bettergovph/project-noah-hazard-maps

Project NOAH Hazard Maps

Dataset Summary

The Project NOAH (Nationwide Operational Assessment of Hazards) Hazard Maps is a comprehensive collection of geospatial datasets covering natural hazard assessments across the Philippines. The dataset includes three major hazard types:

1. Flood Hazard Maps - Flood inundation maps for 5-year, 25-year, and 100-year rainfall return periods
2. Landslide Hazard Maps - Shallow landslide susceptibility, structurally-controlled landslide hazards, and debris flow/alluvial fan delineations
3. Storm Surge Hazard Maps - Storm surge advisory maps for four severity levels based on simulations of 721 historical tropical cyclones (1951-2013)

The dataset covers all 81 provinces of the Philippines with province-level granularity, totaling approximately 23GB of geospatial data in ESRI Shapefile format.

Languages

The dataset documentation and metadata are in English. Geographic feature names are in Filipino and English.

Dataset Structure

Data Instances

The dataset is organized into three main directories:

project-noah-downloads/
├── Flood/
│   ├── 5yr/           # 5-year return period flood maps
│   ├── 25yr/          # 25-year return period flood maps
│   ├── 100yr/         # 100-year return period flood maps
│   └── metadata_flood.txt
├── Landslide/
│   ├── LandslideHazards/      # Merged landslide hazard maps
│   ├── DebrisFlowAlluvialFan/ # Debris flow and alluvial fan maps
│   └── metadata_landslide.txt
├── Storm Surge/
│   ├── StormSurgeAdvisory1/   # SSA 1 (2.01m to 3m)
│   ├── StormSurgeAdvisory2/   # SSA 2 (3.01m to 4m)
│   ├── StormSurgeAdvisory3/   # SSA 3 (4.01m to 5m)
│   ├── StormSurgeAdvisory4/   # SSA 4 (>5m)
│   └── metadata_stormsurge.txt
├── PMTiles/
│   ├── noah_hazard_maps.pmtiles   # Combined all-hazard map (2.5 GB)
│   └── layers/
│       ├── flood_5yr.pmtiles      # 5-year flood return period (486 MB)
│       ├── flood_25yr.pmtiles     # 25-year flood return period (563 MB)
│       ├── flood_100yr.pmtiles    # 100-year flood return period (969 MB)
│       ├── landslide.pmtiles      # Landslide susceptibility (262 MB)
│       ├── debris_flow.pmtiles    # Debris flow and alluvial fan (20 MB)
│       ├── storm_surge_ssa1.pmtiles  # Storm surge SSA 1 (59 MB)
│       ├── storm_surge_ssa2.pmtiles  # Storm surge SSA 2 (65 MB)
│       ├── storm_surge_ssa3.pmtiles  # Storm surge SSA 3 (66 MB)
│       └── storm_surge_ssa4.pmtiles  # Storm surge SSA 4 (64 MB)
└── NOAH_License.pdf

Each province is provided as a separate ZIP archive containing ESRI Shapefiles (.shp, .shx, .dbf, .prj, etc.).

Example shapefile attributes for flood hazard:

{
  "Var": 3,
  "geometry": "POLYGON ((121.0 14.5, 121.1 14.5, ...))"
}

Data Fields

Flood Hazard Maps

• Var: Hazard classification indicator (Integer)
  • 1: Low hazard (0-0.5 meters flood depth)
  • 2: Medium hazard (>0.5-1.5 meters flood depth)
  • 3: High hazard (>1.5 meters flood depth)

Hazard levels consider both flood depth and velocity. Areas with shallow but fast-flowing water may have higher hazard levels than depth alone would indicate.

