Statistical downscaling of sea levels: application of multi-criteria analysis for selection of global climate models

Sea level rise is one of the serious after-maths of global warming on the hydrosphere. The scientific community often depends on global climate models (GCMs) for projection of future sea levels. Numerous GCMs are available; thus, selecting the most appropriate GCM/GCMs is a critical task to be performed prior to downscaling. In this study, multi-criteria decision-making (MCDM) techniques, namely, Preference Ranking Organisation Method of Enrichment Evaluation (PROMETHEE-II), Elimination Et Choice Translating Reality (ELECTRE-II), and compromise programming, were used to identify appropriate GCMs whose projections can be used to downscale sea level projections at Ernakulam, Kerala, India. Support vector machine was employed to statistically downscale the sea level projections from the projections of GCMs. Five statistical metrics, namely, correlation coefficient (r), normalized root mean square error, absolute normalized average bias, mean absolute relative error, and skill score, were adopted in this study as the performance criteria. The weightage of each criterion was computed using the entropy method. Six GCMs (GISS-E2-H, CanESM2, ACCES S1-0, CNRM-CM5, GFDL-CM3, and CMCC-CM) were considered for the analysis based on the availability of predictors. GISS-E2-H, CanESM2, and ACCESS 1-0 occupied the first three positions respectively in all three MCDM techniques.

Automated summary

In this study, multi-criteria decision-making techniques and support vector machine were used to select suitable GCMs for downscaling sea level projections. Five statistical metrics were adopted as performance criteria. GISS-E2-H, CanESM2, and ACCESS 1-0 were found to be the most appropriate GCMs for sea level projections in Ernakulam, Kerala, India.