[표지]
발간사 / 정진승
ABSTRACT
목차
1. 서론 22
1.1. 연구 배경 및 목적 22
1.1.1. 가뭄의 정의 22
1.1.2. 가뭄 피해 및 기후변화의 영향 23
1.1.3. 연구의 목적 31
1.2. 연구 범위 32
1.2.1. 공간적 범위 32
1.2.2. 시간적 범위 33
1.2.3. 내용적 범위 34
2. 원격 탐사를 활용한 가뭄 감시 35
2.1. 전통적 가뭄 감시 35
2.1.1. 전통적 가뭄 지수 36
2.2. 가뭄 관련 변수의 원격 탐사 40
2.2.1. 강수량 41
2.2.2. 증발산량 41
2.2.3. 토양수분 41
2.2.4. 식생 42
2.3. 원격 탐사 기반 가뭄 지수 43
2.4. 변수 융합을 통한 맞춤형 가뭄 감시 45
3. 원격 탐사를 활용한 가뭄 예측 48
3.1. 가뭄 예측 모델링 48
3.1.1. 회귀 모델 49
3.1.2. 시계열 모델 50
3.1.3. 인공 신경망 모델 51
3.1.4. 의사결정분지도 모델 52
3.1.5. 장기 예측 모델 53
3.2. 가뭄 예측을 위한 원격 탐사의 활용 54
4. 자료 및 연구 방법론 55
4.1. 위성 영상 자료 56
4.1.1. 강수량 56
4.1.2. 지표면온도 58
4.1.3. 증발산량 60
4.1.4. 광합성유효복사율 62
4.1.5. 엽면적지수 62
4.1.6. 식생지수 63
4.1.7. 토양수분 63
4.2. 원격 탐사 기반 가뭄 지수 66
4.2.1. 표면반사율 66
4.2.2. 원격 탐사 기반 가뭄 지수의 산정 67
4.3. 관측 자료 67
4.3.1. 종관기상관측자료 67
4.3.2. KoFlux 자료 67
4.3.3. 기후 지수 72
4.3.4. 고랭지 채소 수확량 73
4.3.5. 산불 발생 빈도 및 피해 면적 74
4.4. 관측 자료 기반 가뭄 지수 75
4.4.1. SPI의 산정 75
4.4.2. SPEI의 산정 76
4.5. 지면모델 자료 76
4.5.1. 지면모델 개요 76
4.5.2. NOAH-LSM 81
4.5.3. 지면모델 결과와 위성 영상 자료의 결합 82
4.6. 기계학습을 통한 가뭄 감시 및 예측 모델 개발 82
4.6.1. 인공지능과 전문가시스템 82
4.6.2. 기계학습 83
4.6.3. Random Forests 85
4.6.4. 모델의 개발 86
4.6.5. 모델의 적용 89
5. 결과 및 토의 96
5.1. 위성 영상 추출 변수의 검증 96
5.1.1. 강수량 96
5.1.2. 지표면온도 102
5.1.3. 증발산량 114
5.1.4. 토양수분 124
5.2. 가뭄 감시 및 예측 모델 선정 145
5.2.1. 가뭄 감시 모델의 선정 145
5.2.2. 가뭄 예측 모델의 선정 154
5.3. 가뭄 예측 모델의 검증 168
5.3.1. 고랭지 채소 수확량 168
5.3.2. 산불 발생 빈도 및 피해 면적 172
6. 결론 및 시사점 176
6.1. 연구 결과의 요약 176
6.2. 연구의 결론 및 시사점 180
6.3. 활용 제안 181
6.4. 향후 가뭄 연구 제안 183
REFERENCES 184
[뒷표지] 192
Table 1-1. Drought events in South Korea from 1962~1994 28
Table 2-1. Classification categories of SPI 39
Table 4-1. Weather stations used in the Korean Peninsula 68
Table 4-2. List of KoFlux sites used in this study 71
Table 4-3. Input variables of random forest models for drought monitoring 88
Table 5-1. Statistics between TRMM PRCP and observed monthly precipitation averaged over... 97
Table 5-2. Statistics between MODIS LST and observed monthly temperature averaged over... 103
Table 5-3. Statistics between MODIS PET and calculated PET based on Penman-Monteith... 115
Table 5-4. Statistics between NOAH-LSM PET and calculated PET based on Penman-Monteith... 119
Table 5-5. Statistics between monthly averaged AMSR-E A_SM and KoFlux Soil Moisture for... 126
Table 5-6. Statistics between monthly averaged AMSR-E D_SM and KoFlux Soil Moisture for... 127
Table 5-7. Statistics between monthly averaged AMSR-E SM_MEAN and KoFlux Soil Moisture... 128
Table 5-8. Statistics between daily AMSR-E A_SM and KoFlux Soil Moisture for two KoFlux... 131
Table 5-9. Statistics between daily AMSR-E D_SM and KoFlux Soil Moisture for two KoFlux... 132
Table 5-10. Statistics between daily AMSR-E SM_MEAN and KoFlux Soil Moisture for two... 133
Table 5-11. Statistics between monthly NOAH-LSM Soil Moisture and KoFlux Soil Moisture for... 136
Table 5-12. Statistics between daily NOAH-LSM Soil Moisture and KoFlux Soil Moisture for two... 138
Table 5-13. Statistics between monthly AMSR-E SM_MEAN and NOAH- LSM Soil Moisture 143
