Disaster Damage and Response Estimation: public demonstration

An automated reporting pipeline that turns a cyclone scenario into a costed, sector-by-sector damage and response assessment for every area council in a country. Built in R and Quarto for the Government of Vanuatu by Théophile L. Mouton and Yan Holtz.

View the report

---

Every figure in this repository is simulated

This is a public demonstration of the pipeline, not a publication of its results. The real assessment was produced for the Government of Vanuatu from official statistics and remains with the client.

Everything in data/ here is synthetic data generated from scratch by scripts/simulate_data.R. No real value was read, rescaled or perturbed to produce it: the script reads only the structure of the original inputs (geography, indicator and attribute names, units, sources, years) and then generates values by giving each area council a synthetic population weight and distributing plausible national totals across councils.

The numbers therefore do not describe Vanuatu and must not be cited or reused as if they did. What is real is everything around them: the geography, the indicator taxonomy, the sector coverage, and every line of calculation, aggregation and presentation logic.

scripts/audit_render.R is the check that enforces this. It extracts every embedded data payload from the rendered page and confirms that none of it matches the original report, comparing observed agreement against the agreement expected by chance rather than against an arbitrary threshold.

What the pipeline does

The input is a scenario file: one row per area council, giving the cyclone category (2 to 5) that council experiences.

National,Province,Area Council,Hazard,Intensity
Vanuatu,Shefa,Port Vila,Cyclone,4
Vanuatu,Tafea,North Tanna,Cyclone,3

Change that file, re-render, and the entire assessment recomputes. Nothing else is touched by hand.

For each of ten sectors the report works through the same four stages:

1. Baseline: what exists, per area council
2. Damage: what the scenario destroys, via category-specific multipliers
3. Response: what relief that implies, in physical units
4. Financial: what it costs to replace, in vatu

Sectors covered: Education, Emergency Telecommunications, Energy, Food Security, Gender & Protection, Health, Logistics, Shelter, WASH, and Business.

Results are aggregated from area council to province to national level in a single pass, and presented as interactive sortable tables and choropleth maps, with every table also written to output/ as CSV for downstream use.

Scale

Source document	~6,700 lines of Quarto, 62 R chunks
Geography	71 area councils, 6 provinces
Sectors	10
Indicators	38, across 164 attributes
Interactive tables	44
Choropleth maps	13
CSV exports	46

The report also runs its own quality checks, validating council names against the scenario config and flagging missing baselines before any estimate is computed, then re-checking the outputs at the end.

Reproducing it

Requires R with dplyr, tidyr, reactable, htmltools, readxl, here, sf and leaflet, plus Quarto.

git clone https://github.com/TheophileMt92/vanuatu-disaster-report-demo.git
cd vanuatu-disaster-report-demo
quarto render index.qmd

simulate_data.R regenerates data/ and needs the original inputs, so it is included as documentation of method rather than as a step you can run. The synthetic data it produced is committed, so the render works as-is. It is seeded, so the figures are stable across renders.

Repository layout

index.qmd                        the pipeline
data/
  baseline_indicators.csv        what exists, per council: 38 indicators, 164 attributes
  damage_multipliers.csv         proportion of each asset lost, per cyclone category
  response_resources.csv         relief items issued per affected unit
  unit_costs.csv                 replacement cost per unit, in vatu
  hazard_scenario.csv            the scenario: a cyclone category per council
  council_province_lookup.csv    council to province mapping
  GIS_layers/area_councils.geojson   council boundaries
assets/                          report styling
output/                          generated CSVs, one per table
scripts/
  simulate_data.R                how the synthetic data was made
  patch_qmd.R                    the changes between this demo and the client report
  audit_render.R                 the check that no real figure survives

The scripts/ folder documents how this demonstration was derived from the client project. It is not part of the pipeline: index.qmd reads only data/ and renders without any of it.

Licence and credit

Pipeline and report by Théophile L. Mouton and Yan Holtz. Published with the agreement of both authors. The underlying assessment was commissioned by the Government of Vanuatu; nothing belonging to that engagement is reproduced here.