Urban Flood Susceptibility Zonation of Guwahati Using an AHP-Based Multi-Criteria Model in QGIS

This planning thesis maps where in Guwahati the terrain and land use make flooding most likely, and converts that map into ward-level planning action. Guwahati, the largest city of North-East India with about 9.6 lakh municipal residents at Census 2011, is squeezed between the Brahmaputra and a ring of hills. Every monsoon, hill runoff meets a choked drainage network — the Bharalu the main channel — while the wetlands that once stored flood water, including the Ramsar-listed Deepor Beel and the Silsako and Borsola beels, keep shrinking.

The thesis builds a flood susceptibility model in QGIS from eight spatial factors: elevation, slope, topographic wetness index, drainage density, distance to drainage, monsoon rainfall, land use/land cover and distance to wetlands. Factor weights are derived with the Analytic Hierarchy Process (AHP): a Saaty pairwise-comparison matrix in Excel, eigenvector weights and a consistency ratio below 0.10. A weighted overlay produces a susceptibility map in five classes, which is validated against historical flood locations compiled from Assam State Disaster Management Authority daily flood reports, satellite flood imagery and a post-monsoon field survey, and cross-checked with the NRSC Flood Hazard Zonation Atlas of Assam.

Susceptibility is then averaged per municipal ward and combined with Census 2011 ward population density to identify priority wards. The thesis closes with planning recommendations — wetland protection, channel restoration, zoning controls on low-lying land and drainage maintenance — aligned with the NDMA urban-flooding guidelines and the Master Plan for Guwahati Metropolitan Area 2025.

Guwahati suffers severe waterlogging every monsoon: hill runoff, an overloaded and encroached drainage network, Brahmaputra backflow into the Bharalu and the steady loss of storing wetlands combine to inundate streets and neighbourhoods across the city. Yet plot approvals, road levels and drainage works are decided without a city-scale, evidence-based picture of which areas are physically most susceptible to flooding, and which wards combine high susceptibility with dense populations. Flooding is managed after the event instead of prevented through planning.

The problem this thesis addresses is: which parts of Guwahati are most susceptible to flooding according to terrain, drainage, rainfall and land-cover factors; how closely does the modelled zonation match recorded flood events; and which wards should be prioritised for wetland protection, channel restoration, zoning controls and drainage investment in the next plan revision?

Abstract

Guwahati floods every monsoon, and the damage repeats because planning decisions are taken without a spatial picture of where water actually accumulates. This thesis constructs a GIS-based flood susceptibility zonation of the city using eight factor layers, weights them through the Analytic Hierarchy Process, validates the resulting map against recorded flood locations, aggregates it to wards and derives planning recommendations for the city's development authority.

Introduction

The city's flooding has a clear geography. Rain falling on the hills around the city runs off into low-lying natural depressions and beels; the Bharalu and Mora Bharalu channels carry city water to the Brahmaputra, but when the river runs high their outflow effectively closes and water backs up in the streets. Wetlands such as Deepor Beel — designated a Ramsar site in 2002 — Silsako Beel and Borsola Beel historically buffered these surges, but encroachment, earth filling and construction have reduced their storage. The result is waterlogging that now disrupts entire wards for days at a time.

Literature review and gap

AHP-based flood susceptibility mapping is well established in Indian research, mostly at district or basin scale. Urban-scale studies that (a) validate against an official record of flood events, (b) report the consistency of expert weights, and (c) convert the map into specific zoning and infrastructure recommendations for a named planning authority are far fewer. The NDMA national guidelines on the management of urban flooding (2010) call for exactly this — prevention through land-use planning rather than relief after the event — so the thesis is positioned as the spatial evidence base those guidelines ask planning authorities to prepare.

Existing vs proposed approach

• Existing: flooding is responded to ward by ward after it happens; drainage budgets are distributed by visibility; low-lying plots are approved without a city-scale susceptibility picture; wetland loss is tracked only when it becomes a dispute.
• Proposed: a transparent, reproducible multi-criteria model in open-source GIS, with published weights and a validation against recorded flood events, aggregated to wards so that zoning, drainage and conservation decisions can be prioritised.