Table 5-14. Statistics between converted AMSR-E Soil Moisture and KoFlux Soil Moisture for... 145
Table 5-15. RMSE of random forest models for target variables. 147
Table 5-16. Coefficient of determination (R²) of random forest models for target variables 148
Table 5-17. Relative importance of input variables for each target variable based on observed... 149
Table 5-18. Relative importance of input variables for each target variable based on remotely... 150
Table 5-19. RMSE of random forest models for target variables based on observation data... 156
Table 5-20. RMSE of random forest models for target variables based on remotely sensed... 157
Table 5-21. Coefficient of determination (R²) values of random forest models for target variables... 159
Table 5-22. Coefficient of determination (R²) values of random forest models for target... 160
Table 5-23. RMSE of random forest models for target variables based on observation data... 162
Table 5-24. RMSE of random forest models for target variables based on remotely sensed... 163
Table 5-25. Coefficient of determination (R²) values of random forest models for target variables... 164
Table 5-26. Coefficient of determination (R²) values of random forest models for target variables... 165
Table 5-27. Relative importance of input variables for each target variable based on remotely... 167
Table 5-28. Statistics between crop yield and target variable data 169
Table 5-29. Statistics between observed and estimated target variable data 171
Table 5-30. Statistics between forest fire frequency and target variable data 174
Table 5-31. Statistics between observed and estimated NDWI7 in Gyeongsangnam-do 175
Figure 1-1. Top ten drought disasters from 1900~2014 25
Figure 1-2. Top ten drought disasters from 1985~2014 26
Figure 1-3. Study area–the Korean Peninsula 34
Figure 2-1. Processes contributing to (a) soil moisture-temperature, and (b) soil moisture-precipitation... 46
Figure 2-2. September P-PET percentiles in Upper Namhan-gang basin 46
Figure 3-1. Components for drought forecasting 50
Figure 3-2. Components of a typical Artificial Neural Network model 51
Figure 3-3. A decision-tree expert system with a rule and conditions to test a hypothesis 53
Figure 4-1. Diagram of the TRMM satellite and instruments 57
Figure 4-2. Global Precipitation Measurement Mission Core Observatory Satellite 58
Figure 4-3. Terra spacecraft with instruments 59
Figure 4-4. Sample Terra orbital tracks on 15 February 2005 60
Figure 4-5. Flow chart of the MODIS evapotranspiration algorithm 61
Figure 4-6. Aqua's instruments 65
Figure 4-7. Sample Aqua orbital tracks on 15 February 2005 65
Figure 4-8. Two KoFlux sites used in this study 71
Figure 4-9. Multivariate ENSO Index 73
Figure 4-10. Four SI-DO and 18 SI-KUN-KU provinces with crop yield data 74
Figure 4-11. Illustration of a first-generation land surface model 77
Figure 4-12. Illustration of a second-generation land surface model 78
Figure 4-13. Illustration of a third-generation land surface model 79
