Research reveals how mosquitoes use vision and carbon dioxide to hunt humans

At a glance:

• Researchers from Georgia Tech and MIT used Bayesian inference to build a mathematical model of mosquito flight behavior from 53+ million data points.
• The study found mosquitoes have two flight modes—active exploration at 0.7 m/s and an idle state for landing preparation—and are attracted to dark objects and carbon dioxide.
• Combining visual and carbon dioxide stimuli reduced the approach distance to just 20 cm, with human heads being prime targets due to their darkness and CO2 emissions.

The Largest Dataset of Its Kind

Infectious diseases transmitted by mosquitoes—including malaria, dengue fever, and Zika fever—claim more than 770,000 lives worldwide each year. Understanding how mosquitoes locate and target humans has long been a critical challenge in controlling the spread of these deadly diseases. A research team led by the Georgia Institute of Technology and Massachusetts Institute of Technology has now made a significant breakthrough by automatically deriving a dynamic model governing mosquito flight through the application of Bayesian inference statistical methods to an unprecedented volume of movement data.

The researchers released two female Aedes aegypti mosquitoes into a sealed experimental space and recorded their flight paths in 0.01-second increments using two infrared cameras. The data collected from a total of 20 experiments exceeds 53 million points, with more than 400,000 flight paths recorded. This represents the largest dataset ever assembled for quantitatively measuring mosquito flight behavior. "The big question was, how do mosquitoes find a human target?" explains Cheng-Yi Fei, a postdoctoral researcher at MIT. "There were previous experimental studies on what kind of cues might be important. But nothing has been especially quantitative."

Visual Cues Play a Surprising Role

The experiment began by photographing mosquitoes flying around human subjects dressed in dark-colored clothing. This observation revealed that Aedes aegypti mosquitoes were concentrating their approach on human heads—a fundamental discovery that served as the starting point for the entire study. The researchers then experimented with subjects dressed in black on one side and white on the other. Although carbon dioxide and body odor were emitted equally from both sides of the body, the mosquitoes' flight trajectories were concentrated only on the black side. This result vividly demonstrated that visual stimuli play an important role in target search even in apparently windless environments.

A detailed analysis of mosquitoes flying in a stimulant-free environment revealed that their flight patterns could be broadly classified into two distinct types. One was the active state, in which mosquitoes actively explored the space while maintaining a speed of approximately 0.7 meters per second. The other was the idle state, in which they flew almost without using thrust. The idle state is thought to be a preparation stage for landing and was observed more frequently near the ceiling of the experimental space.

How Mosquitoes Respond to Different Stimuli

Analysis of mosquito responses to visual stimuli revealed that mosquitoes are attracted to dark objects and slow down when they get within about 40 centimeters. However, without additional cues such as body odor, humidity, or heat, mosquitoes often flew away even after approaching their target. This suggests that visual stimuli alone are insufficient to induce landing and blood-sucking behavior.

The response to carbon dioxide sources was entirely different. Mosquitoes that entered within a radius of about 40 centimeters of the carbon dioxide source suddenly slowed to 0.2 m/s and began flying erratically, swaying without a clear direction. Numerical simulations also showed that mosquitoes can detect carbon dioxide concentrations as low as 0.1 percent and that their detection range extends to approximately 50 centimeters from the source.

Furthermore, the mosquito response changed dramatically when visual stimuli and carbon dioxide were presented simultaneously. The mosquitoes began to circle around the target, and significantly more mosquitoes concentrated near the target than when either stimulus was used alone. According to the researchers, this behavior could not be reproduced by a model that simply added the responses to vision and carbon dioxide—in other words, it is highly likely that multiple sensory sources influence each other in the mosquito brain.

Why Human Heads Are Prime Targets

To test the prediction accuracy of the mathematical model, the research team used a subject dressed in white with a black hood as a "black sphere emitting carbon dioxide" to see how well the model could reproduce the actual distribution of mosquitoes. As a result, they succeeded in accurately predicting the mosquito density distribution around the human head. The human head often appears dark to mosquitoes and is also a part of the body that emits a lot of carbon dioxide, making it a place where two types of mosquito-attracting stimuli overlap.

To quantify the risk of mosquito bites, the researchers measured the distance at which 50 percent of their trajectories converged around the target, which was about 65 cm without any stimulus. With visual stimulus alone, the distance was about 40 cm; with carbon dioxide alone, about 25 cm; and with a combination of visual and carbon dioxide, the distance was reduced to about 20 cm. This again demonstrated that mosquitoes tend to approach humans more closely when multiple sensory stimuli are superimposed.

Applications for Mosquito Control

Researchers believe that the mathematical model they have developed will allow for the pre-simulation and optimization of mosquito trap designs on computers. They also hope that it will have applications to other mosquito species, including the Anopheles mosquito, which transmits malaria. "Our work suggests that mosquito traps need specifically calibrated, multisensory lures to keep mosquitoes engaged long enough to be captured," says MIT professor Jorn Dunkel.

The team has also developed an interactive web app that allows users to try out flight models of all the mosquitoes they studied. Using Bayesian inference, the researchers were able to construct a mathematical model that could reproduce experimental results with high accuracy while compressing mosquito behavior to fewer than 30 parameters. This parsimonious approach makes the model practical for real-world applications in vector control and public health planning.