Satellites and Household Surveys Pinpoint Kerala’s Most Dangerous Flood Zones

When the monsoon of 2018 unleashed rains roughly 42 percent above normal across Kerala, the hill district of Wayanad became one of the state’s most haunting landscapes of destruction. Rivers such as the Periyar, Pamba and Chalakudy burst their banks, thirteen districts were flooded, and in villages like Pozhuthana the water tore through homes, roads, bridges, schools and health facilities. Years later, a team of geologists from MES Ponnani College, affiliated with the University of Calicut, has returned to that disaster with a different kind of instrument: not a rain gauge or a river gauge, but a carefully weighted combination of satellite data, digital terrain models and a decision-making mathematics developed in the 1970s. Their goal was deceptively simple and extraordinarily consequential for the village — to draw a map that shows, parcel by parcel, where the water will come next.

The study, published as an open-access case study in the journal Discover Geoscience, focuses on Pozhuthana Grama Panchayath, a 70.29-square-kilometer patch of the Western Ghats in Vythiri Taluk whose very name is said to derive from the phrase Puzha Tanna, meaning the place that rivers gave. The terrain rises to nearly 2,100 meters, and the elevation within the panchayath itself ranges from 624 to 1,505 meters above mean sea level. That dramatic relief is precisely what makes the area treacherous: water that falls on the high ridges converges rapidly into the valleys, and the low-lying valley fills and gently undulating piedmont zones act as temporary reservoirs during heavy rainfall. The researchers chose the site deliberately, because it sits at a scale that most flood studies in India ignore. While susceptibility mapping has been carried out for entire river basins and states, micro-level assessments at the panchayat scale — the level at which Kerala’s Panchayati Raj institutions actually plan and respond — have been conspicuously missing.

To fill that gap, the team built what hydrologists call a multi-criteria decision analysis, executed entirely within the ArcGIS 10.8.2 environment. Eight thematic layers were assembled: elevation and slope extracted from a Shuttle Radar Topography Mission digital elevation model, rainfall interpolated from station data, drainage density derived from the stream network, and geology, geomorphology, soil and land use/land cover compiled from satellite imagery, the Bhukosh portal and field-verified geological maps. Each layer was reclassified onto a common scale of one to five, where five marks the strongest contribution to flood susceptibility. Low elevations, gentle slopes, dense drainage networks, clay-rich soils and built-up or agricultural land all scored high, because each of these conditions either slows the escape of water or accelerates its arrival.

The rainfall surface deserves particular attention, because rainfall is the engine of every flood in the region. The team used inverse distance weighted interpolation, a technique that estimates values at unsampled locations by giving progressively less influence to measurements taken farther away. A power value of two was applied to sharpen that distance decay, and a fixed search radius ensured that every interpolated cell drew on a consistent neighborhood of observations. Data from the Pinangode and Vythiri stations, roughly seven kilometers apart, were blended with field-monitored rainfall measurements collected on site, and the resulting raster surface was validated against the observed moisture variability across the panchayath. The approach is mathematically humble compared with sophisticated geostatistics, but its minimal assumptions make it well suited to data-sparse mountain terrain and to the fast-moving realities of monsoon flooding.

With the eight layers ranked, the researchers turned to the Analytic Hierarchy Process, the pairwise comparison framework devised by Thomas Saaty, to decide how much each factor should count. Using Saaty’s one-to-nine fundamental scale, they judged rainfall to be the dominant driver, assigning it a weight of 25 percent, followed by elevation at 20 percent. Slope, geomorphology, soil, drainage density and land use/land cover each received 10 percent, while geology — which influences flooding indirectly through the low permeability of the region’s charnockite and gneissic rocks — received 5 percent. The pairwise comparison matrix was then normalized by column summation and row averaging to extract the priority weights. The consistency arithmetic that followed was, by the standards of the method, remarkable: the principal eigenvalue came out at exactly 8.0, equal to the number of parameters, yielding a Consistency Index of zero and a Consistency Ratio of 0.00 against a random index of 1.41. In plain terms, the expert judgments embedded in the matrix were perfectly transitive, with no logical contradictions, which is rare in real-world applications where a ratio of 0.10 is usually considered acceptable.

