Assessment of Ground and Drone Surveys of Large Waterbird Breeding Rookeries: A Comparative Study
[ad_1]
1. Introduction
The utilisation of drones in ecological monitoring and conservation efforts has seen a remarkable surge in recent years, showcasing their versatility and effectiveness. However, it is crucial to recognise that while drones offer invaluable insights, they do not always supplant human observation but rather augment it with a distinct and complementary dataset.
When studying large aggregations of animals, drones are particularly useful in their ability to map boundaries, build high resolution maps and collect a fixed record of animals in time and space and have been used to study a wide range of vertebrates including birds, hippos, crocodiles and wombats [1,2,3,4,5]. Due to these added advantages, drones have been used extensively to monitor waterbird rookeries including seabirds [6,7,8] and wetland birds [9,10,11,12]. Drones can be used in conjunction with ground-based methods such as researcher visits, camera traps or acoustic sampling to measure abundance and reproductive success and monitor rookery conditions [13,14].
Waterbird rookeries have unique challenges for monitoring, which include restricted ground access, large numbers of birds (>’000′s) and a large spatial extent that may cover many 10s of hectares [8,12]. Consequently, this makes the accurate estimation of rookery size (both in area and number of birds) difficult to obtain by ground-based observers.
Drones can address many of these challenges. Drones can be flown from the outside of the rookery extent, making access potentially easier and quicker and causing less disturbance to the nesting birds [15,16]. They are useful in locating nests [17,18] and capture imagery that can later be analysed to derive the rookery extent and provide a count of nests and/or birds [11,19].
To further the utility of drones in monitoring waterbird rookeries, we aimed to compare ground-collected reproductive success data with drone-derived data to assess whether drone-captured nest data could replace ground-based nest visits. Similar studies have included the calculation of reproductive success but without ground truthing [9] or with focus on non-colonial nesting species [18], on small accessible colonies [20,21], on coastal nesting species [22] or on non waterbird species such as raptors [23]. We hypothesize that the two data collection methods will produce differing reproductive success values, as detection rates will differ between the methods. We predict these differences will vary across the nesting stage of the birds, reflecting the nuanced behaviours exhibited by birds at each developmental phase, with differences particularly prevalent as chicks get older and become more mobile.
3. Results
In total, data from 490 nests were analysed. At the Block Bank rookery, 93 nests were counted on the ground, and 148 from drone imagery (additional nests were visible and therefore counted in the drone imagery). At the Bala rookery, 122 nests were counted in the ground surveys, whereas 127 were counted from the drone imagery (an additional 5 nests were visible in the selected clumps in the drone imagery). If the analyses were not limited to within the assigned ground counted clumps as a comparison study, many more hundreds of nests could have been seen and therefore counted in the drone imagery.
Rookery counts were aligned with chick development stages, with early surveys dominated by egg counts, followed by chicks and squirters, which then progressed into the highly mobile chick stages of runners, flappers and flyers (Figure 2). Flyers were observed only at the Block Bank rookery as the surveys continued later into the breeding season (Figure 2).
Mean clutch sizes, i.e., the number of eggs per nest, were comparable between the ground-based and drone derived counts, with a mean of 1.95 ± 0.6 in ground counts and 1.92 ±0.6 in drone counts (excluding nests with no eggs), and this was not significantly different between survey methods (χ2 (1, N = 62) = 0.001, p = 0.97).
Similarly, the mean numbers of chicks at each nest clump (χ2 (1, N = 66) = 0.73, p = 0.394), squirters at each nest clump (χ2 (1, N = 66) = 0.74, p = 0.391), flappers (χ2 (1, N = 41) = 0.16, p = 0.687) and flyers (χ2 (1, N = 41) = 2.88, p = 0.08) did not significantly differ between survey types. Runner counts, however, were significantly different between survey methods, higher in ground counts than in drone counts (χ2 (1, N = 59) = 10.5, p = 0.001).
Differences in mean counts per nest clump (accounting for the number of nests) between the drone imagery and ground-based counts were more pronounced at different developmental stages (Figure 3), with the largest differences occurring from the runner stage onward, when the young are highly mobile.
