自适应网格在大气海洋问题中的初步应用

自适应网格法是80年代兴起的通过求解椭圆型方程的边值问题来数值生成网格的一种新方法。它是在任意形状的区域上求偏微分方程的数值解的一种非常有效的工具。该方法抛弃了等距均匀的差分网格，代之以能够自动地适应所研究问题中解的特征的疏密程度不均的曲线网格。如在边界上计算网格与实际边界相重合，在区域内部可任意调节网格点的疏密程度等。本文扼要地介绍了自适应网格的原理及其构造方法。并将其应用于生成南海区域的计算网格以及数值预报台风路径的自适应网格。     

基于移动网格技术的热融湖动态演化过程数值模拟

基于移动网格技术建立了热融湖动态演化有限元数值模型, 研究了青藏高原多年冻土区典型热融湖动态演化过程, 分析了热融湖半径、深度的变化过程及其对湖底及周围多年冻土温度状况的影响. 结果表明:在移动边界热融湖模型中, 热融湖半径以0.7 m·a^-1的速度近线性地增大; 随着下伏高含冰量冻土的融化, 热融湖深度增加先慢后快, 最后逐渐减小趋于稳定. 热融湖深度和半径从5月末至翌年1月末增加显著, 在2-5月间基本保持稳定. 伴随着热融湖的扩展, 地表边界逐渐演变为湖底边界, 热融湖的热影响范围逐渐增大. 在固定边界热融湖模型中, 其热影响会逐渐趋于稳定, 由于初始尺寸大, 其湖底多年冻土退化速率大于移动边界模型, 而远离湖边的多年冻土退化速率要小于移动边界模型. 如果不考虑热融湖边界随时间的变化, 可能会高估热融湖对湖底多年冻土的热影响, 而低估其对附近多年冻土的热影响.     

拖网选择性对鱼类表型性状的选择作用研究

参照渔业生物学、渔业资源解析学、数量遗传学等有关理论模型,并应用数值模拟方法,对长期选择性捕捞（重点为渔具选择性）是否参与造成了渔业资源的生物学变化（体长、性成熟等）进行某些基础性探索。设计了一个具有一定体长遗传特征的模拟鱼类种群,并利用一种具有固定选择性参数的拖网对其进行长期捕捞。结果表明,多年的持续的捕捞压力使得模拟种群中鱼类体长逐渐减小,减小的幅度随开发率的增大而增大,而终止捕捞后种群的体长并没有发生恢复的迹象,说明这种变化很可能伴随种群遗传的变化,具有自然不可逆性。     

Diagnoses of vertical transport in a three-dimensional finite element model of the tidal circulation around an island

A three-dimensional finite element model is used to investigate the formation of shallow-water eddies in the wake of Rattray Island (Great Barrier Reef, Australia). Field measurements and visual observations show that stable eddies develop in the lee of the island at rising and falling tides. The water turbidity downstream of the island suggests the existence of strong upwelling that would be responsible for carrying bed sediments up to the sea surface. We first propose to look at the upwelling velocity and then use the theory of the age to diagnose vertical transport. The water age is defined as the time elapsed since particles of water left the sea bottom, where the age is prescribed to be zero. Two versions of this diagnosis are considered. Although the model predicts upwelling within the eddies, it is not sufficiently intense to account for vertical transport throughout the water column during the life span of the eddies. As mesh resolution increases, this upwelling does not intensify. However, strong upwelling is then resolved off the island's tips, which is confirmed by the results obtained with the age. This study also shows that the finite element method, together with unstructured meshes, performs well for representing three-dimensional flow past an island.     

