4DVAR HF radar assimilation South of Africa 1

The oceanic circulation south of Africa is characterised by a complex 6 dynamics with a strong variability due to the presence of the Agulhas current and 7 numerous eddies. The area of interest of this paper, is also the location of several 8 natural gas ﬁelds under seaﬂoor which are targeted for drilling and exploitation. 9 The complex and powerful ocean currents induces signiﬁcant issues for ship opera- 10 tions at the surface as well as under the surface for deep sea operations. Therefore, 11 the knowledge of the state of the currents and the ability to forecast them in a real- 12 istic manners could greatly enforce the safety of various marine operation. Follow- 13 ing this objective an array of HF radar systems were deployed to allow a detailed 14 knowledge of the Agulhas currents and its associated eddy activity. It is shown 15 in this study that 4DVAR assimilation of HF radar allow to represent the surface 16 circulation more realistically. Two kind of experiments have been performed, a 17 one month analysis and two days forecast. The one month 4DVAR experiment 18 have been compared to geostrophic currents issued from altimeters and highlight 19 an important improvement of the geostrophic currents. Furthermore despite the 20 restricted size of the area covered with HF radar, we show that the solution is 21 improved almost in the whole domain, mainly upstream and downstream of the 22 HF radar’s covered area. We also show that while beneﬁts of the assimilation on 23 the surface current intensity is signiﬁcantly reduced in the ﬁrst 6 hours of the 24 forecast, the correction in direction persists after 48 hours. 25


27
The oceanic circulation south of Africa is characterised by a complex dynamics 28 with a strong variability due to the presence of the Agulhas current and numerous 29 mesoscale eddies from the Mozambique Channel (Penven et al., 2006;Halo et al., 30 availability, stratification and primary productivity in the eastern Agulhas Bank. 48 It as also been shown by Meyer and Niekerk (2016) that implementing an ocean 49 current power plant in this region would outperforms onshore wind power plants 50 and could increase the load carrying capacity of the country.

51
The area of interest of this paper, represented on Figure 1 is also the location 52 of several natural gas fields under seafloor which are targeted for drilling and ex-53 ploitation. The complex and powerful ocean currents induces significant issues for 54 ship operations at the surface as well as under the surface for deep sea operations. 55 Strong ocean currents can also modify the height and direction of ocean waves, 56 causing dangerous sea states (Quilfen et al., 2018). The risk of extreme waves is an 57 important hazard for the shipping activity and off shore industry when crossing 58 the main current systems. Therefore, knowledge of the currents state and the abil-59 ity to forecast it in a realistic manners could greatly enforce the safety of various 60 marine operations. 61 Following this objective an array of HF radar was deployed along the coast to 62 allow a detailed knowledge of the Agulhas currents and its associated eddy activity.

63
The purpose of the present document is to present and evaluate the impact of the 64 4DVAR assimilation of those radar data on ocean model simulation and forecast 65 of the sea surface currents.

66
Data used for assimilation and validation are described in the following section.

67
The model setup and the assimilation procedure are described in a third section 68 while results are presented in section four and further discuss in the conclusion. To monitor the variability of the Agulhas currents during offshore operations, three 71 HF radar were installed on the south coast of South Africa. The location of the 72 radar system (black square) and the averaged area of measurement during April 73 2020 is represented on Figure 2. After a first treatment of the radar data using 74 manufacturer software, radial velocities are combined on a Cartesian grid at 6km 75 resolution using the method describe by Barth et al. (2010) and made available 76 every 30 minutes.

77
Altimeters data used were generated by a processing system including data  The ocean circulation model used in this study was the Regional Oceanic Modeling 89 System described in detail in Shchepetkin andMcWilliams (2003, 2005 Fig. 2 Black squares represent the emplacement of the HF radar installed to monitor the area delimited by the black contour (cf Figure 1). The colored area represents the intensity of the averaged current measured by the radar during April 2020 and the white arrows are representative of the averaged direction of the surface currents during the same period.
horizontal dissipation/diffusion while on the vertical a GLS scheme is used to de- were assimilated using 1-day assimilation windows. The 4D-Var analysis produced at the end of each day was used as initial condition for the next assimilation cycle.

160
In the previous section it has been shown that assimilating HF radar currents 161 allows to improve the geostrophic circulation when compared to satellite derived 162 velocities. Nonetheless, those altimeter data are at low resolution (25km) with daily 163 data only, while HF radar are available at 6km resolution every 30 minutes. Since HF radar currents are assimilated in the model and therefore cannot be used for 165 further validations, some forecast have been made starting from assimilated initial 166 condition and FREE condition. This allow to validate the forecast against the HF 167 radar data and explore the benefits of the assimilation on smaller scales and on 168 the forecast capabilities of the current configuration.

169
Two 48 hours forecast initiated from the 20 of April 00H00 were performed.

170
One forecast were initiated from FREE run and the second one from the 4DVAR.

171
RMSE time series of both forecast against HF radar data are presented on Figure   172 6. They show a strong improvement in intensity (top panel) and direction (bottom 173 panel) when the forecast is initiated from 4DVAR. While the reduction of RMSE 174 of the surface current intensity tend to decrease from 50% to 15% during the 48 hr 175 forecast, the RMSE of the surface current direction is reduced from 73% to 50% 176 during this same period. Therefore the main improvement of the forecast is related 177 to the currents direction. Figure 7 represents the maps of RMSE for surface current 178 intensity (Figure 7a,b,c) and direction (Figure 7d,e,f) along the 48 hours of the 179 forecast. Although it confirms that most of the forecast improvement is related to 180 the current direction it also shows that current intensity and direction were signif-181 icantly improved in the center of the area covered by the HF radar measurement.  radar. This southward shift, seen in the FREE forecast is therefore an artefact of 192 the model which can be corrected by using an assimilated initial condition. 194 In this study the benefits of the 4DVAR assimilation of surface currents issued 195 from HF radar in one of the most highly dynamic region of the world is presented.

196
While the intense dynamics of the region make difficult for most of the numerical 197 oceanic models to realistically reproduce the position and intensity of the Agulhas 198 current and associated eddies, it has been shown that a 4DVAR assimilation of