Introduction

Guwahati, a city located in Assam, is considered the gateway to the North-Eastern Region of India. Behind the high-precision susceptibility modelling of Guwahati lies a complex, highly iterative process of geospatial data integration, geostatistical analysis, spatial problem solving, and empirical validation. Delineating flood vulnerability at a granular ward level across the Guwahati Municipal Corporation (GMC) required the researchers to navigate significant technical hurdles and scale mismatches to turn raw physical data into an actionable tool for urban spatial justice.

1. The Heuristic Expert Panel

While Multi-Criteria Decision Analysis (MCDA) and the Analytic Hierarchy Process (AHP) are mathematically rigorous, they are inherently built on human expert judgment. To establish the initial weights for the study's 16 hydro-geomorphological and anthropogenic parameters, the researchers assembled a specialised panel of four anonymous experts. These individuals were selected based on their regional geomorphological expertise and a minimum of 10 years of experience in urban flash flood dynamics within the Brahmaputra valley, representing institutions such as Bhattadev University, the National Institute of Hydrology (NIH) Roorkee, and the Central University of Karnataka.

The process of translating their collective expertise into a cohesive model was deeply collaborative and iterative:

• Independent Weighting & Conflict Resolution: Each expert initially weighted the parameters independently. Where disagreements arose, the panel engaged in consensus-building discussions. In cases of extreme conflict, they utilised the geometric mean method to synthesise individual pairwise comparison matrices, ensuring the reciprocal property of the AHP matrices remained mathematically intact.
• The Feedback Loop: To ensure the logical consistency of these expert-assigned weights, the Consistency Ratio (CR) was calculated. If the initial matrices yielded a CR above the scientifically acceptable threshold of 0.10, the experts entered a recursive refinement loop. They re-evaluated and adjusted their pairwise scores until achieving a highly stable, logically consistent CR of 0.07.

2. Resolving the "Scale and Resolution Mismatch"

One of the most significant "behind-the-scenes" challenges of this study was the integration of highly disparate spatial datasets into a single, cohesive analysis grid. The researchers synthesised data across vast scale divides: high-resolution Sentinel-2 10-meter to 1:5,000,000 FAO-UNESCO soil texture maps.

To overcome this scale mismatch, the team performed a rigorous standardisation protocol in ArcGIS Pro. They reprojected and resampled all layers to a uniform 30-meter grid. For the coarse 0.25-degree rainfall dataset, they utilised Inverse Distance Weighting (IDW) interpolation to generate a continuous annual average rainfall surface before downsampling it to match the standard grid. To prevent these anthropogenic and natural parameters from introducing statistical redundancy, they ran a multicollinearity assessment using Tolerance and the Variance Inflation Factor (VIF). By keeping the maximum VIF to a mere 3.636 (well below the critical threshold of 10.0), they proved that human-induced drivers (like roads and urbanisation) provided unique, statistically independent insights.

 3. Multi-Stage Validation

To ensure this expert-driven model wasn't simply a subjective mapping exercise, the authors subjected the resulting Flood Susceptibility Index (FSI) to extensive empirical validation.

First, they compiled a historical flood inventory covering the decade 2014–2024. Using Google Earth Engine (GEE), they processed 10-meter resolution Sentinel-1 Synthetic Aperture Radar (SAR) imagery to capture ground-truth water extent during major flood events in 2017, 2020, and 2022. They identified 367 sample points (184 flooded and 183 non-flooded), splitting them into a 70% training subset and a 30% testing subset.

The model demonstrated exceptional accuracy:

• ROC-AUC Curves: The model achieved an Area Under the Curve (AUC) of 0.828 for the success curve (training) and a superior 0.909 for the prediction curve (testing), classifying its predictive power as excellent.
• Cohen’s Kappa: They calculated a Kappa Coefficient of 0.808, signifying "almost perfect agreement" between historical flood occurrences and predicted susceptibility.
• Sensitivity Analyses: They ran both Single Parameter Sensitivity Analysis (SPSA) and Map Removal Sensitivity Analysis (MRSA). MRSA systematically removed one parameter at a time and evaluated the variance. This mathematical stress-test confirmed that removing the rainfall layer caused the highest sensitivity index variation (0.90%), mathematically validating the expert panel's choice to weight monsoonal rainfall as the primary flood trigger (18.20%).

Relevance to Community: Spatial Justice in Action

As geographers and collaborative partners, this research is directly relevant to how we conceptualise welfare geography and moral geography. In the classical tradition of David M. Smith, our core academic inquiry centres on "who gets what, where, and how". This study serves as a direct modern application of that philosophical framework.

In Guwahati, rapid, unplanned urbanisation has converted the natural landscape into an impervious corridor where built-up surfaces now cover 85.22% of the administrative domain. This concretisation has physically encroached upon vital natural wetland basins—known locally as "Beels" (such as Deepor Beel, Silsakoo Beel, and Borsola Beel)—which historically functioned as natural safety regulators to absorb monsoonal discharge.

By shifting the analytical lens from broad, district-level "macro-basins" to granular ward-level dynamics, this study directly addresses local human suffering. It reveals that 47.35% of the GMC area is highly or very highly susceptible to catastrophic flooding. These vulnerabilities are deeply unequal, concentrating in specific low-lying administrative units like Wards 1, 2, 9, 23, and 24.

For our community, this is where geostatistical modelling intersects with moral responsibility. By identifying "Special Inundation Management Zones", this study provides the precise empirical evidence municipal planners need to implement ward-specific, equitable fiscal allocations and localised drainage master plans. Protecting the remaining Beels through strict "No-Development Zones" is no longer just an ecological recommendation—it is a moral imperative to safeguard the homes, assets, and dignity of the urban poor living on the margins of the Global South's rapidly expanding riverine metropolises.