AM 
N. Spangehl et al.: Intercomparing the quality of recent reanalyses for offshore wind farm planning 
Wind speed, FINO1 
45 
10 
35 
30 
25 
20 
15 
10 
(a) 
Fa 
Dataset 
COSMO-REAG, 100 m 
CERRA-fc, 100 m 
CERRA-EDA, 100 m 
ERA5, 100 m 
HoKliSim-De, 100 m - 
NEWA, 100m —- 
FINO1, 102 m 160 
18 
Bi 
z 
I 
©, N 
GC 7 
8 a 
De 2 
Q 
Dr 
7, > 
Q a 
5% 5 
time [day] 
„ 
X 
Q 
% 
sr 
Do, en 
Wind speed, FINO1, 2013-10-28 
n 
45 
40 
{b) 
35 
30 
25 
Pf. 
A 
Dataset 
COSMO-REAG, 100 m —— 
CERRA-fc, 100 m 
CERRA-EDA, 100 m 
ERA5S, 100m — 
HoKliSim-De, 100 m ---- 
NEWA, 100m — - 
FINO1, 102 m 190 
ME 
u 
a zz A 
20 
15 
VL DA 
0 
% 
J HELL Sn 
Ob A907 9a IT RDI 
time [hour] 
Figure 4. Hourly wind speed at 100 m for model based products at FINO1 grid points during windstorm CHRISTIAN. Results are shown 
for (blue solid curve) COSMO-REAG, (purple solid curve) CERRA-Tfc, (purple solid thin curves) CERRA-EDA, (cyan solid curve) ERAS, 
(cyan dashed curve) HoKliSim-De and (red solid curve) NEWA. Results obtained from FINO1 measurements at 102 m (10 min averages) 
are additionally shown (bold black dashed curve). The time period is (a) 25-31 October 2013 and (b) 28 October 2013. 
3.2.1 Distribution of biases 
For the ten-year time period comparison against the Scat- 
;erometer and Model (e5) (CMEMS) dataset the median 
as is nearly zero for ERA5 and COSMO-REAG6 (Fig. 5, 
cop left panel). In 50% of the cases (lower to upper quar- 
üle range indicated by the blue boxes) the bias is within 
—0.2 to 0.2ms”!. Larger biases are found as indicated by 
he whiskers. 
The maximum negative bias of ERA5S is —3.1 ms”! in 
contrast to the maximum positive bias of 0.9m s71! (Table 4). 
A systematic shift towards positive bias values is found for 
CERRA as indicated by the median bias of 0.2ms7! and 
the lower to upper quartile ranging from 0.05 to 0.35 ms! 
Adv. Sci. Res... 20. 109-128, 202: 
Compared to ERA5 a similar maximum negative bias of 
-3.1 ms”! is found compared to a larger maximum posi- 
tive bias of 2.0ms7!. Larger maximum negative and pos- 
itive biases of —3.7 and 2.2ms”7! are found for COSMO- 
REAG6 (Table 4). Contrary to CERRA the bias distribution of 
HoKliSim-De shows a shift towards negative values with the 
median bias of about —0.2 ms”! (Fig. 5, top left panel). 
A slight shift of the bias distribution towards positive 
values is seen for ERAS5 and COSMO-REA6 when using 
the HOAPS dataset as reference (Fig. 5, right panels). For 
CERRA the shift towards positive bias values seen earlier 
is confirmed. HoKliSim-De shows a marginal shift towards 
negative bias values. Results confirm that HOAPS shows 
https://doi.org/10.5194/asr-20-109-2023