A Model Approach toward the Management of Future Uncertainties - PowerPoint PPT Presentation

A Model Approach toward the Management of Future Uncertainties Shunsuke Mori (Tokyo University of Science) ICA-RUS International Symposium, Dec.4-6, 2013, Tokyo

1. Introduction Various “uncertainties” in the climate decision making -> key barrier against policy agreement - Impacts of climate changes - Uneven societal distribution of cost and benefits - Deployment strategy of technology options, etc. Existing method – basically focusing on expected utility, however - “Extreme (tipping) Event” : low probability and high risk - Long tail risk Shutdown of thermohaline circulation (THC), collapse of west antarctic ice sheet, the collapse of Greenland ice sheet, methane outburst, increase of hurricane and cyclones, etc. Dec ecision bas n based on ed on max aximum ex expec pected ut d utility ( (MxU xU) w woul ould d not not be be pr pref eferab able. -> A Alter ernat native m e metho hod d : minimum um regret et strat ategy egy ( (MnR nR)

“Extreme (tipping) Event” : low probability and high risk - Nuclear power : high technological potential but low societal acceptance, at least in Japan. After March 11, 2012, Gigantic earthquake followed by nuclear station accident, people in Japan seriously consider the “unexpected outcomes”. - CCS and geo-engineering : large possibility to mitigate the global warming, but regrettable when warming damage is low. - Other possible tipping events: - shutdown of thermohaline circulation (THC), collapse of west antarctic ice sheet, the collapse of Greenland ice sheet, methane outburst, increase of hurricane and cyclones, etc. Dec ecision bas n based on ed on max aximum ex expec pected ut d utility ( (MxU xU) w woul ould d not not be be appl applicab able. -> A Alter ernat native m e metho hod d : minimum um regret et strat ategy egy ( (MnR nR)

Theme 4: (Leader, S.Mori) Technology Development and Implementation Strategies under Uncertainties: - Potential contribution and constraints? - Potential cost? - Potential trade-offs and synergies among options? → Need for the comprehensive quantitative analysis Structure of the Integrated Research on the Development of Global Climate Risk Management Strategies Project

2. Maximum Expected Utility vs. Minimum Regret - example Optimum fuel mix of P 1 (Kerosene): low expected price but large uncertainty P 2 (Coal based DME): high expected price but small uncertainty Option A: Low expected cost but high uncertainty Realized cost of Option A Regret of Option A Realized cost Realized cost of Option B of Option B Realized cost Regret of of Option B mixed-strategy Realized cost Realized cost of Option A Realized cost of Option B Realized cost Mixed strategy =α × Option A + (1- α) × Option B of Option A of Option A Option B: High expected cost but low uncertainty Option B can be better than Option A with low probability. Option A is better than Option B in many cases. However, Expected cost strategy always chooses Option A only. -> how about risk ? Mixed-strategy focusing on regret

2. Maximum Expected Utility vs. Minimum Regret – example-2 Optimum fuel mix of P 1 (Kerosene): low expected price but large uncertainty P 2 (Coal based DME): high expected price but small uncertainty 180 80 160 70 140 60 120 50 100 40 80 30 60 20 40 10 20 0 0 Jan-09 Jan-10 Jan-11 Jan-12 Jan-13 Jan-14 Jan-15 Jan-16 Jan-17 Jan-18 Jan-19 Jan-20 Jan-21 Jan-22 Jan-23 Jan-24 Jan-25 Jan-26 Jan-27 Jan-28 Jan-29 Jan-09 Jan-10 Jan-11 Jan-12 Jan-13 Jan-14 Jan-15 Jan-16 Jan-17 Jan-18 Jan-19 Jan-20 Jan-21 Jan-22 Jan-23 Jan-24 Jan-25 Jan-26 Jan-27 Jan-28 Jan-29 Figure 1 Kerosine price in ¥/ktoe Figure 2 DME price in ¥/ktoe ( ) ω = α ω + − α ω Optimum fuel mix * P ( t ; ) ( t ) P ( t ; ) 1 ( t ) P ( t ; ) 1 2 Kerosene is always fully selected when the expected value is employed.

2. Maximum Expected Utility vs. Minimum Regret – cont. Conventional minimizing regret of strategy α *(t) { } α = − − * * Re gret ( ( t ) | n ) max P ( t ; n ) P ( t ; n ), P ( t ; n ) P ( t ; n ) { } 1 2 α = α * ( t ) min max . Re gret ( ( t ) | n ) n Problems - Normal distribution → theoretical “maximum value” ? - Strong correlation among variables → theoretical distribution ? - “Spaghetti” simulation results → 95% confidence interval ? Alternative formulation { } θ + ∑ ∑ θ 1 / θ min . w ( n ) P _ UP ( t ; n ) P _ UP ( t ; n ) 1 2 n t → Minkowski generalized distance − = − * P ( t ; n ) P ( t ; n ) P _ UP ( t , n ) P _ LO ( t , n ) 1 1 1 − = − * P ( t ; n ) P ( t ; n ) P _ UP ( t , n ) P _ LO ( t , n ) 2 2 2 θ → ∞ the above converges to the min-max strategy.

