ...

7
  • Systematically exploring uncertainty
  • 81 Carbon cycle variants
  • 3 Land-use-change emissions (Houghton, scaled),
  • 3 CO2 fertilisation of photosynthesis ("beta"),
  • 3 Temperature-soil respiration feedback ("q10"),
  • 3 Ocean mixing rate (eddy diffusivity of
    Bern-Hilda model)
  • 6 Ratios of emissions of different gases
  • Emissions of all gases (including CH4, N2O,
    HFCs, sulphate/carbon aerosol and ozone
    precursors) reduced by same proportion as CO2
    with respect to one of six SRES baselines
  • note atmospheric chemistry feedbacks included,
    but not varied
  • 84 Forcing/Climate Model variants
  • 3 Solar variability radiative forcing
  • 4 Sulphate aerosol radiative forcing
  • 7 GCM parameterisations climate sensitivity,
    ocean mixing/upwelling, surface fluxes (W-R UDEB
    model tuned as IPCC TAR appx 9.1)
  • note for sea-level rise, should add more
    uncertainty in ice-melt

8
Demonstration of JCM
9
Carbon Cycle
Climate Model
Other gases/Aerosols
10
Probability from fit to historical data
  • Relative probability of each set of parameters
    derived from inverse of "error" (model - data)
  • Measured global temperatures (CRU proxies)
  • Measured CO2 concentration (Mauna Loa others)
  • Reject low-probability variants (kept 468 / 6804)
  • Ensures coherent combinations of parameters, e.g.
  • More sensitive climate models with higher
    sulphate forcing
  • High historical landuse emissions with higher
    fertilisation factor
  • Still 2808 curves per plot (including 6 SRES per
    set)So show 10 cumulative frequency bands
    (using probabilities)

11
Carbon Cycle
Climate Model
Other gases/Aerosols
12
Shifting the Burden of Uncertainty
  • On average, all sets of scenarios stabilise at
    the same temperature level of 2C above
    preindustrial level. But their uncertainty
    ranges are very different!
  • A Temperature limit rather than a Concentration
    limit reduces the uncertainty for Impacts/
    Adaptation...
  • (assuming we commit to adjust emissions to stay
    below the limit, as the science evolves)
  • ...however this increases the uncertainty
    regarding emissions Mitigation pathways.
  • Which is better?

13
What CO2 level stabilises Tlt 2C ?
  • note 90 of cum freq means that 90 of variants
    weighted by probability fell below this level
  • note concentrations derived from IPCC-TAR
    science are lower than those from SAR,
    principally due to less sulphate cooling, and
    slightly higher sensitivity
  • note 550ppm "CO2 equivalent" (all gases) would
    bring us close to 2C. However, to keep the
    temperature level, total radiative forcing (and
    hence CO2 equivalent) must decline gradually.
    This is possible while CO2 remains level, due to
    declining CH4 and O3 (which have short
    lifetimes).

14
Is it 'realistic'? check trends change in CO2
emissions per capita per year x-axis from 1950 to
2050, y-axis from 10 to -10 Left
Stabilisation at 450ppm Right SRES A1B
15
Why should CO2 concentrations/temperature
constant?What about forcing (all gases), or
sea-level, or ...?Inertia in the climate
systemStabilising CO2 alone doesn't stabilise
temperature (as below from TARSYR Q6) However
stable CO2 may correspond to stable Temperature
if other gases with shorter lifetimes are also
mitigated to a similar extent.
16
As natural scientist, am not advocating 2C level,
only that derivations from it should be
consistent with latest science...
  • Interpretation of Article 2 needs a global
    dialogue (Article 6)
  • Risk/Value Judgements (including equity
    implications)
  • Impacts Key Vulnerabilities? Acceptable level
    of Change?
  • Risk Target Indicator? Acceptable Level of
    Certainty?
  • (choice of target indicator shifts the burden of
    uncertainty)
  • Such risk/value decisions cannot be made by
    scientific experts alone.
  • There is not yet any global consensus about the
    safe levelhow can modellers help
    citizens/policymakers to explore this?

17
Stabilisation is consistent with sustainable
development, but this is not the only paradigm
for future climate policy...Economists prefer
optimisation (maximising welfare)Suggest
that the optimum level is much higher than EU
policy -e.g. Climneg papers 600-1000ppm (and
continuing to rise).But hidden risk/value
judgements very controversial.And not
considering uncertainty...Others consider
pessimisation (avoiding critical thresholds of
change) -e.g. Tolerable Windows guardrail
approach (ICLIPS)WGBU concludes that we need to
aim well below 450ppm.Yet guardrails too rigid -
also not considering uncertainty!Can these all
be reconciled?gtRisk analysis framework?
18
Integrated assessment (for Climneg
project) Assess balance of mitigation and
impacts/adaptation, etc. Apply to Game
Theory (coalition formation etc.) Economic costs
module added to JCM, applying Climneg formulae.
(abatement costs, damage costs, time-integral
with discount rate, etc.) Abatement costs (RICE /
MACGEM) based on comparison to SRES baselines.
Need to incorporate MACs for each gas.
Combine with probabalistic approach to convert
optimisation to Risk Analysis. But I do not
believe simple climate-impact cost functions from
economists - much more work needed (need your
help!) Also should also consider a set of
world-views for impact valuation / equitable
aggregation / discounting / risk-aversion
etc. JCM could help to make the assumptions
transparent.
19
(No Transcript)
20
Regional Climate Impacts
  • GCM climate patterns
  • scaled instantly to JCM average
  • Latest datasets from IPCC-DDC
  • JCM module to be used in DDC website
  • Climate is multi-dimensional
  • Temperature average, min, max, dtr
  • Precipitation, humidity, clouds, sunlight,
    pressure, wind
  • Seasonal cycle, view monthly animation
  • Compare GCMs to see uncertainty
  • Combine change with baseline climatology
  • Calculate averages for any country / region
  • (now for various region-sets SRES, JCM, RICE,
    CWS15, EDGAR, 50, all nations...)
  • Next challenge is to derive socioeconomic /
    ecological impacts from such data,
  • using regional socioeconomic models to assess
    vulnerability.