Elephants move faster in small fragments of low productivity in Amboseli ecosystems: Kenya

Understanding factors affecting the behaviour and movement patterns of the African elephant is important for wildlife conservation, especially in increasingly human-dominated savanna landscapes. Currently, knowledge on how landscape fragmentation and vegetation productivity affect elephant speed of movement remains poorly understood. In this study, we tested whether landscape fragmentation and vegetation productivity explains elephant speed of movement in the Amboseli ecosystem in Kenya. We used GPS collar data from five elephants to quantify elephant speed of movement for three seasons (wet, dry and transitional). We then used multiple regression to model the relationship between speed of movement and landscape fragmentation, as well as vegetation productivity for each season. Results of this study demonstrate that landscape fragmentation and vegetation productivity predicted elephant speed of movement poorly (R² < 0.4) when used as solitary covariates. However, a combination of the covariates significantly (p < 0.05) explained variance in elephant speed of movement with improved R² values of 0.69, 0.45, 0.47 for wet, transition and dry seasons, respectively.     

A tile-based method for geodesic buffer generation in a virtual globe

Geometric buffers are important for spatial analysis in many applications of geographic information systems (GISs), such as environmental measurement and management, human health, urban planning, etc. Geometric buffer generation algorithms are well studied in the Euclidean space where the buffer distance is measured by Euclidean metrics; however, very few algorithms are available for generating geometric buffers on the terrain surface in a virtual globe where the buffer distance is measured by geodesic metrics. This paper proposes a tile-based method for geodesic buffer generation according to the characteristics of a virtual globe. It extends the vector tile model (VTM) to organize terrain and vector data, and the XYH algorithm is improved to build geodesic distance fields for terrain meshes. Based on the data organization and the improved XYH algorithm, a geodesic buffer is generated via three main steps: selecting and assembling tiles, updating geodesic distance fields and tracing the boundaries of buffer zones. This method is implemented with multi-scale terrain and vector data, and the experimental results show that it is valid and exact and can be applied in practical applications.     

A continuum mechanics‐based framework for boundary and finite element mesh optimization in two dimensions for application in excavation analysis

The determination of the optimum excavation sequences in mining and civil engineering using numerical stress analysis procedures requires repeated solution of large models. Often such models contain much more complexity and geometric detail than required to arrive at an accurate stress analysis solution, especially considering our limited knowledge of rock mass properties. This paper develops an automated framework for estimating the effects of excavations at a region of interest, and optimizing the geometry used for stress analysis. It eliminates or simplifies the excavations in a model while maintaining the accuracy of analysis results. The framework can equally be applied to two‐dimensional boundary and finite element models. The framework will have the largest impact for non‐linear finite element analysis. It can significantly reduce computational times for such analysis by simplifying models. Error estimators are used in the framework to assess accuracy. The advantages of applying the framework are demonstrated on an excavation‐sequencing scenario. Copyright © 2005 John Wiley & Sons, Ltd.     

Simulation of localization failure with strain‐gradient‐enhanced damage mechanics

Strain gradient implies an important characteristic in localized damage deformation, which can be observed in the softening state of brittle materials, and strain gradients constitute the basic behaviours of localization failure area of the materials. The most important point in strain gradient is its damaging function including an internal length scale, which can be used to express the scale effects of mechanical responses of brittle rock mass. By extending the strain gradient theory and introducing an intrinsic material length scale into the constitutive law, the authors develop an isotropic damage model as well as a micro‐crack‐based anisotropic damage model for rock‐like materials in this paper. The proposed models were used to simulate the damage localization under uniaxial tension and plain strain compression, respectively. The simulated results well illustrated the potential of these models in dealing with the well‐known mesh‐sensitivity problem in FEM. In the computation, elements with C₁ continuity have been implemented to incorporate the proposed models for failure localization. When regular rectangle elements are encountered, the coupling between finite difference method (FDM) and conventional finite element method (FEM) is used to avoid large modification to the existing FEM code, and to obtain relatively higher efficiency and reasonably good accuracy. Application of the anisotropic model to the 3D‐non‐linear FEM analysis of Ertan arch dam has been conducted and the results of its numerical simulation coincide well with those from the failure behaviours obtained by Ertan geophysical model test. In this paper, new applications of gradient theories and models for a feasible approach to simulate localized damage in brittle materials are presented. Copyright © 2002 John Wiley & Sons, Ltd.