A deterministic evaluation of the performance of consolidation strategies for multi-method ensemble forecasts to predict monthly SST over the Pacific is presented. The study is based on ensembles from 8 coupled ocean-atmosphere models and one empirical model. The focus is on comparing constrained linear regression approaches to the simple multi-methods average (MM). Given the limited sample to train the regression coefficients, which is typical in current dynamical climate prediction systems, a strategy to gather information from surrounding grid points to the point of analysis, and from each ensemble member is used. Results show that constrained regression methods produce superior forecasts than the MM for most of the tropical Pacific and in some parts of the extratropics.