SeisSol as a Community Code for Reproducible Computational Seismology - CoCoReCS

软件包SeisSol（www.seissol.org）能够在高性能计算平台上以高阶精度模拟计算地震学中的复杂问题。SeisSol具有广泛的应用潜力，对社会和工业有很大的影响，例如模拟诱发地震（例如地热发电厂或CO2储存过程）。一个特别成功的用例是地震的非线性动态破裂过程的多物理模拟，该过程与几何复杂、高度详细的模型中产生的地震波传播相耦合-例如2004年苏门答腊-安达曼大推力地震。2015年，SeisSol已作为开源项目发布（github.com/SeisSol）.虽然有几个试点用户采用了该软件，但已查明以下三个问题妨碍地球科学家广泛采用SeisSol：（1）为SeisSol生成合适的统一几何（CAD）模型表示和高质量离散化网格通常是一项具有挑战性的任务，直到最近才变得可行。(2)虽然SeisSol是开源的（包括一个多功能的软件基础设施和文档），它仍然缺乏安装支持，教程和使用示例，指导潜在的用户，这些用户通常在高性能计算方法方面没有丰富的经验。(3)为了让SeisSol在社区中得到广泛的接受，它需要提供一个开放的平台来托管可重现的模拟场景--从用于代码验证的既定基准，特定建模需求的成功实现到允许按需紧急模拟的生产就绪场景。在这个项目中，我们希望解决所有三个障碍：（1）提供一个记录良好的工作流程，使用商业和开源CAD和网格化软件生成复杂地球结构的网格。(2)为计算地震学社区的潜在用户编制培训材料和（面对面和在线）介绍课程。(3)设计和安装一个基础设施，以维护广泛的可重现的模拟场景，从用于持续软件验证的基准模拟到允许生产运行以进行地震灾害评估的场景。我们将建立的最佳做法对超级计算中心特别有意义，这些中心目前缺乏维护模拟软件以及模型情景和实时计算输入数据的程序。

The software package SeisSol (www.seissol.org) enables the simulation of complex problems in computational seismology with high-order accuracy on high-performance computing platforms. SeisSol has excellent potential for a wide range of applications with high impact on society and industry, such as the simulation of induced earthquakes (by geothermal power plants or CO2 storage processes, e.g.) and physics-based seismic hazard assessment.A particularly successful use-case is the multi-physics simulation of the non-linear dynamic rupture process of earthquakes coupled to the resulting seismic wave propagation in geometrically complex, highly detailed models -- such as the 2004 Sumatra-Andaman Megathrust Earthquake.In 2015, SeisSol has been released as an open-source project (github.com/SeisSol).While several pilot users have taken up the software, the following three issues have been identified to hinder a widespread adoption of SeisSol by Earth scientists:(1) Generating suitable unified geometric (CAD) model representations and high-quality discretisation meshes for SeisSol is, in general, a challenging task, which only recently became feasible.(2) While SeisSol is available as open-source (incl. a versatile software infrastructure and documentation), it still lacks installation support, tutorials and usage examples that guide potential users which are in general not highly experienced in high performance computing approaches. (3) For a widespread acceptance of SeisSol in the community it needs to provide an open platform to host reproducible simulation scenarios -- from established benchmarks for code validation, successful implementation of specific modelling requirements to production-ready scenarios allowing on-demand urgent simulations.In this project, we want to tackle all three roadblocks:(1) Provide a well-documented workflow to generate meshes of complex Earth structures, using commercial and open-source CAD and meshing software.(2) Establish training material and (face-to-face and online) introduction courses for potential users in the Computational Seismology community. (3) Design and install an infrastructure that maintains a wide range of reproducible simulation scenarios, ranging from benchmark simulations for constant software validation to scenarios that allow production runs for seismic hazard assessment. The best practices we will establish are of particular interest for supercomputing centres, which currently lack procedures for maintaining simulation software together with model scenarios and input data for real-time computation.