Riku Suutari, Markus Amann, Janusz Cofala, Zbigniew Klimont Wolfgang Sch

1
Riku Suutari, Markus Amann, Janusz Cofala,
Zbigniew Klimont Wolfgang Schöpp
A methodology to propagate uncertainties through
the RAINS scenario calculations A study
funded by the UK Department of Environment, Food
and Rural Affairs
2
Uncertainty analysis through error propagation
Emission model
Dispersion model
Critical load model
Transfer matrices
Regional emissions
Critical load functions
Protection isolines
Deposition estimate
Environmental impact
3
Assumed uncertainties in input parametersCoeffici
ents of variation
SO2 NOx NH3
Activity rates 1990 2010 0.06-0.200.12-0.30 0.06-0.200.12-0.30 0.05-0.50 0.10-1.00
Emission factors 0.05-0.10 0.075-0.15 0.15-0.40
Removal efficiencies 0.005-0.05 0.015-0.10 0.05-0.15
Atmospheric transfer coefficients 0.10-0.15
Critical loads CLmax 0.22 0.12 0.12
4
Uncertainties of intermediate results95
confidence intervals
SO2 NOx NH3
Emissions 13 13 15
Deposition 14-17
Critical loads excess(area of protected ecosystems) -5 - 2.5 -5 - 2.5 -5 - 2.5
5
Probability for protecting ecosystems Gothenburg
Protocol 2010
6
Conclusions General

• A methodology for uncertainty treatment is now
  available, but conclusions are still incomplete
  and preliminary
• Quantifications of uncertainties are the most
  uncertain elements in uncertainty analyses
• Correlations and distributions difficult to
  quantify
• Uncertainties only for specific model output, but
  not for general model

7
Conclusions Uncertainties in the RAINS model
chain

• The potential for error compensation determines
  uncertainties of emission and deposition
  estimates
• Spread of ecosystems sensitivity determines
  uncertainties in ecosystems protection
• Uncertainties in ecosystems protection are
  smaller than many uncertainties of input
  parameters
• Uncertainties are determined by data availability
  and can be influenced by model design
• Overall largest uncertainties due to sub-grid
  deposition and limited data about critical loads

8
ConclusionsUncertainty results in a policy
context

• Do these type of conclusions (quantified
  uncertainties) help the decision makers?
• Implications for target setting! Is a new
  approach required?
• Error propagation useful for scenario analysis
• For optimization different approach (robustness!)
  necessary