Rn

V. Maurer et al.: Evaluation of coupled and uncoupled simulations

reflect the bias of NEMO against the Copernicus observaons
 with too low SSTs in winter and too high SSTs in the
Atlantic in summer. The differences between ROAM-NBS
and ICON-CLM (Fig. 4b) over the ocean are identical to the
ROAM-NBS/ERAS5 differences by design, as ICON-CLM
uses ERA5 SSTs as a lower boundary condition over the
ocean. Over land, the differences are small (i.e. < +-0.25K)
compared to the biases against ERAS5.
As mentioned in Sect. 2.1, identical parameter settings
as in ICON-CLM-UDAG were used in the ICON-CLM and
ROAM-NBS simulations, apart from the increased minimum
diffusion coefficient for heat. The higher minimum diffusion
 coefficient increases vertical mixing, especially in stable
 boundary layers, causing downward mixing of warmer
air and an increase in near-surface temperatures. This adaptation
 was made to compensate for the wintertime SST cold
bias of NEMO-NBS, which can also influence the air temperatures
 directly downstream of the ocean regions. However,
 from the evaluation of air temperature at 2m (tas)
against the E-OBS dataset (Cornes et al., 2018), which is
available over land only, it is obvious that too low diurnal
maximum temperatures (tasmax) are still prevailing (Figs. 5c
and A2b) in ROAM-NBS. At the same time, minimum temperatures
 (tasmin) have a small warm bias compared to E-OBS
 (Figs. 5b and A2a). For the diurnal average tas, the
[{CON-CLM-UDAG simulation is up to 0.7K too cool in
winter for the entire E-OBS domain (Fig. 5a). The diurnal
ninimum (tasmin) in ICON-CLM-UDAG is too high in almost
 all regions and all months, with maximum biases of
about 0.45 K in summer for the whole E-OBS domain average
 (Fig. 5b). At the same time, the diurnal maximum is up
to 1.2 K too low in ICON-CLM-UDAG (Fig. 5c). Thus, the
amplitude of the diurnal cycle is too low on average.
The increased minimum diffusion in ICON-CLM decreases
 the diurnal mean temperature bias in winter (DJP),
while it is slightly increased from May to September. Accordingly,
 minimum and maximum temperatures are increased,
 which means that the already positive minimum
temperature bias is getting larger. The largest positive minimum
 temperature bias of 0.8 K can be observed in February
in Scandinavia.
To sum it up, the wintertime cold bias of ICON-CLM-UDAG
 is reduced by the adapted tuning parameter in ICON-CLM
 by about 0.2 K, at the cost of an increased warm bias of
che diurnal minimum by about the same order of magnitude.
However, the absolute values of the negative tasmax bias are
still larger than those of the positive tasmin bias, apart from
July and August.
Comparing ROAM-NBS and ICON-CLM, we can say that
che positive diurnal mean and minimum bias are slightly
higher in ROAM-NBS than in ICON-CLM, with a very simılar
 bias for the diurnal maximum. As expected, the winterüme
 SST cold bias of the ocean is also slightly reflected in
the temperatures over land.

3.2.2 Precipitation and flux differences

To give an overview of precipitation and fluxes over the
ocean, for which good measurement products are not available
 or are not at a sufficient spatial resolution to adequately
evaluate the NBS region, the differences between the coupled
and uncoupled simulations are analyzed here. For completeness,
 seasonal mean bias maps are provided for precipitation
and surface net longwave radiation in Fig. Al in the Appendix.
 As the SST is prescribed for ICON-CLM, the SST
differences between ROAM-NBS and ICON-CLM also reflect
 the bias against observations, which does not directly
mean that the precipitation and flux differences between both
simulations, as shown here, reflect biases. However, they can
give a good indication of the reaction of the coupled model
to SST biases.
For precipitation, the differences between ROAM-NBS
and ICON-CLM reflect the differences in evaporation or latent
 heat flux (hfls) over the ocean (Fig. 6): Precipitation
over the ocean is higher by up to 0.15 mmd”7! in ROAM-NBS
 in regions where evaporation is higher, such as over
the Atlantic ocean and North Sea in summer or the Baltic
Sea in winter. Vice versa, precipitation is lower by up to
—0.15mmd”! when evaporation is lower, as over the Atlantic
 ocean and the North Sea in summer. However, these
differences are small compared to the absolute precipitation
biases over land, which exceed +0.6 mm d7! in various regions
 (Fig. Alb and d; note the adapted color scale compared
to Fig. 6).
The mean values of precipitation and evaporation differences
 between ROAM-NBS and ICON-CLM are also
provided for different ocean domains in Fig. 7, together
with the differences for sensible heat flux (hfss), 10 m wind
speed (sfc Wind) and surface temperature (Tsfc). Tsfc over
the (ice-free) ocean in ICON is equivalent to the SST. The
absolute differences clearly show that the sign of the Tsfc
difference determines the sign of the flux, sfcWind and pr
differences. For example, for a lower Tsfc of ROAM-NBS as
in the Open Atlantic, Atlantic South and North Sea boxes in
DJF and the Baltic Sea box in JJA, also hflis, hfss, sfc Wind,
and pr are lower than in ICON-CLM. Accordingly, all differences
 in the Open Atlantic, Atlantic South and North Sea
boxes in JJA are positive. The only exception is the Baltic
Sea in DJF, where the spatially averaged Tsfc differences
are very small (below —0.05 K). The absolute area-averages
for ICON-CLM (given as numbers in Fig. 7) show that the
fluxes, and therefore, also the flux differences become higher
over a warmer ocean surface. Looking at the Atlantic South
and the North Sea in DJIF, for example, the Tsfc difference
is very similar in both regions (—0.45 and —0.43 K, respectively),
 but the heat flux difference is larger for the Atlantic
South (— 10.57 W m”? for the sum of hfls and hfss compared
to —9.13 W m”? for the North Sea). But the Atlantic South
is, on average, warmer than the North Sea (286.1 K compared
to 280.8 K). However, the absolute temperatures of the ocean

Geosci. Model Dev... 19. 543578. 2026

https://doi.ore/10.5194/smd-19-543-2026