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

The risk of mosquito-borne diseases is expected to increase due to climate change, and thus it is needed to forecast how mosquito abundance would vary in response to dynamic climatic conditions. Both short-term prediction and long-term projection are essential for supporting timely mosquito control and mitigating vector-borne disease risk. This study developed a machine learning–based forecasting framework to support an early warning system for short-term mosquito abundance prediction and to project long-term mosquito abundance. Short-term prediction was conducted using machine learning models (MLMs) with short-term meteorological forecast data. Long-term projection was made by future climate data (~2100) under SSP1–2.6 and SSP2–4.5. Results showed that MLMs using short-term meteorological forecast data produced mosquito abundance comparable to those using observational data (R 2 difference of 0.03), demonstrating the reliability of short-term predictions. In addition, it outperformed the current monitoring system with R 2 improvements of 0.20–0.48 across all landscape type. These results indicate that reliable forecasts of mosquito abundance can be provided in advance, enabling public health agencies to implement mosquito control measures before mosquito abundance increases. SHapley Additive exPlanations analysis indicated that the previous day's mosquito abundance and maximum temperature were the most influential ecological and meteorological predictors. Under climate change scenarios, mosquito abundance was expected to decrease due to higher temperatures. In contrast, the rate of increase in mosquito abundance during spring is expected to rise, indicating a temporal shift in mosquito occurrence. These seasonal shifts suggest that mosquito surveillance and vector control may need to begin earlier to mitigate mosquito-borne disease risk. Overall, the proposed framework provides a practical decision-support tool for mosquito surveillance, vector-borne disease risk mitigation, and One Health preparedness under climate change.