3.2 Country production and collaboration
The analysis of country performance shows the development path of bird-related research in the main countries. The top 5 countries in terms of total number of publications are given in Table S1. It can be seen that the USA ranks first in terms of total number of publications (356), total citations (7357) and h-index (44), indicating the high academic influence in the field of BIAD. Partners in Flight (PIF) (Hagan and Johnston, 1992) was established in the USA in 1990 to carry out conservation work for migratory birds, followed by the American Bird Conservancy (ABC) established in 1994. The interest in migratory birds led to an explosion of research over a long period of time, and indirectly to the exploration and rapid development of BIAD techniques in the USA (Faaborg, 2005). Research began early in the USA (Fig. 2a) and the findings of the bird radar detection technology were applied in practice. In the early 1990s, the National Weather Service upgraded the USA weather surveillance radars, replacing the previous (WSR-57) radar network with 151 new Doppler weather surveillance radars (WSR-88D) (Gauthreaux and Belser, 2003). Numerous studies on Doppler radar for bird detection have subsequently emerged and have proved that Doppler radar is useful for studies of bird migration (Diehl et al., 2003; Gauthreaux et al., 2001).
For China, research in the field of BIAD technology started late and is relatively weak (Fig. 2b). The focus of environmental protection was mainly on pollution control and reduction, due to the large-scale industrial development that China had previously experienced (Zhang and Wen, 2008). In 2011, the Ministry of Ecology and Environment (formerly the Ministry of Environmental Protection) explicitly proposed the construction of a national biodiversity observation network (China Biodiversity Observation Network, China BON). Bird observation and ecological construction have gradually become the focus of research, which has led to the construction of a standardized and systematic bird diversity observation system on a national scale. After 2015, China entered a phase of rapid development with great potential for growth.
Bird movements are global, often involving multiple regions or countries (Alerstam et al., 2007; Jahn et al., 2020). The study of national co-operation patterns helps to understand research participation and co-operation in frontier areas of ornithology. We apply the SNA to analyze academic collaboration between the top 22 countries in terms of the number of publications (Fig. 2c). The size of each dot represents the number of publications in that country, while the line between the dots indicates there is collaboration between the two countries, with thicker lines indicating closer collaboration. This interpretation is based on the fact that all countries or institutions benefit from the scholarly results of collaborative research (Hood and Wilson, 2001). The USA has the highest number of collaborative publications with 119 articles (Table S1), representing approximately one third of its total number of publications and the majority of publications in the UK are also collaborative (74 articles, or approximately 81%). The USA study is more comprehensive, focusing more on the ecological and conservation aspects of the birds (Fig. S1). The pattern of cooperation in this field is not limited to neighboring countries or within the same continents. It also includes cross-continental and cross-regional cooperation. USA is at the center of this collaborative network (Fig. S2), with the closest collaboration with UK (30 co-authored articles), followed by with China (22) and with Canada (20). The overall results show that the United States has taken good initiative in BIAD exploration and has provided a wide range of collaborative opportunities for other countries.
3.3 Keywords analysis
3.3.1 Keywords clusters
Author keywords highly summarize the core content of an article. We extracted and analyzed the author keywords (occurrence greater than or equal to 6 times) of the 952 articles mined from 2003–2022 and generated an author keyword network map (Fig. 3). Keywords were divided into 5 clusters with each color representing one cluster and the larger nodes representing the higher frequency. The five clusters consist of one research content block (i.e. bird behavior) and four technology blocks (i.e. image detection, radar detection, acoustic monitoring and other combinable technologies).
Cluster #1: Image Processing
Image detection techniques are mainly used for bird species detection (Huang and Basanta, 2019). Information on bird species is obtained by inputting photos captured by the camera and subsequently extracting from the image information and comparing with the trained dataset, i.e., from input to output (Rai et al., 2022). The accuracy and speed of the image recognition results largely depends on the volume of the data for deep learning training (Tsai and Tseng, 2021). Keywords in clustering 1 (object detection, feature extraction, and image segmentation) are important stages of image processing, which are the focus of scholars' research for more accurate and faster detection of targets. Feature extraction, which mainly includes the extraction of morphology, appearance, color, etc., is an input process that extracts important elements of the original image for fine-grained recognition (Rai et al., 2022). Distinguishing the background color of the natural environment from the color of birds is a challenging task (Roslan et al., 2017), that requires precise image segmentation techniques. Image segmentation is a complex and difficult step in digital imaging and is a hot research area in image recognition technology. Bird image recognition, as a typical example of fine-grained image recognition classification, is not only of great academic research value, but also of practical significance in solving practical problems.
