Flux Adjustment on Seasonal-Scale Sea Surface 1 Temperature Drift in NICOCO 2

Abstract


Introduction
models (Cubasch et al., 1992;Voss et al., 1998). As the integration period was relatively 149 short in this study, our intention was to achieve a realistic seasonal SST evolution as the 150 ensemble mean by adjusting the surface fluxes, rather than adjusting the equilibrium state.

151
In this framework, each ensemble member represents a possible realization that is wobbling 152 around the ensemble mean seasonal evolution. To minimize artificial intervention, flux 153 adjustment was applied only to surface heat fluxes given to the ocean surface; hence, there 154 were no adjustments applied to momentum fluxes, freshwater fluxes, and surface heat 155 fluxes to the atmosphere. 156 In this study, the flux adjustment amount was designed to adjust the SST evolutions in (NICAM-THF) was obtained. The flux adjustment amount (hereafter F(x,y,t), where x,y,t 169 indicate longitude, latitude, and time, respectively) was determined as the difference 170 between COCO-THF and NICAM-THF. Note that the flux adjustment is distinct from the 171 nudging of SST toward a reference state. In the nudging, the F is evaluated during the  This simple method is advantageous because any arbitrary parameters, such as relaxation 177 constants, are unnecessary. Weaver et al. (1996) argued that some typical flux adjustment 178 methods, including the one employed in the present study, converge to the same flux 179 adjustment amount. Therefore, the results in the following sections are likely to be insensitive 180 to choice of the method, while there may be a better method which requires only smaller 181 amount of adjustment fluxes (Weaver et al. 1996). 182 To examine the importance of the temporal resolution in F(x,y,t), we conducted two sets of 183 flux-adjusted NICOCO integrations. In one integration, F(x,y,t) is averaged over the analysis 184 period beforehand and added as a temporary constant term, while retaining its spatial 185 variation. In the second experiment, F(x,y,t) was updated daily. well-known feature of ocean models with a quarter-degree resolution or coarser (Choi et al., 196 2002;Nakano et al., 2008). We confirmed that these biases are improved in uncoupled 197 COCO integrations with a 0.1° resolution, which will be described in a separate paper. The with the negative total surface heat flux for almost the entire period (black line in Fig. 7b).

291
The NICOCO free experiment (red line in Fig. 7a) also exhibited linear cooling, but the 292 negative slope was insufficient, resulting in a warming drift. In the NICOCO experiments with 293 constant flux adjustment (green line in Fig. 7a), the slope was modified to be more negative 294 owing to the negative F(x,y,t), which corresponds to the sign reversal in Fig. 2c. As expected, 295 the SST time series with a daily updated flux adjustment (orange line in Fig. 7a) was almost 296 similar to those of the uncoupled COCO.

297
Within the tropical domain (southern black boxes in Fig. 1), the SST evolution in the Although the NICOCO free experiment exhibited a steady warning throughout the integration 309 (red line in Fig. 7c), the SST evolution was modified to be nearly constant by the constant 310 flux adjustment (green line in Fig. 7c). The ensemble mean SST of the NICOCO experiments 311 with daily updated flux adjustment (orange line in Fig. 7c) was similar to that of the uncoupled 312 COCO, as the heat flux adjustment exhibits a rapid decrease to be strongly negative (orange 313 line in Fig. 7d).

314
Thus, it has been demonstrated that simple flux adjustment can successfully achieve