Figure 1
From: Structural decomposition of decadal climate prediction errors: A Bayesian approach
Structural decomposition and Bayesian analysis of decadal climate prediction errors. (a) (Top panel) grey lines: empirical hindcast errors in monthly mean spatially-averaged SST in the Angola-Benguela front region; colored lines/shading: associated posterior marginal estimates of systematic error components including drift/bias (red), annual seasonal bias (green) and semiannual seasonal bias (blue); (mid panel) posterior estimate of the stochastic trend; (bottom panel) posterior residual errors. Lines are posterior marginal medians, dark/light shading are 5th–95th and 1st–99th percentile ranges. (b) (Left) trace plots of the Monte Carlo Markov chains for the three model parameters; (right) posterior (black) versus prior (grey) distributions, shown as logarithms for better visualization. The probability distributions are plotted using 25 bins.