When calculating reproductive success using drone-derived or ground-based data final, the estimates were similar. Final ground-based reproductive estimates were higher for both rookeries at 21% at the Bala rookery and 32% at the Block Bank, compared to 17% and 19%, respectively, in the drone surveys (Figure 4). The difference between the two methods was smaller at the Bala rookery (±5%) than at the Block Bank (±13%) (Figure 4).
4. Discussion
The measurement of nesting metrics such as numbers of nests, eggs, chicks, etc., associated with large waterbird rookeries comes with many challenges; the most difficult challenge is often related to the expanse and accessibility. Consequently, the use of drones to survey rookeries has become a natural next step in the development of survey techniques [9]. The ability of drones to collect similar data to traditional ground surveys, however, has been debated [33] and is an area of research that has required further exploration [10]. Our study had the advantage of obtaining both ground and drone data over the same nesting areas in the same survey, allowing us to directly compare the collected counts in both methods. This work further highlights the utility of drones in monitoring waterbird breeding rookeries. However, there are key differences between the survey techniques that suggest both have their unique challenges and advantages.
Perhaps the most significant advantage of the use of drones in waterbird rookeries is the ability to access them from afar. Many wading waterbird species are dependent on inundation regimes to breed and require nests to be built over water [34,35]. Further, they can be sensitive to human disturbance [36], and as a result of a combination of these requirements, they are often found in remote wetland landscapes. This is particularly apparent in Australia, where many major waterbird breeding colonies are located in large floodplain wetlands in the arid and semi-arid zones of Australia [37,38]. Further, their vegetation requirements for nesting often make their nests difficult to see into (particularly tree nesters) or difficult to approach when limited by dense shrubbery [29,39]. As such, drones are a logical approach to gain visual access into these difficult areas.
When aiming to capture information similar to that of ground surveys, however, drones must be flown at low altitudes (~15 m above nests, camera dependant) to obtain the necessary resolution needed to accurately identify and count eggs and determine chick development stages. As drone camera technology improves, flights can be higher above nests. Drone flights over colonial waterbird rookeries are, by definition, in areas of high bird density. This poses a risk to the birds and the drone and requires the drone pilot to have very clear sight of the drone to avoid collisions with birds in the air. The research on drone impacts on birds and other wildlife is growing, with overall low evidence of disturbance to birds [12,40], even when flying at altitudes as low as 12 m [41] and up to a 4 m distance from the birds [42]. This does, however, come with caveats, as some species, particularly some seabirds and swifts, appear to be more sensitive to drones [26,27,43,44]. Amongst waterfowl, some species show more sensitivity to drones than others, but overall behaviour is largely unaffected [45]. Smaller drones that produce less noise have been shown to result in even lower levels of disturbance to wildlife than larger drones [46]. Despite the minimal disturbance of drones to many species of birds, researchers looking to use drones should first check the species-specific literature, watch carefully for signs of disturbance and monitor flights closely to avoid impacts. As such, in many cases, an observer must still be present in the rookery, negating the distance advantage of the drone when the goal is to calculate reproductive success.
An advantage of using drones is the ability to collect large amounts of data in a short amount of time, particularly when flying directly from nest clump to nest clump. The bird’s-eye view of the imagery facilitates the counting of a greater number of nests within the same survey area, particular in areas with dense nesting [47]. Ground-based surveys cover fewer nest clumps due to the difficulties in moving through wetlands and detectability. This was demonstrated in our nest count result at the Block Bank rookery (148 nests captured by the drone vs. 93 in the ground survey). This is particularly important when wanting to increase the number of data points to estimate accurate numbers of eggs for clutch size estimates, or even to assess the number of nests per clump. The ability to see into the areas of the rookery with a drone that are difficult to access can reduce sampling bias, which naturally arises as surveyed nests can be biased toward those that are easily accessible. These nests may be misrepresentations of the entire rookery due to other factors such as higher levels of predation or greater exposure to adverse weather [48,49].
