1. COMBAT: COMBination of AlTimetry and HF radar
observations for coastal data assimilation
Caballero, Ainhoa (1)*; Ayoub, Nadia (2); Mulet, Sandrine (3); Rubio, Anna (1); Davila, Xabier (2); Dufau, Claire (3); Mader, Julien (1); Marina Chifflet (1)
1: AZTI, Spain; 2: LEGOS, France; 3: CLS, France

2. Mean Dynamic Topography of coastal areas
South-eastern Bay of Biscay
Mean Dynamic Topography of coastal areas
CNES-CLS13 MDT
MDT is a key reference surface needed for estimating the Absolute Dynamic Topography and for the assimilation of altimetry data in ocean modelling systems.
Since 2004, global MDT solutions combining altimeter, gravimeter and in-situ data have been routinely calculated.
These MDT solutions provide an accurate estimate of the MDT at spatial scales larger than 100 km.
In coastal areas, the global MDT solutions calculated are often less accurate than in the open ocean, because:
In-situ measurements are sparse.
Errors on altimetry measurements are higher due to the limitations of altimetry.
HFR

3. LAND-BASED HF RADARS
LAND-BASED HF RADARS
Measure total surface currents
The spatial and temporal scales resolved, depend mainly on the operation frequency and available bandwidth
High Frequency radar (HFR) is a land-based remote sensing instrument offering an unique insight to coastal ocean variability, by providing synoptic, high frequency and high resolution data at the ocean atmosphere interface. HFR, utilized for oceanographic purposes, typically operate in the band between 8 and 37 MHz corresponding to wavelengths of 37 to 8 m, and thus providing measure of ocean current at the typical scales shown in the table. The system at the SE Bay of Biscay operates at ~4.5 MHz, providing good quality data up to 150 km from the coast (the maximum range is higher but this is the mean range where we obtain good quality currents) at 1 hour temporal resolution and 5km range resolution (total currents are mapped in a 5km regular grid).
Typical scales
Temporal resolution
Spatial resolution
Max. range
Integration depth
10 min - 3 hours
100 m - 12 km km
10 cm - 3 m
Rubio et al., 2017 Front. Mar. Sci. 4:8. doi: /fmars

4. COMBINED USE OF ALTIMETRY AND HF RADARS
(Left) Location of the IBIROOS HFR operational (green) and future (yellow) sites and their theoretical radial range (represented by the circles, following Rubio et al., 2017a). (Right) Jason-2 altimeter tracks in IBIROOS area.
Simultaneous assimilation of multiple data sources significantly improves model’s simulations (e.g. Breivik and Saetra 2001, Barth et al., 2008, 2011; Iermano, 2016; Stanev et al., 2015).
Assimilation of HF radar + altimetry + in-situ data -> improved approach for coastal monitoring.
Increasing number of HF radar systems in the world ocean coasts.

5. OBJECTIVES

6. Combining the high spatial and temporal resolution of the HF Radar velocities with other in situ and remote sensing measurements
Main objective
To improve the interface between coastal monitoring and modelling systems by providing a new, improved coastal MDT including a 7 year-long HF radar data set, for the SE Bay of Biscay.
Objective 1
Objective 2
Better understanding the physical contents of altimetry and HF radar measurements.
Providing a methodology to obtain a new MDT improved with HF radar observations, which will contribute to a more effective constraint of the model’s hydrodynamics in coastal strips.

7. Local/regional MDT Altimetry & in situ data TOTAL CURRENTS METHODS
GEOSTROPHIC
CURRENTS
MDT COMPUTATION
From a coastal HF radar system
For isolating the geostrophy component
From a coastal HF radar system
NUMERICAL SIMULATIONS
For testing different methods

8. WORKPACKAGES
WP1 -> Data and model-data comparisons will be done in terms of surface circulation (currents and elevation) from tidal to seasonal scales, in order to:
Test the consistency between the simulated fields and the observations
Characterize and compare the content in terms of spatial and temporal scales of the simulated and observed surface currents
Use the simulations to estimate the geostrophic component from the total velocity field and to test a method to filter out the ageostrophic component from the total velocity in preparation for WP2
COMBAT PROJECT

9. DATA

10. LAND-BASED HF RADAR IN THE SE BAY OF BISCAY
Coastal HF radar antennas measure total surface currents
The SE-BoB radar provides, since 2009, maps of hourly total surface currents.
The resolved scales, depend mainly on the operation frequency
Central frequency/bandwidth
4.5 MHz/40-kHz
Time period
ongoing
Frequency of the data
Hourly
Typical resolution
5 km
Typical range
150 km
Depth of the Current
2-3 m
Ferrer et al., 2015
Rubio et al., 2017