Cluster #2: Radar detection
The advantage of radar bird detection over bird image detection techniques is that it is not limited by weather and distance, and can achieve a 24/7 observations to obtain information on the flight height, direction and speed of birds. However, it is currently not possible to obtain accurate species information of birds through radar detection (Larkin et al., 2002). Therefore, radar is often applied in the study of bird migration and airport bird detection (Bruderer, 1997; Schmaljohann et al., 2008). Cluster 2 reveals that, there are more articles on radar studies of bird migration, and that bird detection radar can be used to obtain bird flight information through S-band horizontal scanning radar (coverage) and X-band vertical scanning radar (altitude) (Gong et al., 2019). Doppler weather radar is more resistant to clutter because it can be analyzed in terms of radial velocity and spectrum width (Dokter et al., 2011; Larkin and Diehl, 2002), and clutter signals can be distinguished and removed during analysis. Nonetheless, refinement of the radar echo signal is a prerequisite for improving detection and identification performance. With a wider detection range, Doppler weather radar can analyze bird information while acquiring weather information. Therefore, it is widely used to study the influence of weather factors on bird migration and to build bird migration prediction models by combining into machine learning and other methods (Hamer et al., 2021; Van Doren and Horton, 2018).
Radar bird detection techniques are popular means of detection at airports. There have been studies using radar to successfully distinguish different characteristics of the birds. Gong and co-workers accurately determined the flight pattern (wingbeat or glide) of birds by using radar echoes (Gong et al., 2020b), giving new insights on radar identification to identify bird species. Gong's team also used the variation in radar echoes to accurately identify the size of birds (Gong et al., 2020a). Since bird mass is an indicator to quantify the relative size of the hazard to the aircraft and the size of birds is generally proportional to their mass, then the identification of bird size based on radar echoes corresponds to the level of bird strike hazard as a size class and is re-quantified (Zakrajsek and Bissonette, 2001). This method provides a new way of thinking for airport bird management.
Cluster #3: Acoustic monitoring
Acoustic signals are an important medium for birds to communicate in their natural environment, and acoustic monitoring tools have become one of the key methods used by ecologists to assess biodiversity (Budka et al., 2022). As it can be seen from the keywords in Cluster 3, Passive Acoustic Monitoring (PAM) has been the focus of scholarly attention, with studies often using individual Autonomous Recording Units (ARU) for recording. The large storage capacity of the digital recording devices used for acoustic monitoring allows long recording times, and the recorded data can be used for subsequent studies and re-evaluation (Budka et al., 2022). Acoustically recorded bird sound data usually requires pre-processing operations such as bias correction to ensure accurate results. Acoustic monitoring can be used for a wide range of purposes, including bird identification (Wang et al., 2022), monitoring bird species diversity for avian and ecological conservation (Bradfer-Lawrence et al., 2020), in addition to monitoring the nocturnal migratory behavior of birds (Sanders and Mennill, 2014).
The acoustic signals emitted by birds during different ecological processes (mating, breeding, migration, etc.) are generally different, which makes acoustic monitoring of birds more difficult and at the same time presents an extremely high research value. The variable vocal behavior of birds has an impact on each acoustic monitoring method, therefore, high quality data sets should be trained or for example ensure longer monitoring periods (Perez-Granados et al., 2021), which could become an important future research direction.
Cluster #4: Other combined technologies
The three main techniques of BIAD discussed above are often used in combination with each other or with some other technical means, such as machine learning, deep learning, Unmanned Aircraft Vehicle (UAV) detection, and computer vision processing, as shown in Cluster 4. It is worth noting that artificial intelligence techniques such as deep learning have been widely used as a necessary aid in the field of BIAD. It is undeniable that the development and exploration of the field of BIAD can lead to technological advances in areas such as artificial intelligence.
Cluster #5: Bird behavior
What appears in Cluster 5 are keywords on behavioral aspects of birds, which are also the target of studies using the techniques described above. The behavior of birds emphasizes an adaptive process in which birds tend to produce various responses, i.e., a reaction to stimuli, when facing intrinsic and extrinsic conditions (Driggers et al., 2008). There are many types of bird behaviors, including signaling and communication, predation and prey, foraging, breeding, community, and rhythmic behavior.
3.3.2 Keywords evolution trends
We analyze the trends in the evolution of author keywords over time (Fig. 4). We selected the top 30 author keywords in terms of frequency of occurrence and then divided them into three phases: 2003–2011, 2012–2017 and 2018–2022. The higher the value, the higher the frequency of the word's occurrence over the whole-time span. Table S2 shows the changes in the top 29 author keywords over the three phases, where 'Bird' is the core theme term of research and is therefore not discussed in the analysis.
