Application of a finite element model (TELEMAC) to computing the wind induced response of the Irish Sea

Initially a brief overview of the problem of computing the wind-induced circulation on the west coast of Britain is reviewed together with storm surge modelling. To date this work has primarily been performed with finite difference models. However, here new work is presented using a finite element model with a range of mesh refinements in shallow water regions to examine the influence of mesh resolution upon the wind-induced circulation off the west coast of Britain. Steady state current fields are computed for uniform westerly and southerly winds and compared with a uniform grid (of order 7 km) finite difference model solution. Calculations show that in deep water regions away from the coastal influence, the large-scale circulation features in the finite element solution are in good agreement with those found in the finite difference model. This suggests that they can be adequately resolved on a 7 km mesh. In the nearshore region and within estuaries a significantly finer mesh is required, with the variable mesh finite element model showing significant small scale variability in the nearshore area. Refining the mesh in the Mersey and using an accurate topographic data set, shows that although the larger scale features in the estuary can be resolved in the coarser mesh model, accurate topography is required to model their exact location. In addition smaller scale features are found that were not resolved in the coarser mesh models. Due to the effects of “wetting and drying” and the importance of non-linear processes in shallow regions difficulties occurred in de-tiding the full solution in order to determine the wind forced residual. Determining the wind forced solution in shallow water from a calculation in which wind and tidal forcing are included poses problems as to how to “de-tide” the solution in such a highly non-linear region. An approach based upon the harmonic analysis of the total solution, rather than subtracting a “tide only” solution is shown to be most effective and has implications for storm surge prediction.General and specific conclusions on the importance of highly accurate bathymetry, good mesh resolution and de-tiding method upon the accuracy of the wind forced solution in nearshore regions are summarized in the final part of the paper. The implications for storm surge prediction together with suggestions for future research to enhance the accuracy of storm surge prediction, namely “the way forward” are given at the end of the paper.     

Storm surge computations for the west coast of Britain using a finite element model (TELEMAC)

An unstructured mesh finite element model of the sea region off the west coast of Britain is used to examine the storm surge event of November 1977. This period is chosen because accurate meteorological data to drive the model and coastal observations for validation purposes are available. In addition, previous published results from a coarse-grid (resolution 7 km) finite difference model of the region and high-resolution (1 km) limited area (namely eastern Irish Sea) model are available for comparison purposes. To enable a “like with like” comparison to be made, the finite element model covers the same domain and has the same meteorological forcing as these earlier finite difference models. In addition, the mesh is based on an identical set of water depths. Calculations show that the finite element model can reproduce both the “external” and “internal” components of the surge in the region. This shows that the “far field” (external) component of the surge can accurately propagate through the irregular mesh, and the model responds accurately, without over- or under-damping, to local wind forcing. Calculations show significant temporal and spatial variability in the surge in close agreement with that found in earlier finite difference calculations. In addition, root mean square errors between computed and observed surge are comparable to those found in previous finite different calculations. The ability to vary the mesh in nearshore regions reveals appreciable small-scale variability that was not found in the previous finite difference solutions. However, the requirement to perform a “like with like” comparison using the same water depths means that the full potential of the unstructured grid model to improve resolution in the nearshore region is inhibited. This is clearly evident in the Mersey estuary region where a higher resolution unstructured mesh model, forced with uniform winds, had shown high topographic variability due to small-scale variations in topography that are not resolved here. Despite the lack of high resolution in the nearshore region, the model showed results that were consistent with the previous storm surge models of the region. Calculations suggest that to improve on these earlier results, a finer nearshore mesh is required based upon accurate nearshore topography.     

On the sensitivity of computed higher tidal harmonics to mesh size in a finite element model

A finite element model of the Irish and Celtic Sea regions with a range of grid resolutions is used to examine the influence of resolution upon the higher harmonics of the tide in the region. Comparisons are also made with published results from finite difference models of the area, and observations. Calculations using fine near-shore elements with non-zero water depths in coastal regions were found to be more accurate and less time consuming than those using a zero coastal water depth. A detailed examination of the spatial variability of the higher harmonics in near-shore regions of the eastern Irish Sea particularly the Solway and Morecambe Bay showed significant small-scale variability. This together with the variation in higher harmonics in the eastern Irish Sea and adjacent estuaries, clearly shows the need for an unstructured grid model of the region that can include the estuaries. To match the high resolution of the model in near-shore regions accurate high-resolution topography is required.     

