Critical-Link Identification and Flood-Disruption Resilience of the Hyderabad Road Network (OSMnx)

HydNet-Critical is a graph-analytics study of Hyderabad's road network that identifies the links whose loss hurts the city most, and quantifies by how much. The street network is constructed from OpenStreetMap with OSMnx as a directed, weighted drive graph for the GHMC area, with every construction and simplification choice — network filter, intersection consolidation tolerance, edge directionality, speed-to-travel-time conversion — documented, because each one changes the criticality answer.

Criticality is measured per link and aggregated per corridor using betweenness centrality (how many shortest travel-time paths use a link) and closeness centrality on the travel-time graph, producing a ranked list of Hyderabad's most critical links and corridors — expected to concentrate on Musi river crossings, radial arterials and the inner ring.

A scenario engine then removes edges to simulate disruptions: individual bridge closures across the Musi, closure of elevated-corridor segments such as the PVNR Expressway, surface-road diversions along metro-corridor stretches, and a combined monsoon-flood scenario that takes out a documented set of low-lying crossings at once. For every scenario the study measures accessibility deltas against the undisrupted baseline — travel-time inflation, detour-ratio distributions and disconnected origin–destination pairs — on the same origin–destination sample, so deltas are paired and comparable.

Where citable ground truth exists, the ranking is sanity-checked against it: corridor classifications and volume shares from the HMDA Comprehensive Transportation Study, and the flood-affected crossings documented after recent monsoon events. The deliverable is a reproducible, scripted pipeline and a dissertation-ready results chapter.

Hyderabad's road network carries rapid growth with limited redundancy: a handful of river crossings, radial arterials and elevated corridors connect zones that have no practical alternative route. During monsoon flooding, low-lying crossings of the Musi and city nalas close with little warning; metro and corridor works divert traffic for years at a stretch. Yet there is no public, reproducible ranking of which links are structurally critical to the city, and closure planning cannot cite measured accessibility consequences of losing a given bridge or corridor — decisions rest on intuition and complaint volumes. This study fills that gap: build the city's drive network from OpenStreetMap with fully documented construction choices, rank links and corridors by betweenness and closeness criticality on travel time, simulate bridge, elevated-corridor and metro-corridor closures as explicit edge-removal scenarios, and measure paired accessibility deltas against the baseline, validating the ranking against citable transport-study and flood-impact observations where they exist.

Abstract

This dissertation studies the structural resilience of Hyderabad's road network using open data. A directed drive-network graph of the metropolitan area is built from OpenStreetMap with OSMnx under documented filtering and simplification; betweenness and closeness criticality identify the links and corridors that carry the largest share of shortest travel-time paths; and an edge-removal scenario engine measures the accessibility consequences of bridge, elevated-corridor and metro-corridor disruptions, including a combined monsoon-flood scenario. Results are travel-time deltas, detour-ratio distributions and disconnected-pair counts per scenario, validated where possible against HMDA transport-study corridor evidence and documented flood impacts.

Introduction

Hyderabad concentrates population, employment and infrastructure on a road network that monsoons repeatedly stress: low-lying crossings of the Musi and the city's nalas flood and close, elevated corridors and metro construction divert traffic for years, and closures cascade because alternatives are few. Planners discuss critical links qualitatively; this study makes the question quantitative and reproducible on open data, so closure planning and mitigation investment can cite measured numbers.

Existing system and literature gap

Network-science methods — shortest paths, centrality measures, percolation-style edge removal — are mature, and OSM-based street-network studies exist for many world cities. Published Indian work, however, typically (a) does not document network construction and simplification choices, which materially change centrality answers; (b) reports global city-level metrics without the corridor-level aggregation a planner can act on; or (c) omits measured accessibility deltas per disruption scenario. No public, reproducible critical-link study for Hyderabad combines documented OSMnx construction, corridor-level criticality and scenario deltas with a stated validation attempt against transport-study evidence.

Proposed work

• Pin a dated OSM snapshot (Geofabrik India extract) and build the drive graph with OSMnx, documenting filters, simplification and travel-time assignment.
• Compute node and edge betweenness (with a declared exact-or-sampled policy) and closeness on the travel-time graph; roll link criticality up into named-corridor rankings.
• Run a construction-sensitivity analysis showing how rankings shift under alternative filters, tolerances and speed assumptions.
• Encode a scenario matrix — individual Musi bridge closures, a PVNR-type elevated-corridor closure, metro-corridor surface-road disruption, and a combined monsoon-flood set — as versioned edge-removal lists with sourcing notes.
• Measure per-scenario accessibility deltas: mean and 95th-percentile travel-time change, detour-ratio distribution, disconnected origin–destination pairs and zone-level accessibility change.
• Validate corridor rankings against HMDA Comprehensive Transportation Study corridor classifications and documented flood-affected crossings, reporting agreement and disagreement honestly.

