Comparative reproductive ecology of Old and New World Trogons, an order in decline across the world

Many tropical species show declining populations. The pantropical order Trogoniformes has 76% of its species ranked as declining, reflecting a world-wide problem. Here we report on the reproductive ecology and life history traits of the declining and near-threatened old world Whitehead’s Trogon (Harpactes whiteheadi), the declining new world Collared Trogon (Trogon collaris) and the stable Masked Trogon (T. personatus). We also reviewed the literature on reproductive ecology and life history traits of trogons to assess possible commonalities that might help explain population declines. We found that the declining Whitehead’s and Masked Trogons had reasonable nest success (32% and 25%, respectively), while the stable Masked Trogon had poor reproductive success (9%), all contrary to population trends. However, the limited literature data suggested that poor reproductive success may be common among trogons, which may contribute to population declines. Parents fed young at a low rate and had long on-bouts for incubation and nestling warming that reduced activity at the nest, as favored by high nest predation risk over evolutionary time. We found that young fledged from the nest with poorly developed wings, as also favored by high nest predation risk. Evolved nestling periods among trogon species suggests that poor wing development is likely common. Wing development has been shown to affect juvenile survival after leaving the nest. The poor wing development may be an important contributor to population declines that deserves more attention. Evolved life history traits are important to recognize as creating population vulnerabilities in a changing world.

Methods
Whitehead’s Trogon was studied in Kinabalu Park, Sabah, Malaysian Borneo (6° 05'N, 116° 33'E), a 754 km2 protected area of primary forest. Research was conducted during the 2009–2020 breeding seasons from early February to mid-June. Seven study plots were established at elevations of 1,450–1,950 m. These plots were contiguously located and included ca. 560 ha, with each plot ca. 60–70 ha in size (Martin & Mouton, 2020).

Collared and Masked Trogons were studied in the northern Andes in Yacambú National Park, a 269 km2 area in Lara State, western Venezuela (9°38′N 69°40′W). The fieldwork was restricted to primary cloud forest habitat between 1400 and 2000 m, encompassing a similar elevation range to our study in Borneo. Data was collected during seven breeding seasons from 2002 to 2008 and from late February to early July. Research was conducted on seven study plots similar in size (ca. 60-70 ha) to those on the Borneo site (Martin and Mouton 2020). These trogons were not focal study species, such that we did not collect as comprehensive data as for the Whitehead’s Trogon.

In general, the same standardized data collection methods were used in both Borneo and Venezuela studies, described as following. We located nests by observational cues of breeding pairs and systematic search (Martin & Geupel, 1993; Şahin Arslan & Martin, 2019; Şahin Arslan, Muñoz, & Martin, 2023), and measured the nest and nest-substrate heights using clinometers. We obtained the elevation of the nest location with a GPS device (Garmin, Olathe, Kansas, USA) for Whitehead’s Trogon.

A nest initiation date was specified as the day the first egg was laid in a nest, and the egg-laying season was characterized by the distribution of nest initiation dates. Nests were checked daily during egg-laying and the first two days of incubation to obtain the exact day the last egg was laid to ascertain the start day of incubation. If a nest was found during incubation and was of unknown age, we checked the nest daily until hatch. Nests were also checked daily or twice daily near hatching and fledging to obtain exact timing of transitions for measuring incubation and nestling period lengths (Martin, Oteyza, Boyce, Lloyd, & Ton, 2015; Martin, Oteyza, Mitchell, Potticary, & Lloyd, 2015; Şahin Arslan et al., 2023). Otherwise, nests were generally checked every other day in Borneo, but from 1-4 days in Venezuela, to determine status and predation (Martin & Geupel, 1993). Clutch size was only used from nests located during building or egg-laying. We did not include nests observed later to ensure no partial loss was included (Martin et al., 2006). The incubation period was defined as the number of days between the last egg laid and last egg hatched (Martin, Auer, Bassar, Niklison, & Lloyd, 2007; Nice, 1954). The nestling period was defined as the days between the last egg hatched and the last nestling fledged and only used for nests where the last egg laid and hatch days were observed within 24 h of precision (Martin, Lloyd, et al., 2011).