Landslide Hazard Maps

• HAZ: Hazard classification indicator (Integer)
  • 1: Low hazard - Build only with continuous monitoring
  • 2: Medium hazard - Build only with slope protection and intervention; continuous monitoring
  • 3: High hazard - No dwelling zone

Storm Surge Hazard Maps

• HAZ: Hazard classification indicator (Integer)
  • 1: Low hazard (0.2m < max depth < 0.5m, and 0 < max depth × velocity < 0.5 sq.m/s)
  • 2: Medium hazard (0.5m < max depth < 1.5m, or 0.5 < max depth × velocity < 1.5 sq.m/s)
  • 3: High hazard (max depth > 1.5m, or max depth × velocity > 1.5 sq.m/s)

Storm Surge Advisory (SSA) levels correspond to peak storm tide heights:

• SSA 1: 2.01m to 3m
• SSA 2: 3.01m to 4m
• SSA 3: 4.01m to 5m
• SSA 4: More than 5m

Known Data Gaps

Flood/100yr/TawiTawi.zip is present in the dataset but is an empty ZIP archive (22 bytes, no files inside). No flood hazard data is available for Tawi-Tawi province at the 100-year return period.

Data Splits

The dataset is organized by hazard type and scenario rather than traditional train/validation/test splits:

Hazard Type	Scenarios	Provinces	Description
Flood	3 (5yr, 25yr, 100yr)	80	Return period-based flood maps
Landslide	2 (Hazards, Debris Flow)	82	Susceptibility and runout maps
Storm Surge	4 (SSA 1-4)	68	Advisory level-based inundation maps

PMTiles

The dataset includes pre-built PMTiles files — a single-file, cloud-optimized vector tile format — derived from all shapefiles. They are stored in PMTiles/ via Git LFS.

noah_hazard_maps.pmtiles is a combined file with 9 named vector tile layers:

Layer	Hazard	Attribute	Values
flood_5yr	Flood 5-year return period	Var	1 / 2 / 3
flood_25yr	Flood 25-year return period	Var	1 / 2 / 3
flood_100yr	Flood 100-year return period	Var	1 / 2 / 3
landslide	Landslide hazard zones	HAZ	1 / 2 / 3
debris_flow	Debris flow and alluvial fan	HAZ	1 / 2 / 3
storm_surge_ssa1	Storm surge advisory 1 (up to 2 m)	HAZ	1 / 2 / 3
storm_surge_ssa2	Storm surge advisory 2 (up to 3 m)	HAZ	1 / 2 / 3
storm_surge_ssa3	Storm surge advisory 3 (up to 4 m)	HAZ	1 / 2 / 3
storm_surge_ssa4	Storm surge advisory 4 (above 4 m)	HAZ	1 / 2 / 3

Values correspond to hazard levels: 1 = Low, 2 = Medium, 3 = High.

Individual per-layer files are also available under PMTiles/layers/ as checkpoints from the conversion process.

Viewing

Browse the combined PMTiles file in the PMTiles viewer.

Usage with MapLibre GL JS

Load the file directly from HuggingFace using the pmtiles protocol plugin for MapLibre GL JS.

<script src="https://unpkg.com/maplibre-gl/dist/maplibre-gl.js"></script>
<script src="https://unpkg.com/pmtiles/dist/pmtiles.js"></script>

const protocol = new pmtiles.Protocol();
maplibregl.addProtocol("pmtiles", protocol.tile);

const map = new maplibregl.Map({ container: "map", /* ... */ });

const PMTILES_URL =
  "pmtiles://https://huggingface.co/datasets/bettergovph/project-noah-hazard-maps/resolve/main/PMTiles/noah_hazard_maps.pmtiles";

map.on("load", () => {
  map.addSource("noah", { type: "vector", url: PMTILES_URL });

  // Flood attribute: Var (1=Low, 2=Medium, 3=High)
  // source-layer options: flood_5yr, flood_25yr, flood_100yr
  map.addLayer({
    id: "flood-100yr",
    type: "fill",
    source: "noah",
    "source-layer": "flood_100yr",
    paint: {
      "fill-color": ["match", ["get", "Var"], 1, "#93c5fd", 2, "#3b82f6", 3, "#1d4ed8", "transparent"],
      "fill-opacity": 0.65
    }
  });

