인공신경망을 이용한 겨울철 접합 대순환 모형 결과 보정 연구 = Correction of Winter Coupled General Circulation model result Using Artificial Neural Network

다국어 초록 (Multilingual Abstract)

Recently, climate disaster has happened because of natural variation such as El Niño and La Niña, and global warming. Therefore need for long term prediction has increased rapidly. Coupled General Circulation Model(CGCM) is useful tool for future prediction and climate changes. CGCMs, however, have errors owing to uncertainties in initial condition and boundary condition, model dynamics and insufficiency of physical processes. These errors make long term prediction difficult. Therefore we must correct model output to improve predictability using statistical methods. So far, linear methods like linear regression, principle component analysis have been used in Meteorology and Oceanography studies. Since the late 1980s, Artificial Neural Network (ANN) methods have become popular.
In this study, hindcast experiment was performed from September to February of 1972~2001 using CGCM. The correlations between CGCM raw outputs and observations are low. Also, global maps of temperature and precipitation show model errors.
The model errors are removed using ANN. Cross-validation is performed from December to the next year February in this study using ANN. Predictands are surface temperature and precipitation, predictors are surface temperature, precipitation, 850hPa temperature, 500-1000hPa thickness, 500, 700hPa vertical velocity. First of all, one point correlation is calculated for these variables. Next, the highest correlated grid points are selected and these enter into input layer of ANN.
The correction results by ANN show that predictands are improved greatly. For instance, the correlation maps and scatter plots, and verification scores reveal that the correction between corrected model results and observations are highly correlated.
This study suggests that ANN can be a superior tool for the correction of errors of CGCM output.

목차 (Table of Contents)

• 목차
• List of Tables = ⅱ
• List of Figures = ⅲ
• 제 1 장 서론 = 1
• 제 2 장 접합 대순환 모형을 이용한 장기 예측 실험 = 6
• 2.1 접합 대순환 모형의 구조 = 6
• 2.2 접합 대순환 모형을 이용한 겨울철 장기 예측 실험 방법 = 9
• 제 3 장 인공 신경망 모델을 이용한 보정 실험 = 11
• 3.1 인공 신경망을 이용한 보정 방법 = 11
• 제 4 장 모형의 실험결과와 보정 결과 = 17
• 4.1 접합 대순환 모형의 결과분석 = 17
• 4.2 인공 신경망을 이용한 보정 결과 분석 = 26
• 제 5 장 결론 및 토의 = 46
• 참고 문헌 = 48
• ABSTRACT = 52
• 감사의 글 = 54