Data-Driven Interpolation of Sea Level Anomalies Using Analog Data Assimilation

Lguensat, Redouane; Viet, Phi Huynh; Sun, Miao; Chen, Ge; Tian, Fenglin; Chapron, Bertrand; Fablet, Ronan

doi:10.3390/rs11070858

Cited by 27 publications

(34 citation statements)

References 46 publications

(85 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Beyond the reconstruction of gapfilled SST fields, we believe that the reported experiments illustrate the potential of PB-AnDA models for the reconstruction of geophysical products from remote sensing data, especially other sea surface tracers such as SSH (Sea Surface Height), SSS (Sea Surface Salinity) and ocean colour, as well as atmospheric variables. It might be noted that our recent application on the interpolation of altimeter-derived SSH fields further supports this potential [27]. For such applications, the relevance of PB-AnDA models is expected to strongly depend on one hand on the availability of large-scale simulation or observation-driven datasets to build representative catalogs of exemplars of the range of space-time scales of interests, and, on the other hand, on the validity of the assumption that state dynamics are locally-linear with respect to the considered regressors.…”

Section: Discussionmentioning

confidence: 54%

“…The lack of large-scale dataset along with the computational complexity of analog methods has long limited their applicability. In this context, we recently introduced the analog data assimilation (AnDA) and demonstrated its relevance for the reconstruction of complex dynamical systems for partial observations, including sea surface dynamics [26,27]. Here, as stated in the next section, we further explore and evaluate AnDA schemes for the reconstruction of cloud-free SST fields from satellite-derived measurements.The remainder of the paper is organized as follows.…”

mentioning

confidence: 99%

See 1 more Smart Citation

Spatio-Temporal Interpolation of Cloudy SST Fields Using Conditional Analog Data Assimilation

et al. 2018

Self Cite

View full text Add to dashboard Cite

The ever increasing geophysical data streams pouring from earth observation satellite missions and numerical simulations along with the development of dedicated big data infrastructure advocate for truly exploiting the potential of these datasets, through novel data-driven strategies, to deliver enhanced satellite-derived gapfilled geophysical products from partial satellite observations. We here demonstrate the relevance of the analog data assimilation (AnDA) for an application to the reconstruction of cloud-free level-4 gridded Sea Surface Temperature (SST). We propose novel AnDA models which exploit auxiliary variables such as sea surface currents and significantly reduce the computational complexity of AnDA. Numerical experiments benchmark the proposed models with respect to state-of-the-art interpolation techniques such as optimal interpolation and EOF-based schemes. We report relative improvement up to 40%/50% in terms of RMSE and also show a good parallelization performance, which supports the feasibility of an upscaling on a global scale.filled and the fine-scale variability of the SST fields. Physically-driven data assimilation models [20] may outperform OI if relevant dynamical priors can be defined [21]. The trade-off to be considered between the complexity and genericity of this physical prior remains however complex, especially when considering the assimilation of a single sea surface tracer as SST.Besides model-driven schemes, the ever increasing availability of satellite-derived data and of simulation data from high-resolution ocean models has paved the way for the development of data-driven methods. EOF-based models were among the early and perhaps most popular data-driven methods applied to the reconstruction of SST fields from cloudy SST data [14,15,22] as well of other sea surface tracers such as ocean colour [22]. EOF-based approaches are particularly appealing for ocean remote sensing as they relate to a model of the covariance structure of the considered fields and may adapt to any type of geometry of missing data and interpolation grid. Their use is also motivated by their ability to decompose the spatiotemporal variability of the sea surface fields according to different modes, which may be interpreted geophysically. A renewed interest can also be noticed for analog schemes and applications to forecasting and assimilation issues [23,24]. Analog schemes, proposed a long time ago in geoscience [25], rely on the idea that the dynamics of a given system may repeat to some extent. Given a set of previously observed or analysed data, one may retrieve examples similar to a current state in this set, such that the future of this current state may be forecasted from the known evolution of these similar situations. The lack of large-scale dataset along with the computational complexity of analog methods has long limited their applicability. In this context, we recently introduced the analog data assimilation (AnDA) and demonstrated its relevance for the reconstruction of complex dynamical systems fo...

