본문 바로가기 주메뉴 바로가기
22대 대한민국 국회 국회정보길라잡이 국회의원검색
회원가입 로그인
HOME

정보검색

소장정보 검색

국회도서관 홈으로 정보검색 소장정보 검색

결과 내 검색

동의어 포함

고급검색

인명/단체명

저자정보 상세정보
인명/단체명을 입력하세요.

키워드

  • 대표어

  • 외국어

-

목차보기

[표지]

연구 보고서

목차

요약문 27

제1장 서론 29

제1절 과제의 배경 및 필요성 29

제2절 과제의 범위 30

제3절 과제 추진 체계 32

제2장 1차년도(2022년) 연구결과 33

제1절 정형·비정형 자료 활용 및 융합 기초기술 개발 33

1. 정지궤도기상위성 관측영상 시간(10분간격), 공간해상도 및 영역(동아시아) 일치 33

2. 레이더 바람장 융합을 위한 전처리 기술 개발 41

3. 지상관측 및(정지·극궤도 위성) 데이터 기반 격자화 강수데이터 생산 방안 연구 45

제2절 한반도 강수 유형별 학습데이터셋 생산 57

1. 한반도 강수유형 분류, 레이블링 및 검증 방안 연구 57

2. 수치모델 지형정보 산출 70

제3절 3차원 고해상도 자료 생산 기술 연구(제주지역 중심으로) 71

1. 제주지역 날씨유형, 유형별 핵심인자, 임계치 분석 71

2. 지상·고층 지점과 원격탐사(위성, 레이더 등) 관측자료 융합 고해상도 격자자료 산출 기술 기초설계 79

제3장 2차년도(2023년) 연구결과 96

제1절 정형·비정형 자료 활용 및 융합 기초기술 개발 96

1. 정지궤도기상위성 관측영상 시간(10분간격), 공간해상도 및 영역(동아시아) 일치 자동화 96

2. 레이더 관측 및 바람장, 수상체 등 산출정보 융합 전처리 기술 개발 107

3. 지상관측 및 정지·극궤도 위성데이터 기반 격자화 강수데이터 생산 알고리즘 개발 113

4. 레이더 관측자료 시·공간 해상도·영역 일치 및 관측공백 보간 기법 연구 122

5. 격자 기반 재해정보 레이블링 데이터셋 생산 방안 연구 137

제2절 한반도 강수 유형별 학습데이터셋 생산 142

1. 인공지능 강수 학습, 테스트, 검증 단계별 데이터 표준화 정의 142

2. 한반도 강수 유형별 데이터(수치모델, 위성, 레이더) 산출 및 레이블링 145

3. 강수유형 분류 자동화 및 자료 제공 방안 연구 154

4. 학습자료 생산 자동화 기능 통합 기술 기초설계 156

제3절 3차원 고해상도 자료 생산 기술 연구(제주지역 중심으로) 160

1. 관측자료의 공간해상도 상세화 기술 개발 160

2. 제주지역 3차원 고해상도 실황분석장 생산 방안 연구 167

제4장 3차년도(2024년) 연구결과 177

제1절 정형·비정형 자료 활용 및 융합 기초기술 개발 177

1. 레이더 관측 및 바람장, 수상체 정보 기반 학습데이터셋 개발 및 생산 자동화 177

2. 지상관측 및 정지·극궤도 위성데이터 기반 격자화 강수데이터 생산 자동화 188

3. 레이더 관측공백 보간 기법 개발 198

4. 격자 기반 재해정보 레이블링 데이터셋 생산 220

제2절 한반도 강수 유형별 학습데이터셋 생산 222

1. 한반도 강수 유무 및 유형 분류 과정 자동화 완료 222

2. 위성, 레이더, 수치모델자료 융합 기반 학습자료 표준화 232

3. 학습자료 자동화 통합 및 감시·수동수행 기능 개발 243

제3절 3차원 고해상도 자료 생산 기술 연구(제주지역 중심으로) 249

1. 제주지역 3차원 고해상도 모델기반 분석장 생성 249

2. 제주지역 3차원 고해상도 관측기반 분석장 생성 267

제5장 요약 및 결론 285

1. 정형·비정형 자료 활용 및 융합 기초기술 개발 285

1.1. 정지궤도 기상위성 관측영상 시간(10분 간격), 공간해상도 및 영역 일치 285

1.2. 레이더 관측 및 바람장, 수상체정보 기반 학습데이터셋 개발 및 생산 자동화 286

1.3. 지상관측 및 정지·극궤도 위성데이터 기반 격자화 강수데이터 생산 자동화 287

1.4. 레이더 관측공백 보간 기법 개발 290

1.5. 격자 기반 재해정보 레이블링 데이터셋 생산 292

2. 한반도 강수 유형별 학습데이터셋 생산 293

2.1. 위성, 레이더, 수치모델 자료 학습자료 표준화 294

2.2. 학습자료 자동화 통합 및 감시·수동 수행 기능 295

3. 3차원 고해상도 자료 생산 기술 연구(제주지역 중심으로) 297

3.1. 제주지역 3차원 고해상도 모델기반 분석장 생산 297

3.2. 제주지역 3차원 고해상도 관측기반 분석장 생성 298

참고 문헌 302

부록 1. 주요 성과 304

[뒷표지] 305

List of Tables

Table 1.1.1. Comparison of characteristics of numerical and observational data 29

Table 2.1.1. Average statistical values for each GK-2A channel applied to the model 38

Table 2.1.2. Coefficients of the first-order regression equation and number of... 39

Table 2.1.3. Regression coefficients and number of data points used for the SWIR... 39

Table 2.1.4. Comparison of statistical values for model application results before... 40

