13.3 Network Analysis: Geometric Networks, Utility Networks & Routing

13.3 Network Analysis: Geometric Networks, Utility Networks & Routing

Quick Summary: Linear infrastructure systems—such as municipal street networks, river systems, water distribution lines, and electrical grids—are modeled using network topology. Rooted in mathematical graph theory, a network consists of a topologically connected framework of edges (linear conduits) and junctions (nodes/vertices). Network models bifurcate into two distinct paradigms: Directed Networks (Utility and Geometric Networks), where commodities flow along physical pressure gradients or systemic rules toward sinks, and Undirected Networks (Transportation Networks), where mobile agents navigate street systems subject to turn restrictions, speed limits, and one-way rules. Understanding graph-traversal algorithms like Dijkstra's and $A^*$, configuring non-planar elevation grade separations, and selecting appropriate network solvers is essential for solving transportation and logistics problems.

---

1. Graph Theory Foundations & Network Topology

In GIS, a network is a system of topologically interconnected linear features and points abstracted as a mathematical graph $G = (V, E)$:

• Vertices ($V$) / Junctions: Point locations where linear segments terminate or intersect (e.g., street intersections, utility valves, pipeline fittings, dead-end cul-de-sacs).
• Edges ($E$) / Links: Linear segments representing physical conduits along which flow or travel occurs (e.g., street centerlines, water mains, rail tracks).

   Junction (Node 1)               Edge (Link e1)               Junction (Node 2)
         ( V1 )======================================================( V2 )
           ||                                                          ||
           || Edge (Link e2)                                           || Edge (Link e3)
           ||                                                          ||
         ( V3 )======================================================( V4 )
   Junction (Node 3)               Edge (Link e4)               Junction (Node 4)

Connectivity Policies

How edges connect at junctions is governed by strict topological connectivity policies:

• End-Point Connectivity: Edges only connect and permit flow at their physical start and end vertices. An edge passing across another line without an explicit end-point vertex does not permit turning or traversal. This policy is essential for modeling grade-separated overpasses and bridges.
• Any-Vertex Connectivity: Edges connect at any coincident vertex along their length. Commonly applied to simple pedestrian sidewalk networks or local street systems where cross streets always intersect.

Grade Separation & Non-Planar Networks (Elevation Fields)

In standard 2D vector GIS, two lines that cross each other create a geometric intersection. In reality, a highway passing over a local road via an overpass bridge does not allow vehicles to turn off the overpass onto the road below.

To represent 3D grade separations within a 2D network model without corrupting topology, network engines utilize Elevation Fields (often designated F_ELEV / T_ELEV for From-Elevation and To-Elevation, or Z-levels):

   Bridge Overpass (Level 1):     Node A (F_ELEV=1) ---------------- Node B (T_ELEV=1)
                                                             |
                                                   [ NO CONNECTION! ] (Crossing in 2D,
                                                             |         different Z-levels)
   Local Surface Road (Level 0):  Node C (F_ELEV=0) ---------------- Node D (T_ELEV=0)

If the bridge edge and the surface road edge have different elevation values at their apparent 2D crossing point (e.g., $1$ vs. $0$), the network topology engine recognizes that they are vertically separated, preventing the solver from generating illegal turn transitions between them.

Network Cost (Impedance) & Turn Features

• Impedance / Cost Attributes: The measure of resistance required to traverse an edge. Common costs include length (meters, miles), travel time (seconds, minutes), financial cost (tolls), or energy expenditure. Costs are directional:
  • FT_Minutes: Cost traversing From the From-Node To the To-Node.
  • TF_Minutes: Cost traversing From the To-Node To the From-Node.
  • A one-way street sets one direction to a valid travel time and the opposing direction to $-1$ or Restricted.
• Turn Feature Classes & Turn Tables: Intersections are rarely frictionless. Left turns across oncoming traffic take longer than right turns, and some maneuvers are prohibited entirely. Turn features store explicit directional maneuver penalties (e.g., adding a 25-second delay to a left turn) or enforce temporal restrictions (e.g., No Left Turn between 4:00 PM and 6:00 PM).

---

2. Directed vs. Undirected Networks

Geospatial networks are structurally divided into two fundamentally different data models based on how movement is governed:

   DIRECTED NETWORK (Utility / Flow)             UNDIRECTED NETWORK (Transportation)
   =================================             ===================================
   • Flow direction determined by physics,       • Travel direction decided by agent
     pressure, gravity, or pump state.             (driver, pedestrian, cyclist).
   • Unidirectional at any given instant.        • Bidirectional traversal permitted
   • Elements: Sources, Sinks, Barriers.           (unless marked one-way).
   • Tracing: Upstream, Downstream, Isolation.   • Solvers: Shortest Path, VRP, Service Area.
   
