Spots 
Scientists at the Institute of Advanced Virology (IAV) in Keralam are developing a dengue forecasting system that integrates epidemiological and weather data on a machine-learning platform to predict dengue incidence trends in advance.
The aim is to help public health managers plan resource allocation and preventive measures proactively, rather than responding only after outbreaks occur.
Climate factors are important early warning indicators for dengue because they directly influence the biology of the Aedes mosquito and the replication of the dengue virus.
Warm temperatures accelerate mosquito development, shorten the virus’s extrinsic incubation period—the time required for the virus to become transmissible inside the mosquito—and can increase transmission efficiency. Moderate to high humidity can prolong mosquito survival, while moderate rainfall creates breeding habitats. Torrential rain, however, can flush out mosquito breeding sites.
Keralam, which remains warm and wet for much of the year, therefore has climatic conditions favourable to Aedes proliferation and dengue transmission. This has prompted considerable interest in developing forecasting models that combine meteorological and epidemiological data to enable disease surveillance to stay ahead of outbreaks.
“Our machine learning model integrates six years of epidemiological data (March 2020 to February 2026) with climatic variables, including rainfall, temperature and average humidity, to generate weekly dengue forecasts,” says Abhinand C.S., scientist in the Department of Virus Genomics, Bioinformatics and Statistics at IAV, who developed the model.
The Dengue Early Warning System (DEWS), the current version of the system, is accessible at https://dews.iav.res.in, and uses epidemiological data from the State Surveillance Unit (SSU) of the Directorate of Health Services and meteorological data from the India Meteorological Department.
The dengue case trends predicted by DEWS showed a positive correlation with actual case data from the Health department, Dr. Abhinand says. Further evaluation using the SSU data for March-July 2026 has indicated the model’s potential for early prediction of district-wise dengue trends.
The model incorporates daily epidemiological and climatic parameters, including temperature, rainfall and humidity, allowing a more granular representation of disease dynamics. Each district is treated as an independent unit, taking into account district-specific geographical and environmental characteristics, including waterbodies and drainage systems. This enables the model to be trained and optimised for individual districts.
For easier interpretation by public health managers, predicted disease burden is classified into four risk levels—very high, high, moderate and low, based on predefined prediction ranges. This provides a district-wise indication of the expected risk.
Scientists at the Institute of Advanced Virology (IAV) in Keralam are developing a dengue forecasting system that integrates epidemiological and weather data on a machine-learning platform to predict dengue incidence trends in advance.