Table 2.1.5. Thresholds for 3D wind quality control 42

Table 2.1.6. Period and resolution of secured HSR and AWS data 50

Table 2.1.7. Period and resolution of secured GK-2A data 51

Table 2.1.8. D2D learning results based on changes in thresholds for training... 56

Table 2.2.1. Details of the precipitation type classification experiment 58

Table 2.2.2. Defining precipitation types using SOM techniques 60

Table 2.2.3. Spatial patterns of precipitation type based on cluster analysis by... 62

Table 2.2.4. Defining precipitation types using som, k-means, and hca techniques 62

Table 2.2.5. An element-by-element vector value comparison graph for each type of case 65

Table 2.2.6. Precipitation distribution and sea level pressure spatial patterns by type 65

Table 2.2.7. Define precipitation type by variable characteristics 66

Table 2.2.8. contingency table 67

Table 2.2.9. Summer +0-144h ERA5 analysis field labeling and ECMWF... 68

Table 2.2.10. PC results by forecast time in summer 68

Table 2.2.11. Winter +0-144 h ERA5 analysis field labeling and ECMWF... 69

Table 2.2.12. PC results by forecast time in winter 69

Table 2.2.13. Status of numerical model of korea meteorological administration 70

Table 2.3.1. Number and resolution of ground observation data points on... 80

Table 2.3.2. Domain information for 3-dimensional wind field 82

Table 2.3.3. Channel wavelength(unit), spatial resolution, and observation... 87

Table 2.3.4. GK-2A Atmospheric Motion Vector(AMV) information 87

Table 2.3.5. Used weighting coefficient options of various cost functions 93

Table 3.1.1. Averaged statistical result of the model application by GK-2A channels 101

Table 3.1.2. Linear regression coefficients and number of data for regression 101

Table 3.1.3. Quadratic regression coefficients and number of data for regression 102

Table 3.1.4. Averaged statistical result of the model application by COMS channels 105

Table 3.1.5. R3D quality control conditions and threshold values for each step 110

Table 3.1.6. Radar-AWS and satellite data information 114

Table 3.1.7. Common area information 115

Table 3.1.8. Averaged value by each evaluation index of analysis models 121

Table 3.1.9. Selected single bands and band differences as input domain 131

Table 3.1.10. Composition ratios and details of rainfall types in the dataset 132

Table 3.1.11. Averaged value by each evaluation index of analysis models 133

Table 3.1.12. Disaster data information for the production of administrative... 137

Table 3.1.13. Information on landslide data for the production of administrative... 138

Table 3.1.14. The number of administrative districts(Si, Gun, Gu) where 100... 141

Table 3.1.15. The number of administrative districts(Eup, Myeon, Dong) where... 141

