Geometry / geography
Geometry operates in a coordinate plane; geography models the curved earth. The choice changes distance, area, and containment behavior.
N · Native Spatial Intelligence deep dive
Reason about where, when, movement, and networks
Treat coordinates as measured observations—not magic truth—and combine them with topology, constraints, jurisdiction, time, and uncertainty.
Core concepts
A map is a view; spatial intelligence is decision context
CRS
A coordinate reference system defines how coordinates map to the earth. Unlabeled or mismatched CRS values can create plausible but wrong answers.
Accuracy and precision
Accuracy is closeness to reality; precision is measurement detail. More decimal places do not guarantee a correct location.
Topology
Rules about connectivity, adjacency, overlap, containment, and shared boundaries—essential for networks and valid spatial relationships.
Geofence
A spatial boundary whose entry, exit, or dwell can trigger assessment; uncertainty and boundary hysteresis prevent noisy alerts.
Network routing
Finds feasible paths over connected edges using cost, direction, closures, vehicle limits, schedules, and turn restrictions—not straight-line distance.
Temporal spatial data
Position and boundaries are valid at a time; moving objects need event time, trajectory, speed, and confidence.
Spatial privacy
Exact location can expose people and sensitive sites; minimize precision, audience, retention, and linkage according to purpose.
Logistics example
“Nearest warehouse” is not one calculation
| Naive question | Required context | Safe answer |
|---|---|---|
| Which warehouse is closest? | Road-network travel time, closure, truck height/weight, cargo restrictions, hours, dock capacity, inventory, jurisdiction, weather | Nearest feasible warehouse for this truck, shipment, time, and policy. |
| Is the truck inside the closure? | CRS, location age/accuracy, closure geometry validity, boundary rule, road segment, event time | Containment result plus confidence and fallback requirement. |
| Can we reroute? | Connected topology, turn restrictions, prohibited roads, driver hours, ETA/SLA, risk zones | Ranked alternatives with constraints, uncertainty, and approval status. |
Technology map
GIS, spatial SQL, streaming maps, and routing
| Need | Open / specialized | Databricks | Snowflake | Microsoft Fabric / Azure |
|---|---|---|---|---|
| Spatial storage/query | PostGIS, DuckDB Spatial, Apache Sedona, GeoParquet | ST geospatial SQL functions; Apache Sedona/Mosaic where justified | GEOGRAPHY/GEOMETRY types and geospatial SQL functions | Fabric Warehouse/Lakehouse data plus KQL geospatial functions; Azure SQL spatial |
| Desktop/enterprise GIS | QGIS, GeoServer | Integrate governed lakehouse data with GIS tools | Connect spatial tables to GIS/BI clients | ArcGIS, Azure Maps, Power BI maps, Fabric Maps |
| Python analysis | GeoPandas, Shapely, PyProj, Rasterio, NetworkX/OSMnx | Python notebooks/jobs with governed tables | Snowpark Python and geospatial SQL | Fabric notebooks/Spark, Python geospatial libraries |
| Routing/networks | pgRouting, Valhalla, GraphHopper, OSRM | External routing engine or custom graph/network processing | External routing service; SQL pre/post-processing | Azure Maps Route service or ArcGIS Network Analyst |
| Real-time movement | Kafka/Flink/Sedona, moving-object stores | Structured Streaming plus spatial functions | Snowpipe/Streams plus geospatial processing | Eventstreams, Eventhouse/KQL and Fabric Maps |
Do not confuse mapping with reasoning: visualization shows where an object appears. Certification needs authoritative sources, CRS, time, accuracy, constraints, edge-case tests, and safe behavior when confidence is low.
Implementation
How to achieve the N layer
Frame
Identify decisions where location, route, proximity, boundary, jurisdiction, movement, or spatial risk materially changes the answer.
Source
Approve geometry/position sources; record CRS, units, time, accuracy, precision, lineage, license, owner, and permitted use.
Model
Define geometry types, topology, networks, temporal validity, boundaries, geofences, indexes, and canonical identifiers.
Constrain
Encode physical, vehicle, safety, legal, environmental, schedule, privacy, and jurisdiction rules.
Validate
Benchmark known routes and test antimeridian/poles where relevant, CRS transforms, invalid geometry, edge boundaries, stale positions, and outages.
Operate
Monitor source freshness, positional error, infeasible routes, geofence false alerts, service availability, drift, and human overrides.
Evidence
- Spatial source catalog and metadata contract
- CRS/unit/time/accuracy records and transformation tests
- Topology, routing, geofence, and privacy rules
- Golden-route benchmarks, edge cases, outages, fallbacks, dashboards
Acceptance
- Known routes satisfy every physical and policy constraint.
- Low-confidence or stale location causes safe fallback.
- Boundary and CRS cases match authoritative expected results.
- Every spatial action records source, time, precision, constraints, and confidence.