NoraResearchLab/world-petroleum-assessment-units

World Petroleum Assessment Units

A modernized, AI/API-ready GeoParquet dataset of USGS World Petroleum Assessment 2000 Assessment Units.

---

Overview

The World Petroleum Assessment Units dataset is a modernization of the U.S. Geological Survey's World Petroleum Assessment 2000 (WPA 2000) Assessment Units — a globally distributed collection of petroleum assessment units defined by USGS geoscientists based on geological knowledge, exploration and production history, and scientific literature.

This project converts the legacy shapefile into a clean, validated GeoParquet dataset, preserving the original geological information, assessment attributes, spatial relationships, and provenance — without requiring downstream users to wrangle the aging shapefile format directly.

This is a sibling project to AGDFS (Automated Geological Data Fetching System) and the GLiM API Backend, following the same design philosophy: unify large, awkward-to-host geoscience datasets behind a simple, modern format so downstream AI training and analysis pipelines never need to touch the raw source files directly.

---

Data Source

Dataset	World Petroleum Assessment 2000 — Assessment Units
Authors	U.S. Geological Survey (USGS) World Energy Assessment Team
Citation	U.S. Geological Survey World Energy Assessment Team (2000). World Petroleum Assessment 2000. U.S. Geological Survey Digital Data Series DDS-60. https://pubs.usgs.gov/dds/dds-060/
License	Public Domain — U.S. Government Work
Coverage	Global
Records	270 Assessment Units
Original format	Esri Shapefile (legacy GIS format)
Modern format	GeoParquet

Note: Though USGS data is public domain, retaining the citation above is good practice for provenance and reproducibility in any derived product.

---

Pipeline

Legacy Shapefile (.shp)
        │
        ▼
Schema Normalization — field names, types, and units standardized
        │
        ▼
Geometry Validation — invalid / self-intersecting polygons repaired
        │
        ▼
GeoParquet Conversion — geometry + attributes written to columnar format
        │
        ▼
Metadata Enrichment — provenance, CRS, and field descriptions attached
        │
        ▼
AI/API Ready

The goal is to make historically valuable geological and petroleum-resource information easier for modern computational systems, researchers, APIs, and AI applications to access and use.

---

Production Files

Unlike larger sibling datasets (e.g. GLiM's ~1.24M polygons), the World Petroleum Assessment Units dataset is small enough — 270 records — that no spatial/attribute split is required for memory optimization:

File	Contents	Purpose
world_petroleum_assessment_units.parquet	assessment_unit_id, geometry, and all original attribute fields (province, AU name, resource estimates, geologic province, etc.)	Single GeoParquet file — geometry and attributes together, loaded directly at API startup.

---

How It Works

Incoming request (lat, lon)  OR  assessment_unit_id
        │
        ▼
Spatial query — geopandas .sindex.query(point, predicate="intersects")
        │
        ├── Match found ──────────────► assessment_unit_id
        │
        └── No match (outside all defined AUs)
                │
                ▼
        Return empty result — WPA 2000 coverage is not fully global
        │
        ▼
Attribute fetch — row pulled directly from world_petroleum_assessment_units.parquet
        │
        ▼
Response: { assessment_unit_name, province, resource_estimates, ..., assessment_unit_id }

1. Spatial query — the incoming point is tested against the in-memory R-tree built from world_petroleum_assessment_units.parquet's geometry column via sindex.query(point, predicate="intersects").
2. No-match handling — WPA 2000 coverage is limited to assessed petroleum provinces, so points outside any Assessment Unit correctly return an empty result rather than a forced nearest-neighbor match.
3. Attribute fetch — since geometry and attributes live in the same file, the matched row's full attribute set is returned in a single lookup, with no second file to query.

---

API Endpoints

Endpoint paths below reflect the intended workflow described above — adjust to match your actual route implementation.

Method	Path	Description
GET	/assessment-units	List all assessment units, optionally filtered by region or province.
GET	/assessment-units/{id}	Single assessment unit lookup by assessment_unit_id.
GET	/assessment-units/point?lat={lat}&lon={lon}	Point lookup — returns the assessment unit(s) containing a given coordinate.
GET	/health	Service health check.
GET	/mcp	MCP layer for agent/tool integration (if enabled, matching the AGDFS/GLiM pattern).

---

Tech Stack

• GeoPandas + Pyogrio — shapefile reading and GeoParquet I/O
• Shapely — geometry validation and repair
• PyArrow — GeoParquet storage engine
• FastAPI — API layer (recommended, consistent with AGDFS/GLiM)

---

Project Structure

world-petroleum-assessment-units/
├── data/
│   ├── raw/
│   │   └── wpa2000_assessment_units.shp   # original USGS shapefile (+ .dbf/.shx/.prj)
│   └── processed/
│       └── world_petroleum_assessment_units.parquet
├── app/
│   ├── main.py                            # FastAPI app + endpoints
│   ├── convert.py                         # shapefile → GeoParquet pipeline
│   └── config.py
├── requirements.txt
└── README.md

---

Getting Started

git clone https://github.com/Nora-Research-Lab/<repo-name>.git
cd <repo-name>
pip install -r requirements.txt
uvicorn app.main:app --reload

---

Data Hosting

Given the specialized geospatial format, the processed dataset is hosted externally rather than committed to the repo — following the same pattern as AGDFS's Hugging Face dataset and the GLiM API Backend data:

https://huggingface.co/datasets/NoraResearchLab/world-petroleum-assessment-units

The API downloads/caches world_petroleum_assessment_units.parquet from this location at startup or build time.

---

Deployment Notes

• The dataset's small size (270 records) keeps memory and compute requirements minimal compared to larger sibling datasets like GLiM.
• The GeoParquet file — geometry and attributes together — is loaded once into memory at startup and reused across all requests.
• Original shapefile provenance (CRS, source fields) is preserved in metadata so results remain traceable back to the USGS source.

---

License

Code: add your preferred license here. Data: USGS World Petroleum Assessment 2000 data is a U.S. Government work and is in the public domain in the United States — see citation above for provenance.

---

Maintainer

NORA Research Lab

GitHub: https://github.com/Nora-Research-Lab

Hugging Face: https://huggingface.co/NoraResearchLab

LinkedIn: https://www.linkedin.com/company/nora-research-lab

X: https://x.com/noraresearchlab

---

Part of the NORA Research Lab geoscience API suite.