期刊
ECOLOGICAL MODELLING
卷 418, 期 -, 页码 -出版社
ELSEVIER
DOI: 10.1016/j.ecolmodel.2020.108937
关键词
Species distribution models; Scales; MaxEnt; Regularization multiplier
类别
资金
- German Bauer-Stiftung zur Forderung von Wissenschaft und Forschung
- German Rudolf and Helene Glaser-Stiftung [T0237/28255/2016/KG]
- 2015-2016 BiodivERsA COFUND
- Spanish Ministry of Economy, Industry and Competitiveness(project ODYSSEUS) [BiodivERsA3- 2015-26, PCIN-2016-168]
Species Distribution Models (SDM) are frequently used in ecological research, but the effects of model extent and granularity on model outcomes are rarely addressed. In freshwater SDMs, two different approaches are commonly used to define the granularity of the models, i.e. to subdivide entire river systems into appropriate spatial modelling units: river reaches on the actual river network or subcatchments. We built maximum entropy (MaxEnt) SDMs for barbel (Barbus barbus) and grayling (Thymallus thymallus) in the River Ruhr catchment in Germany with identical environmental predictor data and analysed the difference between these two approaches. All models performed well (AUC > 0.9) and geographic predictors (e.g. distance to the river source) dominated over hydromorphological predictors in explaining species distributions. There was high agreement between the two model setups in river parts of low and high habitat suitability but considerably lower agreement in river parts of medium habitat suitability. In these medium suitable river parts, model results were spatially very heterogeneous and alternated at fine spatial scales especially in models based on river reaches. Increasing model regularization, a setting to control overfitting, had a smoothing effect on the environmental variables in the river reach models similar to the coarser subcatchment granularity. A restriction of the spatial extent led to a shift in predictor contributions to the model and an increase of the importance of hydromorphologic predictors by ca. 45 %. Restricting the model extend to the natural core distributional area of a given fish species might therefore be considered beneficial for the application of SDMs in a management context. By decreasing the weight of fixed geographic predictors in the model, predictors relating to hydromorphological river structures (i.e. which are accessible for restoration projects) gain importance. We conclude that SDM setups based on river reaches and subcatchments can both give valuable complementary information about the distribution of species. The high-resolution model based on river reaches might better discover individual local habitat features, whereas the subcatchment model might better account for the minimum spatial requirements of a fish population.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据