Article / Shiao-Jun, Chen
Interviewees / Wei-Cheng, Yeh, Park Ranger, Conservation and Research Section, Kenting National Park Headquarters Hong-Qiao, Chen, Technical Specialist, Conservation and Research Section, Kinmen National Park Headquarters Mao-Chang, Chen, Technical Specialist, Public Works Section, Kenting National Park Headquarters Chun-Fen, Kuo, Section Chief, Conservation and Research Section, Yushan National Park Headquarters
About 20 years ago, in order to initiate the remote observation and monitoring of coral reef ecosystems, the Biodiversity Research Center of Academia Sinica and several collaborating units established Taiwan’s first undersea image monitoring system in the waters of South Bay at Kenting National Park. This project ushered in an era of protecting life in national parks through the use of technology.
Recording animal tracks with infrared automatic cameras or video cameras, deploying drones to observe the effectiveness of reforestation and to vaccinate Formosan sika deer, the transition to “eco-lighting” equipment in recent years to protect the living environments of nocturnal animals, and the implementation of artificial intelligence to analyze the parks’ birdsong—these are all examples of applying emerging digital technology to overcome challenges in ecological conservation

When walking around the Sheding tribal village in Kenting, one often encounters groups of charming creatures with large eyes that sparkle like obsidian and brown fur adorned with white spots, which only adds to their cuteness. These are Formosan sika deer, which were once extinct in the wild in Taiwan. It was not until 1984 that Kenting National Park Headquarters undertook the important task of restoring and conserving this species. Through 14 rounds of rewilding from 1994 to 2009, the herd of Formosan sika deer has spread from Sheding to Baishami in Manzhou, Jiupeng Base, and even Chuhuo in Hengchun.
From Restoration to Birth Control
However, the natural conditions at each of these locations vary significantly, including the presence of sufficient food resources in the wild and the suitability of the local living environment and climate. All of these factors have impacted the success of restoration and the distribution density of deer herds in the region. After approximately 30 years, the Formosan sika deer herd near Sheding currently has the highest density. According to a survey commissioned by Kenting National Park Headquarters and conducted by the research team from National Pingtung University of Science and Technology, if the tracking range is extended to Hengchun, the total number of Formosan sika deer is currently estimated to be more than 2,000, which is vastly higher than the local ecological carrying capacity of 1,500.
In recent years, farmers near Sheding and Manzhou have occasionally filed reports with Kenting National Park Headquarters regarding crop damage caused by Formosan sika deer invading their fields. There have also been reports of traffic accidents caused by Formosan sika deer suddenly rushing onto roads or people stopping to view them. Water sports operators have even occasionally witnessed Formosan sika deer jumping into the sea and swimming around while being chased. Evidently, the “human–deer conflict” caused by the herd’s high density is a problem that Kenting National Park Headquarters urgently needs to solve.
Over the years, Kenting National Park Headquarters has continued to cooperate with neighboring schools and research units to track the distribution and estimated population of the Formosan sika deer using various methods. “Based on the literature and data left behind from the early days, we discovered that the original researchers and teachers had a hard time because the experimental equipment and facilities were not sophisticated enough. They could only collect rough statistics by observing the behavior of the animals themselves or their biological remains,” explained Wei-Cheng, Yeh, the park ranger of the Conservation and Research Section at Kenting National Park Headquarters, as he looked through the early records.
Every spring, from April to May, is the season when male Formosan sika deer change their antlers, shedding the old ones and developing new buds. As the new antlers harden, the deer constantly seek tree bark to rub against to help with the shedding and alleviate the itchiness. “Early researchers would collect biological statistics by examining markings on worn out bark or Formosan sika deer’s footsteps and droppings. They employed techniques similar to those utilized by indigenous people in the past when hunting.
This method was used until 2016, when Kenting National Park Headquarters switched to using infrared sensor-type automatic cameras to count the frequency and distribution of deer herds in order to standardize data collection and improve accuracy. The goal was to achieve comprehensive ecological strategy planning and promotion, effectively control the growth rate of wild Formosan sika deer herds in areas where human–deer conflicts were more serious, and balance the overall ecological development. “Birth control” using injections of contraceptive vaccines was one of the milder methods considered; therefore, Kenting National Park Headquarters started to develop such vaccines.





