Environmentally Impaired Property Transaction Analysis Combining Decision Trees and Monte Carlo Simulation Timothy Havranek and Poh Boon Ung October 2007 Agenda • Background Information • Modeling Process • Decision Tree • Results Project Background and Objective • Estimate probable remediation costs • Work in support of purchase negotiations of an environmentally impaired property • Client is a property development company: – Would acquire liability for historical environmental impacts – Seeking strategic plan for least cost remediation in light of uncertainties – Planned to use estimate in support of lowering purchase offer • Current property owner had filed Chapter 11 bankruptcy procedures which complicated the negotiation process Site Background • Property was former metals processing plant • Soil, groundwater, surface water, and wetlands impacted by lead, arsenic and chromium • Groundwater also impacted by petroleum hydrocarbons • Property is over 150 acres in size and located near a major river • Property could be redeveloped for industrial, commercial, or municipal use • Current owner under a consent order to perform environmental cleanup Decision Analysis Process Decisi on Proces s Evaluation & Framin g Modelin g Sensitivity Analysis Decisio n Action Plan Framing Meeting for QDA Site Background Stakeholder Analysis Action Items Strategies Shared Vision of Path Forward Chance Events Multi-Criteria Analysis Current Policies Decisions & Choices Potential Remedial Alternatives Groundwater Soil Surface Water Free Product Wetlands Impermeable Wall Soil / 6" Asphalt Cap Dredge Basin Trench Skimmer System Restoration Reactive Barrier Wall Engineered Cap Dual Phase Extraction System Mitigation Model Structure • Capital and Operations & Maintenance (O&M) costs were estimated using standard engineering costs forms – Pert Distributions used to estimate uncertainties in quantities, unit prices, installation year, and O&M durations • Engineering costs sheets were linked to 30-year cash flow model for each remedial technique • Net Present Value for each remedial technique linked decision tree and Monte Carlo simulation model Benefits of Decision Tree • Visual representation of available choices • Valuable communication tool • Helps organize alternatives • Provides a working map of the project strategy Estimating Individual Cost Components Item Description Unit Quant. MIN Quant. ML Quant. MAX Distribution Unit Price MIN Unit Price ML Unit Price MAX Distribution TOTAL Capital Costs 1 Dewatering Control Slurry Walls Reactive Gates - Reactive Media Reactive Gates - Sand T&D of excavated material Misc. Disposal Day 50 100 150 100 $2,000 $2,000 $2,000 $2,000 $200,000 SF 18000 52500 67500 49250 $20 $25 $30 $25 $1,231,250 Ton 500 1500 1750 1375 $400 $600 $800 $600 $825,000 Ton 550 1650 1925 1513 $11 $11 $11 $11 $16,643 Ton 70 200 250 187 $120 $160 $200 $160 $29,920 LS 1 1 1 1 $10,000 $10,000 $10,000 $10,000 $10,000 Operation and Maintenance (O&M) Costs Unit Duration MIN Duration ML Duration MAX Distribution Annual Cost Annual Cost Annual Cost Distribution 10 Replacement of PRB YR 1 1 1 $1,612,000 $1,612,000 11 Monitoring YR 30 30 30 $60,000 $55,000 2 3 4 5 6 7 8 9 Total Capital Costs 12 13 $2,312,813 YR YR $1,612,000 $1,612,000 30 $50,000 $55,000 TOTAL Root of Decision Tree: Groundwater and Soil Components Offer Soil /Asphalt Cap 30.0% Yes $3.3 TRUE Soil Remediation $14.5 Engineered Cap FALSE $7.2 TRUE Impermeable Wall Impermeable Wall Fails $3.8 $12.2 70.0% No $0 Groundwater Remediation $12.2 Reactive Barrier FALSE $5.3 Link to Soil Rem $12.7 Link to Soil Rem $11.2 Middle Portion of Tree: Soil and Surface Water Remediation TRUE Dredge Basin $0.7 40.0% Yes, Install $4.5 Offer Soil /Asphalt