The first phase (2003–2011) was the initial stage of research in this field. It was mainly characterized by the initial exploration and application of combining information technology with bird observation. It was gradually realized that human observation was limited by many factors and could no longer meet the needs of accurate bird information acquisition and bird ecological research. As it can be seen in Table S2 in terms of technology, radar and remote sensing techniques were already used to study bird behaviors such as migration (Fig. 4a). Since radar systems were maturely developed in the early 21st century and had already laid some foundation for radar ornithology (Gauthreaux and Belser, 2003). As ecological awareness increased, the development of various technologies had been promoted by the increased research on birds, while the research on each technology remained at a relatively preliminary stage.
The second phase (2012–2017) explored particular technical aspects in greater depth than the first phase. PAM and ARU in acoustic monitoring began to be investigated or used. It is worth noting that individual recognition of bird species has matured at this stage (Fig. 4b). Image detection and its related research attracted more attention in this phase, with the main focus on birds as the subject of study and the continuous improvement of techniques such as artificial intelligence. Overall, phase two is more like a bottleneck period, where researchers have overcome more technical challenges, paving the way for an explosive growth in phase three.
The third phase (2018–2022) is a period of bursting growth, where BIAD technologies are relatively mature and most researches focus on the deeper integration of the technologies into technological innovation. "Deep Learning", which did not appear in either of the first two phases, has risen to the second position in phase three showing a continuous growth. Most of the research content has matured at this phase, and AI-related technologies have become the main technical means in BIAD, with a clear growth trend (Fig. 4b). The field of BIAD is entering a period of booming development that will probably continue for a long time. In addition, more research on birds in this phase is turning to the study of human activities (e.g., pollution emissions, night time lights, etc.) or the interaction of man-made structures (e.g., aircraft, buildings, wind turbines, etc.) with birds, and the exploration of the equilibrium between human and bird coexistence (Canney et al., 2022; Dickman, 2010; Nyhus, 2016). Human activities are a fundamental threat to the survival of birds and wildlife, it also constitutes the basis for solving the problem (Dickman, 2010; Schultz, 2011).
3.3.3 Technical prediction
To a certain extent, invention patents are more representative of the practical application of technologies than academic articles. We extract information on patents related to BIAD technologies from the Derwent database for the period 1985–2022.
As it can be seen in Fig. 5, all three mainstream technologies are still in the growth phase. According to the predictions (Fig. 5a), the technology will see faster development in the next six years and is likely to reach maturity in 2028, with a slower growth rate and a gradual decline after 2044. In the future, scholars should focus on building better image learning models, and seeking more accurate and real-time automated means of photography. Also, work on collecting and training better datasets (Li et al., 2019; Zhang et al., 2014), and continuing to move towards digital and intelligent detection in conjunction with the booming field of artificial intelligence. In addition to bird identification, future researchers should focus on the identification of habitat elements in bird images and study the relationship between birds and habitat (Wang et al., 2021; Yousif et al., 2019).
Figure 5b shows that acoustic monitoring has shown a strong vitality during its developmental phase, and the model predicts that it will reach maturity in 2030 and gradually reaching saturation after 2046. Automated acoustic monitoring reduces observer's bias and temporal bias in long-term bird monitoring programmes (Borker et al., 2014). Similar to the environmental disturbance factor of image processing methods, acoustic monitoring means are subject to interference from environmental noise. Future technologies should consider how to eliminate noise without losing information, especially in the target frequency range (Priyadarshani et al., 2018). It is undeniable that in the future this technology sector will continue to improve towards the integration of recording, detection and analysis digitally, reducing human involvement in it, thus reducing the consumption of material and financial resources, while also ensuring accurate and fast detection. Researchers should also look to reduce costs by developing inexpensive, low-maintenance and widely available devices (Zwerts et al., 2021).
Compared with the other two technologies, bird detection by radar systems has a long nascent period in its early stages, after which the technology will show rapid growth, maturing around 2028 and saturation in 2040, before entering a period of decline (Fig. 5c). As a branch of radar detection system, bird detection radar technology has a small number of patents. The number of saturated patents is expected to be around 1857, and an earlier saturation time. One of the difficulties for this technology is to filter out the characteristic waves of birds from the many echo signals. The technology is now widely used in airport bird detection and has the potential to be integrated with a variety of technologies. Bird detection radar can be integrated with optoelectronic technology and combined with deep learning techniques to identify and classify flying birds (Rozantsev et al., 2017). At the same time, for airports, combined bird detection and repulsion equipment can be developed to improve the effectiveness of bird repulsion. As for the technology itself, a new system of bird detection radar with general detection capability should also be developed to create good conditions for the refined processing of echo information of targeted flying birds.
