Ocean Initialization Issues for Transient Climate Change Ensemble
Ocean Initialization Issues for Transient Climate Change Ensemble Experiments Frank Bryan & Gokhan Danabasoglu NCAR
Assumptions & Biases • Our interest in these experiments include the simulation of the climate through the full 20 th century as well as future projections. • By initialization we mean designing a sampling strategy to extract initial states from a pre-industrial control integration of the coupled system. • My comments reflect a personal research interest in the long-term behavior of thermohaline circulation.
Recommendations for Transient Experiment Initialization Independent of Ensemble Size • Plan and budget for very long (1000+ year) pre-industrial control integrations sufficient to establish a period of stationary statistics that completely overlaps all planned transient integrations • Complete control integration(s) and characterize the statistics of their variability prior to starting climate change experiments.
Long-Term Behavior of Control MOC Nakashiki et al (Ocean Modeling, submitted)
Characterize the Variability of the MOC 1990 Controls CCSM 3 -T 85 x 1 CCSM 3 -T 42 x 1 CCSM 2. 2 -T 42 x 1 CCSM 3 -T 31 x 3
1990 T 85 x 1 Control MOC EOFs 17 -25 years G. Danabasoglu
Initializing Transient Experiments 1990 Control 1% CO 2
20 th Century Ensemble Members and P-I Control PI CONTROL
Ensemble Spread in Heat Storage Gent et al (J. Clim. , 2006)
Further Recommendations • Cross-WG research program over next 1. 5 years to better characterize the global manifestations and physics of the MOC decadal mode. – What are the dynamical mechanisms of the decadal oscillations of the MOC? – Why does it appear to depend on model resolution? – Does the amplitude of the oscillation depend on the mean state? – What are the regional and global impacts of the variability? – What are the effects on predictability?
A Resource Conservative Initialization Strategy • Assuming dominant decadal mode exists: – Sample within a single oscillation period (20 to 50 years) of pre-industrial control. – Minimum 4 -5 members for 20 th century experiments – Consider dropping members for larger amplitude forcing and/or stabilization scenarios. • Samples phase space • Minimizes length of control run required
- Slides: 11