Table 3.2.1. Building a training dataset process 142

Table 3.2.2. Define each step of the training dataset 144

Table 3.2.3. Data for the use of training datasets 145

Table 3.2.4. standards for the use of training datasets 146

Table 3.2.5. Standardization process for short to medium-term learning data 147

Table 3.2.6. Standardization process for short-term training data 147

Table 3.2.7. Explanation of methods for spatiotemporal resolution transformation 148

Table 3.2.8. The classification results of precipitation types based on the... 149

Table 3.2.9. Contingency table 150

Table 3.2.10. Summer +0-144 h ERA5 analysis field labeling(1-hour intervals)... 151

Table 3.2.11. Verification results of ECMWF 1-hour standardized data labeling 151

Table 3.2.12. The results of precipitation type classification based on spatial... 152

Table 3.2.13. Contingency table for SOM technique precipitation type... 153

Table 3.2.14. Example of classification results for short-mid-term precipitation types 155

Table 3.2.15. Example of classification results for short-term... 155

Table 3.2.16. Investigation contents by data for short-mid-term equipment 157

Table 3.2.17. Investigation contents by data for short-term equipment 157

Table 3.2.18. Standardization criteria grid and drafting information file 158

Table 3.2.19. Numeric model training dataset variable configuration auxiliary file 159

Table 3.3.1. WRF model configuration 174

Table 4.1.1. Rainfall cases for calculating each test result 177

Table 4.1.2. Formula for each test 179

Table 4.1.3. Adjusted threshold for each test 181

Table 4.1.4. The name of the log file according to the data type 186

Table 4.1.5. Summary of work to automate precipitation data production 189

Table 4.1.6. The name of the log file according to the log file type 191

Table 4.1.7. Weight of satellite 200

Table 4.1.8. Verification result - 2023.05.05 204

Table 4.1.9. Verification result - all case 205

Table 4.1.10. Input data for training and application of nighttime precipitation... 207

Table 4.1.11. Selected input bands and band differences for each target band 208

Table 4.1.12. Dataset details for VIS, NIR generation experiments and... 210

Table 4.1.13. The average statistical values for the test set by generated band 213

Table 4.1.14. The average value of the test set for each evaluation metrics of M1,... 215

Table 4.1.15. Average value and rate of change of M3 and M4 for test set by... 215

Table 4.1.16. Data structures for generated data of HSR and overseas radar... 219

Table 4.1.17. Information on forest fire data for the production of administrative... 220

Table 4.1.18. Output file information of the disaster information dataset 221

Table 4.1.19. Resolution of each grid and size of the output file 221

Table 4.1.20. Examples of disaster occurrence rates and damage information output at... 221

Table 4.2.1. Cluster element pattern by precipitation case 223

Table 4.2.2. Number and ratio of cluster types by precipitation case 223

Table 4.2.3. Precipitation type classification methodology by numerical model 224

Table 4.2.4. KIM 5 cluster type classification results 225

Table 4.2.5. UM 5 cluster type classification results 225

Table 4.2.6. UM 7 cluster type classification results 227

Table 4.2.7. verification of precipitation type classification results by numerical model 228

Table 4.2.8. +6-144h UM, KIM, ECMWF analysis field labeling(6-hour... 228

Table 4.2.9. UM, KIM, ECMWF PC verification results by forecast time(6 hour intervals) 229

Table 4.2.10. Short-mid-term precipitation type classification system operation status 230

Table 4.2.11. Example of classification results for short-term precipitation types 231

Table 4.2.12. Short-term precipitation type classification system operation status 231

Table 4.2.13. Example of classification results for short-mid term precipitation types 231

Table 4.2.14 standards for the use of training datasets 232

Table 4.2.15. Standardized system log example 234

Table 4.2.16. Verification data resolution comparison 236

Table 4.2.17. Temperature verification results by resolution during the... 237

Table 4.2.18. Temperature distribution chart of difference between raw data and... 237

Table 4.2.19. Temperature CDO interpolation results by resolution 20220810 00UTC 238

Table 4.2.20. Example of Temperature CDO interpolation results for each... 238

Table 4.2.21. Temperature distribution chart of difference between raw data and... 239

Table 4.2.22. Temperature CDO interpolation results by resolution 20221219 00UTC 240

Table 4.2.23. Precipitation verification results by resolution during the... 240

Table 4.2.24. Precipitation distribution chart of difference between raw data and... 241