Feasibility

• Technical: a metropolitan drive graph simplifies to tens of thousands of nodes; betweenness with a sampled origin–destination set and standard NetworkX/OSMnx routines runs in hours on a laptop, and scenarios reuse precomputed path structures.
• Data: OSM is free and complete for Hyderabad; the Geofabrik India extract gives a citable, versionable snapshot.
• Economic: no licences; every library is open-source.
• Operational (research): every figure and table regenerates from scripts and the pinned snapshot with a single command chain.

In scope

• The GHMC-area drive network from one pinned OSM snapshot; travel times from OSMnx's speed/tag assumptions, stated as such.
• Betweenness and closeness criticality (exact where feasible, sampled where not, with the policy and seed declared); corridor-level aggregation by name and spatial buffer.
• Edge-removal disruption scenarios (single bridge, elevated corridor, metro-corridor surface roads, combined flood set) with paired accessibility deltas on the same graph.
• Sensitivity analysis of construction choices; validation against citable HMDA and flood-impact observations.

Out of scope

• Dynamic traffic assignment, congestion propagation and microsimulation; the graph uses static travel times.
• Hydrodynamic flood modelling; flood scenarios use documented affected-crossing sets, not simulated flood extents.
• Economic-loss, property-value and emergency-response optimisation analyses.
• Multimodal (metro, bus, MMTS) network modelling beyond using corridor closures as scenarios.
• Other cities — offered as future work.

The study follows a quantitative simulation-and-sensitivity research design with a fixed analysis protocol: define the graph, freeze construction choices, compute criticality, then run scenarios — never re-tuning choices after seeing scenario results. An M.Tech-appropriate phase plan spans 24 weeks.

Graph construction (declared in advance). One Geofabrik India extract is pinned by download date; the drive subgraph is extracted with OSMnx (network_type = drive), intersections are consolidated at a stated tolerance, and travel time is computed from OSM speed tags using OSMnx's routines. Every choice is recorded next to the alternative rejected and why — because the sensitivity phase will quantify exactly how much these choices move the answer.

Criticality computation. Betweenness uses shortest travel-time paths. Where the exact computation is too slow for the full graph, a sampled origin–destination set with a declared size and seed approximates it, and an exact run on the core sub-area checks the approximation against ground truth. Corridor roll-up sums link criticality within named corridors matched by OSM way name and a spatial buffer.

Scenario engine and validation. Each scenario is a versioned list of OSM way IDs with its sourcing documented (published flood reports and news-documented closures for the flood set; corridor geometry for elevated and metro sets). The engine removes edges, recomputes travel-time metrics on the identical origin–destination sample, and diffs against the baseline, so every delta is a paired measurement. Validation runs only after the main results and can confirm or contradict the ranking — it can never tune it.

HydNet-Critical is a scripted pipeline whose artefacts — snapshot, graph, criticality tables, scenario definitions, deltas — are versioned files, so any dissertation number can be traced and regenerated.

flowchart TD
  A["Geofabrik India OSM extract<br/>pinned by download date"] --> B["OSMnx graph construction<br/>drive filter, simplify, travel time"]
  B --> C[("Versioned graph<br/>nodes, edges, travel_time")]
  C --> D["Criticality module<br/>betweenness, closeness, corridor roll-up"]
  C --> S["Construction sensitivity runs<br/>filters, tolerance, speed assumptions"]
  S --> D
  F["Scenario definitions<br/>bridge, elevated, metro, flood sets"] --> E["Scenario engine<br/>edge removal, same OD sample"]
  C --> E
  D --> E
  E --> G["Accessibility deltas<br/>travel time, detour ratio, disconnections"]
  H["HMDA CTS corridors and<br/>documented flood closures"] --> V["Validation and robustness checks"]
  G --> V
  D --> V
  V --> R["Results chapter:<br/>tables, maps, scenario report"]

Three design decisions carry the study's credibility. First, construction transparency: the network-build report lists every filter, tolerance and speed assumption next to the alternative considered, because centrality answers are only as trustworthy as the graph they run on — and the sensitivity module quantifies that trust. Second, paired scenario evaluation: every scenario reuses the same origin–destination sample as the baseline, so accessibility deltas are paired measurements rather than two unrelated runs. Third, separated validation: the comparison against HMDA CTS corridor evidence and documented flood closures lives in its own module and its own chapter — it runs after the main results and can only confirm or contradict the ranking, never tune it.

The scenario engine is deliberately simple — remove edges, recompute, diff — which keeps the mathematics inspectable and the viva defensible: every delta in the dissertation is a shortest-path computation on a graph the examiner can see and reproduce.

• Population-weighted accessibility using census and gridded-population data, so deltas weight people rather than zones equally.
• Coupling with flood-extent layers to generate scenarios from exposure modelling rather than documented closures alone.
• Multimodal resilience: adding metro and bus networks to study substitution when road links fail.
• Sequential and adversarial disruption (nth-link failure, percolation curves) beyond the fixed scenario matrix.
• Probe-data calibration of link speeds, and replication for other Indian metros to test whether criticality patterns transfer.