Linear Topographic Position Index: Variable Cross-Section Profiling for Micro-Segments

Beyond the Vector Map: Building Spatial Knowledge Graphs

Conventional routing engines excel at finding the shortest distance between two points, but they often struggle when applied to complex, real-world outdoor environments. Treating every kilometre of route equally fails to account for critical environmental nuances like steep elevation gains, exposure on open ridge lines, or local topographic landmarks.

By combining PostGIS for heavy spatial processing with Graph to manage network topology, static geographic vectors can be transformed into a dynamic spatial knowledge graph. Modelling route networks as connected Segment and Fell nodes enables pathfinding algorithms to evaluate rich, multi-dimensional properties in real time. This article explores how to build a hybrid spatial-graph pipeline from vector preprocessing and Topographic Position Index (TPI) edge enrichment to sub-second route traversal and RFC 7946-compliant GeoJSON delivery.

What Is Graph Technology?

At its core, graph technology is a way of organising data around connections rather than tables. If you imagine a traditional database as a collection of separate spreadsheets, a graph database is more like a map pinned to a wall, with strings physically tying related items together. The “dots” on the map represent entities (such as cities, people, or mountain summits), while the “strings” represent the explicit relationships between them (such as a highway connecting two towns, a friendship, or a route segment).

In conventional databases, connecting different pieces of information requires calculating where rows in one table match rows in another a process that becomes exponentially slower as networks grow larger and questions become more complex. Graph technology eliminates this bottleneck by making relationships a fundamental part of the data storage itself. When every item already knows exactly what it connects to, navigating through dozens of interconnected steps happens almost instantaneously.

This relationship-first design brings profound advantages to real-world applications like logistics, social networks, and spatial routing. By preserving the natural shape of connected systems, graph databases deliver exceptional query speed, total flexibility as data evolves, and rich contextual understanding allowing applications to answer complex, real-world network questions in milliseconds.

• Relationships as First-Class Citizens: Unlike traditional databases that store data in isolated tables and rows, graph technology models data as connected networks of points (nodes) and links (relationships).
• Intuitive Data Representation: Graph database structures mirror how humans naturally view the world as interconnected webs of people, places, routes, and ideas rather than flat spreadsheets.
• Instant Path Finding: Because links between data points are stored directly alongside the data itself, navigating complex connections requires zero heavy searching or merging.
• Context and Network Intelligence: Storing how items relate to one another unlocks deeper contextual insights, revealing non-obvious patterns, visual proximity, and natural clusters.
• Adaptability to Change: Graph structures are inherently flexible; new types of entities, properties, or connections can be added continuously without breaking or restructuring existing data.
• Exceptional Traversal Efficiency: Answering queries about complex, multi-step connections remains lightning fast, regardless of how large the overall network grows.

Case Study: Moving Beyond Traditional Routing Algorithms

When designing modern routing systems, standard shortest-path algorithms often fall short. Conventional tools focus purely on minimising distance or time, treating every kilometre of route as equal regardless of terrain, exposure, or scenery.

By modelling route networks inside Graph using discrete Segment nodes linked by relationships, static geospatial vectors are transformed into an intelligent, walkable network. Crucially, linking route segments directly to landmark entity Fell nodes via relationships introduces true geographic context. This dual-layer structure allows algorithms to evaluate multi-dimensional properties such as slope, terrain roughness, and proximity to key mountain summits delivering dynamic, highly contextual route generation at scale.

• Limitations of Standard Shortest-Path: Traditional algorithms like Dijkstra’s or A* typically optimise for a single cost metric (usually distance or time), ignoring environmental nuances and route network topology.
• Complex Spatial Intersections: Real-world networks consist of overlapping routes, split segments, and complex intersections that flat vector files struggle to represent efficiently.
• Segment-Centric Graph Architecture: Modelling route sections directly as Segment nodes connected by relationships allows rich, multi-dimensional properties (elevation profiles, terrain roughness, TPI) to sit directly on segment entities.
• Contextual Domain Enrichment: Integrating domain-specific entity Fell nodes linked via relationships anchors the route network to real-world topographic landmarks and mountain summits.
• Context-Aware Routing: Graph networks allow arbitrary weightings incorporating elevation gain, surface difficulty, and proximity to specific fells into path calculations.
• Dynamic Pathfinding: Using graph traversal algorithms and Graph Data Science (GDS) enables dynamic route generation tailored to user preferences (e.g., ridge-top views vs. sheltered valley paths).
• Scalable Data Network: Decoupling spatial processing from topological traversal allows Graph to query thousands of connected segments and landscape entities in milliseconds.

