Chapter 5
Handling Uncertainty and Sensitivity
of ERA Acute Towards Input Parameters
Abstract Uncertainty evaluation and sensitivity testing of the functions and parameters used in ERA Acute serve two functions. ERA Acute is a deterministic model
which is sensitive to the range of values used for the parameters. Parameters have
inherent uncertainties as to what their true values are, and functions may have varying
strength of knowledge. The individual functions were tested with respect to their
sensitivity towards variation of the parameter values using both deterministic and
stochastic testing. Based on the testing, an uncertainty scoring system was used
to identify and prioritize the most important parameters for reducing uncertainty.
Recommendations for handling the uncertainty and securing comparability in spite
of uncertainty were set up as a conclusion of the studies.
Keywords Uncertainty testing · Uncertainty handling · Sensitivity testing ·
Parameter sensitivity scoring · Spearman correlation coefficient analysis · Partial
Rank Correlation Coefficient analysis
5.1 Sensitivity Testing and Uncertainty Handling
Every model has some inherent uncertainty. A model is a simplified mathematical
description which in a quote often accredited to Albert Einstein should be “as simple
as possible, but no simpler”. Both the simplifications and the detailing of sub-models
and equations carry with them uncertainties.
ERA Acute is a deterministic model where the structure of the functions and
their calculation sequences reflect how we believe that oil spills may harm the VECs
in the different compartments. The output of a deterministic model is completely
determined by the input parameters and structure of the model. A stochastic model
on the other hand, has inherent randomness in the model structure and will not
produce the same result, even given the same parameter value (Helton et al. 2006;
Marino et al. 2008).
The functions are mathematical descriptions of how we understand that the impact
and restoration will occur, and therefore also subject to uncertainty about the model
framework and its scientific soundness (see e.g. Gaber et al. 2009). If our assumptions
© The Author(s) 2021
C. Stephansen et al., Assessing Environmental Risk of Oil Spills with ERA Acute,
SpringerBriefs in Environmental Science,
https://doi.org/10.1007/978-3-030-70176-5_5
87
Handling Uncertainty and Sensitivity
of ERA Acute Towards Input Parameters
Abstract Uncertainty evaluation and sensitivity testing of the functions and parameters used in ERA Acute serve two functions. ERA Acute is a deterministic model
which is sensitive to the range of values used for the parameters. Parameters have
inherent uncertainties as to what their true values are, and functions may have varying
strength of knowledge. The individual functions were tested with respect to their
sensitivity towards variation of the parameter values using both deterministic and
stochastic testing. Based on the testing, an uncertainty scoring system was used
to identify and prioritize the most important parameters for reducing uncertainty.
Recommendations for handling the uncertainty and securing comparability in spite
of uncertainty were set up as a conclusion of the studies.
Keywords Uncertainty testing · Uncertainty handling · Sensitivity testing ·
Parameter sensitivity scoring · Spearman correlation coefficient analysis · Partial
Rank Correlation Coefficient analysis
5.1 Sensitivity Testing and Uncertainty Handling
Every model has some inherent uncertainty. A model is a simplified mathematical
description which in a quote often accredited to Albert Einstein should be “as simple
as possible, but no simpler”. Both the simplifications and the detailing of sub-models
and equations carry with them uncertainties.
ERA Acute is a deterministic model where the structure of the functions and
their calculation sequences reflect how we believe that oil spills may harm the VECs
in the different compartments. The output of a deterministic model is completely
determined by the input parameters and structure of the model. A stochastic model
on the other hand, has inherent randomness in the model structure and will not
produce the same result, even given the same parameter value (Helton et al. 2006;
Marino et al. 2008).
The functions are mathematical descriptions of how we understand that the impact
and restoration will occur, and therefore also subject to uncertainty about the model
framework and its scientific soundness (see e.g. Gaber et al. 2009). If our assumptions
© The Author(s) 2021
C. Stephansen et al., Assessing Environmental Risk of Oil Spills with ERA Acute,
SpringerBriefs in Environmental Science,
https://doi.org/10.1007/978-3-030-70176-5_5
87