Taiwan researchers use AI in study of moth color diversity

Taipei, Oct. 24 (CNA) Researchers at Academia Sinica used a form of artificial intelligence to determine the effects of altitude on the color diversity of moths, the institute said Thursday.

The research, which was published in Nature Communications on Oct. 7, was the first study of animal coloration using the deep learning method, a form of AI, Academia Sinica said.

The study focused on the colors of moths native to Taiwan and how their color traits vary based on the altitudes at which they live. The researchers found that as altitudes rise, moths grow more similar in color, and the body and forewings of moths also grow darker in comparison to the color of the entire specimen.

This gradual decline in color diversity as altitudes increase is likely due to the temperature, the researchers said. Colder temperatures at high altitudes require moths to have better ability to regulate body temperature, and because having darker colors allow this, moths tend to be darker and thus more similar in color at high altitudes, according to the study.

At the press conference, lead researcher Shen Sheng-feng (沈聖峰), an associate fellow at the institute's Biodiversity Research Center, said the use of AI was crucial to the study.

Explaining color variation among animals on a broad geographic scale is usually challenging because human definition of colors and patterns can be very subjective, he said.

The use of AI, however, allowed us to obtain objective descriptions and comparisons of the images, he said.

More than 23,000 images of moth specimens from nearly 2,000 species in Taiwan were analyzed in the study, the institute said.

Most of the images were obtained by 103 volunteers, led by the Council of Agriculture (COA), over a five-year span, according to the institute.

The research team was led by Shen and Chen Sheng-wei (陳昇瑋), a research fellow at Academia Sinica's Institute of Information Science.