  // Landslide / debris flow attribute: HAZ (1=Low, 2=Medium, 3=High)
  // source-layer options: landslide, debris_flow
  map.addLayer({
    id: "landslide",
    type: "fill",
    source: "noah",
    "source-layer": "landslide",
    paint: {
      "fill-color": ["match", ["get", "HAZ"], 1, "#fde68a", 2, "#f59e0b", 3, "#b45309", "transparent"],
      "fill-opacity": 0.65
    }
  });

  // Storm surge attribute: HAZ (1=Low, 2=Medium, 3=High)
  // source-layer options: storm_surge_ssa1, storm_surge_ssa2, storm_surge_ssa3, storm_surge_ssa4
  map.addLayer({
    id: "storm-surge-ssa1",
    type: "fill",
    source: "noah",
    "source-layer": "storm_surge_ssa1",
    paint: {
      "fill-color": ["match", ["get", "HAZ"], 1, "#99f6e4", 2, "#2dd4bf", 3, "#0f766e", "transparent"],
      "fill-opacity": 0.65
    }
  });
});

PMTiles conversion process

The shapefiles were converted to PMTiles using a Docker-based pipeline.

Prerequisites: Docker, ~6 GB free disk space beyond the source data.

# Build the image
docker build -t noah-pmtiles .

# Run the conversion (takes several hours)
docker run --rm -v $(pwd):/data noah-pmtiles shp_to_pmtiles.sh

Stage 1 — Parallel tippecanoe runs: Each of the 9 layers gets its own tippecanoe process running in parallel. Shapefiles are streamed from ZIP archives through named FIFOs directly into tippecanoe; no intermediate GeoJSON files are written to disk. Only one unzipped shapefile lives on disk at a time per layer. Each completed layer is saved as layers/<layer>.pmtiles and acts as a checkpoint — re-running skips layers whose files already exist.

Landslide shapefiles use inconsistent field names across provinces (GRIDCODE, GRID, ALLUVIAL, etc.). The script uses ogrinfo to detect the first numeric field per file and aliases it to a normalised output name via ogr2ogr SQL.

Stage 2 — tile-join merge: tile-join -pk merges all 9 per-layer PMTiles into noah_hazard_maps.pmtiles. The -pk flag bypasses tile-join's default 500 KB tile size limit.

Stage 3 — Verify: pmtiles show runs automatically after the merge and prints the tile type, bounds, zoom range, and layer names.

To resume an interrupted run, delete the checkpoint for any layer you want to regenerate and re-run:

rm layers/landslide.pmtiles
docker run --rm -v $(pwd):/data noah-pmtiles shp_to_pmtiles.sh

To verify an existing output file manually:

docker run --rm -v $(pwd):/data noah-pmtiles -c "pmtiles show /data/noah_hazard_maps.pmtiles"

Expected output includes tile type Vector Protobuf (MVT), bounds covering the Philippines (~116–127°E, 4–21°N), zoom range 0–14, and all 9 layer names.

Dataset Creation

Source Data

This data was sourced from the downloadable products of Project NOAH.

Annotations

Personal and Sensitive Information

This dataset does not contain personal or sensitive information. It consists entirely of geospatial hazard zone delineations at the provincial and municipal level.

Additional Information

Licensing Information

The downloadable products of Project NOAH hosted in this server are open data licensed under the Open Data Commons Open Database License (ODC-ODbL). The full details of the license can be found here: https://opendatacommons.org/licenses/odbl/1.0/.

You are free to download, copy, transmit, redistribute, and adapt our data provided that Project NOAH and its contributors are always properly attributed. Please refer to the accompanying readme file for citations and references on how to use the data.

If you alter or build upon our data, you may only distribute the result under the same license (ODC-ODbL).