OPSIS (Open Platform for Systemic Infrastructure Spatial Analysis)

Repository: University of Oxford / GFDRR / World Bank URL: https://global.infrastructureresilience.org/ GitHub: https://github.com/nismod License: Code: MIT | OSM data: ODbL-1.0 | Other data: CC-BY-4.0 Update cadence: Periodic re-runs against updated OSM extracts and hazard models (not real-time)

---

What This Dataset Is

Global geospatial infrastructure networks (roads, rail, ports, airports, waterways) combined with multi-hazard exposure analysis and damage modelling. This is the primary source for where infrastructure is and how vulnerable it is to climate and natural hazards.

OPSIS is the public face of the Global Infrastructure Resilience Initiative (GIRI). The underlying analysis pipelines come from Oxford's nismod programme.

---

API Feature Record Structure

Every feature returned by GET /features/{id} has this top-level shape:

Field	Type	Description
id	Integer	Auto-increment feature ID
string_id	String	Human-readable ID ({slice}_{index}, e.g. "152_1001")
layer	String	Layer name (e.g. "road_edges_motorway", "road_edges_primary")
sublayer	String/null	Optional sublayer
properties	Object	Asset attributes (mode-specific — see below)
damages_expected	Array	Expected Annual Damage across hazard/climate scenarios
damages_return_period	Array	Damage by specific return period
damages_npv	Array	Net present value damage
adaptation	Array	Adaptation/mitigation measures

---

Data Layers (Transport-Specific)

Roads — properties Schema (31 fields, confirmed from live API)

Field	Type	Example	Description
sector	String	"transport"	Always "transport"
subsector	String	"road"	Always "road"
asset_type	String	"motorway", "primary", "secondary", "tertiary"	Road classification (from OSM highway tag)
_asset_type	String	"road_paved", "road_unpaved"	Internal type for damage curve lookups
osm_way_id	Integer	25452113	Original OpenStreetMap way ID
segment_id	Integer	0, 1	Segment index within a split OSM way
from_node_id	String	"152_1377"	Network graph source node
to_node_id	String	"152_535"	Network graph target node
from_iso_a3	String	"USA", "KEN"	ISO alpha-3 country of source node
to_iso_a3	String	"USA", "KEN"	ISO alpha-3 country of target node
length_m	Float	2069.68	Edge length in metres
lanes	Integer	1, 2	Number of lanes (gap-filled from OSM or defaults)
width_m	Float	6.25, 9.5	Estimated road width in metres
paved	Boolean	true, false	Whether the road is paved
bridge	Boolean	true, false	Whether the segment is a bridge
material	String	"asphalt", "gravel"	Surface material
tag_lanes	Float/String	2.0, ""	Raw OSM lanes tag (empty if unmapped)
tag_surface	String	"asphalt", ""	Raw OSM surface tag
tag_bridge	String	""	Raw OSM bridge tag
tag_maxspeed	Float/String	50.0, ""	Raw OSM maxspeed (km/h)
max_speed_kmh	Float	207.69, 110.0	Maximum speed (gap-filled/estimated)
min_speed_kmh	Float	156.17, 32.075	Minimum speed (gap-filled/estimated)
max_tariff	Float	0.54, 2.4	Maximum transport tariff (USD/ton-km)
min_tariff	Float	0.36, 1.6	Minimum transport tariff (USD/ton-km)
tariff_unit	String	"USD/ton-km"	Unit for tariff fields
max_flow_cost	Float	1.124	Maximum flow cost (USD/ton)
min_flow_cost	Float	0.749	Minimum flow cost (USD/ton)
flow_cost_unit	String	"USD/ton"	Unit for flow cost fields
rehab_cost_USD_per_km	Float	4579435.63, 23545.10	Rehabilitation cost per km

Layers served: road_edges_motorway, road_edges_trunk, road_edges_primary, road_edges_secondary, road_edges_tertiary

Railways

Rail is included in the gri_osm.roads_and_rail_version_1 extract processor. Schema follows the same pattern as roads with expected differences:

Field	Expected Values	Notes
sector	"transport"	Same
subsector	"rail"	Rail-specific
asset_type	"rail", "light_rail", "subway", "tram"	From OSM railway tag
tag_gauge	Track gauge value	OSM gauge tag
tag_electrified	Electrification status	OSM electrified tag
Network topology fields	Same pattern as roads	from_node_id, to_node_id, length_m

Layers: rail_edges, rail_nodes

Ports and Airports

These appear in the visualization platform but are not currently part of the GeoPackage extract service (gri_osm processor covers roads + rail only). Available as point/polygon layers in the web viewer with location, name, country, and hazard exposure.

