OS4.7

Data assimilation techniques and applications in coastal and open seas

Since its inception, data assimilation has proven to be enormously useful in the most varied fields throughout the Earth Sciences. It is certainly essential in meteorology, where the short-range forecasts would otherwise be almost impossible. In oceanography, its development has been slower, partly due to the smaller number of continuous and stable observations, and partly due to the fewer studies that show the importance of ocean forecasts for societal benefit. However, recently, these techniques are used more and more widely, both in operational oceanography and to produce climate reconstructions. Although the techniques are similar to those used in the atmospheric field, they have to deal with particularities due to the different environment, where the boundary conditions, open and closed, have greater importance, and the sparsity of observations poses unique challenges.
In this session, we welcome contributions describing data assimilation techniques, both methodological and case studies, in the oceanographic field. We welcome presentations of new techniques or new types of observations that cover every aspect of data assimilation, including varied applications of data assimilation, both in coastal seas and the open ocean.

Co-organized by GI2/NH5/NP5
Convener: Marco Bajo | Co-conveners: Philip Browne, Matthew Martin, Andrea Storto, Jiping Xie
Presentations
| Wed, 25 May, 10:20–11:50 (CEST)
 
Room 1.15/16

Presentations: Wed, 25 May | Room 1.15/16

Chairpersons: Marco Bajo, Jiping Xie
10:20–10:27
|
EGU22-6451
|
ECS
|
Virtual presentation
Breanna Vanderplow, John Kluge, Alexander Soloviev, Richard Dodge, Jon Wood, Johanna Evans, William Venezia, and Michael Ferrar

Predicting ocean circulation in strong currents remains challenging because of limits in modelling capabilities such as resolution. Coastal ocean circulation models typically have horizontal resolution starting from 1 km. To address this matter, we have developed a high resolution three-dimensional computational fluid dynamics (CFD) model for strong ocean currents such as the Gulf Stream. Our model domain contains three inlets and an outlet and has been verified with field data from the Straits of Florida. For model verification, a 6 ADCP mooring array in a rectangular shape was deployed 8 miles offshore on the Miami Terrace. The data from 5 ADCP moorings were used to produce the inlet boundary conditions, which were updated every 1 minute. The sixth ADCP in the center of the outlet was used for model verification. This approach has demonstrated good predictive ability for ocean circulation in the challenging environment of a strong western boundary current. We anticipate our work to be a starting point for the development of sophisticated prediction models applicable to western boundary currents in the range from small-scales to sub-mesoscales, based on advanced data assimilation techniques.

How to cite: Vanderplow, B., Kluge, J., Soloviev, A., Dodge, R., Wood, J., Evans, J., Venezia, W., and Ferrar, M.: Measurement and modeling of small-scale to mesoscale ocean circulation in the Straits of Florida, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-6451, https://doi.org/10.5194/egusphere-egu22-6451, 2022.

10:27–10:34
|
EGU22-2731
|
ECS
|
Virtual presentation
Zhiqiang Chen, Xidong Wang, and Jian Chen

Accurate knowledge of ocean surface currents is crucial for a gamut of applications. In this study, the way in which merging altimeters composited two-dimensional sea surface height (SSH, 1/4°) with remote sensing combined sea surface temperature (SST, 9km) image improves the surface current estimates is investigated. Based on the surface quasigeostrophic (SQG) theory, we reconstruct the surface current by resolving the large scale motions, the mesoscale dynamics, and the oceanic smaller processes. Its feasibility is validated using the altimeter-derived geostrophic current (GC) and drogued drifters in the South Indian Ocean (SIO) during 2011–2015. Results of the two cases show that the effective resolution of the reconstructed surface current (RSC) has improved to 30 km after merging the high-resolution SST information, compared to 70 km of the GC. Moreover, the RSC outperforms the altimeter-derived GC in reproducing the practical dynamical processes. Over the analyzed period, compared with 841 drifters, the statistical results indicate that the RSC reduces the reconstruction errors of zonal velocity, meridional velocity, and velocity phase by about 14.6%, 45.7%, 27.0% in the SIO relative to the GC, respectively. Our method particularly improves the meridional velocity and velocity phase along the Antarctic Circumpolar Current, Agulhas Retroflection, Greater Agulhas System, and South Equatorial Current. In addition, the lower Lagrangian separation distance and higher skill score of the RSC given by Lagrangian analysis also demonstrate that the proposed method is more promising to provide essential information on ocean surface currents applications, such as water property transports, search and rescue, etc.

How to cite: Chen, Z., Wang, X., and Chen, J.: Improving the Surface Currents from the Merging of Altimetry and Sea Surface Temperature Image in the South Indian Ocean, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-2731, https://doi.org/10.5194/egusphere-egu22-2731, 2022.

