Global experiments

Back to Home.

In this set of experiments, I modified the value of certain ocean coefficients globally (e.g. horizontal eddy viscosity, bottom drag coefficient) and compared the results of these sensitivity experiments to a control run. The control run starts from the year 2014 of the CMIP6 EC-Earth3-Veg r1 historical member (T255-ORCA1), uses a 2000-year constant atmospheric forcing and is 200 years long. The sensitivity experiments all start from the year 2130 of the control run (year 117 if we consider the starting year to be 1) and are 50 years long.

Here are the different model experiments with the corresponding color in the plots below:

t613 (red solid curve, run by K. Wyser) is the CMIP6 EC-Earth3-Veg r1 historical member covering 1850-2014 (only the period 1950-2014 is shown here);
D000 (blue solid curve) is the control run starting from year 2014 of the t613 experiment with 2000-year constant atmospheric forcing; this control run is 200 years long; all the sensitivity experiments start in year 2130 of this control run (year 117 if we consider the starting year to be 1) and are 50 years long;
D002 (green dashed curve) starts in year 2130 of the control run (D000); the only change compared to the control run is the horizontal eddy viscosity, which is 5 times lower;
D003 (red dashed curve) is similar to D002, except that the horizontal eddy viscosity is 50 times lower than the control run;
D004 (gray dashed curve) starts in year 2130 of the control run; the only change compared to the control run is the bottom drag coefficient, which is 2 times higher;
D005 (orange dashed curve) is similar to D004, except that the bottom drag coefficient is 20 times higher than the control run;
D007 (magenta dashed curve) starts in year 2130 of the control run and is only 10 years long; the only change compared to the control run is the vertical eddy diffusivity, which is 2 times higher;
D008 (cyan dashed curve) starts in year 2130 of the control run; the only change compared to the control run is the horizontal eddy diffusivity, which is 2 times higher;
D009 (brown dashed curve) starts in year 2130 of the control run; the only change compared to the control run is the surface minimum value of turbulent kinetic energy (TKE), which is 2 times higher;
D010 (yellow dashed curve) starts in year 2130 of the control run; this experiment combines the changes of experiments D004 (bottom drag x 2), D008 (horizontal diffusivity x 2) and D009 (min. value of TKE x 2).

Results show that:

decreasing the horizontal eddy viscosity leads to reduced mean poleward Atlantic ocean heat transport (OHT) at all latitudes (Fig. 1), although the OHT is increased during some specific years when looking at specific latitudes, e.g. 60N (Fig. 2A) and 70N (Fig. 2B);
decreasing the horizontal eddy viscosity leads to an increase in Arctic sea-ice area and volume in the beginning followed by a decrease afterwards in D002, while these quantities are generally decreased in D003 (Figs. 3 and 4);
increasing the bottom drag coefficient leads to an increase in the Atlantic OHT (Figs. 1 and 2), and leads to different results in terms of Arctic sea ice depending on the magnitude of the increase (decrease in D004 and increase in D005) (Figs. 3 and 4);
increasing the vertical eddy diffusivity does not lead to any change in OHT and sea-ice area/volume (Figs. 3 and 4), which explains why we stopped this experiment after 10 years;
increasing the horizontal eddy diffusivity leads to an increase in Atlantic OHT (except at 70N) (Figs. 1 and 2), and to a marked decrease in Arctic sea-ice area and volume (Figs. 3 and 4);
increasing the surface minimum value of the TKE leads to an increase in Atlantic OHT (except at 70N) (Figs. 1 and 2), and generally to a decrease in Arctic sea-ice area and volume (Figs. 3 and 4);
combining an increase in bottom drag coefficient, horizontal eddy diffusivity and surface minimum value of TKE leads to a slight decrease of Atlantic OHT (Figs. 1 and 2) and a clear decrease in Arctic sea-ice area and volume (Figs. 3 and 4);
the model underestimates the Atlantic OHT compared to the Trenberth and Fasullo (2017) estimate, but is overall quite good (except at 40-45N, where there is a clear underestimation) (compare solid red and black curves in Fig. 1);
the model is relatively good at simulating Arctic sea-ice area compared to satellite observations (compare solid red and black curves in Fig. 3) and overestimates Arctic sea-ice volume by about 5000 km^3 compared to PIOMAS reanalysis (compare solid red and black curves in Fig. 4);
the modeled spatial distribution of Arctic sea-ice concentration is relatively good compared to satellite observations (Fig. 5).

These results show that the effect of these sensitivity experiments is not entirely clear and that the role of internal variability might be high. Further investigation is under way.

Fig. 1: Latitudinal transect of mean Atlantic OHT averaged over 50 years (except 2000-2014 for the CMIP6 r1 member to compare to TF2017); OHT estimates from Trenberth and Fasullo (TF2017) and hydrographic measurements (as in Grist et al., 2018) are plotted for reference; the number in brackets is the difference in mean OHT between the experiment and the control (CTRL)

Fig. 2: Time series of annual mean Atlantic OHT at 60N (A) and 70N (B); the number in brackets is the difference in mean OHT between the experiment and the control (CTRL)

Fig. 3: Time series of Arctic sea-ice area in March (A) and September (B)

Fig. 4: Time series of Arctic sea-ice volume in March (A) and September (B)

Fig. 5: Maps of Arctic sea-ice concentration in March (1st row) and September (2nd row) for the control run D000 (1st column), the CMIP6 r1 member t613 (2nd column) and OSI SAF satellite observations (3rd column)

Important note: These figures were produced by D. Docquier and constitute preliminary results (not published in papers).