21
Flexible region sets at both ends of the chain,
to connect a variety of data-sources and
applications. JCM12, JCM50, RICE, CWS15,
SRES4, TGCIA, IMAGE, EDGAR, CDIAC/Houghton,
All Nations, subdivisions... Idea analyse
sub-regions of large diverse countries (eg US
states, Russian oblasts, Chinese provinces), to
consider potential new coalitions if central
govts can't agree policy?
22
  • Regional Distributions
  • JCM can also be used to explore...
  • Attribution of responsibility
  • Regional climate change patterns
  • Abatement and Impact Costs In combination with
    stabilisation scenarios,and scientific
    uncertainties

23
ClimNeg game-theory of coalition-building /
transfersis based essentially on the victim
pays principle.(the more a region anticipate to
suffer from climate change impacts, the more it
should pay others not to pollute)So what about
the polluter pays principle?
24
Attribution of responsibility for climate change
(Brazilian Proposal)
  • Many potential applications, comparing impacts
    due to emissions from
  • countries (pay for adaptation?)
  • projects(CDM)
  • timeslices (inter-generational equity)
  • gases (replace GWP ?)
  • Calculations considered several
  • gases CO2 fossil, CO2 landuse, CH4, N2O
  • regions (4 SRES, 12 JCM, 15CWS, EDGAR...)
  • indicators concn, forcing, temp, sealevel
  • time-slices and future scenarios(now including
    stabilisation scenarios)
  • methods for attributing non-linear processes and
    Feedbacks

Sensitivity to uncertainties is much less for
relative attribution, compared to absolute(maybe
similar effect applies to other problems in
Climneg? - should test) JCM contributed to
UNFCCC intercomparison ( workshops Hadley centre
2002, Berlin 2003). Next stage of intercomparison
during 2004, report to SBSTA 2005. Process helps
to engage developing countries, who often mention
historical responsibility.
25
  • Interpretation of Article 2 needs a global
    dialogue (Article 6)
  • Risk/Value Judgements (including equity
    implications)
  • Impacts Key Vulnerabilities? Acceptable level
    of Change?
  • Risk Target Indicator? Acceptable Level of
    Certainty?
  • (choice of target indicator shifts the burden of
    uncertainty)
  • Such risk/value decisions cannot be made by
    scientific experts alone.

26
  • The ultimate integrated assessment model
    remains the global network of human heads.
  • To reach effective global agreements, we need an
    iterative global dialogue including citizens /
    stakeholders.
  • The corrective feedback process is more important
    than the initial guess. So let's start this
    global debate!
  • But we still need models to provide a
    quantitative framework for the discussion. JCM
    was developed to make models more accessible and
    transparent.

27
(game -theory, game practice...)Role-play on
Article 2 with students Louvain la Neuve,
Belgium, Dec 2002, as if COP11, 2005,Presented
at COP9 Milano, Dec 200360 university students
grouped in 17 delegations (Belgium, Denmark,
Russia, USA, Australia, Saudi-Arabia, Venezuela,
Brazil, Burkina-Faso, Marroco, Tuvalu, India,
Greenpeace, GCC, FAO, WB/IMF, Empêcheurs)had the
task to agree by consensus in a UNFCCC-style
process a quantitative interpretation of
Article 2, an equitable formula for funding
adaptation.Delegates used Java Climate Model to
explore options / uncertainties. Can "justify"
diverse positions by selecting parameters /
indicators !
28
Conclusions of role-play
  • Equity implications were key aspect of discussion
  • Final compromise between Russia and Tuvalu (after
    US quit)
  • Quantitative interpretation of Article 2
    Temperature rise (lt1.9C 2100-1990)
    Sea-level rise (46cm 2100-1990) Principles for
    Adaptation funds Tax on emissions trading
    Percapita emissions GDP formula Principles
    sufficiency/capacity
  • Such "games" also help us to identify scientific
    issues, e.g. Reconciling multi-criteria climate
    targets (inconsistency maybe realistic in policy
    compromises), Meaning of CO2 "equivalents" in
    stabilisation context

29
  • Future development for global dialogue
  • Could we combine such tools and experiences to
    link groups from all corners of the world?
  • JCM also used for teaching in several
    countriesUniv Cath de Louvain (BE) Open
    University (UK), Univ Bern (CH), Univ Waterloo
    (CA),...
  • Such web models might provide a quantative
    framework for a global dialogue. Model can be
    shared by saving snapshots of model parameters to
    pass to others in asynchronous discussion forum.

30
Experiment with Java Climate Model Try JCM at
jcm.chooseclimate.org Trying to combine research
and outreach Works in web browser, very
efficient/compact Instantly responding graphics
show cause-effect from emissions to
impacts, Based on IPCC-TAR methods / data, New
flexible stabilisation scenarios, Regional
emissions, abatement, costs, responsibility
Regional patterns of climate change Transparent,
open-source code, modular, scriptable, Interface
in 10 languages, 50,000 words documentation JCM
also developed with DEA-CCAT Copenhagen,
UNEP-GRID Arendal, KUP Bern
31
Demonstration of JCM
Write a Comment
User Comments (0)