System Update Log
What has changed, in plain language. Minor fixes and behind-the-scenes work aren't listed here.
Two systems make up the service and they are updated independently, so each is listed separately below. A change to one does not mean the other changed at the same time.
Simulation service
Runs the simulations and serves the API at https://bbpservices.net/v1 .
- Large earthquakes now run to completion. Simulations of the biggest events, particularly with the more expensive methods, could previously be stopped before they finished and reported as failures. They are now given the time they need.
- All seven velocity-model regions are available, including both Japanese regions, so every event in the published validation suite can be simulated on either software version we offer.
- Jobs that lose their compute machine are now picked up and retried rather than waiting out the clock.
- The catalogue no longer offers scenarios the service cannot yet score, so a job that could not produce a usable result is refused up front instead of being accepted and billed.
- Comparison results no longer carry a note about simulation hardware that does not apply to them. It now appears only alongside results it genuinely affects.
- First public prototype. The API accepts simulation jobs, prices them before you commit, runs them on demand and returns the results with the data products they produce.
- Two versions of the Broadband Platform software are offered, so a result can be reproduced against the specific version that produced it.
Website and dashboard
Accounts, the simulation dashboard and the published verification and validation results.
- More published comparisons. Validation results now cover additional earthquakes and all seven simulation methods, with both software versions shown side by side so you can see where they agree and where they differ.
- Results pages state how many realizations a number came from, since a single run and an average of many are not the same claim.
- The dashboard suggests sensible starting values for the method-specific settings that previously offered no guidance.
- First public prototype. Register for an account, build a simulation in the dashboard, see what it will cost before submitting, and download the results.
- Verification and validation results are published as they are produced, including where a simulation disagrees with what seismic stations actually recorded. Bias is reported period by period rather than as a single average, because a simulation can look good on average and still be well off at the period a particular structure responds at.
- Registering gives you access once you confirm your email address, which is also how you get your API key.