Cost of inaction: Abrupt non-linear climate change · • Cost estimates of abrupt climate change should only be used for sensitivity analyses • All assumptions underlying temporal

Hans-Martin FüsselStanford University

International Workshop“Climate Change: Cost of Inaction”

Berlin, Germany10 April 2006

Cost of inaction:Abrupt non-linear climate change

Hans-Martin FüsselDIW, 10 April 2006 2

Outline1. Abrupt climate change in the past2. Abrupt climate change in the future?3. Abrupt climate change in climate policy

analysis– Information needs for costing– Challenges– Alternatives

4. Some recommendations


Abrupt climate changeWorking definition:• Sustained change in regional or global climate• Faster than the responsible forcing• Presents significant challenges for adaptation• Not included: Individual extreme weather eventsRelated terms:• Large-scale climate discontinuities (IPCC TAR)• Low-probability high-impact events• (Imaginable) surprises• Climate catastrophes• Tipping points

Paleoclimate I• Holozene climate

(current warm period) is unusually stable

• Without human intervention, the Holocene would last another 15,000 years.

• Past temperature shifts have generally been abrupt (up to 10°C on decadal scales)

Source: Rial et al., 2004

Greenland temperature


Paleoclimate II• Abrupt climate

shifts have been caused by gradual shifts in forcing

• Example: Transition from wet to dry Sahara

Source: Claussen et al., 1999


Paleoclimate III• Past precipitation

shifts have often been abrupt

• Relatively “small”historical climate shifts had large social impacts

• Example: US Dust Bowl in the 1930s

Source: Alley et al., 2003


The nonlinear climate system• Complex and often chaotic• Driven by competing positive and negative

feedbacks with largely unknown thresholds• Dominated by abrupt changes in response

to steady forcing• Exhibits stochastic oscillations (“flickering”)• Inherently unpredictable

Likelihood of abrupt change increases with the magnitude and rate of forcing change


Climate change-related concerns• Stability of ocean circulations:

North Atlantic Deep Water formation / THC• Stability of monsoon regimes:

West Africa, Asia• Shifts in regional precipitation regimes:

North Africa, central East Asia, Amazon• Regime shifts of interannual climate modes:

ENSO, NOA, PDO, SO• Stability of large ice sheets:

Greenland, West Antarctic

“Tipping points” of the climate system

Source:Schellnhuber &Held, 2001


Abrupt climate change in IA models1. Ignoring:

Use best-guess values only2. Valuing:

Assign conditional probabilities & estimate damages(e.g., through highly non-linear damage functions, jump in damage function, or drop in utility function)

Even small probability of very high damages induces much stronger near-term mitigation (depending on discount factors and assumed inertia of the energy system)

3. Constraining:Apply deterministic or probabilistic constraints


Costing abrupt climate changeData needs to determine expected presentdamages from a given climate discontinuity:

1. Conditional probability and/or2. Timing of discontinuity3. Associated economic impacts over time

(for different socioeconomic scenarios)4. Aggregation of all impacts (in all regions)

for all states of the world to a present value


Data needs: ProbabilityData sources:• Model-based analyses:

Monte Carlo/Latin hypercube analyses based on plausible distributions for uncertain parameters

• Expert elicitations:Delphi exercises, questionnaires, etc.(e.g., Kriegler & Hall, Zickfeld et al.)

Caveats:• All probabilities are subjective• Agreement on “low-probability events” is typically

on the order of magnitude only

Data needs: Timing• Timing can be

very uncertain• For the same

forcing and model, THC breakdown can vary after 250 to 500 years

• Factor 2 in timing translates into factor 1000+ with discounting

Source: Wang et al., 1999


Data needs: Economic impacts IData sources:• Recent extreme climate events:

e.g., ENSO, tropical cyclones, floods, surges• Historic abrupt climate change:

e.g., Sahel drought, US Dust Bowl• Expert elicitations

Caveats:• Adaptation• Knock-on effects


Data needs: Economic impacts IIWhat would be the costs of …• flooding of two thirds of the Netherlands?• a flickering transition to a new ice age?

What are/have been the costs of …• the loss in climate predictability?• the US “dust bowl”? Costs for whom?• the global AIDS (or malaria) epidemic?• World War I and II?


