Developing machine learning models to forecast mosquito abundance for climate change impacts and early warning system

Area of Science:

• Environmental Science
• Public Health
• Data Science

Background:

• Climate change is projected to increase the risk of mosquito-borne diseases.
• Accurate forecasting of mosquito abundance is crucial for timely control measures and disease risk mitigation.
• Existing monitoring systems may not fully capture dynamic climatic influences on mosquito populations.

Purpose of the Study:

• To develop a machine learning-based forecasting framework for predicting short-term mosquito abundance.
• To project long-term mosquito abundance under future climate change scenarios (SSP1-2.6 and SSP2-4.5).
• To provide a decision-support tool for public health agencies and One Health initiatives.

Main Methods:

• Utilized machine learning models (MLMs) for short-term mosquito abundance prediction using meteorological forecast data.
• Employed future climate data up to approximately 2100 for long-term projections.
• Applied SHapley Additive exPlanations (SHAP) to identify key ecological and meteorological predictors.

Main Results:

• MLMs demonstrated reliable short-term predictions, comparable to observational data (R² difference of 0.03).
• The framework significantly outperformed the current monitoring system, with R² improvements ranging from 0.20 to 0.48.
• Higher temperatures are projected to decrease overall mosquito abundance, but spring abundance increases are expected, indicating a temporal shift.

Conclusions:

• The developed framework provides reliable, advance forecasts of mosquito abundance, enabling proactive public health interventions.
• Seasonal shifts in mosquito occurrence necessitate earlier initiation of surveillance and vector control efforts.
• This tool supports mosquito surveillance, disease risk mitigation, and One Health preparedness in the context of climate change.