Figure - available from: Frontiers in Marine Science
This content is subject to copyright.
Impacts of ocean observation data evaluated by OSE studies using the latest operational ocean reanalysis system in ECMWF. Plots of normalized RMS differences of upper 700-m column-averaged temperature between the control experiment, in which all in-situ observations are assimilated, and OSEs with removal of mooring data (NoMooring), XBT, MBT, and CTD data (NoShip), Argo data (NoArgo), and all in-situ observations (NoInsitu). Statistics are computed using monthly-mean anomaly data over the 2008–2014 period after removal of the seasonal cycle information, then normalized against the temporal standard deviation of temperature over the same period in the control experiment.
Source publication
This paper summarizes recent efforts on Observing System Evaluation (OS-Eval) by the Ocean Data Assimilation and Prediction (ODAP) communities such as GODAE OceanView and CLIVAR-GSOP. It provides some examples of existing OS-Eval methodologies, and attempts to discuss the potential and limitation of the existing approaches. Observing System Experim...
Citations
... Data assimilation can be used to aid observing system design and evaluation (Fujii et al., 2019), and as a component of "smart" observing systems (Ford et al., 2022), in the following ways: similar assessment can be performed using forecast sensitivity-based observation impact (FSOI), which assesses the impact each observation type has on improving forecasts in a single experiment (Eyre, 2021). ...
In the last two decades UK research institutes have led a wide range of developments in marine data assimilation (MDA), covering areas from the MDA applications in physics and biogeochemistry, to MDA theory. We review the progress over this period and formulate our MDA vision for both the short-term and the longer-term future. We focus on identifying the MDA stakeholder community and current/future areas of impact, as well as the current trends and the future opportunities. This includes rapid growth of machine learning (ML) / artificial intelligence (AI) and digital twin applications. We articulate the MDA needs for future types of observational data (whether planned missions, or hypothetical) and what should be the response of the MDA community to the increase in computational power and new computer architectures (e.g. exascale computing). Although the specifics depend on the MDA area, we advocate for balanced redistribution of the new computational capability among increased model resolution, model complexity, more sophisticated DA algorithms and uncertainty representation (e.g. ensembles). We also advocate for integrated approaches, such as strongly coupled DA (ocean/atmosphere, physics/biogeochemistry, ocean/sea ice) and the use of ML/AI components (e.g. for multivariate increment balancing, bias-correction, model emulation, observation re-gridding, or fusion).
... This has motivated a large research effort both in terms of simulation studies (Uchida et al., 2022), observational effort Villas Boas et al., 2019), and data assimilation methods (Moore et al., 2019;Storto et al., 2019). Regarding the latter aspect, state-of-the-art approaches mostly rely on the one hand on optimal interpolation approaches and on the other hand on data assimilation schemes combined with ocean general circulation models (Baaklini et al., 2021;Benkiran et al., 2021;Fujii et al., 2019). As mentioned above, both approaches still show limitations in the ability to retrieve fine-scale patterns, whereas both observation-driven and theoretical studies evidence the interplay between fine-scale sea surface dynamics and some observed processes such as sea surface tracers (Ciani et al., 2021;Isern-Fontanet et al., 2006) and drifters' trajectories (Sun et al., 2022). ...
... Importantly, these interpolation schemes only retrieve geostrophic sea surface velocities and cannot recover ageostrophic components. By contrast, the assimilation of satellite altimetry and satellite-derived SST observations, possibly complemented by other data sources, in ocean general circulation models (OGCM) (Benkiran et al., 2021;Fujii et al., 2019) aim at reconstructing the whole ocean state, including the total sea surface currents. In such data assimilation schemes, operator Φ M in Equation 7 implements the time stepping of the OGCM and observation operators H m typically encodes the masking operator of the different sources of gappy satellite-derived and in situ data. ...