Data needs: AggregationHow to aggregate costs (and benefits) across different

regions/individuals and across time?• The spatial and temporal aggregation of impacts is not

possible without value judgements• The cost estimates can be very sensitive to the choice

of aggregation method and numeraire• Historical observations are largely irrelevant for

discounting large-scale events in the distant future

How to aggregate non-marginal losses and gains?• Risk aversion requires use of certainty equivalents• Different methodological approaches re. discounting

lead to vastly different estimates


The debate on unbounded damages• Marginal damages can be infinite when climate

change may lead to a “poverty trap” (Tol, 2003)• BUT: Strictly speaking, the expected (logarithmic)

utility of most real-life activities is minus infinity• Unbounded damages are always model artefacts

(e.g., logarithmic utility, endogenous Ramsey discounting for independent regions)

• Alternative decision-analytical approaches:– Excluding policies with “infinitely bad” outcomes

(cost effectiveness, tolerable windows analysis)– Analysis of sequential decision strategies– Risk management: “How safe is safe enough?”

Valuing non-marginal costs and benefits

Left: Discounted GWP for various initial and final stabilization targets.Difference between 400 ppm and 900 ppm final target: 0.025%

Endogenous growth discounting produces misleading cost (and benefit) estimates if applied in non-marginal contexts.Example: Hedging analysis by Yohe et al. (2004)

PVOYohe

Right: Expected discounted GWP for various initial carbon tax levels. Difference between best and worst initial strategy: 0.0004%

Finaltarget

PVOYohe


Challenges of costing abrupt CC• Very large prediction uncertainty

(about likelihood, timing, and magnitude)• Lack of experience re. social impacts

(e.g., adaptation, knock-on affects)• Disagreement about valuing non-market

impacts (e.g., human life / health, culture)• Normative disagreement about aggregation

(e.g., time discounting, equity weighting)• Inadequate methodology

(e.g., for valuing non-marginal damages)


Costs in climate policy analysis• Provide estimates in a metric that appears

relevant to most decision-makers• Required for determining welfare-optimal policies

BUT:• Cost estimates for abrupt climate change are

very highly uncertain• Aggregated costs are crucially dependent on

normative assumptions• Determining costs is not the ultimate goal —

providing useful policy advice is• Costs are just one metric for expressing impacts


Economists on welfare optimization• “The bottom line of all this is that it seems as if the

uncertainty about climate change is too large to apply cost-benefit analysis. […] climate change tests decision analytic tools to the extreme. The results in this paper show that economic analyses of climate policy should be interpreted with more than the usual care.” (Tol, 2003)

• “We have added one more element to our list of reasons why it is inappropriate to use the expected value of discounted net benefits to judge mitigation policy, especially when regional differences are captured in the analysis.” (Yohe, 2003)


Economists on welfare optimization“After 500 years, [global average temperature] is projected to increase 6.2°C climate [in the optimal policy path identified by the DICE-94 model]. While we have only the foggiest idea of what this would imply in terms of ecological, economic, and social outcomes, it would make most thoughtful people —even economists— nervous to induce such a large environmental change. Given the potential for unintended and potentially disastrous consequences, it would be sensible to consider alternative approaches to global warming policies.'' (Nordhaus, 1997)


Summary• Past climate change has been mostly abrupt and flickery;

it is reasonable to assume similar behaviour in the future• Costing abrupt climate change involves huge scientific

and methodological uncertainties; it also raises difficult normative questions

• Current cost estimates for abrupt climate change should be regarded as order of magnitude estimates

• Due to very large uncertainty, assigning costs to potential abrupt change cannot identify “optimal” climate policies

• Recent analyses agree that considering the risk of abrupt climate change calls for significantly stronger mitigation efforts (depending on assumptions about discounting)

• Adaptation is extremely limited in the case ob abrupt climate change; relying only on adaptation is thus an extremely risky strategy to address climate change


Recommendations• Climate policy analysis needs to consider abrupt

climate change• Cost estimates of abrupt climate change should only

be used for sensitivity analyses• All assumptions underlying temporal and spatial

aggregation of costs need to be stated clearly; otherwise the analysis is meaningless for policy

• Apply “alternative” decision-making frameworks(e.g., sequential decision analysis, cost-effectiveness / tolerable windows analysis, risk management)

• Assess the vulnerability of current societies to several plausible climate discontinuities(e.g., ATLANTIS and INTEGRATION projects)


Thank you for your attention!

[email protected]

Cost of inaction: Abrupt non-linear climate change · • Cost estimates of abrupt climate change should only be used for sensitivity analyses • All assumptions underlying temporal

Documents