Global Fellow Di Yang Connects Remote Sensing and Citizen Science in Guatemala and Mexico

A satellite can detect the vegetation, heat and moisture that make an area favorable for mosquitoes. What it cannot see is the rainwater inside a discarded tire, a clay water jar behind a house or a puddle that may disappear within days. Many breeding sites are smaller than a satellite pixel.

Di Yang, Ph.D., assistant professor in the University of Florida Department of Geography, is working to close that gap by connecting what satellites detect from space with what people observe on the ground. This summer, with support from the UF International Center Global Fellows Program and sponsorship from the UF/IFAS Center for Remote Sensing (CRS), Yang took her research from domestic testing into communities in Guatemala and Mexico.

The trip marked the first international deployment of EMERGE — Environmental Monitoring through Education, Research and Geospatial Engagement, an initiative led by Yang’s Geospatial Digital Informatics (GeoDI) Lab. In Guatemala, the team trained more than 2,000 students and teachers to collect environmental observations that can help validate satellite-based models of mosquito habitat risk.

What satellites can’t see

A satellite can detect the vegetation, heat, and moisture that make an area favorable for mosquitoes. What it cannot see is the rainwater inside a discarded tire, a clay water jar behind a house, or a puddle that may disappear within days. Many breeding sites are smaller than a satellite pixel.

Satellites can map the conditions that make an area suitable for mosquitoes, including vegetation density, surface temperature, and moisture, but they cannot see the container itself. That gap is where citizen science becomes part of the research.

Through GLOBE Observer’s Mosquito Habitat Mapper, participants photograph and geolocate breeding sites, generating the ground-reference data for Yang’s deep-learning models. Her lab combines those observations with satellite-derived data processed in Google Earth Engine to produce 30-meter habitat probability maps, an approach that has reached more than 85% accuracy at U.S. test sites.

Guatemala offers conditions that cannot be reproduced domestically: endemic dengue, Zika, malaria transmission, and climate gradients ranging from tropical lowlands to highland ecosystems within a compact geographic area.

“A mosquito breeding site can be a bucket, a tire rim, or a clay water jar,” Yang said. “None of those are visible from orbit. What our models actually learn from is people standing in their own neighborhoods, telling us where the water is. That is not outreach attached to the science. It is the measurement the science depends on.”

Adapting the research with local partners

Part of the Guatemala work involved adapting the observation protocol itself. Container types that dominate breeding habitat in Latin America differ from those common in Florida, so the team worked with local partners to revise how participants classify and record what they find, from tire rims to ceramic water-storage vessels. Getting that vocabulary right is critical to producing data that can effectively train the model.

The EMERGE team in Guatemala included Yang, educational advisor Isidro Niño Olivo, project advisor Russanne “Rusty” Low, and Caroline Nickerson of Florida Community Innovation. They worked in partnership with Magdalena Waleska Aldana Segura of Universidad de San Carlos de Guatemala. Students and teachers moved from classroom instruction into the field, learning to connect their observations with satellite imagery and environmental risk models.

For Yang, those relationships are as important to the future of the project as refining the technology. “The Global Fellows Program funded something that grant budgets rarely cover: the time to sit in a room with partners and figure out what we each actually need,” Yang said. “My models were trained in Florida. These collaborations will shape the next five years of this research more than any single algorithm I write.”

Expanding the work in Mexico

The team then traveled to Mexico’s Yucatán Peninsula, partnering with UNAM’s Escuela Nacional de Estudios Superiores Mérida and Museo de la Luz, UNAM’s science museum in the city’s historic center. Working in a science museum also expanded the project’s reach to families, school groups, and other visitors. To help the training continue after the visit, EMERGE developed bilingual (Spanish and English) digital textbooks and facilitator guides covering participatory science, GLOBE Observer data analysis, Python, Google Earth Engine and geospatial machine learning. Museum staff now have these materials to continue offering the training.

The UF/IFAS Center for Remote Sensing (CRS), which sponsored Yang’s participation in the Global Fellows Program, sees the project as research with both international and local relevance. “CRS is proud to support faculty whose research builds meaningful international partnerships and addresses global challenges” said Jasmeet Judge, CRS director. “Di Yang’s project is a great example, combining innovative technology with local engagement and capacity building. We’re excited to see how these partnerships and insights can contribute to environmental challenges both internationally and here in Florida.”

Bringing international research back to Florida

Although mosquitoes are the immediate research problem and Latin America provides testing environments, Yang sees applications closer to home. Connecting satellite observations with finer-scale information is also important to Florida’s agriculture, water resources, forestry and environmental health. For Yang, citizen science offers one way to close that gap, with an approach that can transfer across research problems. “If we can validate mosquito habitat models with community observations in Guatemala, the same architecture works for water quality in the Indian River Lagoon or forage condition on a Florida ranch,” she said. “I want UF to be the place where that approach is developed and taught, and where our graduate students learn to build both halves of the system.”