A comparison of continuous mass‐lumped finite elements with finite differences for 3‐D wave propagation

The finite‐difference method on rectangular meshes is widely used for time‐domain modelling of the wave equation. It is relatively easy to implement high‐order spatial discretization schemes and parallelization. Also, the method is computationally efficient. However, the use of finite elements on tetrahedral unstructured meshes is more accurate in complex geometries near sharp interfaces. We compared the standard eighth‐order finite‐difference method to fourth‐order continuous mass‐lumped finite elements in terms of accuracy and computational cost. The results show that, for simple models like a cube with constant density and velocity, the finite‐difference method outperforms the finite‐element method by at least an order of magnitude. Outside the application area of rectangular meshes, i.e., for a model with interior complexity and topography well described by tetrahedra, however, finite‐element methods are about two orders of magnitude faster than finite‐difference methods, for a given accuracy.     

面向目标自适应海洋可控源电磁三维矢量有限元正演

本文实现了一种面向目标自适应海洋可控源电磁三维矢量有限元方法.为满足三维复杂电性结构模拟的需求,网格剖分采用非结构化六面体.在组装刚度矩阵之后,形成的大型复数线性方程组分解为等价的实数形式,利用带预条件的广义最小残差法进行求解.在获得微分方程的解之后,为提高解的准确性,通过面向目标的自适应误差估计来指示网格细化,重点加密能使观测点数值模拟精度提高的网格.对于大规模三维数据,为了使模型空间的并行计算达到均衡负载的效果,我们使用METIS函数库来进行网格计算任务量的划分.最后,通过对比一维解析解与三维自适应矢量有限元计算结果,验证了程序的正确性;通过自适应过程中误差指示子的分布,验证了面向目标自适应的有效性;通过对三维复杂模型进行均衡负载下的并行计算,测试了程序的可扩展性.     

The use of reduced finite element models in system identification

An a priori, analytical model in system identification of vibrating structures is improved by input and output measurements with least square fitting. The structure is modelled by the finite element method. Finite element models usually need a large number of degrees of freedom to simulate a small number of lower spectrum eigenfrequencies with accuracy. The large finite element model is reduced to a subspace of significant eigenfrequencies. The size of the subspace is chosen with regard to the frequency content of the measured data and the accuracy of the large analytical model. An identification method is formulated for the large analytical model. This procedure improves system parameters in the matrices of the large model by measured input and output data with a least square functional. The objective function is consistently reduced, so that the whole identification procedure can be performed in the small subspace. The proposed reduction method permits very large and accurate analytical models to be used, and it decreases the computational cost of the identification procedure significantly. The computational efficiency is demonstrated on an in situ experiment of a radar tower.     

A model study of tidal distributions in the Celtic and Irish Sea regions determined with finite volume and finite element models

An unstructured mesh tidal model of the west coast of Britain, covering the Celtic Sea and Irish Sea is used to compare tidal distributions computed with finite element (FE) and finite volume (FV) models. Both models cover an identical region, use the same mesh, and have topography and tidal boundary forcing from a finite difference model that can reproduce the tides in the region. By this means, solutions from both models can be compared without any bias towards one model or another. Two-dimensional calculations show that for a given friction coefficient, there is more damping in the FV model than the FE model. As bottom friction coefficient is reduced, the two models show comparable changes in tidal distributions. In terms of mesh resolution, calculations show that for the M₂ tide, the mesh is sufficiently fine to yield an accurate solution over the whole domain. However, in terms of higher harmonics of the tide, in particular the M₆ component, its small-scale variability in near-shore regions which is comparable to the mesh of the model, suggests that the mesh resolution is insufficient in the near-coastal regions. Even with a finer mesh in these areas, without detailed bottom topography and a spatial varying friction depending on bed types and bed forms, which is not available, model skill would probably not be improved. In addition in the near-shore region, as shown in the literature, the solution is sensitive to the form of the wetting/drying algorithm used in the model. Calculations with a 3D version of the FV model show that for a given value of k, damping is reduced compared to the 2D version due to the differences in bed stress formulation, with the 3D model yielding an accurate tidal distribution over the region.     

