From Uncertain Depositions to Uncertain Critical Load Exceedances

1
From Uncertain Depositions to Uncertain Critical
Load Exceedances
Balancing Critical Loads Data Aggregation and
Uncertainties

• Maximilian Posch
• RIVM
• Coordination Center for Effects (CCE/TF MM)

2
From critical load function to protection/exceedan
ce isolines

• Every ecosystem characterized by an acidity
  critical load (CL) function
• There are many CL-functions in a grid square
• From these protection/exceedance isolines can be
  calculated (3-D CDF)
• These are used in the IAMs
• Details see CCE Status Reports 97/99

3
Uncertainty in CL parameters results in
uncertainty band around every isoline
Integration yields uncertain protection
percentages
S and N deposition are also uncertain and
correlated in every grid square
Overlaying (integration) yields probabilities to
protect a certain percentage of ecosystems
4
Ecosystem protection percentages for a selected
confidence level
For a given deposition (e.g. 1990) for every grid
cell the protection percentage (e.g. 50) for any
given confidence level can be computed and mapped
Ecosystem protection percentage for 50
(deterministic, left) and 95 (on safe side)
confidence level
Uncertainty range 5-95
5
From uncertain CLs to uncertain protection (1-D)
Assuming independent uncertainties of CLs within
a grid cell leads to cancellation of
uncertainties (and loss of spatial information)
but widening of range in protection
CDF. Cancellations (narrowing of uncertainty
band) depend the degree of overlap of CLs.
Examples
Realistic example
6
Uncertainty in computation of exceedances
(protection )
Uncertainty depends of difference in means as
well as (ratio of) variances
Protection probability for given deposition
The larger the difference in the means the larger
can be the uncertainty.
7
Dependence of uncertainty on grid size
Case 1
Subdivision of grid does not reduce CL
variability (var-left var-rightvar-tot
self-similar) If also deposition is the same,
no reduction in uncertainty.
Case 2
Subdivision of grid reduces CL variability Possibl
e reductions of uncertainties depends on how
deposition changes!
Needed Screening and classification of grids
where improvements can be achieved by reducing
grid size. (related optimal number of data
points in grid)
8

• Summary and Recommendations
• Tool available to do uncertainty analyses for
  LRTAP IA (though technical improvements
  desirable)
• Including uncertainties requires additional
  choices to be made by decision makers (confidence
  levels, etc.)
• Efforts have to be made to present results in
  comprehensible formats
• Focus on areas (grid cells) where deposition are
  (will be) similar to CLs
• Investigate grid size - (CL) variability - data
  requirement issues
• Parties have to provide good uncertainty
  estimates of input data -- as well as resources
  -- to allow meaningful uncertainty analyses. Not
  only (co-)variances, but especially possible
  biases.