10:34–10:41
|
EGU22-2107
|
ECS
|
Virtual presentation
Zikang He, Xidong Wang, Xinrong Wu, and Jian Chen

This study uses a variational method combined with satellite observations to reconstruct three-dimensional temperature and salinity profiles for the Northern Indian Ocean (NIO). Sensitivity experiments show that sea surface temperature (SST) dominantly improve the temperature reconstruction of upper 100 m; sea surface salinity (SSS) determines salinity estimation in the upper 100 m; sea surface height anomaly (SSHA) dominates the reconstruction of thermocline. The reconstructed temperature fields can be greatly improved in the thermocline by removing barotropic signal from the altimeter SSH data through a linear regression method. Ocean reanalysis and in situ temperature and salinity data are used to evaluate the results of reconstruction. Comparing with Simple Ocean Data Assimilation (SODA) in 2016, the spectral correlation between the reconstruction and the SODA density anomalies show that the reconstruction fields can retrieve mesoscale and large-scale signals better. Moreover, the reconstruction salinity is much more accurate than SODA salinity in the upper ocean over the Bay of Bengal (BoB). Compared with CTD section observations, the reconstruction fields can capture the mesoscale eddy structure in the Arabian Sea (AS) and BoB well, respectively. The long time series of reconstruction along Argo trajectory shows that the reconstruction fields can better reproduce the observed intraseasonal oscillations of thermocline/halocline in the BoB. Compared with the World Ocean Atlas 2013 (WOA13) climatology, the reconstruction fields can better characterize upper ocean water mass variability.

How to cite: He, Z., Wang, X., Wu, X., and Chen, J.: Projecting Three-dimensional Ocean Thermohaline Structure in the North Indian Ocean from the Satellite Sea Surface Data Based on a Variational Method, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-2107, https://doi.org/10.5194/egusphere-egu22-2107, 2022.

10:41–10:48
|
EGU22-4313
|
ECS
|
Virtual presentation
Bo Dong, Keith Haines, and Matthew Martin

We present a post-hoc smoothing algorithm for use with sequentially generated reanalysis products, utilizing the archive of “future” assimilation increments to update the “current” analysis. This is applied to the Lorenz 1963 model and then to the Met Office GloSea5 Global ¼° ocean reanalysis during 2016.  A decay time parameter is applied to the sequential increments which assumes that background error covariances remain spatially unchanged but decay exponentially away from analysis times. Only increments are smoothed so the reanalysis product retains modelled high-frequency variability, e.g., from atmospheric forcing. Results show significant improvement over the original reanalysis in the 3D temperature and salinity variability, as well as in the sea surface height (SSH) and ocean currents. Spatial gap filling from future data is particularly beneficial. The impact on the time variability of ocean heat and salt content, as well as kinetic energy and the Atlantic Meridional Overturning Circulation (AMOC), is demonstrated. 

How to cite: Dong, B., Haines, K., and Martin, M.: Improving High Resolution Ocean Reanalyses Using a Smoother Algorithm, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-4313, https://doi.org/10.5194/egusphere-egu22-4313, 2022.

10:48–10:55
|
EGU22-4741
|
ECS
|
On-site presentation
Marco Stefanelli, Eric Jansen, Ali Aydogdu, Ivan Federico, Giovanni Coppini, and Nadia Pinardi

Eight of the top ten most populated cities in the world are located by the coast. The improvement of the coastal ocean representation is a key topic to understand the  present and near-future ocean state and predict its evolution under climate change conditions.

The coastal ocean is difficult to model due to the presence of complex coastlines, interaction with inland waters, rapid changes in  topography and highly space-time variability of the phenomena involved. Unstructured-grid models are used to partially attenuate this source of errors in cross-scale (from open sea to coastal regions) oceanographic modelling. On the other hand, the data assimilation methodologies to improve the unstructured-grid models in the coastal seas is being developed only recently (e.g., Aydogdu et al., 2018; Bajo et al., 2019) and needs more advancements.  

Here, we show preliminary results from the coastal ocean forecasting system SANIFS (Southern Adriatic Northern Ionian coastal Forecasting System, Federico et al., 2017) based on SHYFEM fully-baroclinic unstructured-grid model (Umgiesser et al., 2004)  interfaced with OceanVar (Dobricic and Pinardi, 2008; Storto et al., 2014), a state-of-art variational data assimilation scheme, adopted for several systems based on structured grid (e.g. regional CMEMS for Mediterranean and Black Seas, marine.cmems.eu).

In OceanVar, Empirical Orthogonal Functions (EOFs) method is used to reduce the dimensionality of computation removing the statistically less significant modes and to correlate observations and model background in the water column;  while the increments are spread horizontally using the recursive filter method. While this method is typically only used to model covariances between neighbouring points in a structured grid, the algorithm has now been generalised and successfully implemented also for unstructured grids.