Tides in the English Channel and Southern North Sea. A frequency domain analysis using model TEA-NL

The frequency-time finite element model TEA-NL is used to investigate tidal propagation and circulation in the English Channel and the southern North Sea, in the context of the Tidal Flow Forum benchmark. Quality of TEA-NL simulations, relative to both field data and simulations by other numerical models is discussed. Sensitivity analyses are performed with regard to the effect on accuracy of bottom friction, and the effect on computational cost of the strategy used to harmonically decompose and iterate non-linear loading terms in the shallow water equations.     

Modelling tide–surge interaction effects using finite volume and finite element models of the Irish Sea

An unstructured mesh model of the west coast of Britain, covering the same domain and using topography and open boundary forcing that are identical to a previous validated uniform grid finite difference model of the region, is used to compare the performance of a finite volume (FV) and a finite element (FE) model of the area in determining tide–surge interaction in the region. Initial calculations show that although qualitatively both models give comparable tidal solutions in the region, comparison with observations shows that the FV model tends to under-estimate tidal amplitudes and hence background tidal friction in the eastern Irish Sea. Storm surge elevations in the eastern Irish Sea due to westerly, northerly and southerly uniform wind stresses computed with the FV model tend to be slightly higher than those computed with the FE model, due to differences in background tidal friction. However, both models showed comparable non-linear tide–surge interaction effects for all wind directions, suggesting that they can reproduce the extensive tide–surge interaction processes that occur in the eastern Irish Sea. Following on from this model comparison study, the physical processes contributing to surge generation and tide–surge interaction in the region are examined. Calculations are performed with uniform wind stresses from a range of directions, and the balance of various terms in the hydrodynamic equations is examined. A detailed comparison of the spatial variability of time series of non-linear bottom friction and non-linear momentum advection terms at six adjacent nodes at two locations in water depths of 20 and 6 m showed some spatial variability from one node to another. This suggests that even in the near coastal region, where water depths are of the order of 6 m and the mesh is fine (of order 0.5 km), there is significant spatial variability in the non-linear terms. In addition, distributions of maximum bed stress due to tides and wind forcing in nearshore regions show appreciable spatial variability. This suggests that intensive measurement campaigns and very high-resolution mesh models are required to validate and reproduce the non-linear processes that occur in these regions and to predict extreme bed stresses that can give rise to sediment movement. High-resolution meshes will also be required in pollution transport problems.     

Adaptive-grid simulation of groundwater flow in heterogeneous aquifers

The prediction of contaminant transport in porous media requires the computation of the flow velocity. This work presents a methodology for high-accuracy computation of flow in a heterogeneous isotropic formation, employing a dual-flow formulation and adaptive gridding. The dual equations, describing the hydraulic head and the streamfunction, are numerically solved through finite element approximations. The application of classic finite-element methods requires a rather large number of nodes to represent suitably the flow in high-contrast formations. We present a mesh-adaptive approach that enhances the accuracy of the numerical flow solution for a given computational effort. We rely on an a posteriori error estimator to identify areas where refinements of the finite element mesh are needed or unrefinements are acceptable. We also demonstrate through numerical experiments that the developed methodology efficiently enhances accuracy through successive mesh adaptation.     

