首页|Spatial species distribution models: Using Bayes inference with INLA and SPDE to improve the tree species choice for important European tree species
Spatial species distribution models: Using Bayes inference with INLA and SPDE to improve the tree species choice for important European tree species
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NSTL
Elsevier
Species distribution models (SDMs) are a standard tool for predicting species occurrence under climate change. Despite its limitations, SDMs have been widely used to assist forest management decisions in their choice of future tree species. The accuracy of SDMs is often affected by heterogeneous occurrence data caused by different forest inventory schemes, historical processes, climate and insect calamities and more. These processes bias the relationships modelled between species occurrence and climate. However, numerous studies have shown that explicit modelling of spatial effects may improve model accuracy. In this study, we applied the Integrated Nested Laplace Approximation (INLA) and Stochastic Partial Differential Equations (SPDE) algorithms as a means to accomplish Bayes inference for Generalized Additive Models (GAMs) with explicit modelling of spatial effects. Our results show that including spatial effects in GAM-based SDMs improved model performance and accuracy leading to more reliable predictions for the species Favorability. Conditional predictions that remove spatial effects from the models allow us to distinguish core and marginal ecological ranges with less bias through forest management, which may support the tree-species choice for climate-resilient forests.
Species distribution modelsGeneralized additive modelsIntegrated Nested Laplace Approximation (INLA)Spatial effectsClimate changeAssisted migrationCLIMATE-CHANGESCOTS PINEFOREST MANAGEMENTRANGEAUTOCORRELATIONACCURACYIMPACTGROWTHSHIFT
NSTL
Elsevier