... This has motivated a large research effort toward the exploitation of other observation sources, alone or combined with satellite altimetry, to retrieve sea surface dynamics, including among others SST (Fablet et al., 2018;Isern-Fontanet et al., 2014;Rio et al., 2016), Ocean Color (Ciani et al., 2021), sea surface drifters (Baaklini et al., 2021;Sun et al., 2022), and SAR observations (Chapron et al., 2005). From a methodological point of view, we may distinguish three main categories of approaches: optimal interpolation schemes (Cressie & Wikle, 2015;Taburet et al., 2019), data-driven approaches Manucharyan et al., 2021), and data assimilation scheme using OGCM (Benkiran et al., 2021;Fujii et al., 2019) or QG dynamical priors (Le Guillou et al., 2020;Ubelmann et al., 2014). The proposed 4DVarNet scheme benefits, on the one hand, from a variational data assimilation formulation to make explicit observation and dynamical priors, especially the expected though unknown relationship between SST and SSC features, and, on the other hand, from the computational efficiency of deep learning schemes, to learn uncalibrated terms and solvers from data in an end-to-end manner. ...
Satellite altimetry offers a unique approach for direct sea surface current observation, but it is limited to measuring the surface‐constrained geostrophic component. Ageostrophic dynamics, prevalent at horizontal scales below 100 km and time scales below 10 days, are often underestimated by ocean reanalyzes employing data assimilation schemes. To address this limitation, we introduce a novel deep learning scheme, rooted in a variational data assimilation formulation with trainable observations and a priori terms, that harnesses the synergies between satellite‐derived sea surface observations, namely sea surface height (SSH) and sea surface temperature (SST), to enhance sea surface current reconstruction. Numerical experiments, conducted using realistic simulations, in a case study area of the Gulf Stream, demonstrate the potential of the proposed scheme to capture ageostrophic dynamics at time scales of 2.5–3.0 days and horizontal scales of 0.5°–0.7°. The analysis of diverse observation configurations, encompassing nadir along‐track altimetry, wide‐swath SWOT (Surface Water and Ocean Topography) altimetry, and SST data, highlights the pivotal role of SST features in retrieving a significant portion of the ageostrophic dynamics (approximately 47%). These findings underscore the potential of deep learning and 4DVarNet schemes in improving ocean reanalyzes and enhancing our understanding of ocean dynamics.
... This is particularly important for extreme events such as tropical cyclones (e.g., Domingues et al., 2021), which can rapidly intensify when traveling over anomalously warm coastal waters (Dzwonkowski et al., 2020). Operational oceanography relies on the integration of in situ and remote sensing observation networks to deliver accurate forecasts and alert systems (e.g., Fujii et al., 2019). In addition, correctly simulating the top of the thermocline, one of the most difficult regions to predict, improves data assimilating ocean forecasts (Santana et al., 2023). ...
... This project has been endorsed by the United Nations Decade of Ocean Science for Sustainable Development (UNESCO-IOC, 2021) and has been tasked to "identify the optimal combination of the different ocean observation platforms through observing system design and evaluation, and to develop assimilation methods which can enable drawing synergistic effects from these combinations." In addition, SynObs continually assesses the impact of observing systems on ocean forecast and reanalysis systems, expedites the design and assessment of new observing systems, identifies "gaps" in the global ocean observing system, and proposes optimized observation arrays on regional to global scales (Fujii et al., 2019). This group also recommends protocols for observation impact assessment (e.g., Observing System Evaluation (OSE)) and common metrics that quantify observation impacts. ...
... The upper 700 m temperature RMSE improvement of TAO experiment was biggest (∼0.25°C) west of the dateline and along the northern edge of the TAO/TRITON moorings, between 5°N-10°N. Fujii et al. (2019) provides a good overview (on behalf of the Observing System Evaluation Task Team) of the recent tropical Pacific ocean OSE work (e.g. (Fujii, Ogawa, et al., 2015;Lea et al., 2014;Xue et al., 2017;Zuo et al., 2019), for NCEP, UK Met Office, JMA and ECMWF OSEs, respectively), and outlines the goals of the SynObs Project. ...