The weighted overlay of the eight reclassified layers produced the study’s centerpiece: a Flood Hazard Zonation map dividing the panchayath into five risk classes. The spatial pattern is strikingly coherent. Very high risk zones, covering 7.47 percent of the area, cluster in the central to north-central basin around Achooranam and Pozhuthana, with high risk zones — 28.86 percent of the land — extending toward the northeast near Pinangode. Low and very low risk zones, together accounting for 38.64 percent, dominate the western, southwestern and southern sectors near Sugandhagiri and Kolichal, where the terrain is higher, slopes are steeper and gravelly soils drain more freely. The moderate class occupies 25.05 percent. Combined, the high and very high categories reveal that just over 36 percent of the panchayath faces meaningful flood exposure, and those hotspots coincide with low-lying ground, clay and loam soils that swell and seal when wet, dense drainage networks that concentrate runoff quickly, and settlements whose impervious surfaces shed rainfall rather than absorbing it.

What elevates this study above many GIS-based hazard maps is the ground truthing. The researchers surveyed 1,120 households across the panchayath, using purposive sampling that deliberately targeted riverbank settlements, low-lying areas and historically vulnerable zones. Of those households, 118 — 10.5 percent — reported being directly flooded during the 2018 disaster, while the remaining 1,002 occupied relatively elevated ground. When the affected houses were plotted against the modeled hazard zones, the correspondence was absolute: every one of the 118 flooded households fell within the very high risk zone. GPS measurements confirmed that flood impacts concentrated at the lowest elevations while higher ground remained largely untouched, exactly as the weighted model predicted.

The statistical validation was equally emphatic. A Receiver Operating Characteristic analysis, the standard test of a classification model’s ability to discriminate between positive and negative cases, produced an Area Under the Curve of 0.97 — a value indicating exceptional discriminatory power, close to the theoretical maximum of 1.0. A confusion matrix built from the household survey showed that the model correctly identified all positive cases, the flooded houses, and all negative cases, the unaffected ones. In practical terms, the combination of rainfall and elevation weights, reinforced by the six supporting layers, captured the hydrological reality of Pozhuthana with near-perfect fidelity, isolating a small, specific high-risk portion of the landscape that contains virtually all of the observed damage.

The implications reach well beyond one village. The 2018 deluge was not an anomaly in Kerala’s history — major floods struck the region in 1341, 1907, 1924, 1961, 1974, 1992, 2003, 2013, 2018 and 2019 — and climate change studies suggest that short-duration, high-intensity monsoon rainfall events are becoming more frequent. Yet most hazard modeling in India operates at basin or district scales, too coarse to guide the decisions of a panchayat council deciding where to permit construction, where to reinforce a bridge, or where to stage evacuation routes. By demonstrating that a weighted overlay model, calibrated with nothing more exotic than a digital elevation model, satellite imagery, station rainfall and a household survey, can achieve near-perfect agreement with observed flood impacts, the study offers a replicable template for other vulnerable rural communities in hilly terrain.

The authors are careful to note that flood simulation and risk assessment cannot prevent floods; they can only reduce the damage they inflict. But the Pozhuthana map turns an abstraction into a planning instrument. It tells the village’s local government, which the study highlights as a critical actor in the disaster preparedness and recovery chain, exactly where the 36 percent of land that matters most lies. It identifies Pozhuthana and Achooranam as the most critical zones, flags the dense drainage convergence areas of the central basin as hydrological flashpoints, and confirms that the communities living on the valley floors — the same households that watched the water rise in 2018 — are the ones the next flood will find first. In an era when a single monsoon season can undo decades of development, a map that knows where the water will go may be the cheapest insurance a mountain village can buy.

Subject of Research: GIS-based flood hazard zonation using the Analytic Hierarchy Process in a Kerala panchayat

Article Title: Flood hazard zonation in Pozhuthana Grama Panchayath of Kerala using geoinformatics and analytic hierarchy process

Article References: Shirin, P. N., Sanjayan, M. S., Brijesh, V. K., & Swetha, R. (2026). Flood hazard zonation in Pozhuthana Grama Panchayath of Kerala using geoinformatics and analytic hierarchy process. Discover Geoscience, 4(1), Article 320. https://doi.org/10.1007/s44288-026-00664-6

Image Credits: AI Generated

DOI: 10.1007/s44288-026-00664-6

Keywords: flood hazard zonation, geoinformatics, Analytic Hierarchy Process, GIS, remote sensing, Kerala floods 2018, Wayanad, weighted overlay analysis, digital elevation model, drainage density, land use land cover, disaster preparedness