Effects of Internal Climate Variability on Historical Ocean Wave Height Trend Assessment

This study assesses the effects of internal climate variability on wave height trend assessment using the d4PDF-WaveHs, the first single model initial-condition large ensemble (100-member) of significant wave height (H-s) simulations for the 1951-2010 period, which was produced using sea level pressure taken from Japan's d4PDF ensemble of historical climate simulations. Here, the focus is on assessing trends in annual mean and maximum H-s. The result is compared with other model simulations that account for other sources of uncertainty, and with modern wave reanalyses. It is shown that the trend variability arising from internal climate variability is comparable to the variability caused by other factors, such as climate model uncertainty. This study also assesses the likelihood to mis-estimate trends when using only one ensemble member and therefore one possible realization of the climate system. Using single member failed to detect the statistically significant notable positive trend shown in the ensemble in some areas of the Southern Ocean. The North Atlantic Ocean is found to have large internal climate variability, where different ensemble-members can show trends of the opposite signs for the same area. The minimum ensemble size necessary to effectively reduce the risk of mis-assessing H-s trends is estimated to be 10; but this largely depends on the specific wave statistic and the region of interest, with larger ensembles being required to assess extremes. The results also show that wave reanalyses are not suitable for analyzing H-s trends due to temporal inhomogeneities therein, in agreement with recent studies.

Automated summary

This study assesses the effects of internal climate variability on wave height trend assessment and finds that the trend variability arising from internal climate variability is comparable to the variability caused by other factors. Using a single ensemble member may lead to mis-estimation of trends, while larger ensemble sizes can reduce this risk. Furthermore, wave reanalyses are not suitable for analyzing wave height trends.