LB 28-240 - How Will Extreme Climate Affect South Korea's Vegetation Distribution?

Climate serves as one of the key factors in vegetation distribution, and its volatility is increasing due to rapid climate change. The impact of extreme climate phenomena is getting bigger as well, and it is expected to appear more frequently in the future. However, many studies on vegetation distribution modelling usually use only long-term average climate values as indices, so it is necessary to quantify the effect of extreme climate on vegetation. In this study, we presented a method of comparing the extent to which average and extreme climates affect the vegetation distribution in South Korea by setting up two Random Forest models: 1) one with only the average climate and 2) the other with average and extreme climate together. We used national vegetation data and selected the top 20 vegetation types by area as dependent variable. For the independent variables, we used WorldClim’s average climate indices(1970-2000; Fick & Hijmans, 2017) and extreme climate indices developed from Korea Meteorological Administration (2000-2017). Additionally, we used non-climate elements including soil type, depth of soil, soil component, slope, aspect, and elevation to increase the model accuracy. After selecting the modeling variables by removing multicolinearity, we compared the testing result of two Random Forest models. Moreover, we compared the class error and balanced accuracy for all vegetation types by calculating the difference between the models and calculated the variable importance in the models. Model with only the average climate values showed 0.86 in overall accuracy and 0.85 in Kappa coefficient, whereas adding extreme values showed higher classification accuracy, 0.90 and Kappa coefficient, 0.89. Both class error difference and balanced accuracy difference revealed that adding extreme climate indices in predicting vegetation distribution reduces error of the model, which showed extreme climate indices had a great impact on vegetation distribution modelling. However, data on extreme climates are limited, so the development of this information will lead to a better understanding of vegetation distribution.