About this project

Background

Heatwaves have become an increasingly serious concern in India because of rising temperatures, rapid urbanisation and changing climatic conditions. Several parts of the country have recently experienced prolonged extreme heat, and the diverse climatic character of the subcontinent means that a single national threshold cannot describe heat stress everywhere at once.

Advances in artificial intelligence and the Internet of Things create an opening. AI-driven spatio-temporal analysis can surface complex temperature patterns across regions and seasons, while IoT-enabled Automated Weather Stations supply the continuous localized observation that a gridded model cannot provide on its own. This platform brings the two together on top of the long-period IMD record.

Objectives

1. Analyse spatio-temporal temperature patterns and heatwave behaviour through region-wise and seasonal analysis of meteorological observations.
2. Develop AI-driven models for region-wise maximum temperature forecasting, heatwave prediction, hotspot identification and severity classification.
3. Design and deploy IoT-enabled Automated Weather Stations for localized monitoring, assessment of local heat conditions and validation of region-wise forecasts.
4. Develop visualization, decision-support and early warning mechanisms for effective heatwave monitoring and dissemination of advisories.

Data

Historical maximum-temperature observations are drawn from the India Meteorological Department, Pune gridded archive published at imdpune.gov.in. The GRD files carry timestamp, latitude, longitude and maximum observed temperature, which the pipeline extracts, quality-checks and segments into seven IMD homogeneous regions and four IMD seasons before any model is fitted.

The figures shown across this demonstration site are representative climatological values used to exercise the interface. A production deployment reads the live IMD archive and the live AWS ingest stream in place of the bundled reference data.

Operational challenges

• Continuously retraining models on new observations without degrading forecast skill.
• Choosing AWS sites that genuinely represent the region they are meant to validate.
• Keeping field hardware alive through heat, monsoon, power cuts and network gaps.
• Scaling to more regions, larger archives, more stations and more concurrent users.