Starting in the early 1990's, the Tropical Atmosphere Ocean (TAO)/TRIangle Trans Ocean buoy Network (TRITON) array has been the pervasive source for observing large‐scale equatorial wave propagation which is key for El Nino/Southern Oscillation (ENSO) predictions. However, removal of western TRITON moorings, the plan to reorganize the array (i.e., TPOS 2020), and availability of other sources of in situ data (e.g., Argo) have highlighted the need to rigorously assess the impact of TAO/TRITON data on ENSO predictions. Therefore, we evaluate TAO/TRITON array using data denial assimilation experiments and assess the impact on coupled atmosphere/ocean predictions of the big 2015 El Niño. Validation of the CONTROL (assimilates all available data) and NOTAO (withholds TAO/TRITON data) reanalyses shows that assimilating TAO/TRITON data generally improves comparisons versus gridded and pointwise in situ observations. This is especially true across the entire basin above and in the eastern half of the Pacific just below the thermocline for temperature. Even with relatively few observations, salinity is generally improved except near 120°W near the surface. To evaluate the impact of TAO/TRITON data on ENSO initialization, seasonal forecasts were initialized from the CONTROL and NOTAO experiments. For the 9‐month forecasts which were initialized in January, July, and October 2015, both the amplitude and the accuracy of the ensembles initialized with TAO/TRITON data were closer to observations. Through the analysis of Kelvin and Rossby waves, we show that the impact of TAO/TRITON is to generally shoal the mixed layer depth, leading to amplification of the El Niño downwelling signal, and improving the amplitude of the ENSO signal.
... Data assimilation technology [75][76][77] combines observation data with numerical model data, corrects numerical model results with observation data [78][79][80], and employs short-term analysis results as the initial value of model prediction to optimize numerical model parameters and improve prediction accuracy [81,82]. The data assimilation principles [83,84] are based on the statistical estimation theory such as optimal interpolation and ensemble Kalman filter) and optimal control or variational theory, such as three-dimensional and four-dimensional variational. ...
... Other DA systems with significant use of OSEs/OSSEs include the European Centre for Medium-Range Weather Forecasts' (ECMWF) Earth System, which includes DA capabilities for ocean and ice-cover models, ESA's PyOSSE project, and the ocean science community's projects on optimization of observing systems [Fujii et al., 2019]. OSSEs are used regularly for new instruments or missions, e.g. the NOAA's GeoXO successor program to GOES-R (for GeoXO's advanced atmospheric sounder, see OSSEs here: https://www.ssec.wisc.edu/geo-irsounder/osse/). ...
... Numerous studies based on numerical experiments have investigated the impact of existing or future in situ observations in ocean analysis and forecasting systems (Fujii et al., 2019). For instance, the complementarity between tropical mooring, Argo and altimetry data has been demonstrated for global ocean analysis (Balmaseda et al., 2007;Turpin et al., 2016) and seasonal forecasting (Balmaseda and Anderson, 2009;Balmaseda et al., 2013;Fujii et al., 2015). ...
... Other studies have also focused on specific regions, like the Tropical Pacific (e.g., Zhu et al., 2021), the Australian coast (e.g., Jones et al., 2012;Aydogdu et al., 2016), or the abyssal Ocean (e.g., Gasparin et al., 2020;Levin et al., 2021). However, a large part of impact studies was dedicated to existing in situ observations and/or did not consider the integrated value of the global ocean observing system (e.g., no assimilation of altimetry; Fujii et al., 2019). In addition, usual evaluation metrics, mostly based on box-averaged statistics, make it difficult to separate observation impacts depending on specific space and temporal scales. ...
... Further developments of data assimilation systems might be needed to make a better use of these observations in models. Unlike many other data assimilation systems, the benefits of in situ ocean observations are estimated here as a complementary information to satellite data sets (Fujii et al., 2019). The number of experiments, including simulations that assimilate observing system components separately, is a central point of this study and demonstrates the complexity of impact studies in a multivariate system. ...
