Lithosphere removal (delamination) following continental collision: shape and complexity of observables predicted from 3D numerical models

Upload contains most essential data files from the manuscript. The raw data amount to several TB and cannot be uploaded.

1    Model data

1.1 Models covered

Relatively frequent output is available for the reference model (R in paper, see Table S2), prefixed dp10. Limited output is uploaded for models with geometric variations (OSM -> dp19, TSM->dp26).

1.2 Types of data

Most files are visualisation files (ending *.vtr). These are intended for loading into the visualisation software Paraview. Files starting with comp_ contain the lithological composition field; other visualisation files hold the fields of selected physical parameters. Note that as of the design of the study, the maximum file size allowed for loading into Paraview is ca. 2 GB. This limits the number of fields available according to model size - larger models (geometrical variations) hold therefore less fields.

Binary files with ending *.prn are saved model states, from which programs can be rerun.

Time information mapping step numbers to years is provided in times...txt.

2   Software for reproduction

The numerical code is not in the public domain, and its use is restricted. A pre-compiled binary is provided and allows reproduction of the main/reference model. We present the input files; however, these cannot be changed freely (equivalent to software distribution), and changes will lead to termination of the program.

2.1  Requirements, preparation and use

The code is compiled on CentOS with gcc 4.8.2. There is no library dependency, apart from the intrinsic OpenMP and glibc (>= 2.17). The following requirements must be satisfied in order to run it:

Linux OS (tested on CentOS and Fedora)

> 160 GB shared memory

The following provisions are recommended:

16 cores

high bandwidth storage

storage space O(TB)

To prepare, simply unzip the provided snapshot clsd_reproduce_gcc_CentOS.zip in an appropriate directory (cluster). The input file init.t3c sets the initial conditions; the input file mode.t3c controls solver and file output. file.t3c is the pointer to the last output step; if a model snapshot is available as binary dump (*.prn file), setting the pointer to its number allows restarting.

For normal usage, create initial condition with executable in3mg, and subsequently run i3mg. This will create snapshots and *.vtr visualisation files.

High-performance computing facilities, runtime (months), proper software environment, and training in use of the code, or in the use of the visualisation software, are not provided.

本次上传的数据包包含论文配套的绝大多数核心数据文件。原始数据总规模达数TB，无法直接上传。

1 模型数据
1.1 覆盖模型
参考模型（论文中记为R，详见补充表S2）的输出文件较为充足，文件名前缀为dp10。带有几何变体的模型（OSM -> dp19，TSM->dp26）仅上传了有限的输出文件。
1.2 数据类型
大部分文件为可视化文件（后缀为*.vtr），可加载至帕拉维（Paraview）可视化软件中使用。以comp_开头的文件包含岩性组分场；其余可视化文件则存储选定物理参数的场数据。需注意，根据本研究的设计规范，可加载至帕拉维的单文件最大尺寸约为2 GB。这一限制会根据模型尺寸调整可输出的场数量——尺寸更大的几何变体模型所能包含的场数更少。
后缀为*.prn的二进制文件为保存的模型状态文件，可用于程序的复现运行。
文件times...txt提供了模拟步数到实际年份的时间映射信息。

2 复现用软件
本研究使用的数值代码未公开，使用权限受到限制。本次提供了预编译的二进制文件，可用于复现主模型/参考模型的运行结果。同时提供了配套输入文件，但此类文件无法随意修改（修改行为等同于软件分发），否则将导致程序运行终止。
2.1 运行要求、准备流程与使用方法
本代码基于CentOS系统与gcc 4.8.2编译完成，仅依赖内置的OpenMP与glibc（版本≥2.17），无其他第三方库依赖。运行该代码需满足以下要求：
- Linux操作系统（已在CentOS与Fedora上完成测试）
- 共享内存容量不低于160 GB
推荐配置如下：
- 16核CPU
- 高带宽存储设备
- 存储空间需求达TB级
准备运行的步骤十分简便：只需将提供的压缩包clsd_reproduce_gcc_CentOS.zip解压至合适的目录（如集群节点）即可。其中，输入文件init.t3c用于设置初始条件；mode.t3c用于控制求解器与文件输出策略；file.t3c为指向最后一次输出步数的指针：若存在二进制格式的模型快照（*.prn文件），将该指针设置为对应快照的编号即可实现重启运行。
常规运行流程为：先通过可执行文件in3mg生成初始条件，随后运行i3mg，该程序将生成模型快照与*.vtr格式的可视化文件。

本次服务不提供高性能计算设施、数月量级的运行时长、适配的软件环境，也不提供代码或帕拉维可视化软件的使用培训。