Feature-based data assimilation in geophysics

Morzfeld, Matthias; Adams, Jesse; Lunderman, Spencer; Orozco, Rafael

doi:10.5194/npg-25-355-2018

Cited by 20 publications

(8 citation statements)

References 59 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…An important feature to mention and not often addressed by DA studies (Morzfeld et al, 2018, is a nice exception), is sensitivity of data assimilation methods to the observations chosen for assimilation. In the case of fSCA assimilations a melt period is defined as this is when the observations provide information about the snow depletion curve (e.g.…”

Section: Melt Period Definitionmentioning

confidence: 99%

Hyper-resolution ensemble-based snow reanalysis in mountain regions using clustering

Fiddes¹,

Aalstad²,

Westermann³

2019

Preprint

View full text Add to dashboard Cite

Abstract. Spatial variability in high-relief landscapes is immense, and grid-based models cannot be run at spatial resolutions to explicitly represent important physical processes. This hampers the assessment of the current and future evolution of important issues such as water availability or mass movement hazards. Here, we present a new processing chain that couples an efficient subgrid method with a downscaling tool and data assimilation method with the purpose to improve numerical simulation of surface processes at multiple spatial and temporal scales in ungauged basins. The novelty of the approach is that while we add 1–2 orders of magnitude of computational cost by ensemble simulations, we save 4–5 orders of magnitude over explicitly simulating a high-resolution grid. This approach makes data assimilation at large spatio-temporal scales feasible. In addition, this approach utilises only freely available global datasets and is therefore able to run globally. We demonstrate marked improvements in estimating snow height and snow water equivalent at various experimental scales using this approach. We propose this as a suitable method for a wide variety of operational and research applications where surface models need to be run at large scales with sparse to non-existent ground observations and with the flexibility to assimilate diverse variables retrieved by EO missions.

show abstract

Section: Melt Period Definitionmentioning

confidence: 99%

Hyper-resolution ensemble-based snow reanalysis in mountain regions using clustering

Fiddes¹,

Aalstad²,

Westermann³

2019

Preprint

View full text Add to dashboard Cite

show abstract

“…A second approach consists in designing f (d) to extract specifically information linked to the structural parameters s using domain knowledge, leading to the so-called insight-driven features (Morzfeld et al, 2018). For instance, Hermans et al (2015) applied an insight-driven approach favoring a combination of inversion and multidimensional scaling (MDS) to extract relevant features for the geological scenario, while Scheidt et al (2015a) used a wavelet transform on seismic reflection data in combination with an L 2 -norm distance as insight-driven feature to update different uncertain geological parameters.…”

Section: Designing Data Features To Inform On Structural Parametersmentioning

confidence: 99%

“…In the latter case, the noise model for the data is no longer valid for the features. Instead of handling this using "perturbed" observations (Hermans et al, 2016;Morzfeld et al, 2018), adaptive KDE can deal with this directly because it works for heteroscedastic and multimodal distributions.…”

Section: Kde and Cross-validation Approachmentioning

confidence: 99%

A cross-validation framework to extract data features for reducing structural uncertainty in subsurface heterogeneity

Lopez-Alvis

Hermans

Nguyen

2019

Advances in Water Resources

View full text Add to dashboard Cite

Spatial heterogeneity is a critical issue in the management of water resources. However, most studies do not consider uncertainty at different levels in the conceptualization of the subsurface patterns, for example using one single geological scenario to generate an ensemble of realizations. In this paper, we represent the spatial uncertainty by the use of hierarchical models in which higher-level parameters control the structure. Reduction of uncertainty in such higher-level structural parameters with observation data may be done by updating the complete hierarchical model, but this is, in general, computationally challenging. To address this, methods have been proposed that directly update these structural parameters by means of extracting lower dimensional representations of data called data features that are informative and applying a statistical estimation technique using these features. The difficulty of such methods, however, lies in the choice and design of data features, i.e. their extraction function and their dimensionality, which have been shown to be case-dependent. Therefore, we propose a cross-validation framework to properly assess the robustness of each designed feature and make the choice of the best feature more objective. Such framework aids also in choosing the values for the parameters of the statistical estimation technique, such as the bandwidth for kernel density estimation. We demonstrate the approach on a synthetic case with cross-hole ground penetrating

show abstract

“…Moreover, the synthetic likelihood approach involves re-creating the likelihood for every new model parameter value, 90 which would require excessive CPU times in our setting. Recent work by Morzfeld et al (2018) describes another feature-vector approach for data assimilation. For more details and comparisons among these approaches, see the discussion below in Section 2.1. https://doi.org/10.5194/gmd-2020-350 Preprint.…”

mentioning

confidence: 99%

“…These methods also require long-time integration of the forward model for each candidate parameter value θ, rather than integration for only one epoch. Morzfeld et al (2018) also discuss several ways of using feature vectors for inference in geophysics. A distinction of the present work is that we use 210 an ECDF-based summary statistic that is provably Gaussian, and we perform extensive Bayesian analysis of the parameter posteriors via novel MCMC methods.…”

mentioning

confidence: 99%

Efficient Bayesian inference for large chaotic dynamical systems

Springer¹,

Haario²,

Susiluoto³

et al. 2020

Preprint

View full text Add to dashboard Cite

Abstract. Estimating parameters of chaotic geophysical models is challenging due to these models' inherent unpredictability. With temporally sparse long-range observations, these models cannot be calibrated using standard least squares or filtering methods. Obvious remedies, such as averaging over temporal and spatial data to characterize the mean behavior, do not capture the subtleties of the underlying dynamics. We perform Bayesian inference of parameters in high-dimensional and computationally demanding chaotic dynamical systems by combining two approaches: (i) measuring model-data mismatch by comparing chaotic attractors, and (ii) mitigating the computational cost of inference by using surrogate models. Specifically, we construct a likelihood function suited to chaotic models by evaluating a distribution over distances between points in the phase space; this distribution defines a summary statistic that depends on the attractor's geometry, rather than on pointwise matching of trajectories. This statistic is computationally expensive to simulate, compounding the usual challenges of Bayesian computation with physical models. Thus we develop an inexpensive surrogate for the log-likelihood via local approximation Markov chain Monte Carlo, which in our simulations reduces the time required for accurate inference by orders of magnitude. We investigate the behavior of the resulting algorithm on model problems, and then use a quasi-geostrophic model to demonstrate its large-scale application.

show abstract

Feature-based data assimilation in geophysics

Cited by 20 publications

References 59 publications

Hyper-resolution ensemble-based snow reanalysis in mountain regions using clustering

Hyper-resolution ensemble-based snow reanalysis in mountain regions using clustering

A cross-validation framework to extract data features for reducing structural uncertainty in subsurface heterogeneity

Efficient Bayesian inference for large chaotic dynamical systems

Contact Info

Product

Resources

About