基于自适应网格的仿真型有限差分地震波数值模拟

在复杂山地和复杂海底条件下,地表和海底的剧烈起伏对地震波数值模拟提出了更高的要求.常规有限差分法采用矩形网格对模型进行网格剖分,由于矩形网格自身的限制,起伏地表或起伏海底只能由一系列阶梯状折线代替,从而引起人为虚假绕射波.此外,在模拟液-固界面的反射波时,如果界面与网格线不一致,则需要更密的网格才能得到精确的结果.为了解决上述问题,本文将自适应网格生成技术引入到起伏海底速度模型的网格剖分中,采用高阶仿真型有限差分法(MFD)对曲线坐标下的声波方程波进行了数值模拟.利用自适应网格生成技术对速度模型进行网格剖分不仅可以准确地描述模型边界,而且可以有效消除虚假绕射波.高阶仿真型有限差分法可以有效压制频散提高计算精度.模型试算结果表明,本文方法对复杂海底模型具有很好的适应性.     

深度均匀采样梯形网格有限差分地震波场模拟方法

由于重力引起的岩石压实效应,一般来说,地震波传播速度由浅入深整体逐渐增大.梯形坐标系设计可耦合速度由浅入深逐渐增大的变化,该坐标系中均匀网格采样所对应的物理直角坐标系网格由浅入深逐渐增大,也即浅部低速区对应细网格,深部高速区对应粗网格.在梯形坐标系表征波动方程后利用有限差分求解,本文实现一种深度均匀采样、横向采样间隔随深度增加逐渐线性增大的有限差分地震波模拟方法.梯形坐标系波动方程离散后,仍采用常规均匀网格有限差分算法对其求解.由于横向网格大小由浅入深线性增加,本方法可避免不同大小网格区域过渡所产生的虚假反射.梯形坐标系波场模拟浅层精度高,深层横向响应范围广,可有效减少有限差分网格数量.本文提出的方法是在更广义的坐标系下利用有限差分求解波动方程,正交坐标系仅为该梯形坐标系之特例.本文旨在为大速度动态范围深地高效高精度地震波场模拟提供一种思路.     

Dynamic responses of coastal structures during earthquakes including sediment–sea–structure interaction

The effects of the interaction among sea water, sediment, backfill-soil and coastal structures (embankments) were included in the present study. The formulation is derived from fundamental theories in various fields, including marine hydrodynamics, flow in porous medium, and structural dynamics. The hybrid finite-difference and finite element methods were used in the analysis. The finite-difference method was used to calculate the nonlinear hydrodynamic pressures of sea water as well as the pore water in the sediment acting on the coastal embankment faces by seismic-wave actions. The fluid-filled solid mixture was used to model sediment and back-fill soil and the corresponding dynamic responses were also evaluated by finite difference method. The dynamic response of the coastal structures was calculated by finite element method. The numerical results are presented for various water depths and ground motion intensities. The significant dynamic forces on coastal structures were calculated during earthquakes and the possible sliding of the coastal embankment will occur and the special foundation treatment should be made.     

Using the extended finite element method for simulation of transient well leakage in multilayer aquifers

The extended finite element (XFEM) is applied to the problem of transient leakage from abandoned or free-flowing artesian wells in perforated aquifer-aquitard systems. To more accurately capture the singularities in potentiometric head at the wells, the standard linear finite element basis is locally augmented with asymptotic analytical solutions which enable more accurate calculations of leakage rates between aquifers. Highly accurate flux estimates are obtained without the need for higher mesh resolution near wells. Simulations are carried out to test both the accuracy and convergence properties of the XFEM implementation, and the XFEM results are compared to those of a high-resolution standard finite element model. It is seen that for the type of singularity-driven problem posed here, the standard FEM is unable to resolve leakage rates without very fine discretization, but that the XFEM performs robustly with fewer degrees of freedom. The impact of aquifer geometric heterogeneity on leakage rates is assessed and seen to be an important factor in determining total leakage. It is demonstrated that the XFEM may be a valuable tool in many water resources applications where small-scale effects can impact global system behavior.     

