기계학습 기반 최근 기후변화에 따른 지표수종의 서식처 변화 규명 (Machine Learning Application for Identifying Habitat Suitability Changes of Indicator Tree Species against Recent Climate Change)

서지정보

· 발행기관 : 한국기후변화학회
· 수록지 정보 : 한국기후변화학회지 / 11권 / 6호 / 793 ~ 805페이지
· 저자명 : 임철희, 김현준

초록

In this study, a recent climate change phenomenon and impact were identified by predicting a time-series suitable region of indicator species of warm-temperate and subalpine forests, in which climate change effects are prominent, using machine learning models and recent climate information. In the recent bioclimatic indices, temperature and seasonality of precipitation have increased. The suitable habitat regions of the past warm-temperate and subalpine forests predicted through the machine learning-based random forest model and the bioclimatic indices were similar to the actual natural forest distribution. In the last 18 years, in the warm-temperate forest, though the annual deviation was high, a clear increasing trend has been observed.
On average, the potentially suitable habitat areas increased more than three times. In subalpine forests, the suitable habitat area decreased significantly and is very limited in the southern sub-alpine area. After 2013, little suitable area was present, with an average of only 23% compared to the past period. These results are related to the group death of subalpine forest in the 2010s and indicate that the accumulated climatic non-suitability has caused death of the subalpine forest. Climate change impacts on indicator species are both a risk and an opportunity, depending on the species, and we expect wise adaptation and measures to create more opportunities.

영어초록

In this study, a recent climate change phenomenon and impact were identified by predicting a time-series suitable region of indicator species of warm-temperate and subalpine forests, in which climate change effects are prominent, using machine learning models and recent climate information. In the recent bioclimatic indices, temperature and seasonality of precipitation have increased. The suitable habitat regions of the past warm-temperate and subalpine forests predicted through the machine learning-based random forest model and the bioclimatic indices were similar to the actual natural forest distribution. In the last 18 years, in the warm-temperate forest, though the annual deviation was high, a clear increasing trend has been observed.
On average, the potentially suitable habitat areas increased more than three times. In subalpine forests, the suitable habitat area decreased significantly and is very limited in the southern sub-alpine area. After 2013, little suitable area was present, with an average of only 23% compared to the past period. These results are related to the group death of subalpine forest in the 2010s and indicate that the accumulated climatic non-suitability has caused death of the subalpine forest. Climate change impacts on indicator species are both a risk and an opportunity, depending on the species, and we expect wise adaptation and measures to create more opportunities.

참고자료

· 없음