Feasibility

• Data: every input is free — SRTM DEM from USGS EarthExplorer, Sentinel-2 imagery from the Copernicus Data Space, IMD 0.25 degree gridded rainfall, Census 2011 ward tables, and the Master Plan 2025 published by the Guwahati Metropolitan Development Authority.
• Software: QGIS and Excel are free; the workflow runs on a standard laptop with 8 GB RAM.
• Time: data collection and layer preparation take about six weeks; the AHP, overlay and validation fit into another six; recommendations and the report complete the remaining eight, with field visits after the monsoon.

In scope

• The Guwahati Municipal Corporation area with the metropolitan fringe for context, over one monsoon-season analysis.
• Susceptibility as a physical-geographic model (terrain, drainage, rainfall, land cover) — not hydraulic simulation of water depths.
• Five susceptibility classes; ward-level aggregation; population density as an exposure overlay.
• Planning recommendations at the scale of zones, channels and wards.

Out of scope

• Hydrodynamic (1D/2D) modelling of flow depths and velocities.
• Real-time flood forecasting or early-warning systems.
• Detailed engineering design of drains, pumping stations or embankments.
• Legal adjudication of encroachments on individual plots.

The thesis follows a quantitative spatial research design with multi-criteria decision analysis, executed over 20 weeks.

Analytical methods: AHP (Saaty's 1–9 pairwise scale, principal-eigenvector weights, consistency ratio), weighted overlay, supervised LULC classification, zonal statistics, and descriptive comparison of modelled susceptibility with recorded flood events. A noted caveat: IMD's 0.25 degree grid is coarse relative to one city, so rainfall enters the model as a monsoon context factor, and the spatial discrimination comes mainly from terrain, drainage and land cover — this limitation is stated openly in the report.

The study design below is reproduced in the methodology chapter of the report.

flowchart TD
  A["Data acquisition: SRTM DEM, Sentinel-2 imagery, IMD gridded rainfall, drainage network, Census 2011, Master Plan 2025"] --> B["Pre-processing in QGIS: clip, reproject, resample, fill DEM sinks"]
  B --> C1[Elevation]
  B --> C2[Slope]
  B --> C3[Topographic wetness index]
  B --> C4[Drainage density]
  B --> C5[Distance to drainage]
  B --> C6[Monsoon rainfall]
  B --> C7["Land use and land cover"]
  B --> C8[Distance to wetlands]
  C1 --> D["Reclassify each factor into five ranked subclasses"]
  C2 --> D
  C3 --> D
  C4 --> D
  C5 --> D
  C6 --> D
  C7 --> D
  C8 --> D
  D --> E["AHP: Saaty pairwise comparison of the eight factors"]
  E --> F{"Consistency ratio below 0.10?"}
  F -- no --> E
  F -- yes --> G["Weighted overlay: susceptibility index = sum of weight x class rank"]
  G --> H["Susceptibility zonation map in five classes"]
  I["Historical flood points: ASDMA reports, satellite flood imagery, field survey"] --> J[Validation]
  H --> J
  H --> K["Ward-wise aggregation with Census 2011 population density"]
  J --> K
  K --> L[Priority wards for planning action]
  L --> M["Planning recommendations aligned to NDMA guidelines and Master Plan 2025"]
  M --> N[Report, map set and final jury]

Factor set and rationale

Validation logic

Each historical flood point is checked against the class it falls in; the map is judged credible when the majority of recorded events fall in the high and very-high classes and misses can be explained (for example, a flood caused by a local drain failure rather than terrain). No numeric accuracy claim is reported without this explanation.

• Add Sentinel-1 SAR-based flood-extent mapping for validation independent of reports and news.
• Pilot a 1D/2D hydrodynamic model (HEC-RAS) on one catchment, such as the Bharalu, to move from susceptibility to simulated depths.
• Extend the analysis from susceptibility to a full risk index by adding building and infrastructure exposure with vulnerability weights.
• Integrate the draft GIS-based Master Plan 2045 data so recommendations feed directly into the new plan's land-use proposals.
• Repeat the analysis every few years to track how beel loss shifts susceptibility — a simple monitoring dashboard for the development authority.