稀疏存储的显式有限元三角网格地震波数值模拟及其PML吸收边界条件

有限元法是复杂介质地震模拟的有力工具,它能比较客观地反映地震波的传播,比较细致地再现地震图像.但是,为了获得较精确的结果,有限元法模拟地震波的传播需要的网格点数多,具有计算量大和消耗内存多的缺点.针对上述缺点,本文对刚度矩阵采用压缩存储行(CSR)格式,以减少计算量并节省内存;采用集中质量矩阵得到对角的质量矩阵以提高有限元法(显式有限元)的计算效率;时间离散采用保能量的Newmark算法以提高有限元法的计算精度;采用变分形式(弱形式)的PML吸收边界条件对人工截断边界进行处理.通过与高精度的数值方法--谱元法的数值试验的对比表明,上述方法的引入可使有限元法在计算精度和计算效率方面均可取得比较显著的改进.为了获得相当的计算精度,相比于7阶谱元法,显式有限元法需要更精细的网格.然而,显式有限元法的计算速度比前者快近2倍,而内存需求仅为谱元法的1/4~1/6.     

On the sensitivity of tidal residuals off the west coast of Britain to mesh resolution

An irregular mesh model of the west coast of Britain is used to examine the sensitivity of tidal residuals to mesh resolution in the region. Computed residuals are compared with earlier published results determined with a high resolution (1 km grid) finite difference model of the eastern Irish Sea. Initial calculations show that tidal residuals are largest in nearshore regions particularly in the vicinity of headlands. Local refinement of the mesh in these regions leads to a more detailed picture of the flow field, particularly adjacent to the coast. Although large scale offshore features of the flow can be resolved using the high resolution finite difference model, such an approach leads to a “stair case” representation of the coastal boundary with an adjacent near coastal region of spurious tidal residuals. By using an irregular mesh that follows the coast, this effect is removed. In the Mersey river region the tidal residual is resolved with a mesh resolution of 120 m, although calculations show that its distribution is particularly sensitive to small scale features of the topography. A variable mesh that can accurately represent the lateral variations in river width and details of topography in both the nearshore and estuarine environment appears essential in modelling the coastal spread of freshwater plumes from rivers and pollutants discharged into the near coastal environment.     

An inter-comparison of tidal solutions computed with a range of unstructured grid models of the Irish and Celtic Sea Regions

Three finite element codes, namely TELEMAC, ADCIRC and QUODDY, are used to compute the spatial distributions of the M₂, M₄ and M₆ components of the tide in the sea region off the west coast of Britain. This region is chosen because there is an accurate topographic dataset in the area and detailed open boundary M₂ tidal forcing for driving the model. In addition, accurate solutions (based upon comparisons with extensive observations) using uniform grid finite difference models forced with these open boundary data exist for comparison purposes. By using boundary forcing, bottom topography and bottom drag coefficients identical to those used in an earlier finite difference model, there is no danger of comparing finite element solutions for “untuned unoptimised solutions” with those from a “tuned optimised solution”. In addition, by placing the open boundary in all finite element calculations at the same location as that used in a previous finite difference model and using the same M₂ tidal boundary forcing and water depths, a like with like comparison of solutions derived with the various finite element models was possible. In addition, this open boundary was well removed from the shallow water region, namely the eastern Irish Sea where the higher harmonics were generated. Since these are not included in the open boundary, forcing their generation was determined by physical processes within the models. Consequently, an inter-comparison of these higher harmonics generated by the various finite element codes gives some indication of the degree of variability in the solution particularly in coastal regions from one finite element model to another. Initial calculations using high-resolution near-shore topography in the eastern Irish Sea and including “wetting and drying” showed that M₂ tidal amplitudes and phases in the region computed with TELEMAC were in good agreement with observations. The ADCIRC code gave amplitudes about 30 cm lower and phases about 8° higher. For the M₄ tide, in the eastern Irish Sea amplitudes computed with TELEMAC were about 4 cm higher than ADCIRC on average, with phase differences of order 5°. For the M₆ component, amplitudes and phases showed significant small-scale variability in the eastern Irish Sea, and no clear bias between the models could be found. Although setting a minimum water depth of 5 m in the near-shore region, hence removing wetting and drying, reduced the small-scale variability in the models, the differences in M₂ and M₄ tide between models remained. For M₆, a significant reduction in variability occurred in the eastern Irish Sea when a minimum 5-m water depth was specified. In this case, TELEMAC gave amplitudes that were 1 cm higher and phases 30° lower than ADCIRC on average. For QUODDY in the eastern Irish Sea, average M₂ tidal amplitudes were about 10 cm higher and phase 8° higher than those computed with TELEMAC. For M₄, amplitudes were approximately 2 cm higher with phases of order 15° higher in the northern part of the region and 15° lower in the southern part. For M₆ in the north of the region, amplitudes were 2 cm higher and about 2 cm lower in the south. Very rapid M₆ tidal-phase changes occurred in the near-shore regions. The lessons learned from this model inter-comparison study are summarised in the final section of the paper. In addition, the problems of performing a detailed model–model inter-comparison are discussed, as are the enormous difficulties of conducting a true model skill assessment that would require detailed measurements of tidal boundary forcing, near-shore topography and precise knowledge of bed types and bed forms. Such data are at present not available.     