Table 4.2.25. Precipitation distribution chart of difference between raw data and... 241

Table 4.2.26. Precipitation CDO interpolation results by resolution 20220804... 242

Table 4.2.27. Time required for each short-to medium-term system process 244

Table 4.2.28. Time required for each short-term system process 245

Table 4.2.29. Overview of the learning material production monitoring system 245

Table 4.2.30. Colors displayed based on task execution status on the monitoring webpage 246

Table 4.2.31. Page categorization information for the learning material production... 247

Table 4.2.32. Structure of the integrated log file 247

Table 4.2.33. Monitoring web page function description 248

Table. 4.3.1. WRF model configuration 250

Table. 4.3.2. Case description 251

Table. 4.3.3. MAE and PC of total accumulated rainfall for each case and... 266

Table. 4.3.4. The name of the log file according to the log file type 275

Table. 4.3.5. Cases used in verification 276

Table. 4.3.6. The selected events to valid wind of numerical model output 280

List of Figures

Fig. 2.1.1. Example of alignment of research area for radar HSR, GK-2A, and... 34

Fig. 2.1.2. The data period for train, validation, and application of the model 34

Fig. 2.1.3. Data pair method used in this study 35

Fig. 2.1.4. Structure of Data2Data for this study 36

Fig. 2.1.5. Results of applying GK-2A IR10.5 channel to the generated model 37

Fig. 2.1.6. Scatter plot comparison of each channel between GK-2A and COMS 38

Fig. 2.1.7. Results of model application before and after calibration of the WV channel 40

Fig. 2.1.8. Time series results of applying the temporal resolution enhancement... 40

Fig. 2.1.9. CFAD of WISSDOM wind field u, v, w wind components and... 41

Fig. 2.1.10. Examples of results from performing abrupt change checks and spike checks. 43

Fig. 2.1.11. QC Flag Example(September 6, 2020 22:20 KST) 44

Fig. 2.1.12. (a) Location and interpolation method of AWS(Automatic... 46

Fig. 2.1.13. Four tested precipitation generation algorithm 46

Fig. 2.1.14. Algorithm flowchart of rain data merge 47

Fig. 2.1.15. (a) One-point removal cross-validation error according to case and... 48

Fig. 2.1.16. Example of rain data merge 49

Fig. 2.1.17. Study area(left) and gridded AWS result(right) 51

Fig. 2.1.18. (a) Radar rainfall, (b) AWS rainfall, (c) regression predicted AWS,... 52

Fig. 2.1.19. Results of AWS grid mapping using regression kriging method 53

Fig. 2.1.20. (a) RK AWS, (b) virtual, (c) RK AWS-virtual, (d) map of hit-false-miss,... 54

Fig. 2.1.21. (a) RK AWS, (b) virtual, (c) RK AWS-virtual, (d) map of hit-false-miss,... 54

Fig. 2.1.22. (a) RK AWS, (b) virtual, (c) RK AWS-virtual, (d) map of hit-false-miss,... 55

Fig. 2.1.23. (a) RK AWS, (b) virtual, (c) RK AWS-virtual, (d) map of hit-false-miss,... 55

Fig. 2.2.1. Flowchart of korea's local heavy rain type classification using som algorithm 59

Fig. 2.2.2. Characteristics of spatial distribution by type of local precipitation 60

Fig. 2.2.3. Classification of precipitation types according to cluster analysis. First... 61

Fig. 2.2.4. Precipitation type cluster classification analysis area by variable characteristics 63

Fig. 2.2.5. Dendrogram results of HCA techniques by variable characteristics 64

Fig. 2.2.6. Labeling process flowchart 67

Fig. 2.3.1. Weather categories with exemplary synoptic surface weather map 72

Fig. 2.3.2. (Upper Left) 850 hPa geopotential height(gpm, red contour) and... 73

Fig. 2.3.3. (Upper Left) 850 hPa horizontal gradient of equivalent potential... 74

Fig. 2.3.4. (Upper Left) Difference between the sea surface temperature and 850 hPa... 76

Fig. 2.3.5. (Left) 850 hPa vertical velocity(10⁻¹ ms⁻¹, shading) and horizontal... 77

Fig. 2.3.6. (Left) 850hPa stretching deformation(10⁻⁵ s⁻¹) and (Right) horizontal... 77