show abstract

Section: Discussionmentioning

confidence: 54%

mentioning

confidence: 99%

Spatio-Temporal Interpolation of Cloudy SST Fields Using Conditional Analog Data Assimilation

et al. 2018

Self Cite

View full text Add to dashboard Cite

show abstract

“…Analog methods are one of the first data-driven techniques developed within a data assimilation framework [19]. In our recent study [20][21][22], we proved the relevance of such data-driven approache when addressing the spatio-temporal interpolation of sea surface geophysical tracers. Combining analog data assimilation (AnDA) with a patch-based representation have shown great results with respect to the state-of-the-art OI and EOF-based schemes.…”

Section: Introductionmentioning

confidence: 84%

“…For this reason, a global representation of the spatio-temporal field is most likely to fail due to computational limitations. Following our previous works on analog data assimilation [21,22], we consider a patch-based representation as sketched in Figure 1 (A patch is a P × P subregion of a 2D field with P the width and the height of the patch). This patch-based representation is fully embedded in the considered NN architecture to make explicit both the extraction of the patches from a 2D field and the reconstruction of a 2D field from the collection of patches.…”

Section: Patch-based Nn Architecturementioning

confidence: 99%

Neural Network Based Kalman Filters for the Spatio-Temporal Interpolation of Satellite-Derived Sea Surface Temperature

et al. 2018

Self Cite

View full text Add to dashboard Cite

The forecasting and reconstruction of oceanic dynamics is a crucial challenge. While model driven strategies are still the state-of-the-art approaches in the reconstruction of spatio-temporal dynamics. The ever increasing availability of data collections in oceanography raised the relevance of data-driven approaches as computationally efficient representations of spatio-temporal fields reconstruction. This tools proved to outperform classical state-of-the-art interpolation techniques such as optimal interpolation and DINEOF in the retrievement of fine scale structures while still been computationally efficient comparing to model based data assimilation schemes. However, coupling this data-driven priors to classical filtering schemes limits their potential representativity. From this point of view, the recent advances in machine learning and especially neural networks and deep learning can provide a new infrastructure for dynamical modeling and interpolation within a data-driven framework. In this work we adress this challenge and develop a novel Neural-Network-based (NN-based) Kalman filter for spatio-temporal interpolation of sea surface dynamics. Based on a data-driven probabilistic representation of spatio-temporal fields, our approach can be regarded as an alternative to classical filtering schemes such as the ensemble Kalman filters (EnKF) in data assimilation. Overall, the key features of the proposed approach are two-fold: (i) we propose a novel architecture for the stochastic representation of two dimensional (2D) geophysical dynamics based on a neural networks, (ii) we derive the associated parametric Kalman-like filtering scheme for a computationally-efficient spatio-temporal interpolation of Sea Surface Temperature (SST) fields. We illustrate the relevance of our contribution for an OSSE (Observing System Simulation Experiment) in a case-study region off South Africa. Our numerical experiments report significant improvements in terms of reconstruction performance compared with operational and state-of-the-art schemes (e.g., optimal interpolation, Empirical Orthogonal Function (EOF) based interpolation and analog data assimilation).

show abstract

“…Following our previous works on analog data assimilation [18,15], we consider a patch-based representation as a mean to decompose space-time scales as well as to provide an explicit relationship between the global and local (patch-level) representations. Regarding the mean model F, the proposed architecture proceeds as follows :…”

Section: Proposed Stochastic Dynamical Modelmentioning

confidence: 99%

Learning Stochastic Representations of Geophysical Dynamics

Ouala

Fablet

Herzet

et al. 2019

ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

Self Cite

View full text Add to dashboard Cite

In the last years, Neural Networks have enriched the stateof-the-art in probabilistic modeling. This is principally due to the advances in deep learning which allow a better understanding of complex systems. However, the stochastic representation of spatio-temporal fields is still an open challenge that may benefit from the recent advances in probabilistic modelization. In this work, we explore neural network to derive a stochastic representation of spatio-temporal dynamical systems based on ensemble forecasting. Trough the implementation of our stochastic model in a classical Kalman filtering scheme, we demonstrate the relevance of the proposed architecture in the reconstruction of geophysical fields with respect to the state-of-the-art approaches.

show abstract

Data-Driven Interpolation of Sea Level Anomalies Using Analog Data Assimilation

Cited by 27 publications

References 46 publications

Spatio-Temporal Interpolation of Cloudy SST Fields Using Conditional Analog Data Assimilation

Spatio-Temporal Interpolation of Cloudy SST Fields Using Conditional Analog Data Assimilation

Neural Network Based Kalman Filters for the Spatio-Temporal Interpolation of Satellite-Derived Sea Surface Temperature

Learning Stochastic Representations of Geophysical Dynamics

Contact Info

Product

Resources

About