Daily nest predation rates and daily survival rates were estimated using maximum likelihood estimation via the Mayfield method (Hensler & Nichols, 1981; Mayfield, 1961, 1975). This method is highly correlated with the logistic exposure method (Şahin Arslan & Martin, 2023; Shaffer, 2004) but allows more ready comparisons with the wider availability of Mayfield estimates in the literature. We considered a nest successful if parents were observed feeding young outside the nest or the young left within two days of normal fledging age. If nest contents disappeared earlier, we considered it to be due to predation.

We used an electronic scale with 0.001 g accuracy (ACCULAB, Elk Grove, Illinois, USA) to weigh fresh eggs on the day the last egg was laid or within the first 2 d of incubation. Nestlings were weighed for the first 3 days and then every other day throughout the rest of the nestling period, while also measuring wing chord and tarsus length using calipers (Mitutoyo) with an accuracy of 0.01 mm. As a part of a banding program, some adults were captured using mist-nets, and their mass, wing chord and tarsus lengths were measured.

Parental behavior at nests was recorded using video cameras for Whitehead’s Trogon during both incubation and nestling stages starting near sunrise. We put 30x zoom video-cameras 4–10 m from the nests and camouflaged the cameras to prevent possible disturbance. We generally sought 6 h video recordings of parental behavior at a nest, but they varied from 4–9 h each day of video recording (mean duration during incubation = 5.96 + 0.24 h, N = 27; during nestling period = 6.33 + 0.13 h, N = 97). Parental activity of the two trogon species in Venezuela were not video-recorded. Video recordings were used to quantify incubation nest attentiveness, as well as brooding attentiveness and feeding rates during the nestling period (Martin, Oteyza, Boyce, et al., 2015; Martin, Oteyza, Mitchell, et al., 2015; Şahin Arslan & Martin, 2019). Incubation nest attentiveness was measured as the percent of total video time that a parent sat on the eggs for each day of video recording (Martin, Oteyza, Boyce, et al., 2015). Brooding attentiveness for nestlings was calculated as the percent of video time that a parent sat on the nestlings for each day of video-recording, and feeding rates as the number of feeding trips of both parents to the nest-h for that recording.

Statistics
We conducted all analyses in R.4.2.2 (R Core Team 2022) and we present mean values with standard errors, ranges, and sample sizes. We estimated growth rate constants (K) for mass, tarsus length, and wing chord using the logistic growth model (Remeš & Martin, 2002). The model is based on the equation: W(t) = A/1 + e (−K∗(t−ti)), where W(t) is body mass, tarsus length, or wing chord length, A is the asymptotic size, t is age and ti is the age at the inflection point where growth rate changes from accelerating to decelerating, and K is the maximum rate of growth which is obtained at the inflection point (Martin, 2015). We tested for differences in the growth curves between Whitehead’s and Collared Trogons using the nls function in R, and using nest identity as a random effect, while specifying the above equation and running a model for each species and then testing for model differences between species using anova.

We used generalized linear mixed-effects models through the glmer function in the lme4 package (Bates, Mächler, Bolker, & Walker, 2015) to investigate the fixed effect of nestling age and brood size on feeding rate, with nest identity as a random effect. Brooding behavior changed in a backwards logistic curve (Şahin Arslan et al., 2023) and is described by the same three parameters as for growth rate above, where in this case A = asymptote at hatching day, K = instantaneous rate of change at the inflection time point, t = the inflection time point where the curve changes from accelerating to decelerating. We used the SSlogis function in the nlme package (Pinheiro & Bates, 2023) to describe the relationship and test for differences between brood sizes in slope (K), intercept (A), and inflection time point (t) of brooding behavior by Whitehead’s Trogon while using nest identity as a random effect. P ≤ 0.05 was considered as statistically significant throughout.