Ocean monitoring and forecasting systems combine information from ocean observations and numerical models through advanced data assimilation techniques. They are essential to monitor and report on past, present and future oceanic conditions. However, given the continuous development of oceanic models and data assimilation techniques in addition to the increased diversity of assimilated platforms, it becomes more and more difficult to establish how information from observations is used, and to determine the utility and relevance of a change of the global ocean observing system on ocean analyses. Here, a series of observing system simulation experiments (OSSE), which consist in simulating synthetic observations from a realistic simulation to be subsequently assimilated in an experimental analysis system, was performed. An original multiscale approach is then used to investigate (i) the impact of various observing system components by distinguishing between satellites and in situ (Argo floats and tropical moorings), and (ii) the impact of recommended changes in observing systems, in particular the impact of Argo floats doubling and enhancements of tropical moorings, on the fidelity of ocean analyses. This multiscale approach is key to better understand how observing system components, with their distinct sampling characteristics, help to constrain physical processes. The study demonstrates the ability of the analysis system to represent 40-80% of the temperature variance at mesoscale (20-30% for salinity), and more than 80% for larger scales. Satellite information, mostly through altimetric data, strongly constrains mesoscale variability, while the impact of in situ temperature and salinity profiles are essential to constrain large scale variability. It is also shown that future enhancements of Argo and tropical mooring arrays observations will likely be beneficial to ocean analyses at both intermediate and large scales, with a higher impact for salinity-related quantities. This work provides a better understanding on the respective role of major satellite and in situ observing system components in the integrated ocean observing system.
... (daily, weekly, and monthly) [11]. Ordinary neural networks (NNs) were first used to make predictions, and deep learning networks were then used. ...
The tropical cyclone is one of the most powerful and destructive meteorological systems on Earth. Researchers note tropical cyclone data every few seconds, but utilizing all of the data with the appropriate accuracy values is difficult. In this system, we predict the various elements' status accuracy and loss in the ocean data set. The use of machine learning methods has developed a lot, and the prediction of the value of the ocean data follows the new enhanced term to give the status of the elements in the data. The LSTM (long short-term memory neural network excavation model) of the historical track's helpful information is more profound and more precise. Bi-LSTM goes the both forward and backward directions, and Adam optimizer, two updated machine learning techniques, are utilized to assess the status of the ocean element in the data set. It goes beyond the existing system to offer an opportunity for a different system result. The data set with a large number of values will also perform accurately. The project's ultimate objective is to give oceanographers a tool to anticipate the quality of ocean data in real-time, which can increase the precision of climate models and help with improved ocean-related decision-making.
... A comprehensive set of ocean observations is routinely provided to operational centers that ingest them in the ocean and atmospheric data assimilation systems to support climate monitoring and prediction activities. As a product of the data assimilation systems, a three-dimensional rendition of the ocean state on a regular grid has greatly advanced our understanding of ocean variability and air-sea interactions (Fujii et al. 2019). ...
Climate variability on sub-seasonal to interannual time scales has significant impacts on our economy, society, and the earth’s environment. Predictability for these time scales is largely due to the influence of the slowly varying climate anomalies in the oceans. The importance of the global oceans in governing climate variability demonstrates the need to monitor and forecast the global oceans in addition to the El Niño-Southern Oscillation in the tropical Pacific. To meet this need, the Climate Prediction Center (CPC) of the National Centers for Environmental Prediction (NCEP) initiated real-time global ocean monitoring, and a monthly briefing in 2007. The monitoring covers observations as well as forecasts for each ocean basin. In this paper, we introduce the monitoring and forecast products. CPC’s efforts bridge the gap between the ocean observing system and the delivery of the analyzed products to the community. We also discuss the challenges involved in ocean monitoring and forecasting, as well as the future directions for these efforts.
... Os OSSEs são empregados para apoiar a implantação de novos sistemas observacionais, nos quais dados sintéticos são usados para capacitar o sistema a assimilar novos dados e investigar o possível impacto que esses dados trarão à análise e ao sistema previsor. Por exemplo, a partir de 2023, os dados altimétricos do satélite Surface Water and Ocean Topography (SWOT) aportarão novas informações sobre a circulação submeso e mesoescala no oceano global e há hoje um grande esforço em andamento nos principais centros de oceanografia operacional para realizar OSSEs e preparar os sistemas de assimilação e previsão para receberem esses dados (BONADUCE et al., 2018;FUJII et al., 2019). Os desafios são grandes, pois o número de dados altimétricos a serem assimilados aumentará em duas ordens de grandeza e os erros associados estarão correlacionados no tempo e no espaço. ...