Cap TRUE Surface Water Remediation $12.8 Approved by Regulators $14.5 No, Engineered Cap 60.0% $7.2 Soil Remediation $14.5 Engineered Cap FALSE $7.2 Link toSurface Water Rem 15.5 Link to Surface Water Rem $15.54 Tree Terminal Nodes: Wetland Issues Yes Yes - Peform Mitigation 50.0% $0.0 Mitigation Approved? $12.8 No - Perform Restoration Wetlands Remediation $12.8 Restoration FALSE 0.00% $0.3 $12.9 $0.3 $13.1 75.0% 3.04% $0.0 $12.7 $12.8 No TRUE 1.01% Restoration Req. after Mit. $0.2 Mitigation 25.0% 50.0% 4.05% $0.3 $12.9 PrecisionTree / @RISK Settings Net Present Value: Frequency Histogram * Distribution based on values of one sampled path per iteration and decisions follow current optimal path X <=$9.0 M 5% 10% X <=$17.6 M 95% Mean = $12.4 Million 9% 8% Probability 7% 6% 5% 4% 3% 2% 1% 0% $6 $8 $10 $12 $14 Cost in $ Millions $16 $18 $20 $22 Net Present Value: Risk Profile * Distribution based on values of one sampled path per iteration and decisions follow current optimal path X <=$9.0 M 5% 100% X <=$17.6 M 95% Mean = $12.4 Million 90% Cumulative Probability 80% 70% 60% 50% 40% 30% 20% 10% 0% $6 $8 $10 $12 $14 Cost in $ Millions $16 $18 $20 $22 Output Statistics (cost in millions) *Results based on values of one sampled path per iteration and decisions follow current optimal path Mean Standard Deviation Mode 5.0% 10.0% 15.0% 45.0% 50% (Median) 55.0% 85.0% 90.0% 95.0% $12.4 $2.5 $12.3 $9.0 $9.3 $9.6 $12.0 $12.2 $12.4 $15.3 $16.0 $17.6 Net Present Value: Frequency Histogram * Distribution based on values of one sampled path per iteration and decisions may change (based on expected values) 60% X <=$11.2 M 5% X <=$13.7 M 95% Mean = $12.4 million 50% Probability 40% 30% 20% 10% 0% $10 $11 $12 $13 $14 Costs in $ Millions $15 $16 $17 Output Statistics (cost in millions) * Results based on values of one sampled path per iteration and decisions may change (based on expected values) Mean Standard Deviation Mode 5.0% 10.0% 15.0% 45.0% 50% (Median) 55.0% 85.0% 90.0% 95.0% $12.4 $0.8 $12.1 $11.2 $11.4 $11.6 $12.2 $12.3 $12.4 $13.2 $13.4 $13.7 Net Present Value: Frequency Histogram * Distribution based on expected value of model per iteration and decisions follow current optimal path X <=$11.3 5% 14% X <=$13.8 95% Mean = $12.5 Million 12% Probability 10% 8% 6% 4% 2% 0% $10 $11 $12 $13 Cost in $ Millions $14 $15 $16 Output Statistics (cost in millions) * Results based on expected value of model per iteration and decisions follow current optimal path Mean Standard Deviation Mode 5.0% 10.0% 15.0% 45.0% 50% (Median) 55.0% 85.0% 90.0% 95.0% $12.5 $0.8 $12.5 $11.3 $11.5 $11.7 $12.3 $12.4 $12.5 $13.3 $13.5 $13.8 Sensitivity Analysis on Impermeable Wall Failure Probability Expected NPV $ Millions $15 $14 $13 $12 $11 $10 10% 20% 30% 40% 50% 60% 70% Probability of Impermeable Wall Failure 1 : Impermeable Wall 2 : Reactive Barrier 80% 90% 13.5 13-13.5 13 12.5-13 12-12.5 12.5 11.5-12 12 11-11.5 11.5 81% 63% 46% 28% 10% 11 28% 10% Cap Approval Probability 46% 63% 10.5 81% Expected NPV $ Millions Two Way Sensitivity Analysis on Probabilities 10.5-11 Impermeable Wall Failure Probability Results and Conclusions • Important to apply appropriate settings in modeling process – Internal management and understanding – Negotiation process • Sensitivity analysis useful to identify decision break-points • Results are being used in negotiations process; client is actively using the results in negotiations to buy the site • Decision tree analysis assisted in: – Identifying optimum strategy – Communicating path forward – Determining the response and potential effects to chance outcomes • Combination of Monte Carlo simulation and decision tree analysis provides additional insights into range of potential outcomes