There is an increasing need for ecosystem-level distribution models (EDMs) and a better understanding of which factors affect their quality. We investigated how the performance and transferability of EDMs are influenced by 1) the choice of predictors and 2) model complexity. We modelled the distribution of 15 pre-classified ecosystem types in Norway using 252 predictors gridded to 100 × 100 m resolution. The ecosystem types are major types in the ‘Nature in Norway' system mainly defined by rule-based criteria such as whether soil or specific functional groups (e.g. trees) are present. The predictors were categorised into four groups, of which three represented proxies for natural, anthropogenic, or terrain processes (‘ecological predictors') and one represented spectral and structural characteristics of the surface observable from above (‘surface predictors'). Models were generated for five levels of model complexity. Model performance and transferability were evaluated with data collected independently of the training data. We found that 1) models trained with surface predictors only performed considerably better and were more transferable than models trained with ecological predictors, and 2) model performance increased with model complexity, levelling off from approximately 10 parameters and reaching a peak at approximately 20 parameters, while model transferability decreased with model complexity. Our findings suggest that surface predictors enhance EDM performance and transferability, most likely because they represent discernible surface characteristics of the ecosystem types. A poor match between the rule-based criteria that define the ecosystem types and the ecological predictors, which represent ecological processes, is a plausible explanation for why surface predictors better predict the distribution of ecosystem types. Our results indicate that, in most cases, the same models are not well suited for contrasting purposes, such as predicting where ecosystems are and explaining why they are there.

中文翻译：

---

生态系统分布模型的预测变量和复杂性的选择：对性能和可转移性的影响


人们越来越需要生态系统级分布模型 (EDM) 以及更好地了解哪些因素影响其质量。我们研究了 1) 预测变量的选择和 2) 模型复杂性如何影响 EDM 的性能和可转移性。我们使用 252 个网格化至 100 × 100 m 分辨率的预测变量对挪威 15 个预分类生态系统类型的分布进行了建模。生态系统类型是“挪威自然”系统中的主要类型，主要由基于规则的标准定义，例如是否存在土壤或特定功能组（例如树木）。预测因子分为四组，其中三组代表自然、人为或地形过程的代理（“生态预测因子”），一组代表从上方可观察到的表面的光谱和结构特征（“表面预测因子”）。模型的生成具有五个级别的模型复杂性。使用独立于训练数据收集的数据来评估模型性能和可转移性。我们发现，1) 使用表面预测器训练的模型比使用生态预测器训练的模型表现得更好，并且更容易迁移，2) 模型性能随着模型复杂性的增加而增加，从大约 10 个参数开始趋于平稳，并在大约 20 个参数时达到峰值，而模型的可移植性随着模型复杂度的增加而降低。我们的研究结果表明，表面预测因子增强了 EDM 性能和可转移性，很可能是因为它们代表了生态系统类型的可辨别的表面特征。 定义生态系统类型的基于规则的标准与代表生态过程的生态预测因子之间的不匹配，可以合理地解释为什么表面预测因子可以更好地预测生态系统类型的分布。我们的结果表明，在大多数情况下，相同的模型不太适合对比目的，例如预测生态系统在哪里并解释它们为什么在那里。