Drones Carry Injection of Contraceptive Vaccines
Each person on the research team of about 10 people led by Professor Chen-Chih, Chen, Institute of Wildlife Conservation of National Pingtung University of Science and Technology, had their own duties, which included researching the ecology and distribution of the Formosan sika deer herd, developing drones, and developing and test vaccines. “At present, this is the world’s first contraceptive vaccine development and injection method. In principle, we have temporarily set up work objectives for a period of ‘half a year’, so the overall progress of research and development appears quite compact, and each work item is interconnected.
“Since we started the project 1 year ago, according to data from blood tests on semi-captive Formosan sika deer in the restoration area, the antibody titer from the contraceptive vaccine is very good, which means that the contraceptive effect is excellent. The next step is to add an adjuvant that can strengthen the effect, and then put the vaccine into the wild Formosan sika deer herd to conduct the clinical trial.”
Consequently, the progress of drone development is extremely important. Yeh mentioned that in the past year, “We have been trying to find the most suitable altitude for the drone to fly at, and discovered that the Formosan sika deer are alerted to the drone at 15 to 20 meters. We use a pneumatic launch device equipped with carbon dioxide cylinders, which not only reduces the drone’s load but also generates less noise during operation, so it can effectively lower the degree of disturbance to the deer herd. These are the experiences gained from losing four drones, and now our first prototype drone has finally been successfully developed and is currently being tested.”
One of the challenges facing the research team is how to mark running wild Formosan sika deer while injecting contraceptive vaccines. “Currently, we are considering marking the deer with glow-in-the-dark, fluorescent, or special light-sensitive pigments,” said Yeh, noting that this will definitely increase the challenge of drone development. However, the project continues to move forward, and each experience raises team members’ expectations and stokes their imaginations. “In order to see the markings clearly, the vaccination process has to be carried out at night. The first obstacle to overcome is that the drone needs to be equipped with night vision and night focusing capabilities. Another difficult challenge is how to accurately calculate the weight under the influence of external forces such as wind power and speed so that the drone can carry both the vaccine and the dye at the same time.”



Monitoring Reforestation Effectiveness with Camera Drones
In Kinmen, camera drones have been deployed to save the nocturnal habitat of great cormorants at Ci Lake!
In the past, over ten thousand great cormorants would travel south to Kinmen every year, nesting through the winter from October to April. At Tai Lake, Yangming Lake and Ci Lake in Kinmen, or Lingshui Lake and Xihu Lake in Little Kinmen, you could see large groups of great cormorants perched on horsetail trees along the lakeshore, stretching their wings in the morning dew, taking to the air and circling the lake, foraging for food, and flying away to frolic. At sunset, they would gradually appear as distant black dots, then fly closer to the shoreline, beautifully silhouetted against the golden glow of the setting sun. Finally, they would land on the branches of tree tops and wait for night to come. Among these birds, the nocturnal great cormorants that inhabited Ci Lake were the largest group.
According to the monitoring records of Kinmen National Park Headquarters, the number of great cormorants wintering in Kinmen peaked at 13,054 in 2019, but has declined year by year, and is now less than 10,000. Based on analysis, Kinmen National Park Headquarters believes that one of the reasons for the great cormorants’ decline is that after Typhoon Meranti blew down 5million trees in Kinmen in 2016, the horsetail tree forests in various areas, including the shoreline of Ci Lake, were severely damaged, destroying the great cormorants’ habitat. Therefore, in 2023, Kinmen National Park Headquarters started a reforestation project for the horsetail tree forests at Ci Lake.
In Kinmen where the sea breeze is strong and the soil is shallow, large trees are difficult to grow. Horsetail trees, known for its drought resistance and salt tolerance, has become an important tree species for afforestation promoted by the military. Over decades, these trees have grown into large trees, attracting the attention of great cormorants migrating south for the winter, which has created a unique ecological landscape every winter. However, through long-term monitoring and records, Kinmen National Park Headquarters has discovered that the great cormorants' preferred night roosting sites have been gradually shifting in recent years. One of the possible reasons for the decline in the group in Ci Lake is related to the devastation of the horsetail tree forests in the area caused by Typhoon Meranti in 2016. As a result, Kinmen National Park Headquarters launched an afforestation program for horsetail trees in Ci Lake in 2023.