11. SPATIAL RESOLUTION (km) PERIOD (Availability)
MODEL SIMULATIONS
Regional simulations from NEMO (CMEMS) & Coastal simulations from SYMPHONIE
NUMERICAL MODELS
NAME
SYMPHONIE
NEMO (IBI)
CONFIGURATION
BOBSHELF
ATLANTIC-IBERIAN BISCAY IRISH OCEAN PHYSICS REANALYSIS (IBI_REANALYSIS_ PHYS_005_002)
VARIABLES
SSH + Surface currents
SSH +Surface currents
FREQUENCY
Hourly-mean
SPATIAL RESOLUTION (km)
<1.5 ~8
PERIOD (Availability)
( )
DATA ASSIMILATION NO
SST, Sea Level, In-Situ TS Profiles

12. MODEL SIMULATIONS Coastal simulations from SYMPHONIE
Regional simulations from NEMO (CMEMS)
Toublanc et al., 2018

13. Spatial resolution (km)
ALTIMETRY DATA
Coastal altimetry product for Jason-2’s 213 and 248 tracks:
1 Hz X-TRACK.
CMEMS reprocessed global ocean along-track L3.
Altimetry
Along track
CTOH
CMEMS
Variables
SLA
Frequency
10 days
Spatial resolution (km)
~7 (along-track)
Period
Manso-Narvarte et al., 2018

14. RESULTS

15. COMPARISON OF COASTAL ALONG TRACK
ALTIMETRY DATA WITH HF RADAR AND IN SITU DATA
From the comparison of these data both along the track and for a crossing point it has been concluded that:
The altimetry data shows higher variability near the coast and it decreases at deeper waters.
The best altimetry-HF Radar agreement is observed in the slope where the geostrophic component is stronger due to a seasonal slope current (IPC current, winter maximum).
Removing the wind forcing by using an Ekman approximation the altimetry-HF Radar agreement increases.
HF radar surface currents & altimetry SLA geostrophic currents
Manso-Narvarte et al., 2018

16. (HF RADAR, IBI AND SYMPHONIE MODELS)
TOTAL CURRENTS
(HF RADAR, IBI AND SYMPHONIE MODELS)
Models have a limited agreement with the HF radar, both agree in the intensity of winter slope current.
Symphonie model shows more intense small-mesoscale structures in the study area (e.g. cyclonic eddy).
The observed and modelled temporal variability from daily to seasonal scales are in agreement.

17. (HF RADAR, IBI AND SYMPHONIE MODELS)
SEMIDIURNAL
INERTIAL
DIURNAL
SPECTRAL ANALYSIS
(HF RADAR, IBI AND SYMPHONIE MODELS) SD SD f f D
2 models solve reasonably well the main peaks of variability: diurnal, semidiurnal and inertial. Period:
Symphonie model describes better than the IBI model the very high frequencies (f> semidiurnal).
At periods > 1 day: IBI is underestimating the level of energy, while Symphonie is overestimating it.
The underestimation by IBI is more marked for the zonal component, while the overestimation by Symphonie is more marked for the meridional component. D SD f f SD D D

18. ESTIMATIONS OF GEOSTROPHIC CURRENTS from the models outputs
The models provide:
hourly total surface current (TC)
hourly Sea Surface Height (SSH)
Tides and inertial motions are filtered out by applying a Butterworth filter on the total current with a 48h cut-off frequency – as done for the HF radar current
Geostrophic currents are computed from the derivative of the daily detided Sea Surface Height (SSH)
 geostrophic currents of reference = GCref

19. CONTRIBUTION OF THE GEOSTROPHIC CURRENTS TO THE TOTAL CURRENTS
(IBI AND SYMPHONIE MODELS)
We analyze the contribution of geostrophy to the total currents in both models
Ratio (%) geostrophic/total currents
IBI AND SYMPHONIE MODELS
IBI model locates the greatest contribution of geostrophy over the shelf; whilst Symphonie also in the open ocean due to some eddy activity.
The winter slope current is mainly geostrophic (GC/TC >90 % in both models)
In summer, the ratio decreases, except over the shelf. The relatively large values over the shelf may be due to relatively small (total) currents there.
In the deeper area, the ratio in IBI simulation reaches < 20-30% whereas it varies between 30-80% in Symphonie (marked spatial heterogeneity).

20. CONTRIBUTION OF THE GEOSTROPHIC CURRENTS TO THE TOTAL CURRENTS
(IBI AND SYMPHONIE MODELS)
During winter slope current events the current is mainly geostrophic.
When the total current is weak, the ageostrophic proportion is higher than for strong currents.
This is also the case for periods, such as summer, where the currents have a small amplitude but a high variability.

