Constraining ocean diffusivity from the 8.2 ka event
Greenland ice-core data containing the 8.2 ka event are utilized by a model-data intercomparison within the Earth system model of intermediate complexity, CLIMBER-2.3 to investigate their potential for constraining the range of uncertain ocean diffusivity properties. Within a stochastic version of t...
Main Authors: | , , , |
---|---|
Format: | Article in Journal/Newspaper |
Language: | English |
Published: |
Berlin : Springer Verlag
2010
|
Subjects: | |
Online Access: | https://oa.tib.eu/renate/handle/123456789/5451 https://doi.org/10.34657/4080 |
Summary: | Greenland ice-core data containing the 8.2 ka event are utilized by a model-data intercomparison within the Earth system model of intermediate complexity, CLIMBER-2.3 to investigate their potential for constraining the range of uncertain ocean diffusivity properties. Within a stochastic version of the model (Bauer et al. in Paleoceanography 19:PA3014, 2004) it has been possible to mimic the pronounced cooling of the 8.2 ka event with relatively good accuracy considering the timing of the event in comparison to other modelling exercises. When statistically inferring from the 8.2 ka event on diffusivity the technical difficulty arises to establish the related likelihood numerically per realisation of the uncertain model parameters: while mainstream uncertainty analyses can assume a quasi-Gaussian shape of likelihood, with weather fluctuating around a long term mean, the 8.2 ka event as a highly nonlinear effect precludes such an a priori assumption. As a result of this study the Bayesian Analysis leads to a sharp single-mode likelihood for ocean diffusivity parameters within CLIMBER-2.3. Depending on the prior distribution this likelihood leads to a reduction of uncertainty in ocean diffusivity parameters (e. g. for flat prior uncertainty in the vertical ocean diffusivity parameter is reduced by factor 2). These results highlight the potential of paleo data to constrain uncertain system properties and strongly suggest to make further steps with more complex models and richer data sets to harvest this potential. © The Author(s) 2009. |
---|