NSHM Hazard processing

NSHM Hazard processing

flowchart TD
    classDef nshm stroke:lightgreen, stroke-width:3px
    classDef AWS stroke:orange, stroke-width:3px
    classDef SVC stroke:powderblue, stroke-width:3px   

    subgraph P1[" "]
        %% direction TD
        RHR[/run hazard realisations/]:::nshm
        RHA[/run hazard aggregations/]:::nshm
        RDR[/run disagg realisations/]:::nshm
        RDA[/run disagg aggregations/]:::nshm
    end

    RHR -->|then| RHA -->|then| RDR -->|then| RDA
    ENV{any of}
    AWS[NSHM AWS Batch]:::AWS
    HPC[HPC Cluster]:::SVC
    LOCAL[Local Workstation]

    %% links
    P1 -.-> |using| ENV
    ENV --> AWS
    ENV --> HPC
    ENV --> LOCAL   

NSHM Hazard processing has four main processes, which each depend on each other in the following sequence:

• Produce Hazard Realisations

• Produce Hazard Aggregrations

• Produce Disagg Realisations

• Produce Disagg Aggregations

Each process may be run on several compute environments:

• an NSHM AWS Batch cloud compute environment

• an HPC cluster compute environment¹

• a Local Workstation eg 16-32 cores, 128 GB RAM, 1TB available disk.

See DiCaprio, et. al. 2024 for a description of the hazard calculation methodology and terminology, some of which is used in these documents.

Info

The full NZSHM model, with around 4000 hazard sites, will currently take months to process unless multiple jobs can be run concurrently. Details of the approximate resource and time requirements can be found in each process description above.

The NZSHM team are working with other hazard software developers to explore ways to reduce the necessary computational costs.

---

1. HPC environment support is currently limited, but is on the CWG work program. ↩