诸多热带物种种群呈现下降趋势。泛热带分布的咬鹃目（Trogoniformes）中有76%的物种种群被列为衰退类，反映出全球性的种群危机。本研究针对种群衰退且近危的旧世界白头咬鹃（Harpactes whiteheadi）、种群衰退的新世界白领咬鹃（Trogon collaris）以及种群稳定的面具咬鹃（T. personatus），报道其繁殖生态学与生活史特征；同时梳理了咬鹃类繁殖生态学与生活史特征的相关文献，以评估可能存在的共性特征，进而为种群衰退现象提供解释。
研究发现，种群衰退的白头咬鹃与面具咬鹃的巢成功率分别为32%与25%，表现尚可；而种群稳定的面具咬鹃繁殖成功率仅为9%，表现较差，这一结果与种群趋势完全相悖。不过，有限的文献数据表明，繁殖成功率偏低可能是咬鹃类的普遍特征，这或许会加剧种群衰退。亲鸟育雏投喂频率较低，且孵化与暖雏的持续时长较长，这会降低巢内活动频率——这是长期演化中应对高巢捕食风险的适应性策略。本研究还发现，幼鸟离巢时翅膀发育尚未完全，这同样是应对高巢捕食风险的适应性特征。咬鹃类物种的育雏期演化特征表明，翅膀发育不全可能是其普遍现象。已有研究证实，翅膀发育状况会影响幼鸟离巢后的存活率。因此，翅膀发育不全可能是导致种群衰退的重要因素，值得进一步关注。在快速变化的全球环境中，演化形成的生活史特征可能会让物种种群面临脆弱性，这一点亟需得到重视。

## 研究方法
白头咬鹃的研究地点位于马来西亚婆罗洲沙巴州的基纳巴卢公园（6°05′N，116°33′E），该保护区面积754 km²，以原生林为主。野外工作于2009–2020年的繁殖季开展，时间为每年2月初至6月中旬。研究共设置7个样地，海拔范围1450–1950 m，样地彼此相连，总面积约560 ha，单个样地面积为60–70 ha（Martin & Mouton, 2020）。
白领咬鹃与面具咬鹃的研究地点位于委内瑞拉西部拉腊州的亚坎布国家公园（9°38′N，69°40′W），该保护区面积269 km²。野外工作仅在海拔1400–2000 m的原生云雾林开展，该海拔范围与婆罗洲的研究站点相近。数据采集于2002–2008年的7个繁殖季，时间为每年2月末至7月初。研究同样设置7个样地，单个样地面积60–70 ha，与婆罗洲站点的样地规模一致（Martin and Mouton 2020）。由于这两种咬鹃并非本研究的核心对象，因此相较于白头咬鹃，我们未对其开展系统性的数据采集工作。
总体而言，婆罗洲与委内瑞拉的两项研究采用了统一的标准化数据采集方法，具体如下：通过观察繁殖配对行为以及系统性搜索来定位鸟巢（Martin & Geupel, 1993; Şahin Arslan & Martin, 2019; Şahin Arslan, Muñoz, & Martin, 2023），并使用测高仪测量巢高与巢址基质高度。对于白头咬鹃，我们使用GPS设备（Garmin，美国堪萨斯州奥拉西）记录巢址的海拔坐标。
产卵初始日期定义为巢内产下首枚卵的当日，产卵季则通过所有巢的产卵初始日期分布来界定。在产卵期与孵化的前两日，我们每日检查鸟巢，以确定最后一枚卵的产下日期，进而明确孵化起始时间。若在孵化期发现未知孵化阶段的鸟巢，则每日检查直至幼鸟孵出。在临近孵出与离巢阶段，我们每日或每两日检查一次鸟巢，以精确记录孵化与育雏期的时长（Martin, Oteyza, Boyce, Lloyd, & Ton, 2015; Martin, Oteyza, Mitchell, Potticary, & Lloyd, 2015; Şahin Arslan et al., 2023）。其余时段，婆罗洲的研究每两日检查一次鸟巢，委内瑞拉的研究则每1–4天检查一次，以确定巢的状态与是否被捕食（Martin & Geupel, 1993）。窝卵数仅取自筑巢期或产卵期发现的鸟巢，我们未纳入后期发现的鸟巢，以避免部分卵丢失的情况（Martin et al., 2006）。孵化期定义为最后一枚卵产出于最后一枚卵孵出之间的天数（Martin, Auer, Bassar, Niklison, & Lloyd, 2007; Nice, 1954）。育雏期定义为最后一枚卵孵出于最后一只幼鸟离巢之间的天数，仅用于那些最后一枚卵的产卵与孵出日期被精确记录（误差在24 h内）的鸟巢（Martin, Lloyd, et al., 2011）。
我们采用梅菲尔德法（Mayfield method）通过极大似然估计来计算每日巢捕食率与每日存活率（Hensler & Nichols, 1981; Mayfield, 1961, 1975）。该方法与逻辑暴露法（logistic exposure method）相关性极高（Şahin Arslan & Martin, 2023; Shaffer, 2004），但更便于与文献中大量已发表的梅菲尔德估计结果进行对比。若观察到亲鸟在巢外投喂幼鸟，或幼鸟在正常离巢年龄的两日之内离巢，则认定该鸟巢成功。若巢内内容物提前消失，则判定为被捕食。
我们使用精度为0.001 g的电子秤（ACCULAB，美国伊利诺伊州埃尔克格罗夫），在最后一枚卵产下当日或孵化前两日对新鲜卵进行称重。幼鸟在出壳后的前3日每日称重一次，之后在整个育雏期内每两日称重一次；同时使用精度为0.01 mm的游标卡尺（Mitutoyo）测量翅弦与跗跖长度。作为环志项目的一部分，部分成鸟被雾网捕获，并测量其体重、翅弦与跗跖长度。
针对白头咬鹃，我们在孵化期与育雏期于日出前后开始使用摄像机记录巢内亲鸟行为。摄像机设置在距离鸟巢4–10 m处，并进行伪装以避免干扰。我们通常希望获取6 h的亲鸟行为录像，但实际时长为4–9 h/天（孵化期平均时长为5.96 ± 0.24 h，N=27；育雏期平均时长为6.33 ± 0.13 h，N=97）。委内瑞拉的两种咬鹃未进行录像记录。录像数据被用于量化孵化期的巢出勤率，以及育雏期的暖巢出勤率与投喂频率（Martin, Oteyza, Boyce, et al., 2015; Martin, Oteyza, Mitchell, et al., 2015; Şahin Arslan & Martin, 2019）。孵化期巢出勤率定义为单日录像中亲鸟抱卵的时长占总录像时长的百分比（Martin, Oteyza, Boyce, et al., 2015）。幼鸟的暖巢出勤率定义为单日录像中亲鸟暖雏的时长占总录像时长的百分比，投喂频率则为单次录像中双亲往返巢区的投喂次数之和。

