Sea Ice
- About
- Imprint
- Scenarios
- Arctic Marine Transportation by 2030
- Introduction
- Aim of this Study
- Key Factor Classification
- Definitions of Key Factors and Future Projections
- 1. Climate
- 2. Legal framework
- 3. Global Trade Dynamics – Global economic growth
- 4. Safety of other Routes
- 5. Socio-economic impact of global climate change
- 6. Oil Price
- 7. Major Arctic Shipping Disasters
- 8. Windows of Operation
- 9. Maritime Insurance Industry
- 10. Collaboration in resource extraction by China, Japan and Russia
- 11. Transit fees
- 12. Conflict between indigenous and commercial use
- 13. Arctic Enforcers
- 14. Energy sources for propulsion
- 15. New resource discovery
- 16. World Trade Patterns
- 17. Regulation in the Arctic
- Consistency matrix
- Scenarios
- Suggest Wild Cards
- Suggest Key Factors
- References
- Glossary
- Yakutat Community Energy Scenarios
- Introduction to Scenario-Management
- The Consistency and Robustness Analysis
- 1. Key Factors and their Future Projections
- 2. Assigning plausibility values to future projections
- 3. Projection Bundles
- 4. Assigning consistency values
- 5. Obtaining overall consistency values for the projection bundles
- 6. The combinatorial problem of the consistency analysis
- 7. The Robustness of a projection bundle
- Disruptive event analysis – Wild Cards
- ScenLab v1.7 Client download
- Arctic Marine Transportation by 2030
7. The Robustness of a projection bundle
To make the evaluation of projection bundles easier evolve:IT has developed the
robustness value. As mentioned before the robustness of a projection bundle is a
combination of the consistency value c, the plausibility value p and the number of
inconsistencies I. The robustness R is derived as follows
R = (p · c / (1 + I))^(1/2),
where the factors p and c are dynamically scaled to keep them in a range between
0 and 1 and the 1 in (1 + I) is added to avoid division by zero.
In contrast to c, p varies over several orders of magnitude. Thus, to obtain a well
balanced robustness value p is not scaled linearly to avoid too heavy dependence
of R on c.
Why introduce another value that just consists of values already in use? There
are two reasons: (i) as discussed above a good projection bundle not only needs to
have a high consistency value, a high plausibility value or a low number of partial
inconsistencies, but all three. Therefore the introduced robustness combines these
three components in a convenient way and takes care of the problem of the very
small plausibility values, i.e. it makes it easier for the user to interpret the quality
of a projection bundle.
(ii) The fitness function of the genetic algorithm (see ??) needs to evaluate the
quality of the projection bundles (individuals). It can do this by only considering
the consistency value, but this would neglect the other features of a good projection
bundle. By using the robustness as measure of fitness the genetic algorithm is able
to search for projection bundles with a favorable combination of good features.
- No comments yet.