来自 Georgia Institute of Technology 和 Massachusetts Institute of Technology 的一个研究团队量化了蚊子如何定位人类，目标是改进对抗蚊媒疾病的救命陷阱，例如疟疾、登革热和寨卡热，这些疾病合计每年在全球造成超过 770,000 人死亡。研究团队使用对一个非常庞大的飞行资料集进行贝叶斯推断，建立了一个参数少于 30 个的蚊子行为动态模型，解决了先前主要以定性方式而非精确量化建模研究的问题。

实验在一个密封舱内，利用 2 台红外线摄影机追踪 2 只雌性 Aedes aegypti 蚊子，在 20 次实验中每 0.01 秒记录一次移动，累积超过 53 million 个资料点与超过 400,000 条飞行路径。蚊子会集中在人的头部，当受试者穿著黑白衣物时偏好黑色一侧，并呈现 2 种广泛的飞行模式：一种约 0.7 m/s 的 सक्रिय模式，以及一种接近天花板的闲置模式。深色物体会吸引牠们，且牠们会在约 40 cm 内减速，但若没有额外线索，牠们常在落地前转身离开；二氧化碳则产生不同反应，使蚊子减速至 0.2 m/s、飞行变得不规则、能侦测到低至 0.1 percent 的浓度，并可在约 50 cm 外做出反应。当视觉线索与二氧化碳结合时，蚊子会更强烈地围绕目标盘旋并聚集，远比单独任何一种线索更明显，显示这些讯号彼此互相作用，而不只是简单相加。

模型验证显示，若将穿白衣、戴黑色头套的受试者视为深色的二氧化碳来源，它就能重现蚊子在人头周围的密度，支持头部之所以具吸引力，是因为它同时结合了视觉上的深色与呼出的气体。当没有任何刺激时，50 percent 的轨迹收敛距离约为 65 cm；只有视觉刺激时约为 40 cm；只有二氧化碳时约为 25 cm；两种线索同时存在时约为 20 cm。研究人员表示，这个模型可用于在电脑上模拟并优化陷阱，也可能扩展到其他物种，包括 Anopheles；但他们也提醒，有效的陷阱很可能需要经过精细校准的多感官诱引，才能让蚊子停留足够久以便捕捉，而且他们也已发布一个可互动的网页应用程式供测试飞行模型。

A research team from Georgia Institute of Technology and Massachusetts Institute of Technology quantified how mosquitoes locate humans, aiming to improve life-saving traps against mosquito-borne diseases such as malaria, dengue fever, and Zika fever, which together contribute to more than 770,000 deaths worldwide each year. Using Bayesian inference on a very large flight dataset, the team built a dynamic model of mosquito behavior with fewer than 30 parameters, addressing a problem that had previously been studied qualitatively rather than with precise quantitative modeling.

The experiments tracked 2 female Aedes aegypti mosquitoes in a sealed chamber with 2 infrared cameras, recording movement every 0.01 seconds across 20 experiments, yielding more than 53 million data points and over 400,000 flight paths. Mosquitoes concentrated on human heads, favored the black side when subjects wore black and white clothing, and showed 2 broad flight modes: an active mode at about 0.7 m/s and an idle mode near the ceiling. Dark objects drew them in, and they slowed within about 40 cm, but without extra cues they often turned away before landing; carbon dioxide produced a different response, making mosquitoes slow to 0.2 m/s, fly erratically, detect concentrations as low as 0.1 percent, and respond from roughly 50 cm away. When visual cues and carbon dioxide were combined, mosquitoes circled targets and clustered much more strongly than with either cue alone, indicating that the signals interact rather than simply adding together.

Model validation showed it could reproduce mosquito density around a human head by treating a white-clad subject with a black hood as a dark carbon dioxide source, supporting the idea that heads are attractive because they combine visual darkness and exhaled gas. The distance at which 50 percent of trajectories converged shifted from about 65 cm with no stimulus to about 40 cm with visual stimulus alone, about 25 cm with carbon dioxide alone, and about 20 cm when both cues were present. The researchers say the model could be used to simulate and optimize traps on computers and may extend to other species, including Anopheles, but they caution that effective traps will likely need carefully calibrated multisensory lures to hold mosquitoes long enough for capture, and they have also released an interactive web app for testing the flight models.