Separating the Global Pattern of Externally Forced Sea Level Rise from Natural Variability in the Short Climate Record

Global mean sea level rapidly increased during the 20^th century, at a rate that doubled in the past few decades. Global satellite altimetry records, which have only been available since 1993, have additionally shown that the recent rise in sea level is neither spatially uniform nor linear in time. However, this change in sea level over such a short period likely convolves the externally forced climate signal with natural climate variability, and separating these is critical for coastal planners and policymakers to account for sea-level impacts on their communities. Previous studies have demonstrated that the “least damped (eigen)mode” (LDM) of a Linear Inverse Model (LIM) can effectively identify both sea surface temperature and sea level trend patterns in long records, even when they bear some similarity to patterns of natural climate variability, but that this approach becomes problematic for shorter records. In this study, we show that applying a Gram-Schmidt orthonormalization to the LIM’s eigenmodes adjusts the LDM so that it can identify the trend pattern even for record lengths of a few decades. We first test the technique by applying it to output from large ensembles of historical simulations made by two climate models, NCAR’s CESM2 and GFDL’s SPEAR: For record lengths as short as a few decades, our technique successfully identifies the forced response, as estimated by the ensemble mean, from any single ensemble member. Finally, we determine the forced sea level rise signal from observations, both on global and regional ocean scales as well as for coastal regions as measured by a gauge network, over the satellite observational era, and show how it differs from simple linear or quadratic trend estimates.