| dc.description.abstract | One of the major contributors to sea level rise along the Dutch coast during the 21st century is Ocean Dynamic Sea Level (ODSL). ODSL is defined as the sea level anomaly due to ocean currents, wind stresses and local thermosteric and halosteric effects. Along the Dutch coast, climate models project an ODSL rise between 0 - 35 cm at the end of the century, depending on the emission scenario. Hence, this accounts for a large part of the total projected sea level rise of 30 - 120 cm along the Dutch coast. It is important that climate models are able to model ODSL correctly. However, the current state-of-the-art climate models from CMIP5 and CMIP6 show a large spread in ODSL projections and ODSL rise at the Dutch coast increased significantly between CMIP5 and CMIP6. One hypothesis for the latter is the larger increase in global mean temperature in CMIP6. This increase is larger than the assessed ranges provided in the Sixth Assessment Report (AR6) by the Intergovernmental Panel on Climate Change. In this study, we aim to improve the quality of ODSL projections from CMIP5 and CMIP6 by better understanding the processes that influence ODSL change along the Dutch coast. First, we use linear regression models to identify the reason for a model’s ODSL change. The processes we consider as explanatory variables for ODSL change are global surface air temperature (GSAT), global mean thermosteric sea level (GMTSL), and the Atlantic meridional overturning circulation (AMOC) strength. Using only GSAT as a predictor variable works for the CMIP5 ensemble but does not explain all long-term changes for CMIP6. Including an additional predictor variable improves the model. More specifically, we find that the model using predictor variables GSAT and AMOC performs best at predicting ODSL change at the Dutch coast for both CMIP5 and CMIP6. For most individual models, we find that an increase in GSAT, and a weakening of the AMOC, relate to an increase in ODSL along the Dutch coast. The regression analysis results are combined with Monte Carlo sampling to generate probabilistic ensembles of ODSL projections consistent with the AR6 assessed ranges of GSAT and GMTSL. This method enables us to correct the ODSL change for the high temperature bias in CMIP6 models. We find that the effect of GSAT is too small to explain the difference in ODSL between CMIP5 and CMIP6. However, we see that the sensitivity to GMTSL and AMOC increased in CMIP6 which could point to a difference in model dynamics between CMIP5 and CMIP6. Furthermore, we find that the location of deep convection is important for ODSL along the Dutch coast. We see that models that show a deep mixed layer in the Greenland Sea for the period 1975 - 2004, project a larger rise in ODSL at the Dutch coast for both ensembles. | |