21. ESTIMATION OF GEOSTROPHIC CURRENTS FROM TOTAL CURRENTS
1st APPROACH
1st approach = apply a moving average on the total current
For each model, two estimates are tested:
running average over 10 days (10d RA)
running average over 5 days (5d RA)
These estimates (GC) are compared with the geostrophic currents computed using the simulated SSH (GCref).
The results of this approach, indicate that the 5d window give better results (as described in the report)

22. ESTIMATION OF GEOSTROPHIC CURRENTS FROM TOTAL CURRENTS
2nd APPROACH
Second approach: subtract the surface Ekman current to the total current
Ekman current estimated following the methodology by Rio et al. (2014, GLR), with the parameters (𝛽 𝑎𝑛𝑑 𝜃) estimated for the study area:
𝑢 𝑤 (𝑧)=𝛽(𝑧) 𝑒 𝑖𝜃(𝑧) 𝜏
Rio et al. (2014, GLR)

23. SUMMARY OF THE RESULTS FROM THE TWO APPROACHES FOR ESTIMATING GC FROM TC
Correlation between mean GC and GCref increase applying a 5d-RA to the GC, and even more removing Ekman.
Removing Ekman, the correlation in the open ocean is high (>0.8), but is lower over the shelf.

24. SPATIAL MEAN OF THE RMSD (cm/s)
SUMMARY OF THE RESULTS FROM THE TWO APPROACHES FOR ESTIMATING GC FROM TC
RMSD between mean GC and GCref decreases applying a 5d-RA to the GC, and even more removing Ekman.
Removing the Ekman current, ageostrophic component is reduced significantly.
RMSD spatial mean < 4 cm/s > 4 cm/s locally over the shelf.
Better agreement for winter than summer (the contrary for the 5d RA method)
 SPATIAL MEAN OF THE RMSD (cm/s) TC
U/V
GC (5d RA)
GC (Ekman C.)
GCref
4.87 / 5.57
4.61 / 4.87
3.89 / 3.59

25. TWO APPROACHES FOR ESTIMATING GC FROM TC
CONCLUSIONS
TWO APPROACHES FOR ESTIMATING GC FROM TC
IBI and Symphonie models solve reasonably well the main peaks of surface currents variability in the SE Bay of Biscay. The later model describes better higher frequencies (f> semidiurnal).
Geostrophy in the study area contributes more to the total current in winter (winter slope current signal) than in summer.
For most of the study area, the resulting current after applying the RA is close to the geostrophic current. The lowest error is reached with the 5d-RA for both models. This highlights the short-time scales of the winter slope current variability.
Removing Ekman the results significantly improves in all the area, with estimated error on the approximation varies between 2 – 4 cm/s (> 4 cm/s over the shelf), and high correlations mainly in the open ocean.
This last approach for extracting the geostrophy from the total currents is the candidate to be applied to the HF radar data, before including in the local MDT computation.

26. RESULTS OF THE APPLICATION OF THE 2nd APPROACH TO THE HF RADAR FIELDS
EKMAN CURRENTS
𝑢 𝑤 (𝑧)=𝛽(𝑧) 𝑒 𝑖𝜃(𝑧) 𝜏
Ekman currents
mean
V variance U variance
V variance U variance
Wind data:
WRF model provided by MeteoGalicia
Native resolution of 12 km
Reproduces the offshore wind fields of the SE-BoB (Ferrer et al., 2010) with reasonable accuracy

27. RESULTS OF THE APPLICATION OF THE 2nd APPROACH TO THE HF RADAR FIELDS
GEOSTROPHIC CURRENTS
HF RADAR, mean
SYMPHONIE, mean
HF RADAR, mean
V variance U variance

28. RESULTS OF THE APPLICATION OF THE 2nd APPROACH TO THE HF RADAR FIELDS
GEOSTROPHIC CURRENTS
GCref from Symphonie mean
GC from HF Radar mean
GC from HF Radar mean
GC from MDT CNES-CLS_2013

29. RESULTS OF THE APPLICATION OF THE 2nd APPROACH TO THE HF RADAR FIELDS
ERROR ESTIMATION
Needed before applying the OI method with the HF Radar data, in the new MDT computation.
The errors could come from:
The instrumental error of the HF radar.
The geometrical error of the construction of the current from the radial velocities.
The error associated to the calculation of the geostrophic currents (obtained by filtering and applying Ekman model).
Instrumental error of the HF Radar
HF Radar data are compared with surface drifter clusters or ADCPs whose uppermost bins are not deeper than 5m, RMSDs typical values range between 3 and 12 cm.s−1 (e.g., Ohlmann et al.,2007; Molcard et al.,2009; Liu et al.,2010; Kalampokis et al., 2016). SE BoB error: between 8 and 10 cm/s.
Error of the 2nd approach for computing surface geostrophic currents
The error of the method for computing the surface geostrophic currents from total ones, is assumed as the spatial mean of the RMSD for U and V between GC and Gcref.