        [Source] ---> (Valve) ---> [Sink]             [Origin] <==================> [Dest]

Directed Networks (Utility & Geometric Networks)

Directed networks model physical commodities—such as potable water, wastewater, stormwater, natural gas, crude oil, and electricity—moving through pressurized pipes, gravity sewers, or copper conductors:

• Flow Mechanics: Flow is governed by hydraulic head, gravity gradients, voltage potential, or mechanical pumps. The commodity cannot decide which branch to take; it flows strictly along the network gradient.
• Network Elements:
  • Sources: Locations where resources enter the network (water treatment plants, electrical generating stations, headwater springs).
  • Sinks: Locations where resources exit the network (household service meters, sewage treatment outfalls, drainage basin mouths).
  • Barriers / Switches / Valves: Devices that can be opened or closed to alter or stop flow.
• Network Tracing Solvers:
  • Downstream Trace: Identifies all pipes and consumers that will be affected by a contaminant spill at a specific upstream point.
  • Upstream Trace: Traces backward against flow to locate the source of an illicit discharge detected at a stormwater outfall.
  • Isolation Trace: Identifies which shutoff valves must be physically closed to isolate a ruptured water main while minimizing service disruptions to surrounding customers.

Undirected Networks (Transportation Networks)

Transportation networks model streets, pedestrian paths, railways, and bicycle trails:

• Agent Autonomy: Movement is controlled by autonomous travelers who make routing decisions at each junction based on personal goals, vehicle restrictions, and cost optimization.
• Network Hierarchy: Road segments are assigned hierarchical levels (e.g., 1 = Interstate Highways, 2 = Arterials, 3 = Local Streets). Solvers leverage hierarchy to optimize long-distance routing: they transition to higher-tier highways as quickly as possible and only drop down to local roads near the final destination, accelerating solve times across massive national networks.

---

3. Pathfinding Algorithms: Dijkstra & $A^*$ Heuristic Search

The mathematical core of network routing is calculating the least-cost path across a weighted topological graph.

Dijkstra's Shortest Path Algorithm

Formulated by Edsger W. Dijkstra in 1959, Dijkstra's algorithm solves the single-source shortest path problem on graphs with non-negative edge weights.

   Dijkstra: Explores radially in ALL directions     A* Search: Directed toward destination by Heuristic
             (Wavefront Expansion)                               (Focused Search Beam)
                    * * *                                               
                  * * * * *                                            
        (Start) * * * * * * * (Target)                 (Start) =======> =======> (Target)
                  * * * * *                                            
                    * * *                                              

Algorithmic Execution Steps:

1. Assign every node a provisional tentative distance: set the start node to $0$ and all other nodes to infinity ($\infty$).
2. Mark all nodes as unvisited. Set the start node as the current node.
3. For the current node, calculate the tentative distance to all its unvisited neighbors: $\text{Distance} = \text{Current_Distance} + \text{Edge_Cost}$.
4. If this calculated distance is less than the neighbor's previously recorded distance, update the neighbor's distance.
5. Once all neighbors of the current node are evaluated, mark the current node as visited (visited nodes are never re-evaluated).
6. Select the unvisited node with the smallest tentative distance as the new current node and repeat.
7. Terminate when the destination node is marked visited.

[!NOTE] Dijkstra's algorithm is mathematically guaranteed to find the globally optimal shortest path. However, because it evaluates nodes radially in all directions like an expanding circular wave, it consumes significant memory and processing time when applied to large, continental-scale street graphs.

$A^*$ (A-Star) Heuristic Search Algorithm

The $A^*$ search algorithm enhances Dijkstra's algorithm by introducing a heuristic function that guides the search directionally toward the target, dramatically pruning unnecessary node evaluations.

$f(n) = g(n) + h(n)$

• $g(n)$: The exact, accumulated cost from the start node to the current candidate node $n$.
• $h(n)$: The estimated remaining cost from node $n$ to the goal destination, evaluated using an admissible heuristic (typically straight-line Euclidean distance divided by maximum theoretical speed limit).
• $f(n)$: The total estimated least-cost path passing through node $n$.

