Abstract Artificial Light At Night (ALAN) is one of the most important anthropogenic environmental components that affects biodiversity worldwide. Despite extensive knowledge on ALAN, being a measure of human activity that directly impacts numerous aspects of animal behaviour, such as orientation and distribution, little is known about its effects on density distribution on a large spatial scale. That is why we decided to explore by means of the Species Distribution Modelling approach (SDM) how ALAN as one of 33 predictors determines farmland and forest bird species densities. In order to safeguard study results from any inconsistency caused by the chosen method, we used two approaches, i.e. the Generalised Additive Model (GAM) and the Random Forest (RF). Within each approach, we developed two models for two bird species, the Black woodpecker and the European stonechat: the first with ALAN, and the second without ALAN as an additional predictor. Having used out-of-bag procedures in the RF approach, information-theoretic criteria for the GAM, and evaluation models based on an independent dataset, we demonstrated that models with ALAN had higher predictive density power than models without it. The Black woodpecker definitely and linearly avoids anthropogenic activity, defined by the level of artificial light, while the European stonechat tolerates human activity to some degree, especially in farmland habitats. What is more, a heuristic analysis of predictive maps based on models without ALAN shows that both species reach high densities in regions where they are deemed rare. Hence, the study proves that urbanisation processes, which can be reflected by ALAN, are among key predictors necessary for developing Species Density Distribution Models for both farmland and forest bird species.

中文翻译：

---

表达为“夜间人工光”的人为活动改善了鸟类种群的预测密度分布

摘要 夜间人造光（ALAN）是影响全球生物多样性的最重要的人为环境成分之一。尽管 ALAN 是一种直接影响动物行为的许多方面（例如方向和分布）的人类活动量度，但尽管对 ALAN 有广泛的了解，但对其在大空间尺度上对密度分布的影响知之甚少。这就是为什么我们决定通过物种分布建模方法 (SDM) 探索 ALAN 作为 33 个预测因子之一如何确定农田和森林鸟类的密度。为了保护研究结果不受所选方法引起的任何不一致的影响，我们使用了两种方法，即广义加性模型 (GAM) 和随机森林 (RF)。在每种方法中，我们为两种鸟类开发了两种模型，黑啄木鸟和欧洲石头聊天：第一个有 ALAN，第二个没有 ALAN 作为额外的预测因子。在 RF 方法中使用了袋外程序、GAM 的信息理论标准以及基于独立数据集的评估模型，我们证明了具有 ALAN 的模型比没有它的模型具有更高的预测密度能力。黑啄木鸟明确地、线性地避免了人造光水平所定义的人为活动，而欧洲石嘴鸟则在一定程度上容忍人类活动，尤其是在农田栖息地。更重要的是，对基于没有 ALAN 的模型的预测图的启发式分析表明，这两种物种在它们被认为稀有的地区达到了高密度。因此，研究证明，ALAN 所反映的城市化过程，