Zhang Zhixin, Kass Jamie M, Bede-Fazekas Ákos, Mammola Stefano, Qu Junmei, Molinos Jorge García, Gu Jiqi, Huang Hongwei, Qu Meng, Yue Ying, Qin Geng, Lin Qiang
CAS Key Laboratory of Tropical Marine Bio-Resources and Ecology, Guangdong Provincial Key Laboratory of Applied Marine Biology, South China Sea Institute of Oceanology, Chinese Academy of Sciences, Guangzhou, China.
University of Chinese Academy of Sciences, Beijing, China.
Conserv Biol. 2025 Aug;39(4):e70015. doi: 10.1111/cobi.70015. Epub 2025 Mar 24.
Species distribution models (SDMs) are important tools for assessing biodiversity change. These models require high-quality occurrence data, which are not always available. Therefore, it is increasingly important to determine how data choice affects predictions of species' ranges. Opportunistic occurrence records and expert maps are both widely used sources of species data for SDMs. However, it is unclear how SDMs based on these data differ in performance, particularly for the marine realm. We built SDMs for 233 marine fish species from 2 families with these 2 occurrence data types and compared their performances and potential distribution predictions. Opportunistic occurrences were sourced from field surveys in the South China Sea and online repositories and expert maps from the International Union for Conservation of Nature Red List database. We used generalized linear models to explore drivers of differences in prediction between the 2 model types. When projecting to distinct regions with no occurrence data, models calibrated using opportunistic occurrences performed better than those using expert maps, indicating better transferability to new environments. Differences in marine predictor values between the 2 data types accounted for the dissimilarity in model predictions, likely because expert maps included large areas with unsuitable environmental conditions. Dissimilarity levels among fish families differed, suggesting a taxonomic bias in biodiversity data between data sources. Our findings highlight the sensitivity of species distribution predictions to the choice of distributional data. Although expert maps have an important role in biodiversity modeling, we suggest researchers assess the accuracy of these maps and reduce commission errors based on knowledge of target species.
物种分布模型（SDMs）是评估生物多样性变化的重要工具。这些模型需要高质量的出现数据，但这些数据并非总是可得。因此，确定数据选择如何影响物种分布预测变得越来越重要。机会性出现记录和专家地图都是物种分布模型广泛使用的物种数据来源。然而，基于这些数据的物种分布模型在性能上如何不同尚不清楚，特别是对于海洋领域。我们用这两种出现数据类型为2个科的233种海洋鱼类构建了物种分布模型，并比较了它们的性能和潜在分布预测。机会性出现数据来自南海的实地调查和在线数据库，专家地图来自国际自然保护联盟红色名录数据库。我们使用广义线性模型来探究这两种模型类型在预测上存在差异的驱动因素。当将模型投影到没有出现数据的不同区域时，使用机会性出现数据校准的模型比使用专家地图的模型表现更好，这表明其对新环境具有更好的可转移性。这两种数据类型在海洋预测变量值上的差异导致了模型预测的不同，可能是因为专家地图包含了大面积环境条件不适宜的区域。不同鱼类科之间的差异程度不同，这表明数据源之间的生物多样性数据存在分类学偏差。我们的研究结果突出了物种分布预测对分布数据选择的敏感性。尽管专家地图在生物多样性建模中具有重要作用，但我们建议研究人员评估这些地图的准确性，并根据目标物种的知识减少误判误差。