However, the shore of Ci Lake where great cormorants live is a sandy area with overgrown thickets , intertwined vines, and weeds covering the ground. "To gain access, you must first open a path, which is time-consuming and labor-intensive. By using drones for preliminary surveys can improve efficiency. Aerial views allow for a more comprehensive assessment of the environmental conditions, while also enabling rapid and extensive image documentation, providing a reference for future monitoring of seedling survival rates after afforestation projects," said Hong-Qiao Chen, Technical Specialist of Conservation and Research Section and licensed drone operator at Kinmen National Park Headquarters.
For example, the initial horsetail tree planting area for the restoration of the great cormorant habitat at Ci Lake was 2.71 hectares, with a survival rate of over 90% after 1 year, which is quite effective. As a result, an additional 1.04 hectares were planted in 2024, with plans to follow up for 2 years in the hope that the survival rate will also exceed 90%. On the other hand, the invasive alien species Spartina alterniflora grows in muddy intertidal areas that are difficult to walk on. Kinmen National Park Headquarters also uses aerial cameras to take pictures and calculate the spread area of Spartina alterniflora to facilitate planning of removal work. In the Zhaishan and Andong areas where the lakes and reservoirs are being dredged, Kinmen National Park Headquarters has also promoted reforestation work in recent years, and also uses aerial cameras to track the results before and after afforestation.
“After all, Kinmen’s military and civil aviation control areas are relatively large, so the altitude and regions where camera drones can fly are relatively limited. Coupled with the influence of the strong northeastern monsoon and sea winds, it’s important to grasp any opportunity to fly,” added Chen Hongqiao, hoping that in the future, more colleagues at Kinmen National Park Headquarters will obtain drone operator licenses. There is a chance that after upgrading the equipment of the camera drones, the camera drones can be used in more applications.



Additional Applications of Camera Drones
Although there are no precipitous mountains in Yangmingshan National Park, visitors do get lost from time to time. Yangmingshan National Park Headquarters has researched the possibility of using camera drones to conduct search and rescue. However, in actual cases of search and rescue, due to dense tree canopies and plants such as silver grass, the camera drones must fly at a very low level to see the ground conditions clearly; therefore, their use is limited. It is still necessary to combine ground search and rescue with the camera drones to get better results.

Smart Streetlights, Light on When Cars Come
In national parks, the desires for “night lighting” and a “dark sky sanctuary” often create a tug-ofwar between nighttime driving safety and ecological protection, especially in areas near highways or communities. Striking a balance between these needs is often a headache for the National Park Headquarters. To solve this problem, in late 2020, Kenting National Park Headquarters set up 20 smart streetlights with built-in sensors on a major road stretching one kilometer between Hengchun West Terrace and the Taiwan Strait.
At night, when the auto-focus infrared cameras installed at both ends of the road detect via artificial intelligence (AI) that a car has entered the section between the Shanhai and Wanlitong communities, they send group control dimming commands to the 20 smart streetlights in the area through the cloud monitoring platform, and the brightness can be adjusted to between 0% and 100%. Each smart streetlight uses its own communication chip to send changes in time and traffic flow to a cloudbased monitoringplatform for scheduling, and then the AI interprets and controls the activation and deactivation of the streetlight. This allows the street lights to be lit sequentially as vehicles approach,reach full brightness at a certain distance before the vehicles pass, and dim to the lowest brightness one by one after the vehicles have left.
“In the early phase of the trial, we thought this was a very innovative device because the duration and brightness of night lighting directly affect biological rhythms and activities. If the smart dimming system can automatically reduce the color temperature to 100 W and the brightness to 3000 K, it can decrease the number of hours of night lighting, which is a great help in maintaining the ecological environment at night,” said Mao-Chang, Chen, TechnicalSpecialist of Public Works Section at Kenting National Park Headquarters. “But,” he continued with some disappointment, “we later found that the effect was not as great as we had imagined.” It turns out that because the area is a major traffic route, there is heavy traffic flow before midnight, so the smart street lights are almost always fully illuminated. It is only after midnight that the automatic dimming effect can be brought into full play.
More importantly, the engineering unit responsible or the tender previously was not familiar with maintenance techniques for AI and information and communications technology, and the cloud monitoring platform was also set up at the original winning bidder. If these systems cannot be transferred to local vendors or the Headquarters’ IT department, then after the 5-year trial period, not only will maintenance be a problem, even connecting to the cloud monitoring platform and conducting AI interpretation will become challenges to overcome. Until a solution can be negotiated and achieved, an all-lit approach has been temporarily adopted to cope with the situation.