## 统计分析
所有分析均在R 4.2.2中完成（R Core Team 2022），结果以均值±标准误、范围与样本量的形式呈现。我们使用逻辑斯蒂生长模型估算体重、跗跖长度与翅弦的生长速率常数（K）（Remeš & Martin, 2002）。该模型基于公式：W(t) = A/[1 + e^(−K∗(t−ti))]，其中W(t)为体重、跗跖长度或翅弦长度，A为渐近值，t为日龄，ti为生长速率由加速转为减速的拐点日龄，K为拐点处的最大生长速率（Martin, 2015）。我们使用R中的nls函数检验白头咬鹃与白领咬鹃的生长曲线差异，以巢身份作为随机效应，分别为两个物种拟合模型，再通过anova检验模型间的差异。
我们使用lme4软件包中的glmer函数构建广义线性混合模型（generalized linear mixed-effects models），以检验育雏日龄与窝雏数对投喂频率的固定效应，以巢身份作为随机效应。暖巢行为呈反向逻辑斯蒂曲线（Şahin Arslan et al., 2023），其参数与生长模型一致：A为出壳日的渐近值，K为拐点处的瞬时变化速率，t为曲线由加速转为减速的拐点日龄。我们使用nlme软件包中的SSlogis函数描述暖巢行为的变化，并以巢身份作为随机效应，检验白头咬鹃暖巢行为的斜率（K）、截距（A）与拐点日龄（t）在不同窝雏数间的差异。本研究中P ≤ 0.05被认定为具有统计学显著性。