Architecture & Data Pipeline: From Spatial Vectors to Graph Topology

Building a graph-powered routing system requires a robust data pipeline to convert raw geographic features into structural graph entities. We use PostGIS to execute heavy spatial operations, such as splitting route polylines into uniform micro-segments and identifying exact topological intersections.

Once vector geometry is transformed into discrete nodes and connected edges, the dataset is loaded into a Graph database. Each Segment node in the graph represents a physical route micro-segment, linked to adjacent segments via relationships, enriched with spatial properties computed in PostGIS. When a routing request is received, user coordinates are mapped to the nearest graph nodes, creating a seamless bridge between geographic space and topological graph space.

• PostGIS as Spatial Foundation: PostgreSQL with PostGIS serves as the primary spatial engine for processing raw vector geometry, splitting segments, and performing geospatial joins.
• Micro-Segment Generation: Continuous line strings are divided into uniform micro-segments to attach precise local attributes without losing fine-grained geographic detail.
• Constructing Node Topology: Intersection points and segment endpoints are extracted, rounded, and deduplicated to generate clean topological nodes with distinct coordinate identities.
• Importing into Graph: Spatial nodes and relationships are ingested into Graph via Dockerised instances, establishing bidirectional or directed route edges.
• Spatial Indexing & Docking: GeoJSON coordinates from external location sources are docked onto the nearest graph node using spatial proximity queries (e.g., PostGIS ST_Distance or Graph spatial procedures).
• Properties Enrichment: Each graph edge is enriched with computed spatial metadata including length, elevation gain, topographic index, and surface profile.

Graph Algorithms in Action: Customising Route Traversals

With the network fully represented as a property graph, pathfinding becomes an exercise in custom graph algorithm execution. Rather than relying on rigid, pre-calculated paths, we project subgraphs into memory using Graph Data Science (GDS) and execute algorithms like Weighted Shortest Path.

Because route edges store comprehensive physical attributes, cost functions can be customised dynamically. For example, a routing request can adjust node/segment weights based on elevation gain, terrain roughness, or visual prominence. Cypher queries dynamically traverse the graph, returning sequentially ordered, multi-segment routes that balance physical exertion against user preferences in milliseconds.

• Graph Data Science (GDS): Leveraging GDS projected graphs to execute memory-efficient pathfinding algorithms across millions of route segments.
• Weighted Shortest Path Calculations: Applying custom cost formulas to edges that balance total distance against physical effort, elevation profiles, and terrain roughness.
• Avoiding Penalty Traps: Dynamically penalising undesirable route conditions such as steep, exposed ridge climbs during bad weather by adjusting node/segment weights on the fly.
• Graph Cypher Queries: Writing clean Cypher queries to navigate graph structures, filter route types, and extract ordered segment paths.
• Topological Re-ordering: Graph traversals naturally yield ordered lists of connected segments, eliminating messy spatial sorting or manual line merging downstream.
• Sub-Second Route Generation: Offloading network traversal to native graph algorithms enables sub-second routing across extensive outdoor networks.

Enriching Graph Edges: Topographic Position Index (TPI)

While distance and total elevation gain provide foundational routing metrics, they fail to capture whether a route segment traverses an exposed ridge line or winds through a sheltered valley floor. Integrating Topographic Position Index (TPI) directly into Segment nodes provides graph algorithms with vital environmental context. By evaluating local height relative to the surrounding landscape, TPI converts qualitative terrain characteristics into clean, numerical values that weight route segments dynamically.

While Fell nodes give macro-level landmark context (e.g., proximity to a specific peak summit), TPI parameters stored on the Segment nodes supply the micro-level terrain profile along the actual path.

To learn more about how TPI is calculated at scale using pre-aggregated 100m×100m spatial tiles and optimised PostGIS bounding box operations, check out our standalone article: Beyond Elevation: Unlocking Environmental Context with Topographic Position Index.

• Context Over Raw Elevation: TPI measures the elevation difference between a specific location and its surrounding neighbourhood, providing structural terrain context beyond standard sea-level altitude.
• Structural Terrain Signals: Positive TPI values identify exposed hilltops, ridges, and peaks, whereas negative values highlight sheltered valley bottoms, gullies, and ravines.
• Informing Path Weights: Attaching TPI metrics directly to Segment nodes enables routing algorithms to favour or penalise exposed ridges versus sheltered valley paths based on user preference or weather.
• Deep-Dive Reference: For an in-depth breakdown of TPI calculation methods, spatial scale considerations, and high-performance PostGIS tile aggregations, explore our dedicated TPI guide.

