InLCA/LCM Seattle, September Dr. Gontran F. Bage Laurence Toffoletto Pr. Louise Deschênes Pr. Réjean Samson Considering Uncertainties in the Functional Unit: Development of a More Flexible Strategy to Achieve the Goal of an LCA Study CIRAIG École Polytechnique de Montréal Canada

InLCA/LCM Seattle, September Outline Problem: Site remediation, environmental impacts and uncertainties Methodology How to do a probabilistic LCA Model development: METEnvORS Case study: Remediation of a diesel-contaminated site Deterministic approach- Classical LCA Probabilistic approach- Use of METEnvORS Conclusions

InLCA/LCM Seattle, September Site remediation, environmental impacts and uncertainties Site remediation technology Problem Secondary impacts Local and global environmental impact generation from the remediation activities Environmental impact minimization is a must in a sustainable development context Primary impacts Reduction of local contamination (site contamination) LCA approach Contaminated site

InLCA/LCM Seattle, September Site remediation, environmental impacts and uncertainties Problem Classical LCA- Deterministic approach Initial level of contamination: Mean concentrationInitial level of contamination: Mean concentration Main assumption: The selected technology is fully effectiveMain assumption: The selected technology is fully effective Functional unit: Decontaminate a given volume of soil from an initial concentration to a final one using a specific technologyFunctional unit: Decontaminate a given volume of soil from an initial concentration to a final one using a specific technology Mean concentration does not consider spatial variabilityMean concentration does not consider spatial variability The technology’s effectiveness is a function of, in part, the initial and final contamination levelsThe technology’s effectiveness is a function of, in part, the initial and final contamination levels !

InLCA/LCM Seattle, September Site remediation, environmental impacts and uncertainties Problem Probabilistic LCA approach Technology’s effectiveness is variableTechnology’s effectiveness is variable Initial level of contamination: Discrete or continuous probability distribution of contamination rangesInitial level of contamination: Discrete or continuous probability distribution of contamination ranges q(s 1 ;s 2 ;s 3 ) % Contamination ranges s1s1 s2s2 s3s3 Decontaminate a given volume of soil from an initial q with a given technology until the most probable range in the final q respects the remediation goals Functional unit s 1 : Remediation goals Remediation evolution Overall definition of the initial level of contamination

InLCA/LCM Seattle, September How to do a probabilistic LCA Methodology Remediation activities Life cycle assessment Primary impacts Secondary impacts All technologies Total impacts Tech. 1 Total impacts Tech. 2 Total impacts Tech. n … Quantify all impacts associated with the remediation activities Input data Technologies’ effectiveness Possible contamination levels METEnvORS All technologies (annually) One additional year of treatment if remediation goals are not achieved Minimizing the expected environmental impacts Deterministic LCA Probabilistic LCA

InLCA/LCM Seattle, September Model development: METEnvORS Initial State q(s 1,s 2 ) Techno. A Techno. B q(s 1 ) q(s 2 ) q(s 1 ) q(s 2 ) Model for the Evaluation of the Technically and Environmentally Optimal Remediation Strategy q’ 2 P(s 2 |q’ 2,A) q’ 2 (s 1 ) < q’ 2 (s 2 ) q’ 1 P(s 1 |q’ 1,A) q’ 1 (s 1 ) > q’ 1 (s 2 ) End of remediation q’ 3 P(s 1 |q’ 3,B) q’ 3 (s 1 ) > q’ 3 (s 2 ) End of remediation q’ 4 P(s 2 |q’ 4,B) q’ 4 (s 1 ) > q’ 4 (s 2 ) End of remediation q’’ 2 q’ 2 (s 1 ) q’ 2 (s 2 ) P(s 2 |q’’ 2,A) q’’ 2 (s 1 ) < q’’ 2 (s 2 ) q’’ 1 P(s 1 |q’’ 1,A) q’’ 1 (s 1 ) > q’’ 1 (s 2 ) End of remediation q’’ 3 q’’ 4 P(s 1 |q’’ 3,B) P(s 2 |q’’ 4,B) q’ 2 (s 1 ) q’ 2 (s 2 ) q’’ 3 (s 1 ) > q’’3(s 2 ) q’’ 4 (s 1 ) > q’’ 4 (s 2 ) End of remediation Techno. A Techno. B … Stage 2 Stage 1 Step 1- Developing all the possibilities