---

Damage & Adaptation Arrays

Each feature carries nested arrays for damage modelling. Confirmed structure from live API records:

damages_expected[] — Expected Annual Damage

Field	Type	Description
ead_amin	Float	EAD annual minimum estimate (USD)
ead_mean	Float	EAD mean estimate (USD)
ead_amax	Float	EAD annual maximum estimate (USD)
eael_amin	Float	Expected Annual Economic Loss — minimum (USD)
eael_mean	Float	EAEL mean (USD)
eael_amax	Float	EAEL maximum (USD)
hazard	String	Hazard type (e.g. "river")
rcp	String	Climate scenario: "baseline", "rcp4p5", "rcp8p5"
epoch	Integer	Time period: 1980, 2030, 2050, 2080
protection_standard	Integer	Existing protection level (return period years)

damages_return_period[] — Per Return Period

Field	Type	Description
exposure	Float	Asset value at risk (USD)
damage_amin / damage_mean / damage_amax	Float	Damage estimates (USD)
loss_amin / loss_mean / loss_amax	Float	Economic loss estimates (USD)
hazard	String	Hazard type
rcp	String	Climate scenario
epoch	Integer	Time period
rp	Integer	Return period in years (25, 50, 100)

adaptation[] — Adaptation Measures

Field	Type	Description
hazard	String	Target hazard type
rcp	String	Climate scenario
adaptation_name	String	Name/description of measure
adaptation_protection_level	Float	Protection standard provided (return period years)
properties	Object	Additional adaptation attributes

Terminology

Term	Meaning
ead	Expected Annual Damage — average annual cost of damage from a hazard
eael	Expected Annual Economic Loss — indirect losses from disruption
rcp	Representative Concentration Pathway (climate scenario)
epoch	Time period for projections (1980=baseline, 2030/2050/2080=future)
rp	Return period — flood frequency in years
rehab_cost_USD_per_km	Cost to reconstruct per kilometre
_asset_type	Internal classification for damage curve lookups

---

Access Methods

1. REST APIs (No Authentication)

API	Base URL	Purpose
Main API	https://global.infrastructureresilience.org/api/	Feature queries, pixel drilling
Extract API	https://global.infrastructureresilience.org/extract/	Country data package generation
Vector Tiles	https://global.infrastructureresilience.org/vector/	36 MVT layers

OpenAPI specs:

• https://global.infrastructureresilience.org/api/openapi.json
• https://global.infrastructureresilience.org/extract/openapi.json

Key endpoints:

GET  /features/{id}                    -- single feature with damage/adaptation data
GET  /features/sorted-by/{field}       -- paginated features with filters (sector, asset_type)
POST /v1/jobs                          -- submit country extraction (async, returns GeoPackage)
GET  /v1/processors                    -- list available data processors
GET  /pixel-driller/point/{lon}/{lat}  -- all layer values at a coordinate

2. Extract API — Country GeoPackage Bundles

The Extract API generates national-level data packages on demand. The key transport processor:

gri_osm.roads_and_rail_version_1 — Roads and Rail from GRI OSM, including damage estimates. Output: GeoPackage.

249 country boundaries available. Packages stored at: https://irv-autopkg.s3.eu-west-2.amazonaws.com/{iso3}/

Other available processors include hazard data (flood, cyclone, drought, landslide), population, power infrastructure — see the OpenAPI spec for full list.

3. Bulk Downloads

• Country data starter kits on Zenodo (GeoPackage + GeoTIFF)
• Extract API as above

3. Code Pipelines

Repository	Purpose
nismod/open-gira	Core Snakemake analysis pipeline
nismod/infra-risk-vis	Backend + frontend web app
nismod/irv-autopkg	Extract API service
nismod/snail	Spatial network analysis library
nismod/snkit	Network cleaning/assembly

---

Data Formats

Format	Use
GeoPackage (.gpkg)	Primary vector format
GeoJSON	Web-friendly vector
GeoTIFF / COG	Raster hazard data
MVT	Vector tiles
Parquet	Large tabular datasets

---

Coverage & Quality

Region	Roads	Rail	Ports	Hazards
Europe	Excellent	Excellent	Good	Good
North America	Excellent	Good	Good	Good
East Asia	Good	Good	Good	Good
South/SE Asia	Good	Moderate	Good	Good
Sub-Saharan Africa	Moderate	Moderate	Good	Good
Latin America	Good	Moderate	Good	Good
Pacific Islands	Limited	N/A	Moderate	Good

Coverage quality is driven by OSM completeness for networks and global hazard model availability for hazard layers.

---

Key Limitations for Automation

• No traffic volumes — network geometries are well-covered but traffic/demand data is mostly absent
• Static snapshots — not real-time; reflects a point-in-time analysis
• OSM-dependent — quality mirrors OSM completeness, which is uneven
• Modelled damage — EAD and disruption estimates are model outputs, not observed events
• Schema varies by country extraction — different extractions may include different field sets depending on available data

---

Integration Notes

• Join key to PortWatch: Match OPSIS port geometries to PortWatch portid for combined resilience + trade monitoring
• Join key to World Bank: Country ISO codes enable linking OPSIS exposure summaries to WDI indicators
• Python stack: GeoPandas, NetworkX, Rasterio, Snail are the core libraries for working with this data
• Full schema reference: Run open-gira pipeline outputs or query the OpenAPI specs for the definitive field list