GeoJSON Payload & Frontend Integration

Once the optimal path through the graph is resolved, the backend reconstructs the geographic line string by stitching together the micro-segments in sequence. The resulting geometry is assembled into a strict, RFC 7946-compliant GeoJSON FeatureCollection, complete with enriched edge metadata.

Adhering strictly to the RFC 7946 standard ensures zero coordinate ambiguity and guarantees out-of-the-box compatibility with web mapping engines like Mapbox GL and Leaflet, as well as desktop GIS tools like QGIS. Beyond powering interactive web interfaces, emitting clean GeoJSON endpoints enables seamless desktop GIS integration allowing analysts to pull route payloads straight into QGIS for multi-layered spatial inspection, custom cartographic rendering, and field-data validation.

In an upcoming follow-up post, we will explore how to set up a dedicated QGIS workflow to consume, style, and analyse these graph-generated GeoJSON routes in real time.

• Standardised RFC 7946 Compliance: Formatting spatial outputs strictly according to the RFC 7946 GeoJSON specification guarantees universal compatibility, standard WGS 84 coordinate formatting, and seamless parsing across web mapping engines.
• Seamless Payload Structuring: Traversed graph paths are mapped back to spatial geometries and formatted directly into standardised GeoJSON FeatureCollection objects.
• Enriched Feature Payloads: Each line segment in the response GeoJSON carries rich properties (e.g., segment distance, elevation metrics, topographic context) for instant client-side rendering.
• Fast API Delivery: Pre-calculated topological relationships reduce backend query complexity, ensuring low latency for web and mobile frontends.
• Dynamic Map Visualisation: Frontends receive clean, ordered coordinate sequences ready to be plotted directly onto map layers using frameworks like Mapbox GL or Leaflet.
• Machine Learning Pipeline Integration: The enriched GeoJSON outputs double as structured training inputs for downstream machine learning models evaluating route difficulty and scenic quality.

Beyond Routes: Unlocking Spatial Intelligence

While route calculation provides an immediate, high-value application, modelling spatial environments as property graphs unlocks capabilities far beyond standard point-to-point pathfinding. Because graph databases excel at analysing complex topology and multi-layered relationships, the exact same graph architecture used for routing can power advanced spatial intelligence models across outdoor management, environmental science, and safety operations.

By adding new node types such as weather stations, habitat zones, or maintenance logs and defining how they relate to Segment and Fell entities, the database evolves into a unified spatial knowledge graph. Whether predicting how severe weather impacts exposed ridge lines, identifying critical ecological corridors, or helping search and rescue teams model lost-person trajectories, graph technology transforms raw geographic coordinates into an active, decision-making ecosystem.

• Terrain & Landform Classification: Expanding relationships beyond paths allows algorithms to group connected Segment and Fell nodes into broader topographic features like mountain ranges, valley basins, or isolated massifs.
• Ecological & Habitat Corridor Analysis: Graph structures excel at evaluating environmental continuity, mapping wildlife migration paths, or modelling seed dispersal along connected valley segments.
• Microclimate & Weather Risk Propagation: Environmental attributes (e.g., wind exposure, ice buildup) can propagate across connected Segment nodes to dynamically identify weather hazard zones across entire regions.
• Search and Rescue Incident Modelling: Modelling terrain as a graph enables response teams to compute probable lost-person movement vectors, search urgency indexes, and optimal search sector boundaries.
• Infrastructure & Asset Management: Route maintenance teams can query the graph to identify critical bottleneck segments, evaluate cumulative erosion exposure, or prioritise repair crews across complex networks.
• Spatial Recommendation Engines: By connecting Segment and Fell nodes to user activity history, graph algorithms can recommend personalised hikes based on landscape similarity rather than just location proximity.

Key Takeaways

• Beyond Distance Optimisation: Graph technology allows automatic route generation based on dynamic, multi-factor environmental criteria rather than simple distance.
• Hybrid Spatial-Graph Engine: Combining PostGIS for spatial manipulation with Graph for network topology yields a high-performance routing pipeline.
• Micro-Segment Precision: Breaking routes into fine-grained segments provides accurate, localised control over edge weights and path selection.
• Dynamic Edge Weighting: Environmental context, slope, and terrain parameters can be factored into pathfinding calculations in real time.
• Sub-Second Latency: Native graph traversal algorithms outperform traditional spatial joins for complex network routing at scale.
• Production-Ready Visualisations: Directly emitting enriched GeoJSON structures enables instant rendering on interactive client-side maps.