By Michael Audu

Researchers have developed a technique using recurrent neural networks (RNNs) to forecast the population of Ae. aegypti mosquitoes, the primary vectors for Dengue fever, at the neighborhood level. This innovative approach leverages Earth Observation (EO) data as proxies for environmental variables, enabling more accurate predictions of Dengue transmission risks.

Led by Dr. Oladimeji Mudele, a Remote Sensing, Artificial Intelligence and Environmental Health Expert, the study focuses on the interaction between environmental variables and the spread of Dengue, a significant threat to human health in many parts of the world. Factors such as precipitation, humidity, vegetation condition, and land surface temperature (LST) play crucial roles in the development and population density of the female Ae. aegypti mosquito, which thrives in urban areas.

“The spread of Dengue is closely linked to the density of the mosquito vector in a given location, which is influenced by local environmental factors,” according to Dr. Mudele. “Our goal was to develop a model that could accurately predict the population of Ae. aegypti mosquitoes at the neighborhood level, using EO data as inputs.”

The study, supported in part by the European Commission through the Horizon 2020 Research and Innovation Programme, builds on previous research that focused on “nowcasting” Ae. aegypti populations at the municipal level. However, this is the first time a methodology has been developed for spatially disaggregated forecasting at the neighborhood level.

“We used a combination of EO satellite images and RNNs to achieve this,” Dr. Mudele noted. “RNNs, specifically long short-term memory (LSTM) and gated recurrent unit (GRU) models, were chosen for their ability to capture long-term temporal dependencies in data sequences.”

The study employed a clustering step using k-means clustering to simplify the task by grouping mosquito count sequences with similar temporal patterns. This clustering technique allowed the researchers to approximate the distribution of mosquito populations within clusters as a single signal, making the forecasting model more efficient and accurate.

“Our model provides a one-week-ahead forecast of Ae. aegypti mosquito populations at the neighborhood level,” Dr. Mudele said. “This information can be invaluable for public health authorities in planning and implementing targeted vector control measures to prevent Dengue outbreaks.”

The use of EO data and RNNs in this study represents a significant advancement in the field of epidemiological forecasting. By combining advanced AI techniques with satellite imagery, researchers are paving the way for more effective disease surveillance and control strategies.

The findings of this study are detailed in a paper published in the IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, and the researchers are hopeful that their approach will contribute to better understanding and management of Dengue fever and other mosquito-borne diseases.