Clustering as a Tool for Understanding Ocean Regions of Sea Level Covariability and Volume Redistribution

Climate modes are typically defined as large-scale patterns of coupled atmospheric and oceanic variability occurring over interannual to multi-decadal time scales. Identifying ocean regions of covariability allows for an improved understanding of these climate modes and their underlying dynamical processes. Here we used hierarchical clustering on sea level anomalies across observational datasets and dynamical model output. These clusters can be used to assess the fidelity of climate models, specifically with regard to large-scale ocean dynamics. Using ~25 years of monthly sea level anomalies from satellite altimetry data and NOAA GFDL’s coupled (CM4) and ocean-only (OM4) models we implement our hierarchical clustering algorithm and explore model-observational similarities. With this tool, we can consider patterns of ocean volume redistribution and we can relate inner-cluster volume redistribution to climate mode indexes to gain a physical understanding of our hierarchical clusters. These results point towards some spatial discrepancies of climate modes in coupled models, and similarities in the ocean-only models. Sensitivities of this clustering method are also considered and ongoing work explores the applicability of alternative clustering approaches.

Further work will apply clustering techniques to subsurface ocean model fields, revealing patterns of variability that are typically under-sampled observationally. These methods can also be applied to non-sealevel related data and can be used to find other regions of co-variability. This assessment and associated tool can inform model development and ultimately lead to improvements in climate simulations.