Overall, the development of the three technologies is synchronous. By identifying the life cycle of the three technologies and combining the results in this paper, Fig. 5d shows the evolution and future direction of BIAD technologies. The S-curve results show that the BIAD technologies will be developed to maturity by 2030. In addition to achieving technological precision, efficiency and inexpensive, breakthroughs should be made in traditional technologies, such as holographic digital array radar and external radiation source radar (ERSR), superior target detection algorithms in image processing, and precise separation of multi-source audio signals in acoustic monitoring. Moreover, there might be crucial ethical changes in the human society which could cause founding totally new policy regarding non-human creatures' interests, the future may require new BIAD technologies to avoid disturbing birds. Moreover, the birds as a part of the system also could be changed: species can temporary or permanently extinct or breed. A variety of reasons could affect the development of BIAD technologies.
3.4 The BIAD at airports
3.4.1 Analysis of BIAD techniques at airports
Airports are a key area in the prevention of bird strikes, which suffers from a lack of holistic awareness of bird activity making it difficult to provide accurate and comprehensive bird data to ensure flight safety. Therefore, bird detection at airports is of particular importance. This section analyzes the application and development potential of BIAD in airports from multiple perspectives, taking into account the current mainstream BIAD technologies.
Radar detection, has been widely used as the main bird detection system in airports (May et al., 2017). Professional bird detection radar can achieve detection and tracking of small target birds, which is more suitable for bird monitoring in and around airports (Chen et al., 2012). Most of the bird detection radars used in airports today are marine radar systems with horizontal detection distances ranging from 5–20 km (Beason et al., 2013). Airport bird radar systems have the advantage of being weatherproof and can efficiently provide airport managers with all-weather information on bird activity, so that bird repellent arrangements and flight plans can be deployed in advance (Coates et al., 2011) (Table 1). In contrast to bird detection radar, Doppler weather radar can also be used for bird detection, because of its large monitoring range (up to several hundred kilometers), it is often used to study bird migration. In the future, there is a potential for the development of Doppler weather radar to be applied to bird warning systems at airports. Of the studies related to bird detection by radar (146), 37 were related to airports, accounting for 25.3% of the total number (Fig. 6). However, the reality is that airports and the surrounding environment are more complex, while birds have a small effective reflective surface and other obstacles in the air can affect the detection and tracking capabilities of radar, which can result in inaccurate sensing. In addition, current radar technology still has difficulties to distinguish bird species information, which is not favorable to airport bird data and experience accumulation.
In recent years, with the continuous optimization of target detection algorithms and detection accuracy, research about image detection applied to airport bird data acquisition has gradually increased, accounting for about 8.47% of image processing methods (Fig. 6). The camera or video equipment around the airport can monitor bird activity in real time. Leading to a comprehensive overall knowledge of common bird species at the airport, and thus achieve effective bird strike prevention. However, there are few studies and applications of BIAD at airports, mainly because of some unsolved problems and limitations with the method at this stage (Table 1). There are two main difficulties in airport bird image detection: environmental interference and low detection efficiency. It is difficult to quickly and accurately segment images of the background environment, thus forming interference. In addition, birds' flight trajectory and attitude are variable and prone to tracking loss and the detection efficiency of the current algorithm model, still cannot meet the demand for real-time and accurate detection of birds in the actual airport environment.
Owing to the technical shortcomings of traditional manual observation and bird detection radar which is already widely used at airports, and the continued development of image processing technology in recent years, the combination of the three methods has good potential for application. Continuous manual observation provides a preliminary understanding of the temporal and spatial patterns of common bird species around the airport, selective screening of bird species based on observation results, and the creation and efficient training of image detection datasets. The radar equipment is used to search a large area to obtain information on the location of the target bird, and the camera equipment is used to guide the capture of a small area. The information obtained from the radar and images is combined and analyzed to produce a number of indicators on the bird (species information, flight direction, speed, etc.).
Acoustic monitoring techniques have mainly been used for bird diversity surveys, with a focus on ecological conservation studies. While few deployments for bird monitoring at airports and few of such studies have been reported, with only 6 relevant studies, accounting for 1.10% of bird acoustic monitoring studies (Fig. 6). However, bird acoustic monitoring techniques can achieve bird species identification and obtain some bird information. Acoustic monitoring is generally used to monitor bird activity for long periods of time and the data needs to be collated and analyzed at a later stage, which cannot meet the requirements of real-time airport detection. At the same time, acoustic monitoring equipment are expensive and their practical application value is yet explored. The ecosystems around airports are generally important habitats for birds, which have a bigger impact on flight safety (Zhao et al., 2019), while bird strikes outside the airport environment are beyond the scope of the measures implemented by the airport (McKee et al., 2016). Taking into account the characteristics of acoustic monitoring and the fact that aircraft noise can have an impact on monitoring results, the technique is more suitable for deployment in the ecological environment around airports to obtain information on species and their abundance, and to provide recommendations for airport bird management.
