Are trapping data suited for home-range estimation?

AbstractModern home-range estimation typically relies on data derived from expensive radio- or GPS-tracking. Although trapping represents a low-cost alternative to telemetry, there lacks an evaluation of the performance of home-range estimators on trap-derived data. Using simulated data, we evaluate three variables reflecting the key trade-offs ecologists face when designing a trapping study: 1) the number of observations obtained per individual, 2) the trap density, and 3) the proportion of the home range falling inside the trapping area. We compare the performance of five home-range estimators (MCP, LoCoH, KDE, AKDE, bicubic interpolation). We further explore the potential benefits of combining these estimators with asymptotic models, which leverage the saturating behavior of changes in the estimated home-range area as the number of observations increases to improve accuracy, as well as different data ordering procedures. We then quantified the bias in home-range size under the different scenarios investigated. The number of observations and the proportion of the home range within the trapping grid were the most important predictors of the accuracy and the precision of home-range estimates. The use of asymptotic models helped obtain accurate estimates at smaller sample sizes, while distance-ordering improved the precision and asymptotic consistency of estimates. While AKDE was the best-performing estimator under most conditions evaluated, bicubic interpolation was a viable alternative under common real-world conditions of low trap density and area covered. A case study using empirical data from white-tailed deer in Florida and another from jaguars in Belize demonstrated support for the findings of our simulation results. Although researchers with trap data often overlook home-range estimation, our results indicate that these data have the capacity to yield accurate estimates of home-range size. Trapping data can therefore lower the economic costs of home-range analysis, potentially enlarging the span of species, researchers and questions studied in ecology and conservation.摘要：现代家域估计通常依赖于成本高昂的无线电或GPS追踪数据。尽管诱捕采样是遥测技术的低成本替代方案，但目前仍缺乏针对诱捕所得数据的家域估计器性能评估。本研究借助模拟数据，评估了生态学家在设计诱捕研究时需考量的三项关键权衡变量：1）单一个体的观测次数，2）诱捕装置密度，3）家域落入诱捕区域的比例。本研究对比了五种家域估计器的性能，分别为最小凸多边形（MCP）、局部凸包（LoCoH）、核密度估计（KDE）、自适应核密度估计（AKDE）以及双三次插值（bicubic interpolation）。本研究进一步探索了将上述估计器与渐近模型结合的潜在优势：渐近模型可利用观测次数增加时家域估计面积的饱和变化特性以提升估计精度，同时还测试了多种不同的数据排序流程。随后，本研究量化了各研究场景下家域大小的估计偏差。研究结果显示，观测次数与家域落入诱捕网格的比例，是影响家域估计准确性与精确性的最核心因素。渐近模型的使用可在样本量较小时获得精准的估计结果，而基于距离的数据排序则能提升估计的精确性与渐近一致性。尽管在多数测试场景中自适应核密度估计（AKDE）的表现最优，但在诱捕密度低、覆盖范围有限的常见现实场景中，双三次插值（bicubic interpolation）是一种可行的替代方案。本研究借助佛罗里达白尾鹿的实测数据与伯利兹美洲豹的实测数据开展了两项案例研究，结果验证了模拟研究的结论。尽管拥有诱捕采样数据的研究者往往忽略家域估计环节，但本研究结果表明，这类数据可用于获得精准的家域大小估计值。因此，诱捕采样数据可降低家域分析的经济成本，有望拓展生态学与保护生物学中可研究的物种类别、研究主体与科学问题。