2.4. Multi-model ensemble construction

Fig. 1 provides a summary of the overall methodological framework followed in the construction of the multi-model LF ensemble in this study. Ensemble construction was essentially performed in two steps:

The three LF models were individually calibrated and validated for each given data set, each relying on its own model structure and implementation;

The three sets of single-model simulations were then combined using a mean-squared error performance-based weighting scheme as outlined in part C of the SI (Johnson and Swinbank, 2009).