Because $h(n)$ pulls the search frontier directly toward the goal, $A^*$ ignores nodes pointing away from the destination, solving routes orders of magnitude faster than standard Dijkstra while preserving path optimality.

---

4. Network Analysis Solver Typologies

Modern network analysis suites implement specialized analytical solvers designed for distinct logistical, emergency response, and commercial facility planning workflows:

   +---------------------------------------------------------------------------------+
   |                           NETWORK SOLVER SUITE                                  |
   +---------------------------------------------------------------------------------+
   | 1. Route Solver               --> Point-to-point shortest/fastest path          |
   | 2. Closest Facility           --> Incident response routing (dispatch)          |
   | 3. Service Area               --> Isochrone catchment envelopes (drive-times)   |
   | 4. Vehicle Routing (VRP)      --> Fleet logistics with capacity & time windows  |
   | 5. OD Cost Matrix             --> M x N matrix calculation without geometry     |
   | 6. Location-Allocation        --> Optimal facility placement to maximize demand |
   +---------------------------------------------------------------------------------+

Solver Functional Matrix

Location-Allocation Problem Models

Location-Allocation solvers resolve complex spatial supply-and-demand placement problems using several mathematical models:

1. Maximize Coverage: Selects facilities such that the maximum amount of demand falls within a specified impedance cutoff (e.g., locating fire stations so 90% of homes are within an 8-minute drive).
2. Minimize Facilities: Chooses the absolute minimum number of facilities required to cover 100% of demand points.
3. Minimize Impedance (P-Median): Positions facilities such that the sum of all weighted travel distances between all demand points and their assigned facilities is minimized (ideal for public libraries or municipal service centers).
4. Maximize Market Share (Huff Gravity Model): Evaluates competitive retail store placement where customer patronage is modeled as a probabilistic function of store size (attractiveness) and travel distance.

---

5. Practical Geospatial Scenario: Municipal Emergency Response & Water Main Rupture

Scenario A: EMS Closest Facility Dispatch

A county 911 communications center receives a report of a multi-vehicle traffic collision on an interstate bypass. Three ambulances are active in the field:

• Analysis: The dispatcher runs the Closest Facility Solver. The incident location is supplied alongside the live GPS coordinates of the ambulances. The network dataset uses real-time traffic speeds and incorporates dynamic turn restrictions.
• Result: The solver determines that while Ambulance 2 is physically closest in straight-line Euclidean distance (2.1 km away), an active construction closure and a divided highway median prevent it from turning across the corridor. The solver routes Ambulance 1 (3.8 km away via highway), delivering a response time that is four minutes faster.

Scenario B: Water Distribution Main Rupture

A 16-inch high-pressure water main ruptures in an urban business district:

• Analysis: Utility engineers utilize a Directed Utility Network to perform an Isolation Trace.
• Result: The trace travels upstream and downstream from the rupture point along connected pipe edges until it encounters operable isolation gate valves. The solver outputs the exact IDs of the three valves that field crews must close to isolate the leak, alongside an attribute list of every customer meter that will temporarily lose water service.

---

6. Common Exam Traps & Pitfalls

[!CAUTION] Exam Trap 13.3.1: Planar Intersections vs. Grade Separations. An exam question will present a scenario where an overpass bridge crosses a freeway, but the routing solver erroneously instructs a vehicle to execute a 90-degree turn off the bridge directly onto the freeway below. The cause is improper planar topology (failure to implement elevation fields like F_ELEV / T_ELEV or Z-levels). Without non-planar elevation fields, standard 2D line crossings are treated as planar intersections where turns are permitted.

[!CAUTION] Exam Trap 13.3.2: Directed Flow vs. Transportation Networks. Questions frequently test the fundamental difference between geometric/utility networks and transportation networks. Remember: Utility networks are directed networks where flow is governed by physical forces (gravity, pressure) toward sinks, supporting upstream/downstream and isolation tracing. Transportation networks are undirected networks where movement is agent-driven, bidirectional, and analyzed using route, service area, and closest facility solvers.

[!CAUTION] Exam Trap 13.3.3: OD Cost Matrix vs. Route Solver Efficiency. A scenario asks: "An analyst needs to calculate travel times between 500 warehouses and 10,000 retail stores for a freight study. Which solver should be used?" Choosing the Route Solver is incorrect; generating 5,000,000 detailed line geometries will crash or stall the software. The correct choice is the Origin-Destination (OD) Cost Matrix Solver, which calculates tabular travel costs across large $M \times N$ combinations without generating vector route geometries.