1. Page 1© Crown copyright 2004 Impact of climate-carbon cycle feedbacks on emissions scenarios to achieve stabilisation 1.Hadley Centre, Met Office, Exeter 2.Centre for Ecology and Hydrology, Dorset 3.Centre for Ecology and Hydrology, Wallingford Chris Jones (1) Peter Cox (2), Chris Huntingford (3)

2. Page 2© Crown copyright 2004 Outline  Climate-Carbon cycle feedbacks  Uncertainties/intercomparisons  Implications for stabilisation emissions  Results  GCM experiments  Simple “reduced form” model results  Discussion  Uncertainties – between and within models  Reducing uncertainty? Model validation  Defining “optimal” pathways to stabilisation?

3. Page 3© Crown copyright 2004 Climate Carbon Cycle feedbacks  Well known that climate-carbon cycle models predict a positive feedback  Climate change will reduce the carbon cycle’s ability to sequester CO2  Models have consensus on sign (+ve), but not magnitude of feedback (i.e. C4MIP)  Uncertainties in the feedback strength mean large uncertainty in:  Future CO2 levels given an emissions scenario  Permissible emissions to stabilise CO2 at a given level

4. Page 4© Crown copyright 2004 Climate Carbon Cycle feedbacks  If climate change weakens natural carbon sinks then we must reduce emissions by more than previously thought to stabilise atmospheric CO2  Passing mention in TAR but needs to be brought out more  TAR showed range of permissible emissions but didn’t stress impact of climate feedbacks in reducing these  Huge political implications  Plea to AR4 authors – Needs to be given more prominence.  Instead of “managing the carbon cycle” this comes under “being managed by the carbon cycle”

5. Page 5© Crown copyright 2004 WRE scenarios  “WRE” is a family of scenarios of CO 2 level, stabilising at 450, 550, 650, 750 and 1000ppm  Wigley, Richels and Edmonds. ‘Economic and environmental choices in the stabilisation of atmospheric CO 2 concentrations’. Nature, 1996  We run the carbon cycle GCM with the prescribed 550 CO 2 scenario and infer the emissions required to achieve it  Results shown in detail for 550ppm  Summary of results for all levels

6. Page 6© Crown copyright 2004 WRE550 CO 2 emissions  Climate feedbacks imply reduced permissible emissions  Lower peak  Earlier peak  Reduced integral

7. Page 7© Crown copyright 2004 WRE550 cumulative emissions  Similar to previous experiments  Ocean continues to uptake carbon, but at reduced rate  Terrestrial sink saturates and reverses

8. Page 8© Crown copyright 2004 Reduced Form “simple” model  GCM prohibitively expensive!  Simple model has:  Global means  climate in terms of T  Responds instantly to CO2  Carbon cycle calibrated to follow GCM from transient run of Cox et al 2000.  Does good job at matching WRE550 GCM run  Aim is to give broad idea of response – don’t trust exact details…

9. Page 9© Crown copyright 2004 WRE550 CO 2 emissions – simple model No feedbacks With feedbacks ( WRE550 )

10. Page 10© Crown copyright 2004 Permissible Emissions  Without feedbacks, we get close to the WRE result  Climate-Carbon cycle feedbacks significantly reduce the permissible emissions for stabilisation  This is true for stabilisation at any level Total emissions, 2000-2300 WREwithout feedbacks with feedbacks Stabilisation at 550 ppm 1393 GtC1355 GtC1010 GtC

11. Page 11© Crown copyright 2004 Other stabilisation levels Greater reductions at higher stabilisation levels Not surprising given greater level of climate change

12. Page 12© Crown copyright 2004 Uncertainties  Large uncertainties undermine political impact of results  Do we understand them?  Can we reduce them?

13. Page 13© Crown copyright 2004 Sources of uncertainty  The impact of carbon-cycle feedbacks on permissible emissions will depend on:  “Political” uncertainties:  Chosen level of stabilisation (and hence climate change)  Scientific uncertainties:  Climate sensitivity: Greater sensitivity will mean stronger feedbacks for given CO2 level  carbon-cycle parameters  vegetation sensitivity to warming/CO2  Soil sensitivity  Ocean response to climate/circulation changes  All climate-carbon cycle studies to date show future weakening of the natural carbon sink in response to climate change  But significant uncertainty in strength of feedback

14. Page 14© Crown copyright 2004 Other models  UVic model – courtesy of Damon Matthews (in press at GRL)  Stabilisation at 1000ppm  Significant reduction in allowed emissions Without feedbacks With feedbacks

15. Page 15© Crown copyright 2004 C4MIP models Cumulative Emissions Reductions (GtC) UVic Stabilise at 1000ppm by 2350

16. Page 16© Crown copyright 2004 C4MIP models Cumulative Emissions Reductions (GtC) UVic Hadley Stabilise at 1000ppm by 2350

17. Page 17© Crown copyright 2004 C4MIP models Cumulative Emissions Reductions (GtC) UVic (g=0.2) Hadley (g=0.31) C4MIP-min (g=0.04) Stabilise at 1000ppm by 2350

18. Page 18© Crown copyright 2004 Range over C4MIP models Cumulative Emissions Reductions (GtC) C4MIP-mean* UVic * = C4MIP results estimated from gain factors derived from C4MIP transient expts (g=0.14) Stabilise at 1000ppm by 2350

19. Page 19© Crown copyright 2004 Implications of uncertainty  2 main implications of the C4MIP uncertainty  Uncertainty does not span zero  All models agree on positive feedback and hence some degree of reduction in permissible emissions  Required emissions vary greatly  Reductions due to climate feedbacks uncertain by almost an order of magnitude

