Premium
Improving performance of species distribution model in mountainous areas with complex topography
Author(s) -
Liao ChiCheng,
Chen YiHuey
Publication year - 2021
Publication title -
ecological research
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.628
H-Index - 68
eISSN - 1440-1703
pISSN - 0912-3814
DOI - 10.1111/1440-1703.12227
Subject(s) - species distribution , multivariate interpolation , environmental science , habitat , scale (ratio) , physical geography , spatial ecology , spatial distribution , contrast (vision) , climate change , climatology , geography , remote sensing , ecology , cartography , statistics , geology , computer science , mathematics , artificial intelligence , bilinear interpolation , biology
Abstract This study aims to improve performance of species distribution model in mountainous areas with complex topography. Random Forest algorithm was used to evaluate distribution ranges of Trochodendron aralioides Siebold & Zucc. in northern Taiwan to identify the determinants of the plant distributions and to assess the effects of data collections on the spatial bias of model predictions. Gridded climate dataset with spatial resolution of 50 × 50 m 2 for the study area was generated from daily data of 18 meteorological stations and environmental variables were interpolated by inverse distance weighted method and altitudinal adjusted by linear regression model using empirical lapse rates varied among slope inclinations. Model performances were much better when model was run with larger sizes of training datasets generated by randomized resample of presence/absence data records. Surprisingly, distribution ranges of the plant were drastically affected by marginal individuals that is likely to have reached its geographical border as well as climatic limits. Wind speed in winter and temperature differences between warmest and coldest months contributed noticeably to the model predictions. In contrast, index of cloud occurrence contributed little to the model predictions, despite T. aralioides was a characteristic species of cloud zone. The interpolated bioclimatic predictors had precisely reflected the habitat differentiation among topographical positions and accurately captured habitat requirements of plant species at fine scale. Our results presented practical and applicable example of novel interpolation method to generate gridded climate dataset that is useful to improve performance of model predictions in mountainous areas with complex topography.