Toward an early warning system for climate-sensitive vector-borne diseases: insights from visceral leishmaniasis in Brazil
Embargo Date
2027-06-03
OA Version
Citation
Abstract
Climate change is reshaping the geographic distribution and transmission mechanisms of infectious diseases worldwide. Vector-borne diseases (VBD) are highly sensitive to environmental fluctuations and are especially vulnerable to these changes. Even marginal temperature increases associated with climate change enable disease vectors to survive in previously uninhabitable locations and accelerate disease transmission. Changes in precipitation and humidity can influence vector populations by creating or eliminating breeding sites, altering host-seeking behavior, and affecting pathogen viability within the vector. However, the relationship between weather variability and VBD risk is often complex, nonlinear, and delayed, necessitating sophisticated methodological approaches for accurate modeling and prediction. Early warning systems that integrate data on meteorological hazards, environmental conditions, and sociodemographic factors show promise in improving disease prevention efforts and mitigating public health burdens.This dissertation uses visceral leishmaniasis (VL) in Brazil as a case study to develop a robust framework for modeling climate-sensitive infectious diseases. VL is a life-threatening vector-borne disease of significant public health concern in Brazil, yet our understanding of its association with climate-related changes remains insufficient. VL is caused by a protozoan parasite (predominantly Leishmania infantum in South America) and is transmitted to humans by the bite of infected female Phlebotomine sandflies (Lutzomyia spp.). As the most severe form of leishmaniasis, VL has a high fatality rate, exceeding 95% in untreated cases, and remains the second-deadliest parasitic disease globally. While effective treatments exist, they are often toxic, challenging to administer, and can be difficult to access in highly impacted regions. Prevention remains a key public health priority, particularly as the dynamics of the parasite and its vector are tightly linked to meteorological conditions such as temperature, humidity, and precipitation. Given projected shifts in weather patterns and ongoing changes in VL distribution due to urbanization, deforestation, and population mobility, integrating meteorological and environmental data into surveillance systems is becoming increasingly essential for proactive disease control.
With three primary research aims, this dissertation lays the foundation for developing a climate-informed early warning system for VL in Brazil. First, I employ a spatiotemporal modeling framework to examine the association between weather anomalies and VL incidence across Brazil. I assess how these relationships vary by urbanization and land-use changes, particularly deforestation. Second, I develop a machine learning model to predict VL cases several months in advance based on climate and non-climate-related risk factors. By quantifying the relative contribution of each predictor to the overall forecast, this model identifies key environmental and epidemiological drivers of VL transmission across diverse geographic settings. Third, I construct a seasonal and climate-driven mathematical compartmental model leveraging information on the biological mechanisms underlying VL transmission to evaluate counterfactual intervention scenarios, providing insights into potential strategies for disease mitigation.
This research addresses emerging public health challenges as climate change continues to alter vector-borne disease transmission and risk worldwide. Brazil is an ideal case study due to its diverse climatic and ecological conditions, allowing for a comprehensive examination of climate-disease associations. By exploring a combination of data-driven and process-based approaches, this dissertation contributes to the development of scalable and transferable frameworks for modeling and forecasting climate-sensitive infectious diseases. The findings from this work inform the types of models that might, in the future, be most useful in a forecasting framework while directly informing early warning system efforts applicable to VL and other climate-sensitive diseases worldwide.
Description
2025