Disagg Aggregations

Produce Disaggregration Aggregations

flowchart LR
    classDef nshm stroke:lightgreen, stroke-width:3px
    classDef AWS stroke:orange, stroke-width:3px

    RHD[/run Disaggregration Aggregations/]:::nshm
    thp[toshi-hazard-post]:::nshm
    DA[(disagg aggregations)]:::AWS

    RHD -->|with| thp -->|produces| DA

inputs:

• JSON index of disagg realisations produced in prior

• realizations from prior step CSV

• user configuration for the aggregations

outputs:

• produces aggregate disagg curves in THS DynamoDB per site, imt, vs30, poE there is approx 100 times more than equiv hazard.

  approx 40kB per site, imt, vs30, poE, so for 36 sites, 5 periods, 7 PoEs => 50mB

### resource/cost/metrics:

• runs locally or in AWS EC2 M5

• for each VS30 (NSHM has 18) we run either:

  • locally (CDC 32 core machine): ~2m preproc + 1500s per permutation/core => ~ 15 hours per VS30

  • cloud (up to ~600cores) => ~2 hours per VS30.

• cloud scheduling within limits of Account DynamodDB rate limiting (reading rlzs).