Preliminary results show that temperature and salinity observations from Argo profilers improve the ocean state. Future steps will also include sea level assimilation. 

This work is a starting point in order to improve our forecast of local extreme events (e.g. heat waves and storm surge) which are statistically increasing in number and intensity in the Mediterranean region due to climate change.

How to cite: Stefanelli, M., Jansen, E., Aydogdu, A., Federico, I., Coppini, G., and Pinardi, N.: Variational data assimilation for advanced cross-scale ocean modelling., EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-4741, https://doi.org/10.5194/egusphere-egu22-4741, 2022.

10:55–11:02
|
EGU22-732
|
On-site presentation
Christian Ferrarin, Marco Bajo, and Georg Umgiesser

The correct reproduction of sea-level dynamics is crucial for forecasting floods and managing the associated risk. On the other hand, sea-level monitoring through observations can provide a description only of past events and it is challenging and costly, both of time and money. In this context, oceanographic models are increasingly used to describe the sea dynamics, providing a spatial/temporal extension to the observations. The best solution, which merges the observation accuracy and the model spatial/temporal resolution, is the data assimilation analysis, which is particularly important in coastal regions with scarce monitoring resources. In this study, we investigate the benefits of assimilating sparse observations from tide gauges in an unstructured hydrodynamic model for simulating the sea level in the Mediterranean Sea. We use the Ensemble Kalman filter, both to obtain an analysis of the past and to produce accurate forecasts. In the analysis we tested the assimilation in storm-surge simulations, only-tide simulations, and total-level simulations, using the observations in the stations. The results of storm-surge simulations were compared with those of total-level simulations, by adding the tide obtained from harmonic analysis of the observations. RMSE and correlation show improvements for all the components of the sea level and all the stations considered (not assimilated). The averaged-over-station RMSE reduces from 9.1 to 3.4 cm for the total level. The greatest improvements happen when the model without assimilation, due to an error of the wind-pressure forcing, did not reproduce some barotropic free modes of oscillation triggered by an initial surge. The preliminary forecast simulations of storm surge show improvements due to the data assimilation extending up to 5 days of forecasting. Even in this case, the longer improvements seem to happen when a free mode of oscillation is triggered. The results of this study will be used to improve the sea level forecasting system in the Adriatic Sea, developed within the framework of the Interreg Italy-Croatia STREAM project (Strategic development of flood management, project ID 10249186).

How to cite: Ferrarin, C., Bajo, M., and Umgiesser, G.: Sea-level modelling in the Mediterranean Sea using data assimilation, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-732, https://doi.org/10.5194/egusphere-egu22-732, 2022.

11:02–11:09
|
EGU22-2555
|
Virtual presentation
Ye Liu and Lars Arneborg

The objective of this study is to investigate if the assimilation of ocean color data into a complex marine ecosystem model can improve the hindcast of key biogeochemical variables in coastal seas. A localized Singular Evolutive Interpolated Kalman filter was used to make assimilation of the daily fully reprocessed product of Multi-Satellite chlorophyll observations into a three-dimensional ecosystem model of the Baltic Sea. Twin experiments are performed to evaluate the performance of the assimilation with respect to both satellite and in situ observations. Compared to the reference run, the assimilation was found to immediately and considerably reduce the bias, root mean square error, and increase the correlation with the spatial distributions of the assimilated chlorophyll data while this improvement is limited to the upper layer of the water column. This feature is explained by the weak correlation taken into account by the assimilation between the surface and deep phytoplankton. The assimilation scheme used is multivariate, updating all biogeochemical model state variables. The other variables were not degraded by the assimilation. More significantly, the skill metrics for non assimilated variables indicate that the hindcast of the mean data values at L4 was improved; however, improvements in the short-term forecast were not discernable. Our results provide general recommendations for the successful application of ocean color assimilation to hindcast key biogeochemical variables in coastal seas.

How to cite: Liu, Y. and Arneborg, L.: Assimilating the remote sensing ocean color data into a biogeochemical model of the Baltic Sea, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-2555, https://doi.org/10.5194/egusphere-egu22-2555, 2022.

11:09–11:16
|
EGU22-923
|
On-site presentation
Jiping Xie and Laurent Bertino

The second version of the Arctic ocean and sea ice reanalysis is based on the coupled ensemble data assimilation system (TOPAZ4b). Compared to its predecessor (Xie et al. 2017) it has benefited from enhancements to observation, model vertical resolution, and forcing datasets. TOPAZ4 relies on version 2.2 of the HYCOM ocean model and the ensemble Kalman filter data assimilation using 100 dynamical members. A 30-years reanalysis of the Arctic ocean and sea ice has been completed starting in 1991, and made available as the multi-year physical product by the Arctic Marine Forecasting Center (ARC MFC) under the Copernicus Marine Environment Monitoring Service. Contrary to the previous version of the Arctic reanalysis, the systematic errors due to fragmented time series of assimilated observations have been removed by using consistent ESA CCI data. The comparison to in situ profiles shows that the temperature and salinity stratification has been considerably improved by the increased vertical resolution in HYCOM, for example in the East Greenland Sea, the temperature root mean square error (RMSE) from surface to 1400 m has been reduced by 50%. These improvements encourage the use of this Arctic reanalysis for climate studies.

