Plant biodiversity and conservation assessment in South Korea using rarely recorded data and a multi-response species distribution model

Background/Question/Methods

Protected areas are often considered the most important areas to be conserved in most national conservation policies, but many of them are not representative of national biodiversity. We assessed the biodiversity of vascular plants within and outside protected areas in South Korea by using a multi-response species distribution model (SDM). Multi-response SDMs have benefits for applications with large numbers of species that include rarely recorded ones in the data set. Surveyed point data for vascular plants open to public in South Korea total 2,012 points and consist of 289 species. In our study, we used a Multivariate Adaptive Regression Splines (MARS) multi-response SDM model that informs model development for a target species with data from multiple species. We started to run the model with species which have at least two observations, and increased the minimum observations required in the next runs. The environmental predictor data were selected for their relevance to the plant species including Bio-climate data (annual mean temperature, temperature seasonality, mean temperature of warmest and coldest quarter, annual precipitation, precipitation of wettest and driest quarter) and topographic data (elevation, slope, curvature, northness).

Results/Conclusions

The mean AIC values of species modeled in each multi-response SDM run were similar for runs using minimum cutoffs of two-to-five observations. The relative qualities of models based on AIC values improved when using a cutoff for modeled species of than five occurrence data. However, we chose to analyze the model run that used a minimum of two points per species, because it provided outputs for the highest number of species, including some endangered and endemic species that would otherwise have been excluded. Important predictor variables were mean temperature of warmest and coldest quarter, precipitation of wettest quarter, and elevation. We found that current protected areas include habitats for approximately 70% of the vascular plants we modeled. The protected areas include 80% of endangered and 70% of endemic species’ habitats, and also harbor ~70% of montane and mesic-adapted species. However, the current protected areas largely consist of high mountainous areas, which is also true of our observation data. Among our occurrence data, only 13% of the records are of mesic-adapted species. We discuss how to evaluate the model inputs and results. We also present a spatial analysis of diversity patterns for species not currently protected.