Run simulations for validation dataset (400m → 200m/100m)

1. 200m validation (for the NZ-wide crustal and subduction) performance

2. Extension to include subduction

3. Run empirical for all crustal events? (it works!)

4. Run empirical for subduction sources

Jason/Robin

1. currently invetigating the subduction performance and then will run 400m before getting to 200m.
2. In progress (Jason) – test sims are done. Waveform analysis to be done
   
3. Run for both small and moderate magnitudes. prelim analysis in progress.
4. On hold (Viktor's code conversion)

1) Cybershake Subduction (8 faults) : try optimised HF code

2) 20p4 result processing/interrogation

3)Running 200m sims LF only

1. lower south island/1 rel
2. all faults/1 rel
3. lower south island / all rels
4. all faults/all rels

cs20p4 : 200m . Initially join HF from 400m.

2)Jason

1) Jonney

1) LF initial run failed. Should work now. Investigating differences in latest version of code.

2)20p4 SC interrogation - active investigation done until further notice.

3) Faults identified to use ~10k CH for testing

On hold : a bug in KISTI workflow

1. GeoNET code Python3-ize (LP)
2. Generalize hacks for KISTI
3. Interrogate IM calc output / update benchmark data if needed (KISTI)
4. CH estimation based on linear regression
5. Integrate with pre-processing (starting with VM)
   1. Separate db creation out of install
   2. Create generic slurm for the rest of install, make install operate on each fault
   3. Create a generic slurm for VM generation
6. GMT build for KISTI
   
   

1. Jason (background)
2. James

6. Sung

2. Deffered

3. HF verified/Indirectly confirmed IM calc on KISTI works correctly. Will need to generate a new benchmark

4. No progress

5. In progress

6. Done

1. GM Selection for Empirical
2. Automate documentation
3. Front-end
   
   
   1. Milestone #2
      1. authentication.
      2. user access to individual tab controlled by authentication
      3. User data managed by API, and additional user data kept in a separate db
      4. Data integrity and consistency between Auth0 and separate db maintained.

Connection to backend

Connection to GM-selection

Checking routine outputs (pSA etc)

4) Empirical DS - DB calculation

1. Daniel
2. Background task (Jason)
3. Andy

4) Not as straightforward to run on HPC as I thought. Managed to get it working on Mahuika BigMem. Estimating 36,000 CH for 24,000 sites for DS calculation. Will need checkpointing before starting. And additional stability checks. Takes ~40 min per station. Could do some memory analysis - unsure if worth the effort needed.

Roadmap (scientific functionality list)

Production - TODO (longer term tasks)

(LP) 

Viktor

Earthquake Detection

1. NN - GMM
   1. Implement an initial basic pipeline with some NN config + flexible feature selection & preprocessing
2. GM Classifier – see link

1. SimAtlas simulation+animation:
   1. Test auto workflow with batch 4. (total 100 faults)
2. Relocate data to nearline (waveform, retired user, txt based empirical)
3. Continue with Hikurangi geometry
4. Find Hoby's 2011Feb (2016 NZSE paper/2018 BSSA) - isilon (10 different slip models / 50 different rupture models)

1. Sung/Jonney
2. Sung
3. James

2. In progress. Retired users RunFolder data

3. Initial fault deformations started, some math issues

4. Found and passed to Karim

1.SimAtlas simulation+animation

3.Hikurangi surface geometry