Fig. 2.3.7. (Left) 850 hPa gradient of geopotential temperature(10⁻⁵, shading),... 78

Fig. 2.3.8. Location of observation points by observation variables displayed... 81

Fig. 2.3.9. Flow chart of early algorithm for 3-dimensional wind field... 82

Fig. 2.3.10. Radar location and observation radius near Jeju 83

Fig. 2.3.11. Spatial distribution of observations : AWS and ASOS... 85

Fig. 2.3.12. (2023.06.11.0300 UTC) Brightness Temperature of water vapor... 86

Fig. 2.3.13. Domain configuration(9.0, 3.0, 1.0, 0.2km resolution) 88

Fig. 2.3.14. Schematic diagram for constructing gridded PPI 89

Fig. 2.3.15. Generating 10-minute mean data from the KMA's AWS... 91

Fig. 2.3.16. Examples of generating 10-minute average field(wind, temperature,... 91

Fig. 2.3.17. Radial velocity PPI images of CLEANER. 92

Fig. 2.3.18. Validation results of u component of 3-dimensional wind... 94

Fig. 2.3.19. Validation results of v component of 3-dimensional wind... 95

Fig. 3.1.1. Study Area 97

Fig. 3.1.2. Results of applying GK-2A VIS 0.6㎛ channel to the model 98

Fig. 3.1.3. Results of applying GK-2A SW 3.8㎛(day) channel to the model 98

Fig. 3.1.4. Results of applying GK-2A SW 3.8㎛(night) channel to the model 99

Fig. 3.1.5. Results of applying GK-2A WV 6.9㎛ channel to the model 99

Fig. 3.1.6. Results of applying GK-2A IR 10.5㎛ channel to the model 100

Fig. 3.1.7. Results of applying GK-2A IR 12.3㎛ channel to the model 100

Fig. 3.1.8. Results of applying COMS VIS channel to the model. 102

Fig. 3.1.9. Results of applying COMS SWIR(day) channel to the model 103

Fig. 3.1.10. Results of applying COMS SWIR(night) channel to the model 103

Fig. 3.1.11. Results of applying COMS WV channel to the created model 104

Fig. 3.1.12. Results of applying COMS IR01 channel to the created model 104

Fig. 3.1.13. Results of applying COMS IR02 channel to the created model 105

Fig. 3.1.14. Results of improved temporal resolution of COMS IR01 channel 106

Fig. 3.1.15. CFAD image for each R3D radar observation variables 107

Fig. 3.1.16. The flow chart of R3D data pre-processing algorithm 108

Fig. 3.1.17. Normal frequency distribution(black) and cumulative frequency... 109

Fig. 3.1.18. CFAD images before(left) and after(right) applying the R3D... 112

Fig. 3.1.19. Algorithm for grid precipitation using AWS 114

Fig. 3.1.20. WISSDOM area(left) and study area(right) 116

Fig. 3.1.21. ResNet structure utilized in the generator 117

Fig. 3.1.22. Examples of the results of the generation algorithm for rain rate... 117

Fig. 3.1.23. Collocation algorithm flow chart 118

Fig. 3.1.24. Radar-AWS collocation result 119

Fig. 3.1.25. GEO collocation result 119

Fig. 3.1.26. LEO collocation result 119

Fig. 3.1.27. Box-and-whisker plot by evaluation metrics for the test set 120

Fig. 3.1.28. Case analysis(Apr. 26, 2022, 10:00 KST) 121

Fig. 3.1.29. The flowcharts of generation algorithm for gridded... 123

Fig. 3.1.30. Rain merge algorithm flow chart 124

Fig. 3.1.31. R3D reflectivity after spatiotemporal matching(four methods of xemsf) 125

Fig. 3.1.32. Examples of gridded precipitation fields at August 8, 2022 18:00... 126

Fig. 3.1.33. Scatter plot analysis between observation values of AWS and... 127

Fig. 3.1.34. East-Asia domain rain data merging example - 2022.09.0 5. 1340 KST 128

Fig. 3.1.35. Rain merge field scatter plot 129

Fig. 3.1.36. GK-2A(red), radar HSR(blue) data area, and rainfall... 130

Fig. 3.1.37. Case analysis(August 21, 2021, 11:00 KST) 133

Fig. 3.1.38. Stratiform precipitation case analysis(March 04, 2021, 15:00 KST) 134