InLCA/LCM Seattle, September q’ 2 (s 1 ) q’ 2 (s 2 ) P(s 2 |q’’ 2,A) q’’ 2 (s 1 ) < q’’ 2 (s 2 ) q’’ 1 P(s 1 |q’’ 1,A) q’’ 1 (s 1 ) > q’’ 1 (s 2 ) End of remediation Techno. A … Techno. B q(s 1 ) q(s 2 ) q’ 3 P(s 1 |q’ 3,B) q’ 3 (s 1 ) > q’ 3 (s 2 ) End of remediation q’ 4 P(s 2 |q’ 4,B) q’ 4 (s 1 ) > q’ 4 (s 2 ) End of remediation Model development: METEnvORS Initial State q(s 1,s 2 ) q’ 1 P(s 1 |q’ 1,A) P(s 2 |q’ 2,A) q(s 1 ) q(s 2 ) Techno. A Methodology q’ 1 (s 1 ) > q’ 1 (s 2 ) q’ 2 (s 1 ) < q’ 2 (s 2 ) End of remediation q’’ 3 q’’ 4 P(s 1 |q’’ 3,B) P(s 2 |q’’ 4,B) q’ 2 (s 1 ) q’ 2 (s 2 ) q’’ 3 (s 1 ) > q’’ 3 (s 2 ) q’’ 4 (s 1 ) > q’’ 4 (s 2 ) End of remediation Techno. B Decision criterion: Environmental impact minimization at each stage Step 2- Backwards resolution q’’ 1 Techno. A … q’ 3 q’ 4 Techno. B Impact Tech B < Impact Tech A Impact Tech A < Impact Tech B

InLCA/LCM Seattle, September Model development: METEnvORS Methodology Initial State q(s 1,s 2 ) q’ 1 Scenario 1 P(s 1 |q’ 1,A) P(s 2 |q’ 2,A) q(s 1 ) q(s 2 ) Techno. A q’’ 3 q’’ 4 P(s 1 |q’’ 3,B) P(s 2 |q’’ 4,B) q’ 2 (s 1 ) q’ 2 (s 2 ) Techno. B Scenario 3 Scenario 2 Scenario 1: q(s 1 ) * P(s 1 |q’ 1,A) Scenario 2: q(s 2 ) * P(s 2 |q’ 2,A) + q’ 2 (s 1 ) * P(s 1 |q’’ 3,B) Scenario 3: q(s 2 ) * P(s 2 |q’ 2,A) + q’ 2 (s 2 ) * P(s 2 |q’’ 4,B) Probabilities of occurrence Scenario 1: Impact(q’ 1 ) + Impact(Tech.A) Stage 1 Scenario 2: Impact(q’’ 3 ) + Impact(Tech.A) Stage 1 + Impact(Tech.B) Stage 2 Scenario 3: Impact(q’’ 4 ) + Impact(Tech.A) Stage 1 + Impact(Tech.B) Stage 2 Total impact

InLCA/LCM Seattle, September Remediation of a diesel- contaminated site Case study Contaminated volume: m 3 of soil Mean concentration: mg diesel/kg soil Initial state of the site (q): Range < 700 mg/kg (s 1 ) Range 700 – mg/kg (s 2 ) Range > mg/kg (s 3 ) Does not require soil excavationDoes not require soil excavation Not fully effective for this remediationNot fully effective for this remediation Requires soil excavationRequires soil excavation Fully effective for this remediationFully effective for this remediation BioventingBiopile % s1s1 s2s2 s3s3 98,8 1,2 0 Remediation goals: Reaching the s 1 range Maximum remediation duration of three years Available technologies:

InLCA/LCM Seattle, September Decontaminate m 3 of diesel-contaminated soil from an initial contamination probability distribution q(0%; 1,2%; 98,8%) using a combination of bioventing and biopile to reach either a new state of the site for which the s 1 range has the highest probability of occurrence or a maximum of three years of treatment. Case study Functional unit Identify the remediation strategy minimizing both primary and secondary impacts meanwhile reaching the < 700 mg diesel/kg range. Objective Remediation of a diesel- contaminated site

InLCA/LCM Seattle, September Remediation of a diesel- contaminated site Case study

InLCA/LCM Seattle, September Deterministic approach- Classical LCA BiopileBioventing Primary impact Pt Secondary impact Pt Pt Total impact Pt Pt Treatment duration 2 years 3 years Classical LCA for each technology for the remediation of the site from the initial mean concentration to a concentration under 700 mg/kg gives the following results: Following a deterministic approach, a three-year bioventing treatment is environmentally preferred to the biopile. Impact have been assessed using SimaPro 5 and the EDIP97 methodology

6 145 mg/kg Bv mg/kg 980 mg/kg B-C 1,2 % 0,8 % 99,2 % mg/kg 100 % <B 0 % Bv mg/kg <B 1,2 % 100 % mg/kg 840 mg/kg B-C 98,8 % 83,2 % 16,8 % mg/kg <B 9,5 % 100 % mg/kg 767 mg/kg B-C 90,5 % 83,2 % 16,8 % mg/kg <B 8,8 % 100 % mg/kg 770 mg/kg B-C 91,2 % 83,2 % 16,8 % mg/kg 418 mg/kg 815 mg/kg Bv B-C 96,7 % <B 3,3 % 100 % 83,2 % 16,8 % mg/kg 409 mg/kg 854 mg/kg Bv B-C 100 % <B 0 % 100 % 20,6 % 79,4 % mg/kg <B 0 % 100 % mg/kg 860 mg/kg B-C 100 % 19,9 % 80,1 % mg/kg mg/kg mg/kg >C 98,8 % 0,8 % 75,5 % 24,3 % mg/kg mg/kg >C 91,9 % mg/kg 0,3 % 80,4 % 19,3 % mg/kg 927 mg/kg 0,8 % 99,2 % B-C 8,1 % mg/kg 100 % <B 0 % mg/kg 357 mg/kg 892 mg/kg 341 mg/kg mg/kg mg/kg Bv B-C 23,3 % 100 % 0,8 % 99,2 % >C 76,7 % 0,5 % 80,2 % 19,3 % <B 0 % Year 1Year 2Year 3 Same remediation activities = Same secondary impact Different final concentrations = Different primary impact Different total impact Most probable scenario Scenario with less impact (total) # # Scenario stopped by the attainment of the remediation goals Scenario stopped by the time constraint

InLCA/LCM Seattle, September Probabilistic approach- Results The Optimal Remediation Strategy (ORS) is established considering the uncertainties surrounding the real level of contamination and the ease at achieving the remediation goals Both deterministic and probabilistic approaches lead to the selection of the same technology, but ORS is made of 28 scenarios  Probability of occurrence of scenarios with total impact < deterministic approach = 72%

InLCA/LCM Seattle, September Probabilistic approach- Results Deterministic approach Probabilistic approach ORS Most probable scenario Duration 3 years 2.8 years (expected value) (expected value) 3 years Total environmental impact Pt Pt (expected value) Pt Probability of occurrence N/AN/A45.3% Final expected concentration In s 1 range (< 700mg/kg) N/A In s 1 range (< 700mg/kg)

InLCA/LCM Seattle, September Conclusions Using METEnvORS highlights on all the occurrences (scenarios) that a remediation may follow These occurrences are results of uncertainties on the level of contamination and the technology’s effectiveness Since the model is multistage, the decision-maker can, with the information he has collected during a stage, review his technology choice at the beginning of the next stage Knowing the ORS, gives the decision-maker a better picture of the reality Including these uncertainties into a site remediation LCA provides a better applicability of LCA in environmental management

InLCA/LCM Seattle, September Acknowledgements This research has been supported by: Fonds d’action québécois pour le développement durable