Figure Captions
Figure 1. Posterior predictive distributions of time series data on the winter North Atlantic Oscillation (NAO) index. The units are based on the difference of normalized sea level pressure between Lisbon, Portugal and Stykkisholmur/Reykjavík, Iceland. Differently shaded series of data represent different regimes. The shaded area represents the 68% central probability interval (CPI) of the posterior predictive distribution; thus, it includes uncertainty about the mean and the variance. Horizontal lines in each shaded region represents the mean. M is the minimum regime length (in years) and λ is the ‘hazard’ rate in the change point prior, i.e., the expectation of the frequency of a regime shift (a λ value of 20 would imply an expectation that regime shifts occur every 20 years).
Figure 2.  Posterior predictive distribution of juvenile Atlantic cod catch rate (number of juvenile cod per beach-seine haul). See the caption for Figure 1 for explanations of the shaded regions of the data.
Figure 3.  Posterior predictive distributions of data on the abundance of Calanus finmarchicus  as estimated from the Continuous Plankton Recorder Survey (area C1). Data have been pooled for the months of March through August. Keeping  constant at 10 yr, model output is shown for λ values of (a) 10, (b) 20, (c) 25, and (d) 50 years. See the caption for Figure 1 for explanations of the shaded regions of the data.
Figure 4.  Posterior predictive distributions of data on sea surface water temperatures measured at Flødevigen Research Station for the months of January through December at a hazard rate (λ) of 25 years ( = 10 years). See the caption for Figure 1 for explanations of the shaded regions of the data.
Figure 5.  Posterior predictive distributions of estimates of instantaneous rate of fishing mortality (F ) for cod aged 2-4 yr in the North Sea. Hazard rate (λ) is 10, 20, 25, and 50 years;  is held constant at 6 years. See the caption for Figure 1 for explanations of the shaded regions of the data.
Figure 6.  The timing of regime shifts in cod catch rate, NAO, water temperature, and abundance of Calanus finmarchicus . Each arrow identifies the beginning of a regime. The direction of the arrow indicates the change in the mean value of the data following each regime shift. The timing of regime shifts is shown in relation to changes in fishing mortality and spawning stock biomass of North Sea cod (each of which is expressed relative to its respective limit reference point; the limit fishing mortality is 0.54, the limit spawning stock biomass is 107,000 tonnes). (a) This panel identifies regime shifts in climate and environmental indices that do not appear to have influenced regime shifts in cod catch rate (i.e., were not followed within 5 years by a cod regime shift). The water temperature regime in 1988 is for January through June; the regime shift in 2005 is for November temperatures. (b) This panel identifies regime shifts in climate and environmental indices that do appear to have influenced regime shifts in cod catch rate (i.e., were followed within 5 years by a cod regime shift). The water temperature regime shifts are for the months of August (1994), September (1996), July (1997), and October (1999).