Fig. 3.1.39. Trend analysis of evaluation metrics for stratiform precipitation cases... 135

Fig. 3.1.40. Convective precipitation case analysis(July 06, 2021, 12:00 KST) 135

Fig. 3.1.41. Trend analysis of evaluation metrics for convective precipitation... 136

Fig. 3.1.42. Grid map with a 0.1° interval in the range of 33~39°N, 125~131°E 139

Fig. 3.1.43. Confirm the positions of grid points with intervals of 0.1° and 0.01° 139

Fig. 3.1.44. Grid points with a spacing of 0.01° included within a 10x10 range... 140

Fig. 3.1.45. 2021.08.22 landslide, disaster(Typhoon) occurrence amount for grid are 141

Fig. 3.2.1. The purpose of each stage in the training dataset. 143

Fig. 3.2.2. Flowchart of labeling process for short-mid-term precipitation types 150

Fig. 3.2.3. Flowchart for short-term precipitation type labeling process 153

Fig. 3.2.4. Automated rainfall type classification system 154

Fig. 3.2.5. Design proposal for a learning dataset generation system 156

Fig. 3.3.1. (a) Algorithm for generating radar-rain gauge merging precipitation field... 161

Fig. 3.3.2. Observed precipitation vs. LOOCV values from (a) AWS, (b) all rain... 162

Fig. 3.3.3.Comparison of observationvalues and LOOCV values before and... 162

Fig. 3.3.4. Flow chart of improved algorithm for 3-dimensional wind field... 163

Fig. 3.3.5. Change of 3-dimensional wind field following surface input data... 165

Fig. 3.3.6. Comparison of u, v, wind speed(WS) with various options. (a)-(c)... 165

Fig. 3.3.7. Comparison of u, v, wind speed(WS) between 3-dimensional... 166

Fig. 3.3.8. Diagram of radar data preprocessing program 168

Fig. 3.3.9. Converted ASOS and AWS data into Little-R format as input for... 169

Fig. 3.3.10. Converted sea-surface data into Little-R format(2023.07.18.0000 UTC)... 170

Fig. 3.3.11. Converted wind profiler data into Little-R format(2023.07.18.0000 UTC) 170

Fig. 3.3.12. Converted rawinsonde data into Little-R format for WRFDA(2023.07.18.0000 UTC) 171

Fig. 3.3.13. Flow chart of the preprocessing program handling... 172

Fig. 3.3.14. Diagram of ASR data preprocessing program 173

Fig. 3.3.15. Diagram of AMV data preprocessing program 173

Fig. 3.3.16. Data assimilation timeline 174

Fig. 3.3.17. (2021.08.10.2200 UTC, Domain 03) (a) Radar reflectivity(dBZ) observation at... 175

Fig. 3.3.18. (2023.11.05.2000UTC) (a), (d) Radar reflectivity(dBZ) observation at 1.5km,... 176

Fig. 4.1.1. STEP test CFAD image before (left) and after (right) adding the... 178

Fig. 4.1.2. Threshold value determination procedure for each test 179

Fig. 4.1.3. Threshold determination process for STEP(CZ) test. 180

Fig. 4.1.4. Comparison of QC flag CAPPI (h＝2km) image (a) before and (b)... 182

Fig. 4.1.5. Comparison of QC flag CFAD (a) before and (b) after threshold adjustment 182

Fig. 4.1.6. Reflectivity at 2 km CAPPI image (a) before and (b) after threshold... 183

Fig. 4.1.7. Comparison of reflectivity(CZ) CFAD (a) before and (b) after... 184

Fig. 4.1.8. Comparison of 1 km altitude average reflectivity (a) before and (b)... 184

Fig. 4.1.9. Program flowchart with exception handling applied... 185

Fig. 4.1.10. Example of global properties of R3D preprocessing result file 186

Fig. 4.1.11. Example of log file 187

Fig. 4.1.12. Example of error messages 187

Fig. 4.1.13. AWS-Radar precipitation data production automation algorithm flowchart 189

Fig. 4.1.14. Examples of preferences and options in the 'Config.ini' file 190

Fig. 4.1.15. Examples of each message in the unified log file and error log file 191

