Automated continuous verification for numerical simulation

Farrell, Patrick E.; Piggott, Matthew D.; Gorman, Gerard J.; Ham, David A.; Wilson, C. R.; Bond, Tamzin

doi:10.5194/gmd-4-435-2011

Cited by 33 publications

(31 citation statements)

References 37 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Note that the time step is adaptive and will therefore decrease along with the element size such that this test checks both spatial and temporal convergence. This verifies that the model equations are implemented correctly (Farrell et al, 2011). A qualitative comparison shows that the solution matches the analytical solution very well (Fig.…”

Section: Forward Model Verificationsupporting

confidence: 53%

Application of the adjoint approach to optimise the initial conditions of a turbidity current with the AdjointTurbidity 1.0 model

Parkinson¹,

Funke²,

Piggott³

et al. 2017

Geosci. Model Dev.

View full text Add to dashboard Cite

Abstract. Turbidity currents are one of the main drivers of sediment transport from the continental shelf to the deep ocean. The resulting sediment deposits can reach hundreds of kilometres into the ocean. Computer models that simulate turbidity currents and the resulting sediment deposit can help us to understand their general behaviour. However, in order to recreate real-world scenarios, the challenge is to find the turbidity current parameters that reproduce the observations of sediment deposits. This paper demonstrates a solution to the inverse sediment transportation problem: for a known sedimentary deposit, the developed model reconstructs details about the turbidity current that produced the deposit. The reconstruction is constrained here by a shallow water sediment-laden density current model, which is discretised by the finite-element method and an adaptive time-stepping scheme. The model is differentiated using the adjoint approach, and an efficient gradientbased optimisation method is applied to identify the turbidity parameters which minimise the misfit between the modelled and the observed field sediment deposits. The capabilities of this approach are demonstrated using measurements taken in the Miocene Marnoso-arenacea Formation (Italy). We find that whilst the model cannot match the deposit exactly due to limitations in the physical processes simulated, it provides valuable insights into the depositional processes and represents a significant advance in our toolset for interpreting turbidity current deposits.

show abstract

Section: Forward Model Verificationsupporting

confidence: 53%

Application of the adjoint approach to optimise the initial conditions of a turbidity current with the AdjointTurbidity 1.0 model

Parkinson¹,

Funke²,

Piggott³

et al. 2017

Geosci. Model Dev.

View full text Add to dashboard Cite

show abstract

“…Varying amounts of software testing are conducted throughout the cycle, but formal code verification practices (e.g., see D'Silva et al, 2008) are only recently starting to be considered for climate model development (Clune and Rood, 2011;Farrell et al, 2011). Nonetheless, the concentration on sound science, as opposed to software correctness, has led to climate models that contain fewer software defects than other comparably sized projects (Pipitone and Easterbrook, 2012).…”

Section: D Lucas Et Al: Failure Analysis Of Climate Simulation Cmentioning

confidence: 99%

Failure analysis of parameter-induced simulation crashes in climate models

et al. 2013

View full text Add to dashboard Cite

Abstract. Simulations using IPCC (Intergovernmental Panel on Climate Change)-class climate models are subject to fail or crash for a variety of reasons. Quantitative analysis of the failures can yield useful insights to better understand and improve the models. During the course of uncertainty quantification (UQ) ensemble simulations to assess the effects of ocean model parameter uncertainties on climate simulations, we experienced a series of simulation crashes within the Parallel Ocean Program (POP2) component of the Community Climate System Model (CCSM4). About 8.5 % of our CCSM4 simulations failed for numerical reasons at combinations of POP2 parameter values. We applied support vector machine (SVM) classification from machine learning to quantify and predict the probability of failure as a function of the values of 18 POP2 parameters. A committee of SVM classifiers readily predicted model failures in an independent validation ensemble, as assessed by the area under the receiver operating characteristic (ROC) curve metric (AUC > 0.96). The causes of the simulation failures were determined through a global sensitivity analysis. Combinations of 8 parameters related to ocean mixing and viscosity from three different POP2 parameterizations were the major sources of the failures. This information can be used to improve POP2 and CCSM4 by incorporating correlations across the relevant parameters. Our method can also be used to quantify, predict, and understand simulation crashes in other complex geoscientific models.

show abstract

“…Testing is often a secondary consideration to new feature implementation, so it is important that extension of the testing suite is as simple as possible. This can simply be run at the time of a new installation, following the upgrade of required libraries or the operating system, or routinely as part of a commit-hook buildbot with dedicated resources to continuously verify new code pushed to a Shingle development code repository (see, for example, Farrell et al, 2011). Being built on standard libraries, it could further form part of an automated wider system framework validation, for the above climate intercomparison projects, for example, reproducing the entire process from initialisation to post-processing, on demand.…”

Section: Continuous Verificationmentioning

confidence: 99%

Shingle 2.0: generalising self-consistent and automated domain discretisation for multi-scale geophysical models

Candy

Pietrzak

2018

Geosci. Model Dev.

View full text Add to dashboard Cite

Abstract. The approaches taken to describe and develop spatial discretisations of the domains required for geophysical simulation models are commonly ad hoc, model-or application-specific, and under-documented. This is particularly acute for simulation models that are flexible in their use of multi-scale, anisotropic, fully unstructured meshes where a relatively large number of heterogeneous parameters are required to constrain their full description. As a consequence, it can be difficult to reproduce simulations, to ensure a provenance in model data handling and initialisation, and a challenge to conduct model intercomparisons rigorously. This paper takes a novel approach to spatial discretisation, considering it much like a numerical simulation model problem of its own. It introduces a generalised, extensible, selfdocumenting approach to carefully describe, and necessarily fully, the constraints over the heterogeneous parameter space that determine how a domain is spatially discretised. This additionally provides a method to accurately record these constraints, using high-level natural language based abstractions that enable full accounts of provenance, sharing, and distribution. Together with this description, a generalised consistent approach to unstructured mesh generation for geophysical models is developed that is automated, robust and repeatable, quick-to-draft, rigorously verified, and consistent with the source data throughout. This interprets the description above to execute a self-consistent spatial discretisation process, which is automatically validated to expected discrete characteristics and metrics.Library code, verification tests, and examples available in the repository at https://github.com/shingleproject/ Shingle. Further details of the project presented at http:// shingleproject.org.

show abstract

Automated continuous verification for numerical simulation

Cited by 33 publications

References 37 publications

Application of the adjoint approach to optimise the initial conditions of a turbidity current with the AdjointTurbidity 1.0 model

Application of the adjoint approach to optimise the initial conditions of a turbidity current with the AdjointTurbidity 1.0 model

Failure analysis of parameter-induced simulation crashes in climate models

Shingle 2.0: generalising self-consistent and automated domain discretisation for multi-scale geophysical models

Contact Info

Product

Resources

About