Simulating Mosquito Populations through the Integration of Climate and Water Resource Modelling

Climate change is a key determinant of public health, influencing disease patterns and human and environmental well-being. Mosquito-borne diseases such as dengue and West-Nile virus continue to pose significant public health challenges worldwide, particularly in regions where environmental conditions favour mosquito production and spread. In recent years, there has been a resurgence of several vector-borne diseases in Europe, driven by climate change, altered water management, and the expanding distribution of invasive mosquito species. Spain has been increasingly affected by this trend, with repeated outbreaks of West-Nile virus—especially in southern regions—and sporadic locally acquired dengue cases reported since 2018.  

Mosquito population dynamics are largely determined by climatic factors, including temperature and water availability. Therefore, understanding the linkage between climate, local water resources, and mosquito dynamics is crucial for better predicting current and future health risks and informing effective disease control and health management. We investigated how the temporal and spatial distribution of water availability and climatic conditions influenced mosquito populations in the Aiguamolls de l’Empordà, a natural wetland area connected to La Muga and El-Fluvia river basins (Catalonia, Northeast Spain), under current and projected climatic scenarios. To do so, we developed a machine learning based Random Forest (RF) model fed withCulex mosquito abundance data (weekly data from 12 traps), climate (rainfall and temperature), and hydrological simulated data (discharge, actual and potential evapotranspiration, and aridity) from 2001 to 2021.  We use projected daily climate from ensemble projections of climate scenarios of the REMO2015 regional climate model under the RCP2.6 and RCP8.5 scenarios (2031-2060) to project the future abundance of mosquito populations in the study area. Our model comprised 48 environmental predictors and the Culex population as the predictand.

 The Culex mosquito population showed a strong positive correlation with temperature-related variables and a negative relationship with discharge and aridity. The RF model showed reasonably good performance in training (R² = 0.90) and testing (R² = 0.61), showing a well-matched temporal pattern of average condition per trap with observed data. Based on Mean Decrease in Impurity analysis, potential evaporation and temperature were found to be highly important predictors.  According to the climate projection under RCP 8.5, in general, mean annual rainfall over the study area will decrease, while minimum and maximum temperatures will increase in the future (2031-2060) compared to the baseline (1981-2010). Thus, these changes could create more favourable conditions for mosquitoes, resulting in substantial additional risk to public health. These results underscore the mounting risk of mosquito-borne diseases in Europe and the necessity for enhanced surveillance and preventive management. Our results contribute to the project “Infectious Disease Decision-support Tools and Alert systems to build climate Resilience to emerging health Threats (IDAlert)” funded by the European Union.

Keywords: Wetlands, Machine Learning, Health Risk, Climate change, Mosquito-borne diseases