Data from: Biological traits, rather than environment, shape detection curves of large vertebrates in neotropical rainforests

Line transect surveys are widely used in neotropical rainforests to estimate the population abundance of medium- and large-sized vertebrates. The use of indices such as Encounter Rate has been criticized because the probability of animal detection may fluctuate due to the heterogeneity of environmental conditions among sites. In addition, the morphological and behavioral characteristics (biological traits) of species affect their detectability. In this study, we compared the extent to which environmental conditions and species’ biological traits bias abundance estimates in terra firme rainforests in French Guiana. The selected environmental conditions included both physical conditions and forest structure covariates, while the selected biological traits included the morphological and behavioral characteristics of species. We used the distance sampling method to model the detection probability as an explicit function of environmental conditions and biological traits and implemented a model selection process to determine the relative importance of each group of covariates. Biological traits contributed to the variability of animal detectability more than environmental conditions, which had only a marginal effect. Detectability was best for large animals with uniform or disruptive markings that live in groups in the canopy top. Detectability was worst for small, solitary, terrestrial animals with mottled markings. In the terra firme rainforests which represent ~80% of the Amazonia and Guianas regions, our findings support the use of relative indices such as the encounter rate to compare population abundance between sites in species-specific studies. Even though terra firme rainforests may appear similar between regions of Amazonia and the Guianas, comparability must be ensured, especially in forests disturbed by human activity. The detection probability can be used as an indicator of species’ vulnerability to hunting and, thus, to the risk of local extinction. Only a few biological trait covariates are required to correctly estimate the detectability of the majority of medium- and large-sized vertebrates. Thus, a biological trait model could be useful in predicting the detection probabilities of rare, uncommon or localized species for which few data are available to fit the detection function.样线调查（line transect surveys）被广泛应用于新热带雨林，以估算中型与大型脊椎动物的种群丰度。此前，诸如遇见率（Encounter Rate）这类指数的应用饱受诟病：由于不同样地间环境条件存在异质性，动物被检测到的概率可能出现波动。此外，物种的形态与行为特征（生物学性状，biological traits）亦会对其可检测性造成影响。本研究针对法属圭亚那的陆地雨林（terra firme rainforests），对比了环境条件与物种生物学性状对种群丰度估算的偏倚程度。本研究选取的环境条件涵盖物理环境与森林结构协变量，所选生物学性状则包含物种的形态与行为特征。我们采用距离采样法（distance sampling method），将检测概率建模为环境条件与生物学性状的显式函数，并通过模型选择流程确定每一组协变量的相对重要性。研究结果显示，相较于仅产生边际效应的环境条件，生物学性状对动物可检测性的变异性贡献更为显著。可检测性最优的类群为栖息于林冠顶层、以群体活动且带有纯色或干扰斑纹的大型动物；而可检测性最差的则是带有斑驳斑纹的小型独居陆生动物。陆地雨林约占亚马孙与圭亚那地区总面积的80%，本研究结论支持在物种特异性研究中采用遇见率这类相对指数，以开展不同样地间的种群丰度对比。尽管亚马孙与圭亚那地区的陆地雨林外观相似，但仍需确保样地间的可比性，在受人类活动干扰的森林中尤需注意。检测概率可作为物种狩猎易感性的指标，进而用于评估其局域灭绝风险。仅需少量生物学性状协变量，即可准确估算绝大多数中型与大型脊椎动物的可检测性。因此，生物学性状模型可用于预测那些数据匮乏、难以拟合检测函数的稀有、少见或局域分布物种的检测概率。