Fig. 4.1.16. Algorithm flow chart for LEO collocation 192

Fig. 4.1.17. AWS and radar merging precipitation field comparison test... 193

Fig. 4.1.18. Scatter plot comparison for quantitative validation(period : June... 194

Fig. 4.1.19. Summary of error verification results using bar graphs 195

Fig. 4.1.20. Cumulative Density Function plot for satellite data 196

Fig. 4.1.21. F18 rainfield adjustment example - 2023.08.10. 15:40 KST 197

Fig. 4.1.22. Example of log files for LEO collocation algorithm 197

Fig. 4.1.23. Rain data merge algorithm flow chart 198

Fig. 4.1.24. Example of weight calculation 199

Fig. 4.1.25. Example of ground rain field and satellite rain field, merge rain rain... 201

Fig. 4.1.26. Merge rainfield mean rainrate and GTS mean rainrate - 2023.05.05 203

Fig. 4.1.27. Verification result scatter plot - all case 204

Fig. 4.1.28. Example of log files for rain data merge algorithm 205

Fig. 4.1.29. Correlation analysis between target bands 207

Fig. 4.1.30. MI between target data and all 45 IR band and band difference combinations 208

Fig. 4.1.31. Correlation heatmap by MI values for VIS band generation 209

Fig. 4.1.32. Correlation heatmap by MI values for NIR1.3 band generation 209

Fig. 4.1.33. Correlation heatmap by MI values for NIR1.6 band generation 210

Fig. 4.1.34. Analysis of virtual VIS and NIR band data for the case on... 212

Fig. 4.1.35. Box-and-whisker plot of statistical metrics by band 213

Fig. 4.1.36. Trend analysis of M1, M2, and M3 for the test set by evaluation metrics 214

Fig. 4.1.37. Mean value bar plots for IR model(M1), daytime model, and... 216

Fig. 4.1.38. Estimated results and qualitative comparisons of precipitation in IR... 217

Fig. 4.1.39. Model training area(green), radar HSR data area(blue), overseas... 218

Fig. 4.1.40. Ground radar data and regional generation examples for the daytime... 219

Fig. 4.2.1. Spatial distribution of cape variables for each numerical model 229

Fig. 4.2.2. Standardized production automation system for short-mid-term... 233

Fig. 4.2.3. Standardized production automation system for short-term... 234

Fig. 4.2.4. Area average method(simple average(left), weighted average(right)) 235

Fig. 4.2.5. Examples of grids used according to grid resolution(0.05 degrees(left),... 236

Fig. 4.2.6. Data automation integrated production system schematic diagram 243

Fig. 4.2.7. Schematic diagram of the learning material production monitoring system 246

Fig. 4.2.8. Example of monitoring display on the monitoring webpage 248

Fig. 4.3.1. WRF domain configuration 250

Fig.4.3.2. WRF modeling and assimilation timeline 251

Fig. 4.3.3. O-B BT statistics for GK-2A water vapor channels(Ch 8-10)... 252

Fig. 4.3.4. 7.3㎛ Brightness temperature(K) of GK-2A(Left), WRF... 253

Fig. 4.3.5. 30 minute precipitation of(Left) AWS, (Middle) NoDA, (Right) DA at... 254

Fig. 4.3.6. Water vapor mixing ratio(shading, g kg⁻¹) and wind(vector, ms⁻¹)... 254

Fig. 4.3.7. Cross-section of water vapor mixing ratio(shading, g kg⁻¹) and wind.... 255

Fig. 4.3.8. 30 minute precipitation of (Left) AWS, (Middle) NoDA, (Right) DA at... 256

Fig. 4.3.9. Water vapor mixing ratio(shading, g kg⁻¹) and wind(vector, ms⁻¹) at 850 hPa 257

Fig. 4.3.10. Cross-section of water vapor mixing ratio (shading, g kg⁻¹) and wind... 258

Fig. 4.3.11. 30 minute precipitation of (Left) AWS, (Middle) NoDA, (Right) DA... 259

Fig. 4.3.12. Water vapor mixing ratio(shading, g kg⁻¹) and wind(vector, ms⁻¹) at 850 hPa 260

Fig. 4.3.13. Cross-section of water vapor mixing ratio(shading, g kg⁻¹) and wind vector... 261