Table 1
Analysis of the application of three techniques in bird detection at airports.
| Types of technology | Advantages | Limitations in airport applications | Application prospect |
| Bird detection radar | 1.Not subject to weather conditions. 2.It is possible to accumulate and obtain bird data throughout the day, which could help analyze the movement patterns of birds. | 1.Need to remove spurious interference quickly. 2.Insufficient coverage of elevation angle, harder to achieve full range of detection. 3.Inability to accurately identify bird species. | Combining radar, imaging equipment and optoelectronic equipment for airport bird monitoring. |
| Image detection | 1.Wide-area and real-time monitoring of birds with low impact on birds and aircraft. 2.Enabling species identification and contributing to data accumulation. | 1.The current general detection model has low detection efficiency, which cannot meet the requirements of real-time bird detection in actual scenes. 2.Interference with target detection by the complex and variable environment of airports. |
| Acoustic monitoring | Ability to monitor the abundance of birds in the habitat around the airport over a long period of time and data can be stored for analysis. | 1.It cannot meet the requirements of real-time airport detection. 2.Acoustic monitoring equipment systems are expensive. 3.High noise levels at airports have an impact on monitoring equipment. | Deployed in the ecological environment around the airport to advise on airport bird management. |
3.4.2 Future bird detection pattern at airports
At present, various bird detection techniques have been gradually applied in airport bird acquisition. This paper proposes a conceptual model for future bird detection at airports based on the results of technical predictions, combined with the development and characteristics of various technologies (Fig. 7). The model initially divides bird detection at airports into micro-areas, meso-areas and macro-areas. The micro-area is the aircraft mainly passes through during taxiing, take-off and landing phases, i.e., within 8 km of the airport perimeter; the meso-area is the airspace within 8–50 km of the airport's perimeter and the surrounding interacting ecosystems (e.g., lakes, wetlands, farmland, forests, etc.), called ecological blocks; the macro-area is the area from 50 km up to the maximum monitoring level of a weather radar. Figure 7 gives the main means of observation for the different divided areas. In the micro-area the main focus is on real-time determination of bird conditions, where manual observation is the most direct, supplemented by bird detection radar, and the identification of species by image detection. In the meso-area, the detection of flight airspace should be based on bird detection radar, with integrated detection and analysis to track the flight of birds and make real-time determinations. For each ecological block, acoustic monitoring equipment can be set up in advance for timely transmission and analysis, to predict the risk of birds in few hours. Doppler weather radar is used in the macro-area to obtain large-scale bird migration information and to set up a prediction system to forecast bird conditions for the next few days or even a month. The results of the forecasts can provide guidance for flight management and bird repelling at the airport.
In general, the future of bird detection should be digital and intelligent, and develop towards a multi-dimensional and a multi-channel acquisition. Different detection techniques have different advantages, and combining them can achieve complementary advantages and mutual evidence. The combination of bird data obtained by various technologies and manual observation will establish a multi-source information system, and integrate detection and analysis to eventually establish a bird observation system.
3.4.3 Bird strike prevention in airports based on a multi-source detection model
Bird strike prevention at airports is characterized by a comprehensive and specialized approach and is a constant concern (Metz et al., 2021). In the future, airports should efficiently integrate advanced technologies from various fields to carry out accurate bird strike prevention and control based on the reliable bird data obtained. Accurate sensing of the overall bird situation at airports is the basis for bird strike prevention and control. A multi-source detection mode (3.4.2) based on automatic detection obtains comprehensive bird information (bird species information, number of birds, flight paths, flight altitude, etc.) over multiple time periods, and visualizes this bird data for effective dynamic bird assessment. The bird assessment process should effectively combine real-time hazard warnings and future bird forecasts to reduce false alarms. Bird predictions should be based on the characteristics of each airport to develop efficient and accurate prediction models. For real-time data, appropriate and effective bird repellent work can be arranged to realize the integration of detection and repellent, while the ground command of flying aircraft can be carried out based on real-time bird information to ensure safe flight. Flight time and flight arrangements can be reasonably made based on the predicted data, and carry out the environmental management within the airport, as well as the construction of the landscape and ecological pattern around the airport (Fig. 8).
