Data from: Two-species occupancy modeling accounting for species misidentification and nondetection

1.In occupancy studies, species misidentification can lead to false positive detections, which can cause severe estimator biases. Currently, all models that account for false positive errors only consider omnibus sources of false detections and are limited to single species occupancy. 2.However, false detections for a given species often occur because of the misidentification with another, closely-related species. To exploit this explicit source of false positive detection error, we develop a two-species occupancy model that accounts for misidentifications between two species of interest. As with other false positive models, identifiability is greatly improved by the availability of unambiguous detections at a subset of site-occasions. Here, we consider the case where some of the field observations can be confirmed using laboratory or other independent identification methods (“confirmatory data”). 3.We performed three simulation studies to (1) assess the model's performance under various realistic scenarios, (2) investigate the influence of the proportion of confirmatory data on estimator accuracy, and (3) compare the performance of this two-species model with that of the single-species false positive model. The model shows good performance under all scenarios, even when only small proportions of detections are confirmed (e.g., 5%). It also clearly outperforms the single-species model. 4.We illustrate application of this model using a four-year data set on two sympatric species of lungless salamanders: the US federally endangered Shenandoah salamander (Plethodon shenandoah), and its presumed competitor, the red-backed salamander (Plethodon cinereus). Occupancy of red-backed salamanders appeared very stable across the four years of study, whereas the Shenandoah salamander displayed substantial turn-over in occupancy of forest habitats among years. 5.Given the extent of species misidentification issues in occupancy studies, this modelling approach should help improve the reliability of estimates of species distribution, which is the goal of many studies and monitoring programs. Further developments, to account for different forms of state uncertainty, can be readily undertaken under our general approach.1. 在物种占据研究（occupancy studies）中，物种误识别会引发假阳性检测，进而导致严重的估计器偏差。当前所有针对假阳性误差的模型，仅考虑假检测的通用来源，且仅适用于单物种占据分析。 2. 然而，特定物种的假检测往往源于与另一近缘物种的误识别。为利用这一明确的假阳性检测误差来源，我们构建了可考量目标两物种间误识别的双物种占据模型。与其他假阳性模型类似，当部分样地-调查回合获得明确检测结果时，模型的可识别性会大幅提升。本研究考虑了部分野外观测可通过实验室或其他独立鉴定方法进行验证的场景（即“验证性数据（confirmatory data）”）。 3. 我们开展了三项模拟研究：(1) 评估模型在多种现实场景下的性能表现；(2) 探究验证性数据占比对估计器精度的影响；(3) 将本双物种模型与单物种假阳性模型的性能进行对比。结果显示，即便仅验证了少量检测结果（如5%），该模型在所有场景下均表现优异，且显著优于单物种假阳性模型。 4. 我们以为期四年的同域分布无肺螈（lungless salamanders）数据集为例，展示该模型的应用场景：研究对象为美国联邦濒危物种谢南多厄无肺螈（Plethodon shenandoah），以及其推测的竞争者红背无肺螈（Plethodon cinereus）。研究期间，红背无肺螈的占据情况在四年间保持稳定，而谢南多厄无肺螈的森林生境占据状况在各年份间存在显著更替。 5. 鉴于物种占据研究中普遍存在物种误识别问题，本建模方法有助于提升物种分布估计的可靠性，而这正是众多研究与监测项目的核心目标。基于本研究的通用框架，可进一步拓展以考量多种形式的状态不确定性。