AI Birdsong Recognition
In recent years, AI has developed rapidly, and the question of how to appropriately utilize it for biodiversity-related research has become an important issue among the international community. In particular, species maps offer important information for understanding the distribution of species and changes in their populations, including information that researchers desperately need to know, such as threats to living environment, challenges encountered due to climate change, and changes after conservation and management actions have been initiated. Among these species maps, the bird maps produced since 1966 are the most complete.
This is also the reason why Yushan National Park Headquarters—which oversees a park home to 233 species of birds—has signed a Memorandum of Understanding with the Taiwan Biodiversity Research Institute pledging to work together on biodiversity-related research and long-term monitoring purposes. The first step in this endeavor is the Taiwan Bird Atlas Project, which was launched in 2024. The project aims to leverage the experience and power of Taiwan’s thriving civic scientists and capitalize on the explosive growth of automated monitoring and AI identification technologies to complete the spatial information puzzle of Yushan’s bird population.
“Our first contact with the Taiwan Biodiversity Research Institute was in 2020,” said Chun-Fen Kuo, Section Chief of the Conservation and Research Section at Yushan National Park Headquarters, about their initial cooperation with the Taiwan Biodiversity Research Institute. Because the southwestern area of Yushan National Park has been impacted by the 921 earthquake, Typhoon Morakot, and subsequent major and minor storms,heavy rains, and earthquakes, the Southern CrossIsland Highway was partially closed, only open for the general public to travel to Meishan. As a result, there was little recent research data on the area from Meishan at the Southern Cross-Island Highway to Wukou, where the elevation increases from about 1,010 meters to about 2,722 meters. At the time, the Taiwan Biodiversity Research Institute, then known as the Biological Research and Conservation Center, needed to set up recognition models for its Sound Identification and Labeling Intelligence for Creatures (SILIC) AI analysis software. The opportunity for collaboration between the two units thus formed, leading to the use of “soundscape” by Yushan National Park Headquarters to understand the evolution of the habitat and biodiversity in the area, which has experienced multiple natural disasters and is almost free of human interference.
With six sets of automated recording devices installed in the area, the system operated 24 hours a day, recording for 1 minute, then resting for 2 minutes. The recordings were fed to SILIC for analysis on a regular basis every month. Through the use of AI recognition, researchers learned that there are six species of Strigidae, including Taenioptynx brodiei, Otus lettia, Otus spilocephalus, Ninox scutulata, Strix aluco, and Strix leptogrammica, singing in the area. Moreover, the higher the altitude, the simpler the composition of the bird species.



These results were useful for studying the ecology of the park. As a result, Yushan National Park Headquarters further signed a Memorandum of Understanding with the Taiwan Biodiversity Research Institute to draw a map of the birds in Yushan National Park through the use of AI identification technology. This map will be used to study the spatial distribution of birds in Taiwan, their abundance indicators, and the relationship between their breeding status and their habitats. Moreover, Taiwan Biodiversity Research Institute will become a partner in the International Bird Atlas Program so that more bird lovers can be attracted to come to Taiwan for exchanges and bird-watching. Furthermore, it will also provide a record of all the sound-producing animals in Yushan National Park, which will serve as a basis for future research.
Over the past 20 years, digital technology has progressed rapidly. Applications include the latest surveying techniques and tools used to identify changes in topography and geomorphology of national parks, which are vast in size and have great differences in altitude. National parks are also utilizing a variety of newly developed digital tools to guide people into the uninhabited areas of the parks, providing immersive experiences through virtual reality images and allowing guests explore the encyclopedia of nature. More importantly, digital technology is helping national parks overcome the challenges and constraints they have faced in the past. The applications of these innovative technologies for researching issues related to ecological conservation are the best proof that national parks are breaking away from traditional thinking and expanding and deepening their functionality through the use of technology.