How to cite: Xie, J. and Bertino, L.: TOPAZ4b: a new version of the ocean and sea-ice Arctic reanalysis, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-923, https://doi.org/10.5194/egusphere-egu22-923, 2022.

11:16–11:23
|
EGU22-5698
|
On-site presentation
Andrea Cipollone, Deep Sankar Banerjee, Ali Aydogdu, Doroteaciro Iovino, and Simona Masina

Recent intercomparison studies among ocean/sea-ice Reanalyses (such as ORA-IP) have shown large discrepancies in many sea-ice-related fields, despite a rather general agreement in the sea-ice extension. The low accuracy of sea-ice thickness measurements together with the highly non-gaussian distributions of related uncertainty, made multivariate sea-ice data assimilation (DA) strategies still at an early stage, although nearly twenty years of thickness observations are now available. In a standard multivariate scheme, the break of Gaussianity can generate un-realistic corrections due to the poor linear relationship driven by the B matrix.

One approach to solve the problem is the implementation of anamorphous transformations that modify the probability density functions of ice anomalies into Gaussian ones (Brankart et al. 2012). In this study, a 3DVar DA scheme (called OceanVar), employed in the routinely production of global/regional ocean reanalysis CGLORS (Storto et al, 2016), has been recently extended to ingest sea-ice concentration (SIC) and thickness (SIT) data. An anamorphous operator, firstly developed and made freely available within the SANGOMA project (http://www.data-assimilation.net/), has been updated and adapted for the bivariate assimilation of SIC/SIT within the OceanVar framework.

We present the comparison among several sensitivity experiments that were performed assimilating different observation datasets and using different DA configurations at 1/4 degree global resolution. Specifically, we assess the impact of ingesting different SIT products, such as SMOS and CRYOSAT-2 data or the merged product CS2SMOS.

We show that the sole assimilation of SIC improves the spatial representation of SIT with respect to a free run. The inclusion of thickness correction, determined by empirical relations, appears to improve the sea ice characteristics in the Atlantic sector and degrade them in the Siberian region; therefore a refined tuning could probably be beneficial. The spatial error reduces sharply only once CRYOSAT-2 data are assimilated jointly with SIC data. In the present set up, all the experiments generally tend to overestimate the sea-ice volume in the case SMOS data are not assimilated. However, observational errors associated with SMOS data are generally too small, leading to jumps in the volume time series at the beginning of the accretion period if not calibrated correctly.

The proposed approach is suitable to be used for covarying ocean/sea-ice variables in future coupled ocean/sea-ice DA.

Storto, A. and Masina, S. (2016), Earth Syst. Sci. Data, 8, 679, doi: 0.5194/essd-8-679-2016

Brankart, et al. (2012), Ocean Sci., 8, 121, doi: 10.5194/os-8-121-2012

 

How to cite: Cipollone, A., Banerjee, D. S., Aydogdu, A., Iovino, D., and Masina, S.: Bivariate sea-ice assimilation for Global Ocean Analysis/Reanalysis, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-5698, https://doi.org/10.5194/egusphere-egu22-5698, 2022.

11:23–11:30
|
EGU22-6848
|
ECS
|
Presentation form not yet defined
Guokun Lyu and Meng Zhou

As part of the ongoing development of a data assimilation system for reconstructing the Arctic ocean-sea ice state, we incorporated an adjoint of sea ice rheology, which was approximated by free drift assumption due to stability problem, into an adjoint model of a coupled ocean-sea ice model. The adjoint sensitivity experiments show that the internal stress effect, represented by the adjoint rheology, induced remarkable differences in the sensitivities to ice drift and wind stress in the central Arctic Ocean. In contrast, ice is mostly free drift in the marginal ice zone. The assimilation experiments reveal that including the adjoint of ice rheology helps extract observational information, especially the ice drift observations, which improves the estimation of the sea ice decline process in 2012. The results suggested great potentials for further improving the Arctic ocean-ice state estimation in the framework of the adjoint method with the adjoint sea ice rheology included. 

How to cite: Lyu, G. and Zhou, M.: Effects of inclusion of adjoint sea ice rheology on estimating ocean-sea ice state, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-6848, https://doi.org/10.5194/egusphere-egu22-6848, 2022.

11:30–11:50