An improved Multimodel Approach for Global Sea Surface Temperature Forecasts

Monday, 15 December 2014
Rajeshwar Mehrotra1, Mohammad Zaved Kaiser Khan1 and Ashish Sharma2, (1)University of New South Wales, Sydney, Australia, (2)University of New South Wales, School of Civil and Environmental Engineering, Sydney, NSW, Australia
The concept of ensemble combinations for formulating improved climate forecasts has gained popularity in recent years. However, many climate models share similar physics or modeling processes, which may lead to similar (or strongly correlated) forecasts. Recent approaches for combining forecasts that take into consideration differences in model accuracy over space and time have either ignored the similarity of forecast among the models or followed a pairwise dynamic combination approach. Here we present a basis for combining model predictions, illustrating the improvements that can be achieved if procedures for factoring in inter-model dependence are utilised. The utility of the approach is demonstrated by combining sea surface temperature (SST) forecasts from five climate models over a period of 1960–2005. The variable of interest, the monthly global sea surface temperature anomalies (SSTA) at a 50´50 latitude–longitude grid, is predicted three months in advance to demonstrate the utility of the proposed algorithm. Results indicate that the proposed approach offers consistent and significant improvements for majority of grid points compared to the case where the dependence among the models is ignored. Therefore, the proposed approach of combining multiple models by taking into account the existing interdependence, provides an attractive alternative to obtain improved climate forecast. In addition, an approach to combine seasonal forecasts from multiple climate models with varying periods of availability is also demonstrated.