Figure 4-14. Levels of land surface modelling approaches 79
Figure 4-15. VIC-3L model 81
Figure 4-16. Components of rule-based expert system 84
Figure 4-17. Locations of grids used: Input data for 5x5㎞ grids located in highland areas are... 92
Figure 4-18. Forest Fire Risk Prediction System of Korea Forest Service 94
Figure 4-19. Locations of grids used: Input data for 5x5㎞ grids located in forest areas are... 95
Figure 5-1. RMSE between TRMM PRCP and observed monthly precipitation during (a) all... 98
Figure 5-2. Pearson's r between TRMM PRCP and observed monthly precipitation during... 100
Figure 5-3. Time-series of TRMM PRCP and observed monthly precipitation from during... 101
Figure 5-4. RMSE between LSTDAY and observed monthly maximum air temperature during... 104
Figure 5-5. Pearson's r between MODIS LSTDAY and observed monthly maximum air... 106
Figure 5-6. RMSE between LSTNIGHT and observed monthly minimum air temperature during... 108
Figure 5-7. Pearson's r between MODIS LSTNIGHT and observed monthly minimum air... 110
Figure 5-8. Time-series of MODIS LSTDAY and observed monthly maximum temperature in... 112
Figure 5-9. Time-series of MODIS LSTNIGHT and observed monthly minimum temperature in... 113
Figure 5-10. Time-series of MODIS LST_MEAN and observed monthly mean temperature in... 113
Figure 5-11. Pearson's r between MODIS PET and calculated PET during (a) all season, (b)... 115
Figure 5-12. Time-series of MODIS PET and calculated PET in Inje station 118
Figure 5-13. RMSE between NOAH-LSM PET and calculated PET during (a) all season, (b)... 121
Figure 5-14. Pearson's r between NOAH-LSM PET and calculated PET during (a) all season,... 122
Figure 5-15. Time-series of NOAH-LSM PET and calculated PET in Gangneung station 124
Figure 5-16. Time-series of KoFlux Soil Moisture and monthly averaged AMSR-E (a) A_SM,... 129
Figure 5-17. Time-series of KoFlux Soil Moisture and daily averaged AMSR-E (a) A_SM, (b)... 134
Figure 5-18. Time-series of monthly NOAH-LSM Soil Moisture and KoFlux Soil Moisture in... 137
Figure 5-19. Time-series of daily aveaged NOAH-LSM Soil Moisture and KoFlux Soil Moisture... 139
Figure 5-20. Pearson's r between AMSR-E Soil Moisture and NOAH-LSM Soil Moisture during... 141
Figure 5-21. Scatterplot between observed SPI1 and estimated SPI1 based on a random forest... 152
Figure 5-22. Scatterplot between observed SPEI1 and estimated SPEI1 based on a random... 152
Figure 5-23. Scatterplot between observed NDWI6 and estimated NDWI6 based on a random... 153
Figure 5-24. Scatterplot between observed NDWI7 and estimated NDWI7 based on a random... 153
Figure 5-25. RMSE of random forest models for target variables based on observation data... 155
Figure 5-26. Coefficient of determination (R²) values of random forest models for target... 158
Figure 5-27. Scatterplot between observed July NDWI7 and estimated July NDWI7 based on a... 172
Figure 5-28. Scatterplot between observed NDWI7 and estimated NDWI7 for all months based... 175