Source-linked AI summary
Numerical analysis of data assimilation for slightly compressible flow
Aytekin Çıbık, Rui Fang
TL;DR
The paper addresses numerical prediction errors for fluid flow arising from uncertain initial states and model mismatch. It analyzes finite element discretizations of a velocity–pressure nudging formulation for slightly compressible data assimilated into incompressible Navier–Stokes equations. The analysis predicts exponential decay of initial errors and model error O(H + μ1^-1/2), with numerical experiments confirming the predicted rates.
Problem
Uncertain initial states and governing-model mismatch limit accurate long-term fluid-flow prediction, motivating analysis beyond velocity-only data assimilation.
Method
The paper establishes stability and finite element error estimates for semi-discrete and linearized backward Euler discretizations of joint velocity–pressure nudging.
Results
O(H + μ1^-1/2) model-error bounds are established, and numerical experiments confirm the predicted convergence rates.
Takeaways & Limitations
Balancing the observation-resolution and pressure-nudging terms with μ1=O(H^-2) yields the optimal convergence rate.
Takeaways & Limitations
The analysis assumes an L2-projection observation operator and finite element assumptions including discrete inf–sup properties.
Abstract
from arXiv · showhide
Continuous data assimilation improves flow predictions by continually nudging a model toward available observational data. For slightly compressible flow, a recent model addresses the limitations of velocity-only nudging by assimilating both velocity and pressure data and nudging both quantities into the incompressible Navier-Stokes equations [5]; continuous-in-time error estimates and preliminary experiments show that this joint nudging is effective and substantially reduces the model error relative to velocity-only nudging. Motivated by these results, we carry out the numerical analysis of the model and its finite element discretizations. We establish stability and error estimates for the semi-discrete scheme and for the fully discrete, linearized backward Euler scheme. The analysis shows an infinite predictability horizon: the effect of the initial error decays exponentially in time, and the model error is first order in the observation resolution H and of order $μ_1^{-1/2}$ in the pressure nudging parameter $μ_1$. Balancing these two error terms, we choose $μ_1=\mathcal{O}(H^{-2})$, which yields the optimal convergence rate. Numerical experiments confirm the predicted rates.
1 Introduction
The paper analyzes continuous data assimilation for slightly compressible flow, where velocity and pressure observations are nudged into an incompressible Navier–Stokes model. It develops finite element stability and error analyses under observation-density and nudging assumptions.
- Continuous data assimilation combines observations with mathematical models to improve numerical fluid-flow predictions.
- The nudging framework uses an observation operator associated with spatial resolution H to drive the assimilated solution toward observed states.
- Slightly compressible flow couples pressure and velocity, while the model formally approaches incompressible Navier–Stokes equations as c →∞.
- The analyzed formulation assimilates slightly compressible reference data into incompressible Navier–Stokes equations by nudging both velocity and pressure.
- The analysis assumes sufficiently dense observations, sufficiently large velocity nudging χ, and an L2-projection observation operator.
- The paper establishes stability estimates for continuous, semi-discrete, linearized backward Euler, and BDF2-IMEX formulations.
2 The discrete problems
This section formulates the weak, finite element, and fully discrete versions of the velocity–pressure nudging problem. It also introduces stability results for semi-discrete, backward Euler, and BDF2-IMEX schemes.
- The weak formulation defines both the incompressible reference problem and the nudged velocity–pressure system.
- The paper presents linearized backward Euler and BDF2-IMEX fully discrete time discretizations, sharing a pressure equation.
- The stability analysis covers the semi-discrete nudged velocity and pressure solutions and both fully discrete schemes.
- Pressure nudging combines observed-pressure assimilation through μ1 with unresolved-mode regularization through μ2.
- The discrete stability argument uses the L2-projection property and the decomposition of pressure norms into observed and unresolved components.
3 Error estimates
The finite element analysis establishes semi-discrete and fully discrete error estimates for the velocity–pressure nudging formulation. Initial errors decay exponentially, while discretization and modeling contributions determine the attainable accuracy.
- Error structure: Exponential decay removes the effect of initial velocity and pressure errors over time, yielding an infinite predictability horizon.The decay rate is O(min{χ, μ1}) in the continuous-time analysis, with backward Euler factor (1 + β∆t)^−1 per timestep.
- Parameter choice: O(H + μ1^−1/2) modeling error motivates balancing observation resolution against pressure nudging strength.Choosing μ1 = O(H^−2) balances the two contributions and yields the optimal convergence rate.
- Semi-discrete analysis: The analysis handles pressure-related and continuity-equation advection terms, which are central difficulties in the slightly compressible setting.The error equation combines pressure, nonlinear advection, divergence, and velocity-nudging contributions.
- Fully discrete analysis: O(∆t + H + Hh^(m−1) + h^m) is the reduced fully discrete estimate under μ1 = O(H^−2) and practical mesh-resolution conditions.The conditions include hm ≲ H^2 and χ ≲ H^2h^(−2m−2), with the observation mesh coarser than the computational mesh.
4 Numerical tests
Numerical tests validate the finite element analysis: backward Euler achieves the predicted space–time behavior, observation-mesh effects agree with theory, and pressure regularization materially improves accuracy.
- 4.1 Spatial convergence and space–time balance: O(h2) pressure accuracy persists under spatial refinement, while velocity rates fall from 2.34 to 0.01 when fixed-∆t temporal error dominates.With ∆t = 0.005, the resulting error floor is determined by ∆t rather than final time T.
- 4.1 Spatial convergence and space–time balance: ∆t = h2 yields second-order pressure convergence and velocity rates approaching 2, confirming global second-order accuracy despite first-order time discretization.Balancing O(∆t) = O(h2) prevents the fixed-time-step term from obscuring spatial convergence.
- 4.2 Temporal order: Backward Euler temporal slopes are pv ≈ 1.00 across refinements, while pp approaches 1 as ∆t decreases.The successive-difference estimator isolates temporal order by cancelling the ∆t-independent O(H) contribution.
- 4.3 Effect of the observation mesh size H: 2.00 and 1.16 are the fitted log–log slopes for velocity and pressure errors versus H, respectively, with pressure consistent with the predicted linear O(H) term.Velocity decays faster than O(H) in the tested range, so the model-error term is not dominant for velocity there.
5 Conclusion
The analysis establishes finite element stability and error estimates for semi-discrete and fully discrete backward Euler schemes. Initial errors decay exponentially, while modeling errors scale as O(H + µ1^-1/2), motivating µ1 = O(H^-2) and supporting the observed convergence rates.
- 5 Conclusion: Semi-discrete and fully discrete linearized backward Euler error estimates establish numerical stability for the proposed data-assimilation formulation.The fully discrete analysis uses backward Euler, while the paper also discusses stability for BDF2-IMEX.
- 5 Conclusion: Initial velocity and pressure errors decay exponentially in time, with backward Euler decay factor (1 + β∆t)^-1 per timestep.The decay rate is determined by β = O(χ, µ1).
- 5 Conclusion: O(H + µ1^-1/2) bounds the velocity and pressure errors under the stated assumptions, and µ1 = O(H^-2) provides the optimal scaling.The analysis assumes µ1 ≥ µ2.
- 5 Conclusion: Numerical experiments recover optimal Taylor–Hood spatial rates until temporal error intervenes, while balanced ∆t = h2 delivers global second order.The observation-mesh and pressure-parameter studies also support the theoretical error structure and the assumption µ1 ≥ µ2.