20. Page 20© Crown copyright 2004 Reducing that uncertainty?  To what extent does the historical record constrain future behaviour?  Climate sensitivity?  No – can’t be well constrained observationally  Causes large spread in future climate and hence in future feedback strength

21. Page 21© Crown copyright 2004 Climate sensitivity  Uncertainty in historical forcing – especially from aerosols – means large uncertainty in climate sensitivity  TAR shows GCM range from 1.5-4.5, but values up to 8-10K can’t be ruled out completely from observations. Andreae et al, Nature, 2005

22. Page 22© Crown copyright 2004 Reducing that uncertainty?  To what extent does the historical record constrain future behaviour?  Climate sensitivity?  No – can’t be well constrained observationally  Causes large spread in future climate and hence in future feedback strength  Carbon cycle parameters?  Not directly from observations – CO2 record can’t distinguish strong fertilisation/strong respiration from weak fertilisation/weak respiration.  But give different future behaviour

23. Page 23© Crown copyright 2004 Single parameter perturbations  Large ensemble of simple model runs with perturbed parameters  In these runs, NPP sensitivity to climate is most important carbon-cycle parameter  More sensitivity than CO2 fertilisation strength or soil respiration sensitivity to temperature  Similar conclusion to Matthews et al., GRL, 2005.  Climate sensitivity outweighs carbon cycle uncertainty CO2 fert’n Soil resp NPP(T) ∆T 2x, 1.5-4.5∆T 2x, 1.5-10 WRE550 WRE450

24. Page 24© Crown copyright 2004 Multiple parameter perturbations  Varying all these parameters, but still fitting historical emissions, gives only very weak constraint on future permissible emissions  High climate sensitivities lead to requirement for significant NEGATIVE emissions Low climate sensitivity High climate sensitivity

25. Page 25© Crown copyright 2004 Reducing that uncertainty?  To what extent does the historical record constrain future behaviour?  Climate sensitivity?  No – can’t be well constrained observationally  Causes large spread in future climate and hence in future feedback strength  Carbon cycle parameters?  Not directly from observations – CO2 record can’t distinguish strong fertilisation/strong respiration from weak fertilisation/weak respiration.  But give different future behaviour  Model validation?  Maybe – recreating observed behaviour is necessary but not sufficient test of a model  C4MIP phase 1 is essential step!

26. Page 26© Crown copyright 2004 C4MIP phase 1 - validation  Atmosphere only model with observed 20 th century SSTs  Just simulate terrestrial carbon cycle  Validate against range of obs:  Site-specific from flux towers  Regional estimates from inversion studies  Interannual variability – e.g. ENSO  Validation is important if we are to know which C4MIP models to trust  But, ability to get these right doesn’t constrain future feedback size – merely gives us clues about how to interpret the models  See Jones & Warnier report on HadCM3LC at:  http://www.metoffice.com/research/hadleycentre/pubs/HCTN/index.html

27. Page 27© Crown copyright 2004 C4MIP phase 1 - validation  Flux tower validation from CarboEurope data  Assess model sensitivity of GPP, Resp against T, P RE GPP Temp precip

28. Page 28© Crown copyright 2004 C4MIP phase 1 - validation  Comparison with TransCom inversions study (Gurney et al, Nature, 2002)  Regional carbon flux estimates from 1992-96  black = transcom  pink = Hadley C4MIP experiment  Agrees pretty well in most places

29. Page 29© Crown copyright 2004 Other potential issues  How important is time to stabilisation?  Emit soon and reduce strongly? Or more gradual?  Can we define an “optimal” pathway?  Sensitivity studies for stabilisation at 550ppm at different rates:  Idealised profiles with asymptotic approach to stabilisation:  CO 2 = a 0 + a 1 * tanh (a 2 + a 3.τ)  Match CO 2 level and rate of change at 2000  τ =time to (95%) stabilisation. Range from 20-150 years.  Not attempted to quantify likelihood – more illustrative  How do climate-carbon cycle feedbacks affect resulting emissions profiles?

30. Page 30© Crown copyright 2004 ‘Optimal’ pathways to stabilisation  “fast” (τ=30) and “slow” (τ=80) emissions profiles to 550 ppm  Carbon cycle feedbacks reduce emissions in all cases

31. Page 31© Crown copyright 2004 ‘Optimal’ pathways to stabilisation  Total 21 st century emissions (higher may be seen as “desirable”)

32. Page 32© Crown copyright 2004 ‘Optimal’ pathways to stabilisation  Max rate of required emissions reductions (higher may be seen as “undesirable”)

33. Page 33© Crown copyright 2004 ‘Optimal’ pathways to stabilisation “worse” “better”  Open Questions:  Can we convert this into “desirability” somehow?  E.g. Linearly combine “total emissions” and “max rate of reduction”  deliberately simplistic – clearly many more factors to consider Shifted optimum?  “desirability” varies with timescale to stabilisation  How do climate- carbon cycle feedbacks affect our choice of “optimal”?

34. Page 34© Crown copyright 2004 Conclusions  Climate feedbacks on the carbon cycle will reduce future natural carbon uptake  Hence, to stabilise CO 2, significantly greater emissions reductions may be required  This is true regardless of:  Stabilisation level  But higher levels see greater reduction  Model  But large spread of feedback strength between models  Timescale to stabilise  Strength of feedback may alter “optimal” shape of trajectory as well as magnitude

35. Page 35© Crown copyright 2004 Conclusions  Large uncertainties between/within models  Observational record directly offers only weak constraint on future behaviour  Validation of complex carbon cycle models against all available data is lacking  Will prove vital to reducing uncertainty