Oasis LMF Workshops

Hands-on Jupyter notebook workshop series for the Oasis Loss Modelling Framework.

Requires: oasislmf 2.5.4 · ods-tools 5.0.1 · Python 3.12

---

Modules

Module	Topic	Model Used
01 — Model Files & Formats	Anatomy of an Oasis model: events, footprint, vulnerability, damage bins	ParisWindstorm
02 — Exposure Data (OED)	Building OED location/account files, validation with ods-tools, deterministic losses	ParisWindstorm
03 — Building a Model	Construct a hurricane model from scratch: hazard, vulnerability, keys service	HURDAT (US Hurricane)
04 — Running with the MDK	oasislmf model run, analysis settings, ORD output formats, EP curves	ParisWindstorm
05 — Model Adjustment	Sensitivity analysis: scale vulnerability, remove uncertainty, compare outputs	ParisWindstorm

Learning Path

01_model_files_and_formats  →  02_exposure_data_oed  →  03_building_a_model (3a → 3b → 3c)
                                                                     ↓
                                              05_model_adjustment  ←  04_running_with_mdk

Reference Models

Model	Location	Description
ParisWindstorm	models/ParisWindstorm/	Simple 4-cell windstorm model, Paris France. Good for understanding model structure.
PiWind	models/PiWind/	OasisLMF reference UK windstorm model. Full-featured with 1447 events.

Setup

pip install -r requirements.txt
jupyter notebook

Running a Model

cd models/ParisWindstorm
oasislmf model run -C oasislmf.json -r /tmp/paris_run

Previous Workshops

The following historical workshop repositories were used as source material:

• LondonWorkshop2018/ — Impact Forecasting EUWS, Python 2.7
• ZurichWorkshop2018/ — Catrisks MEEQ (Beirut earthquake), Python 2.7
• Workshop2019/ — GEM Dominican Republic earthquake, Oasis API
• Workshop2021/ — Building a hurricane model from HURDAT (original source for Module 03)
• Workshop2022/ — Enterprise platform, Kubernetes/Azure, autoscaling