Vector-borne diseases have been on the rise as urbanization brings more people into contact with disease-spreading insects. This phenomenon is typified by the mosquito Aedes Aegypti, responsible for the spread of Zika, chikungunya, yellow fever, and dengue. A 30-fold rise in incidences over the last half-century demonstrates the urgency of employing efficient and targeted vector control methods that can also be utilized in resource-stretched environs in the Global South. Geospatial innovation, including the methods detailed in this blogpost, are helping advance the effort to control vector-borne disease spread.
At the moment, intervention possibilities to fight the spread of mosquito-borne illnesses include distribution of vaccines and physical obstacles to infection like mosquito nets, but the lack of viable vaccines for dengue, Zika and chikungunya and the daytime activity of the Aedes Aegypti mosquito make both avenues less effective in these contexts. Thus, one of the most effective currently-available responses for these mosquito-borne illnesses lacking viable vaccines are vector control methods- eliminating mosquito populations at their breeding ground.
Traditional implementation of these methods proves costly, necessitating a massive workforce, local knowledge of mosquito populations, and facing the obstacle of regulatory constraints. Given the resource scarcity in many nations in the Global South burdened with widespread distribution of Aedes Aegypti, identifying, developing, and implementing cost-effective and targeted prevention techniques is of utmost importance.
Mapping is a viable solution for planning disease prevention based on the location of mosquito breeding habitats. Specifically, public health authorities use risk maps to consider the vulnerability of areas and populations to infection through a combination of hazard mapping of suitability areas and exposure through overlays of human density during biting hours.
Aedes Aegypti is uniquely challenging for mapping. Mosquito populations display high spatial variability due to the species' short flight range of around 200 meters from its habitat and the use of available small breeding habitats in heterogeneous urban environments. These mosquitoes breed in flowerpots, storm drains, street debris, and water containers on roofs often required by a lack of water pipelines. These factors make breeding grounds difficult to detect, yet scientists can use digital methods and data to map areas in cities with higher general suitability for breeding grounds..
Researchers have explored two main approaches to mapping this variety of mosquitoes: sample-based entomological surveillance systems and modeling mosquito abundance with digital proxies.
The former method yields precise counts of mosquito populations but requires an army of laborers and trained personnel to retrieve such data. High breeding ground spatial variability and costs limit the feasibility of sample-based techniques, but including digital proxy use can fill gaps and enrich datasets where classical approaches cannot produce continuous temporal and spatial coverage.
Consider, for example, the case of entomological indicators from ovitraps, which are valid for a buffer or flight range of 200 meters. In a large city where even thousands of traps cannot cover the entire area, modelling techniques can evaluate proxies in order to augment existing trap data. In areas with no existing data, these methodologies can also be used to predict breeding grounds, a less precise but valuable tool in places without the resources to collect base data to build upon.
Given the promise of spatial proxies for more efficient mosquito population mapping, our collaborative team at GIScience and HeiGIT seeks to address some of the research gaps in such modeling in order to aid in the World Health Organization pillar of action calling researchers to “scale up and integrate tools and approaches for global vector control” and the UN’s Sustainable Development Goal 13 to “Strengthen the capacity of all countries, in particular developing countries, for early warning, risk reduction and management of national and global health risks.”
The GIScience and HeiGIT team is working to fill a major gap in current research: the scalable generation of high resolution proxies that allow interpolation of entomological data. In their projects, the team hopes to provide open-source proxies and methods for Aedes Aegypti mapping. Deep learning object detection models have the promise of capturing small breeding containers with low manual labeling effort from very high resolution (VHR) satellite imagery.
The group applied these methods to the city of Rio de Janeiro and designed an algorithm to detect Aedes Aegypti breeding sites from open and available satellite and street view images, though the approach could be used for everything from modeling entomological survey data to predicting inner-urban occurrence patterns when combines with exposure layers.
The resulting maps for the experimental case study compared favorably to a supervised base model. The method therefore represents a more efficient, less resource-intensive possibility for researchers and practitioners endeavoring to counter the spread of mosquito populations and breeding grounds.
All told, the research and public health community faces a growing risk from Aedes Aegypti, with the World Health Organization estimating that by 2080 over 60% of the world’s population will live under direct risk of this mosquito. Vector controls remain the most effective measure to counter the spread of mosquito-borne illness in the absence of viable vaccines, but innovative solutions using openly-accessible data and modern computing techniques are still required to facilitate resource-sensitive vector control methods.
Cooperation at the nexus of public health and geospatial innovation is both urgent and vital, as shown by workshops held by the GeoEpi team at the recent Center for Geographical Analysis 2023 conference “From Geospatial Research to Health Solutions.” The interchange of ideas and advancements among public health professionals and geospatial researchers will be key to developing the frontier for technology-informed solutions to disease spread, like the approach suggested in the paper “Semi-supervised water tank detection to support vector control of emerging infectious diseases transmitted by Aedes Aegypti,” recently published in the International Journal of Applied Earth Observation and Geoinformation.
The project aligns with Thematic Area 1 (“Innovation and partnerships for better and more inclusive data”) and 2 (“Maximizing the use and value of data for better decision making”) of the United Nations World Data Forum. This work also facilitates the goals and lessons of several sessions in the United Nations World Data Forum, which took place in Hangzhou at the end of April. These sessions include “Increasing the use of innovative sources, methods, and technologies to inform decision-making: Insights from the Data For Now Initiative” and “Overcoming Data Graveyards: Country Insights on Advancing Data Use, Uptake, and Impact.” Open-source methods such as this one, shared with publicly-available data at interdisciplinary venues like the Forum where organizations working in the field can explore novel scientific innovations, offer vital avenues for transforming academic research into lifesaving, cooperative applications to stem the increase in vector-borne disease spread.