Spatial Distribution Characteristics of Species Diversity Using Geographically Weighted Regression Model

Abstract

The objective of this study is to evaluate the spatial distribution patterns of species diversity at different spatial scales, focusing on the Baekdudaegan Protected Area, which is a biodiversity hotspot in the Republic of Korea. The tree species diversity index (Shannon-Weaver index; H′) was calculated using tree species data from a 1:5k forest-type map, and the spatial analysis was performed with a 1 × 1 km 2 grid. Ten factors were selected to estimate the impact of topographic (elevation, slope, northern slope, curvature, wetness, and relief) and geographic (distances from water, road, forest road, and urban area) factors on H′ using the ordinary least squares (OLS) and geographically weighted regression (GWR) models. H′ increased with the spatial scale. Also, the coefficient of determination (R²) of the GWR and OLS models increased proportionally and the R² of the GWR model was higher than that of the OLS model. Corrected Akaike information criterion (AICc) was lower in the GWR model than in the OLS model, which indicates that the GWR model fits the calculated H′ better than the OLS model. Thus, the GWR model is considered to be more practical than the OLS model for understanding the effects of topographic and geographic factors on H′ at different scales.

What Drives the Spatial Variation of Interregional Ancient Trees? A Geoinformatics–Based Approach in Henan, Central China

“…In recent years, many statistical techniques have been developed to probe into spatial variations in the relationships between spatial variables in the fields of geo-statistics and spatial modeling [32,33]. Geographically weighted regression (GWR), as one of the most powerful geoinformatics-based tools for exploring spatial heterogeneity, has been more widely used in geography, ecology, forestry, and other disciplines [34,35]. GWR is an extension of the traditional regression analysis method for local rather than global parameter estimation, which takes the spatial structure of the data into account in the model, so that the regression coefficients become a function of the geographic location of the observation points, then solves this spatial non-smoothness problem [36].…”

Modelling the effects of topographic hetero- geneity on distribution of Nitraria tanguto- rum Bobr. species in deserts using LiDAR-data

“…species in the desert on a small scale, such as within the scale of 100 m, where topographic factors significantly affect species distribution. Existing studies have shown that geographic and topographic predictors can improved the fitting performance of SDMs substantially 36 .…”

Modelling the effects of topographic heterogeneity on distribution of Nitraria tangutorum Bobr. species in deserts using LiDAR-data

“…species in the desert on a small scale, such as within the scale of 100 m, where topographic factors significantly affect species distribution. Existing studies have shown that geographic and topographic predictors can substantially improve the fitting performance of SDMs 37 .…”