Building a cyclone impact forecaster with only 61 data points
Almost every year, India’s east coast—especially Odisha, Andhra Pradesh, and West Bengal—takes a hit from a cyclone. While plenty of tools show where a storm is heading, very few tell you how many people actually live inside the impact zone. Early on, that district-level exposure is what responders really need to prioritize resources. I built Cyclone Impact Forecaster to bridge that gap. It…
India's eastern coast, particularly Odisha, Andhra Pradesh, and West Bengal, frequently experiences cyclones each year. Existing tools can predict a storm's path, but few provide information on how many people live within the impact area. Divyansh0208 developed the Cyclone Impact Forecaster to address this need. The tool combines a 6-hour intensity forecast with a district-level exposure ranking for the North Indian Ocean, displayed on a 3D satellite globe.
Users can access a live demo at https://cyclone-forecaster-s015.onrender.com, and the code is available on GitHub at https://github.com/Divyansh0208/Cyclone.
The system ingests storm data from NOAA's IBTrACS ACTIVE file, using Joint Typhoon Warning Center (JTWC) fixes. It also allows users to replay historical cyclones or input custom storm data manually. The 6-hour intensity forecast is generated using an XGBoost model that predicts wind speed changes based on current conditions, wind patterns, central pressure, and their recent trends.
The district exposure model employs a 2-parameter log-linear model built on 61 historical Indian cyclone records from EM-DAT, along with Census 2011 district populations and areas. This model outputs the expected number of people affected, along with a 10-90% prediction interval for each district within the impact zone.
The Cyclone Impact Forecaster uses MapLibre GL for 3D visualization, with district markers scaling according to the estimated affected population. Clicking a district allows users to zoom in for a closer look. All components rely solely on real datasets—IBTrACS for track data, EM-DAT for historical impact records, and the 2011 Census of India for population figures.
Synthetic data was avoided, and even the testing process used real storm tracks from Cyclone Fani. The system achieved better performance on both intensity and impact models compared to baselines, but acknowledges the modest improvement due to the limited data points.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
