By Yin Yan

In the early morning, birdsong awakens the forest in Xinglong Mountain National Nature Reserve in northwest China's Gansu province. Hidden among the dense foliage, acoustic sensors come online simultaneously, accurately identifying and capturing every call.

This serene yet dynamic scene represents a typical day for the wildlife monitoring and conservation research team at Lanzhou University, revealing the remarkable progress of China's ecological monitoring technologies.

Zhang Lixun, head of the research team and a professorate senior engineer at the university's College of Ecology, recalled that in the 1990s, researchers had to carry bulky tape recorders up the mountains. "Once the tapes ran out, we had to hike back down," he said.

From cassette tapes and MP3 players to professional digital recorders, the team has witnessed a transformation in their tools over the years. In 2022, they achieved a significant milestone with the full deployment of a new generation of domestically developed intelligent acoustic sensors, marking the beginning of an era where artificial intelligence enhances wildlife monitoring and conservation efforts.

This technological leap has made broader and deeper observation possible. The team has established 60 monitoring sites across three environmental gradients in and around Lanzhou, capital of Gansu province: from largely undisturbed ecosystems such as Xinglong Mountain and Liancheng nature reserve, to moderately disturbed county-level areas, and finally to downtown Lanzhou dotted with parks and residential communities.

The vast volume of data collected reveals an encouraging trend. "Ten years ago, most birds seen in the city were just passing through," Zhang explained. "Now, many of them have chosen to settle in Lanzhou."

The number of recorded bird species in Lanzhou has grown from just over 100 to 339. Common species such as blackbirds and azure-winged magpies appearing more frequently, and even the once-rare Bohemian waxwings have begun overwintering in Gansu's Yuzhong county. As seed eaters, Bohemian waxwings are drawn by the abundance of Rosaceae plants and Chinese arborvitae in urban green belts, which now serve as well-stocked food sources.

Teaching machines to "understand" birdsong is no simple task. Sound recognition is far more challenging than image recognition. In the early days, members of Zhang's team had to manually analyze recordings frame by frame -- "like editing a film" --to identify bird species.

Today, AI systems conduct initial screening rapidly, though accuracy still requires verification of experienced field researchers who continuously refine the algorithms. The collected acoustic data are uploaded to the University's dynamic wildlife soundscape monitoring and perception platform.

"The more acoustic footprints of local bird species we have," Zhang elaborated, "the more accurate the AI species recognition models become after big-data training." At present, the recognition rate has exceeded 85 percent.

Unlike traditional methods, acoustic monitoring operates around the clock, providing a wealth of data. "While we have accumulated a substantial amount of data," Zhang acknowledged, "the sample size is still insufficient for comprehensive anaylsis." 

Only with more data, he emphasized, can researchers more precisely assess the long-term impacts of climate change or ecological restoration efforts. Zhang believes acoustic monitoring will offer broader research prospects and a stronger data foundation for future ecological studies.

In a paper published by the team in 2024, acoustic monitoring was extensively used to rapidly assess biodiversity during peak periods of bird activity. The study concluded that secondary forests and primary forests play irreplaceable roles in conserving bird diversity, providing scientific evidence for calls to protect intact primary forests and to optimize ecological restoration strategies in arid regions.

At dusk, egrets glide gracefully over the Yellow River, with bird calls blending with the rustling willows along its banks. These sounds, interwoven with the pulse of the city, are faithfully captured by acoustic sensors and ultimately fed into a database, serving as vital indicators of ecological change. According to Zhang, Lanzhou's ecological environment has continued to improve in recent years, offering stronger safeguards for bird habitats.

On laboratory screens, real-time maps flicker with points of light, each quietly telling a story of how technology and long-term dedication enable humanity to listen more closely to nature.

Looking ahead, the team remains optimistic. "We will certainly use better and more advanced equipment," Zhang said. "And we are confident that more birds will choose to make their home in this city along the Yellow River."

Photo shows the dynamic wildlife soundscape monitoring and perception platform developed by Zhang Lixun's team.

Tourists watch swans at a wetland in Dachuan township, Xigu district, Lanzhou, northwest China's Gansu province, Dec. 7, 2025. (Photo/Wang Hong)