θ=1 θ=2 θ=4 θ=5 θ=6 θ=7 θ=8 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 Jan-09 Jan-10 Jan-11 Jan-12 Jan-13 Jan-14 Jan-15 Jan-16 Jan-17 Jan-18 Jan-19 Jan-20 Jan-21 Jan-22 Jan-23 Jan-24 Jan-25 Jan-26 Jan-27 Jan-28 Jan-29 Figure 3 Optimal fuel mix weights α *(t) corresponding to the θ changes → Share of synthetic fuel DME increases with larger θ while no DME was employed under MxU strategy.

3. Extension of MARIA for the regret based assessment MARIA (Multiregional Approach for Resource and Industry Allocation) - an inter-temporal optimization model integrating top-down macroeconomic activity and bottom-up technology flows Economic Activity Climate Change Labor Capital Stock Ocean - Atmosphere heat exchange Electricity Thermal energy Bern carbon cycle model GDP Consumption Max. Trade Investment Other GHG CO2 Food and Feed demand Oil Coal Gas Land - use Changes Crop production Hydropower Nuclear Geothermal - LWR Biomass Potential Cropland - LWR - Pu Solar power - FBR Wind power Forest area Energy Supply Land - use Change Figure 4 Structure of MARIA model

3-2. Future Uncertain Scenarios Uncertainty: 12 scenarios on loss of GDP when the global atmospheric temperature rises 3.0 Celsius degree from the pre-industry level Scenario-1 0.6% - 0.9% Scenario-2 1.2% - 1.8% <- reference case Scenario-3 1.8% - 2.7% Scenario-4 2.4% - 3.6% Scenario-5 3.0% - 4.5% Scenario-6 3.6% - 5.4% Scenario-7 4.2% - 6.3% Scenario-8 4.8% - 7.2% Scenario-9 5.4% - 8.1% Scenario-10 6.0% - 9.0% Scenario-11 12% - 18% Scenario-12 18% - 27% → Conventional Min.-Max regret strategy refers only Scenario-1 and Scenario-12, two extreme cases.

4. Formulation – conventional Min-max approach Objective function f(SCN) : discounted present value of utility under scenario SCN. Optimum solution under SCN is f * (SCN).   C ( SCN ) ∑ ∑   − = + * t h , t f ( SCN ) max . ( 1 r ) L ln   h , t h t L   h , t X*(SCN) -- Optimal solution of control variables under scenario SCN Regret of strategy X * (SCN) under the realized scenario scn’ is ( ) = = * * f ( SCN | scn ' ) max f scn ' ; X ( t ; scn ' ) X ( t ; SCN ) X ∴ = − * * Re gret ( SCN | scn ' ) f ( scn ' ) f ( SCN | scn ' ) - Min-Max regret solution is basically determined by the extreme assumption regardless of the plausibility . - Optimal “policy mix” cannot be generated.

4-2. Formulation – generalized distance approach ( ) ( ) ( ) ( ) = − = − * * * Re gret [ X ( t ) | SCN ] f X ( t ; SCN ) f X ( t ) D _ UP X ( t ) | SCN D _ DN X ( t ) | SCN { } ( ) ∑ θ 1 / θ × − × t min P ( SCN ) ( 1 d ) D _ DN X ( t ) | SCN (single stage decision) SCN X ( t ) P(SCN) denotes occurrence probability of scenario SCN Expansion – ATL multi-stage decision approach = ≤ 0 X ( m , t ) X ( t ) for t T m: future bifurcation possibilities { } ( ) ( ) θ ∑ ∑ ∑ 1 / θ − × t min ( , ) 1 _ ( , ) | P SCN m d D DN X m t SCN (multi stage decision) m SCN t ( , ) X m t Existing MxU formulation ( ) ( ) ∑ ∑ − t (single stage decision) max P ( SCN ) 1 d f X ( t | SCN ) SCN t ( ) ( ) ∑ ∑ ∑ − t (multi stage decision) max P ( SCN , m ) 1 d f X ( m , t | SCN ) SCN m t

4-3. Decision Strategy under Future Uncertainties A: Three decision stages SS3-1 H M L SS3-12 Three stage decision Single stage decision: Two stage decision: 3 decisions then 12 under 12 scenarios 3 decisions under 12 decisions (SS3) (SS1) scenarios (SS2) B: Two decision basis MxU: Maximizing expected utility MnR: Minizing regret in generalized distance

5. Simulation Results 5-1 perfect information 250 30 1 1 2 2 25 200 3 3 20 4 4 150 5 5 15 6 6 100 7 7 10 8 8 50 9 9 5 10 10 11 11 0 0 12 12 Figure 5 GDP without uncertainty Figure 6 CO 2 emission without uncertainty in trillion US dollars in Gt-C (L-m and L-h overlap each other.) - In the perfect information cases, carbon control strategies bifurcate broadly. - Under future uncertainties, how the policy maker(s) can select single emission path?