A common difficulty in surveying ibis nests is the mobility of chicks once they reach the “runner” stage, when they are approximately 31–35 days old. At this point, chicks flee, hide in the vegetation or dive into the water when human observers approach, making counts at this stage very difficult, inaccurate and variable between observers. This issue is highlighted in our ground count data, with Bala ground counts having much higher runner proportions compared to the drone surveys (Figure 2), resulting in overall elevated mean runner numbers (Figure 3) and significant differences between survey types. These ground counts outnumber the previous development stages of chicks and squirters, so they are likely to be an overestimate (Figure 3). This likely contributed to the higher success rates in the ground surveys (Figure 4). The drone has an advantage at this development stage in the rookery and can help obtain more accurate counts as runner stage chicks are less likely to flee as the pilot can remain at a greater distance from the nest clump while keeping the drone within visual line of sight. In this approach, chicks are not alerted to an observer’s presence and remain in their nests or nest clump. The drone-obtained photographs of a nest clump is a fixed record in time, allowing for comparisons of observer counts and reducing the variation between observers who must make a rapid count before chicks flee. These images can be assessed at a later date. Note that the analyses of images adds time to the data acquisition process.
Once chicks reach the flyer stage, both ground observers and drone surveys face the same challenges; neither can approach the nest clump without the flyer-stage chicks taking flight. Knowing this in advance can allow observers to perform counts of the nest clump from a distance (perhaps through binoculars if there is clear line of sight), avoiding getting too close and alarming the birds. An alternative method with the drone is to approach the nest clumps with the camera at a 45-degree angle, taking photos of the clump upon approach. In future surveys, a drone could be flown much higher during the flyer stage to reduce the risk of alarming the birds while still capturing useful images as the chicks are bigger and easier to identify at this age. It is clear, therefore, that at the flyer stage, methods must be altered for both the ground and drone counts to be accurate. Regardless of the survey method, assessing breeding success in colonies with highly mobile chicks is difficult and will likely be plagued with low levels of precision.
An advantage of ground-based surveys is the ability of observers to use chick behaviour and movement as a way of determining the ages of the young. For example, chicks at the “flapper” stage are called so because they naturally start stretching and flapping their wings (Table 1). These behavioural changes and motions cannot be seen in a still image collected by a drone, which can lead to the assignation of different age groups between the drone and ground surveys (Figure 3). Such differences can be seen in the inconsistencies between the mean counts across the drone and ground surveys for the squirter and runner age groups (Figure 3). One potential rectification to this issue is to collect video above each of the survey clumps, rather than still images. If a drone was to hover above a clump while recording 30 s of video, this may reduce the inconsistencies between chick development stages observed in the drone data and ground data (Figure 3).
Another important component of ground surveys is the ability of observers to record data on a range of other factors while in the rookery, for example, evidence of disease (behavioural symptoms, e.g., limber neck, lethargy), and predation (destroyed eggs, dead chicks), which are all important predictors of reproductive success [50,51]. These human-derived observations of a rookery’s condition cannot be easily replaced with a drone and are often very valuable additions to the holistic interpretation of the rookery.
Our work has described and demonstrated the differences in nest metrics derived from drone and ground-based surveys. Each method has unique advantages, and the methods do produce different success estimates (Figure 4) that require different interpretations and analyses of the data. The bird’s-eye view of the drone significantly increased the detection rates of individuals and provided a permanent record in time, but it does not negate the need for a human observer to be present in the rookery, who can assess conditions using other senses. We believe that both survey methods make valuable contributions to waterbird monitoring and neither can totally eliminate the need of the other.
5. Conclusions
Our study directly compared ground and drone surveys conducted simultaneously over the same breeding areas of large waterbird rookeries. Drones enabled rapid data collection and reduced observer bias, particularly during the “runner” stage when ibis chicks are highly mobile. However, challenges arise when flying at low altitudes within densely populated rookeries.
Ground surveys provide valuable contextual observations and allow for the assessment of additional factors such as water variables, disease symptoms and evidence of predation. They also offer the ability to determine the age chicks based on behavioural cues, which may be challenging to capture with drones.
In conclusion, both drone and ground surveys offer unique insights into waterbird nesting dynamics, and integrating both methods is ideal for comprehensive waterbird monitoring and conservation efforts.
[ad_2]