Modelling estuarine and coastal flows using an unstructured triangular finite volume algorithm

Details are given of the development and application of a numerical model for predicting free-surface flows in estuarine and coastal basins using the finite volume method. Both second- and third-order accurate and oscillation free explicit numerical schemes have been used to solve the shallow water equations. The model deploys an unstructured triangular mesh and incorporates two types of mesh layouts, namely the ‘cell centred’ and ‘mesh vertex’ layouts, and provides a powerful mesh generator in which a user can adjust the mesh-size distribution interactively to create a desirable mesh. The quality of mesh has been shown to have a major impact on the overall performance of the numerical model.The model has been applied to simulate two-dimensional dam break flows for which transient water level distributions measured within a laboratory flume were available. In total 12 model runs were undertaken to test the model for various flow conditions. These conditions include: (1) different bed slopes (ranging from zero to 0.8%), (2) different upstream and downstream water level conditions, and (3) initially wet and dry bed conditions, downstream of the dam. Detailed comparisons have been made between model predicted and measured water levels and good agreement achieved between both sets of results. The model was then used to predict water level and velocity distributions in a real estuary, i.e. the Ribble Estuary, where the bed level varies rapidly at certain locations. In order to model the whole estuary, a 1-D numerical model has also been used to model the upper part of the estuary and this model was linked dynamically to the 2-D model. Findings from this application are given in detail.     

Application of a finite element model to the computation of tides in the Mersey Estuary and Eastern Irish Sea

A variable mesh finite element model of the Irish and Celtic Sea regions with/without the inclusion of the Mersey estuary is used to examine the influence of grid resolution and the Mersey upon the higher harmonics of the tide in the region. Comparisons are made with observations and published results from finite difference models of the area. Although including a high resolution representation of the Mersey had little effect upon computed tides in the western Irish Sea it had a significant effect upon tidal currents in the eastern Irish Sea. In addition the higher harmonics of the M₂ tide in near-shore regions of the eastern Irish Sea particularly the Solway and Mersey estuary together with Morecambe Bay showed significant small scale variability. The Mersey was used to test the sensitivity to including estuaries because high resolution accurate topography was available. The results presented here suggest that comparable detailed topographic data sets are required in all estuaries and near-shore regions. In addition comparisons clearly show the need for an unstructured grid model of the region that can include all the estuaries. Such an unstructured grid solution was developed here within a finite element approach, although other methods in particular the finite volume, or coordinate transformations/curvilinear grids and nesting could be applied.     

Temperature-driven fluid flow in porous media using a mixed finite element method and a finite volume method

We present a numerical scheme for the computation of conservative fluid velocity, pressure and temperature fields in a porous medium. For the velocity and pressure we use the primal–dual mixed finite element method of Trujillo and Thomas while for the temperature we use a cell-centered finite volume method. The motivation for this choice of discretization is to compute accurate conservative quantities. Since the variant of the mixed finite element method we use is not commonly used, the numerical schemes are presented in detail. We sketch the computational details and present numerical experiments that justify the accuracy predicted by the theory.