Impacts of the downscaling strategy on projected climate change signals in the Northwest-European Shelf

Our regionally coupled climate system model with an ocean component zoomed on central European seas (see model setup) has also been used to test various strategies on downscaling anthropogenic climate change for the Northwest-European Shelf (NWES) and in particular the North Sea. The application of regional ocean models with uncoupled or coupled atmospheric forcing as well as different domain sizes, resolutions and forcing global models has lead to substantial spreads in projected sea surface temperature (SST), seasonal stratification and Atlantic nutrient transport to the shelf (e.g. Holt et al., 2012; Gröger et al., 2013; Bülow et al., 2014; Mathis & Pohlmann, 2014; Tian et al., 2016). By modifying or excluding certain model components we mimic common downscaling strategies and assess their ability in simulating key parameters independently of the parent global climate projection.


In this overview we compare the parent global RCP8.5 simulation by MPI-ESM-LR and the reference downscaling by the regionally coupled ocean-atmosphere model MPIOM/REMO with two uncoupled experiments:

MPI-ESM-LR: parent global climate projection by MPI-ESM (CMIP5 version) to be downscaled for the NWES
MPIOM/REMO coupled: reference downscaling by the regionally coupled climate system model MPIOM/REMO
MPIOM/ECHAM6 uncoupled: uncoupled downscaling by MPIOM with the same zoomed grid configuration as for the coupled downscaling but driven by unprocessed atmospheric output of MPI-ESM-LR to test the influence of a higher grid resolution in the ocean
MPIOM/REMO uncoupled: uncoupled downscaling by MPIOM also with the zoomed grid but driven by atmospheric forcing from an atmosphere-only downscaling by REMO. Additionally, outside the North Sea a 3-d restoring of water temperature, salinity and nutrient concentration towards MPI-ESM-LR has been applied. This experiment mimics a stand-alone regional ocean model driven by high-resolution atmospheric forcing and oceanic boundary conditions taken from the parent global climate projection.


Projected SST for these four simulations are compared in Fig. 1. The ocean-only downscaling (Fig. 1c) merely reproduces the signal of the global model (Fig. 1a) because the signal is imposed at the sea surface by the atmospheric forcing. Even an intermediate step with a high-resolution atmosphere model (Fig. 1d) can only affect the signal near the coasts. Only coupled regional ocean models can deliver independent estimates (Fig. 1b).

Fig. 1: SST change signal (RCP8.5 2071-2100 minus 1971-2000) in the North Sea as projected by the global model MPI-ESM-LR (a), the regionally coupled model MPIOM/REMO (b), MPIOM driven by atmospheric forcing from the global model (c) and MPIOM with oceanic restoring outside the North Sea (black lines) towards the global model and atmospheric forcing from an atmosphere-only downscaling (d)

Projected salinity signals are shown in Fig. 2. The higher resolution in the ocean and the inclusion of tides lead to a more realistic circulation which improves the salinity estimate considerably (Fig. 2b), even in ocean-only simulations (Fig. 2c). In the traditionally used domain for regional North Sea models, artifacts at the boundary strongly influence the simulated state of the North Sea (Fig. 2d). Most published downscaling efforts, however, are using this approach. Nevertheless, the effect is independent of the use of coupled or uncoupled atmospheric forcing fields.

Fig. 2: As Fig. 1 but for sea surface salinity

Dependencies of changes in water temperature, salinity, and the ocean circulation on the downscaling strategy are transferred to changes in primary production and the effectivity of the shelf carbon pump. Locally, the downscaling strategy can even be decisive for the sign of the change signal.

Inter-model comparisons of climate projections regionalized for the NWES show large uncertainties in the downscaled change signals. Our experiments indicate that the uncertainties due to different downscaling strategies are of the same order of magnitude as the uncertainties due to different forcing global models and downscaling regional models. Moreover, the resulting change signals reveal to depend stronger on the downscaling strategy than on the model skills in simulating accurate present-day conditions.


More details and further experiments can be found in:

Mathis, M., A. Elizalde, U. Mikolajewicz (2018). Which complexity of regional climate system models is essential for downscaling anthropogenic climate change in the Northwest European Shelf? Climate Dynamics (link)


Bülow, K., C. Dieterich, A. Elizalde, M. Gröger, H. Heinrich, S. Hüttl-Kabus, B. Klein, B. Mayer, H. E. M. Meier, U. Mikolajewicz, N. Narayan, T. Pohlmann, G. Rosenhagen, S. Schimanke, D. Sein and J. Su: Comparison of 3 coupled models in the North Sea region under todays and future climate conditions. KLIWAS Schriftenreihe 27, 270 pp, Bundesanstalt für Gewäasserkunde, Koblenz, 2014.

Gröger, M., E. Maier-Reimer, U. Mikolajewicz, A. Moll and D. Sein: NW European shelf under climate warming: implications for open ocean-shelf exchange, primary production, and carbon absorption. Biogeosciences 10, 3767–3792,, 2013.

Holt, J., M. Butenschön, S. L. Wakelin, Y. Artioli and J. I. Allen: Oceanic controls on the primary production of the northwest European continental shelf: model experiments under recent past conditions and a potential future scenario. Biogeosciences 9 (1), 97–117,, 2012.

Mathis, M. and T. Pohlmann: Projection of physical conditions in the North Sea for the 21st century. Climate Research 61, 1–17,, 2014.

Tian, T., J. Su, F. Boberg, S. Yang and T. Schmitt: Estimating uncertainty caused by ocean heat transport to the North Sea: Experiments downscaling EC-EARTH. Climate Dynamics 46 (1), 99–110,, 2016.