30. NEXT STEPS

31. Computing a new MDT for the SE Bay of Biscay with HF RADAR
Following the method described in Rio et al., 2004, 2007, 2011, 2014.
Altimetry
Gravimetry
MDT=MSS-GEOID HEIGHT
Source: Aviso+
GOCE/GRACE gravimeters resolve
scales larger than 100 km
Introducing in situ currents data for resolving scales shorter than 100 km.
E.g. HF RADAR surface currents, drifters.
« Super observations » of mean geostrophic current from processed drifters
( )
First guess of the MDT
CNES-CLS18
MSS CNES-CLS15 - GOCO05s
MDT CNES-CLS18

32. Computing a new MDT for the SE Bay of Biscay with HF RADAR
Following the method described in Rio et al., 2004, 2007, 2011, 2014.
CNES-CLS13 MDT
MDT CNES-CLS13
HF radar, GC mean
MDT CNES-CLS18 (ref )

33. for the SE Bay of Biscay with HF RADAR
Computing a new MDT
for the SE Bay of Biscay with HF RADAR
First guess of the MDT
CNES-CLS18
MSS CNES-CLS15 - GOCO05s
Tasks:
Computation of the « first guess » = optimal filter of raw difference MSS CNES-CLS15 – geoid GOCO05s (λ > 200 km)
Refine the filter in our specific area: SE BoB
Change reference period:
MSS = MSS <SLA >
Inversion with HF radar observations
Target resolution : 5 km
Refine OI parameters for the BoB area and for the HF radar data
HF radar, GC mean

34. for the SE Bay of Biscay with HF RADAR
2017 drifters
drifters
Computing a new MDT
for the SE Bay of Biscay with HF RADAR
Validate the MDT/ADT
Comparison with
independent drifters
Independent radar HF (2018) vs along track MDT+SLA (sentinel3 and Jason3)
Glider (BBTRANS mission)
ADCP from mooring
Expected improvements:
Resolves better the pattern
Higher spatial resolution (around 5 km in the area)
Radar HF (2018)
Along track SLA (Jason3 and Sentinel3)
Glider and ADCP (2018)

35. ISSUES
&
IMPACTS

36. IDENTIFIED ISSUES Administrative delays…
… to carry out the subcontracting of CNRS by AZTI. As a consequence, the recruitment of a researcher for reinforcing the tasks of the WP1, planned by LEGOS, was delayed. Although main calculations have been carried out from the simulations, additional comparisons and analyses would be necessary to fulfill WP1
Delay in delivering D2.1 and D2.2
We suggest delivering them:
D2.1 (Coastal MDT including HF Radar data) : End of September
D2.2 (Report of WP2): End of October
Nevertheless, it is not expected that this delay in WP2 deliverables, will delay the subsequent ones.

37. Expected impact on CMEMS
Development leading to potential impact on CMEMS
Status
Targeted TACs or MFCs
Sensor synergy approach performed in WP1
Near completion
INSITU or MULTIOBS TACs
New coastal MDT including HF Radar data
In progress
SEALEVEL TAC
Roadmap towards an effective DA of new altimetry products into numerical models
To be developed
MFCs
drifters 2 1 3
Perspective in MDT computation:
MDT including both drifters and HF radar in larger area
BoB  test area to merge drifter and HF radar
IBI  test including other HF radar data
Discussion within TAPAS (link with MFCs)
1- COMBAT
2- BoB
3- IBI

38. CONTRIBUTIONS

40. COMBAT
Combining coastal altimetry data with in-situ and land-based remote data for improving the monitoring of the dynamics in the southeastern Bay of Biscay
Caballero, A.1, Rubio, A.1, Rio, M-H2, Ayoub, N., Mader, J.1, Larnicol, G.2, Manso-Narvarte, I.1 and Dufau, C.2
1 AZTI-Marine Research, Herrera Kaia, Portualdea z/g, Pasaia, Spain
2 CLS Space Oceanography Division, 8-10 Rue Hermès, Ramonville St. Agne, France
3 LEGOS, 14 Ave. E. Belin, Toulouse, France

45. ACCEPTED POSTER

46. THANK YOU FOR YOUR ATTENTION