Fig. 4.3.14. Scatter plot of GK-2A WV channels BT [K] from(Left) observation... 262

Fig. 4.3.15. NoDA(black), DA_Old(green), and DA_New(blue) MAE and... 264

Fig. 4.3.16. NoDA(black), DA_Old(green), and DA_New(blue) MAE and... 264

Fig. 4.3.17. NoDA(black), DA_Old(green), and DA_New(blue) MAE... 265

Fig. 4.3.18. NoDA(black), DA_Old(green), and DA_New(blue) MAE and... 265

Fig. 4.3.19. MAE improvement rate of DA compared to NoDA(Case 1-5) 266

Fig. 4.3.20. PC improvement rate of DA compared to NoDA(Case 1-5) 266

Fig. 4.3.21. Examples of image files and ASCIIfiles generate with Python code 268

Fig. 4.3.22. Modified quality control code section in the function 'make_QC_flag_other' 268

Fig. 4.3.23. Example if wind filed images before and after quality control 269

Fig. 4.3.24. Examples of temperature field images and nonlinearly... 270

Fig. 4.3.25. Modifications to the code of the 'Nonlinear_AT' function(left)... 270

Fig. 4.3.26. Examples of function forms approximated nonlinearly... 271

Fig. 4.3.27. Examples of temperature fields generated with nonlinear profiles for... 272

Fig. 4.3.28. (a) Radar precipitation, (b) precipitation field merged with existing... 274

Fig. 4.3.29. General settings in the configuration file 274

Fig. 4.3.30. Output option in the configuration file 275

Fig. 4.3.31. Examples of the production images of the ground observation... 276

Fig. 4.3.32. Scatter plot analysis for Case 1 277

Fig. 4.3.33. Examples of the production images of the ground observation... 277

Fig. 4.3.34. Scatter plot analysis for Case 2 278

Fig. 4.3.35. Flow chart of the developed algorithm for 3-dimensional... 279

Fig. 4.3.36. Photographs during installation of CAPPI algorithm in 5th super... 281

Fig. 4.3.37. Photographs during installation of algorithm for 3-dimensional wind... 282

Fig. 4.3.38. (a) Example of log file when algorithm for 3-dimensional... 283

Fig. 4.3.39. Photographs during installation of algorithm for image... 283

Fig. 4.3.40. Plots of 3-dimensional wind field data about 20 : 00 UTC 5 November 2023. 284

AI 100자 요약 서비스

AI 100자 요약 서비스
인공지능이 요약한 요약문입니다. 본문과 일치하지 않은 내용이 있을 수 있으니, 원문을 확인하시기 바랍니다.
본 연구는 인공지능 기반 기상 예측 정확도 향상을 목표로, 2022년부터 2024년까지 3년간 기상 데이터 전처리 기술 개발에 매진했습니다. 정지궤도 기상위성 영상, 지상/레이더 관측 자료, 수치 예보 모델 데이터 등 다양한 기상 데이터의 품질 관리부터 공간・시간적 일치 처리, 결측값 보완 과정까지 전반적인 개선 작업을 진행했습니다. 특히 동아시아 지역에 대한 고해상도 위성영상 자동 조정 기술과 레이더 관측 공백 채움 알고리즘을 발전시켜 데이터 활용도를 높였으며, 한반도 강수 유형에 따른 자동화된 학습 데이터셋 구축 및 제주 지역 중심의 3차원 고해상도 자료 생산 기술 연구를 병행했습니다. 또한 천리안 위성과 GK-2A 위성 자료를 활용하여 시간 해상도를 개선하고, 여러 관측 자료 간의 융합을 통한 데이터 완성도를 높였습니다. 이러한 노력을 통해 딥러닝 기반 강수 추정 모델의 정확도를 향상시키고, 자동화된 자료 생산 시스템을 구축하여 기상 연구 및 예보 서비스의 효율성을 증대시키는 성과를 달성했으며, 향후 AI 기반 기상 서비스 발전에 필요한 핵심 토대를 마련하는 데 기여할 것입니다.

※ 출처: 상단의 서지정보에서 확인
※ 원저작물에 적용된 공공누리 유형: 제2유형

이용현황보기

기상-AI 데이터 전처리 기술 개발. 3 이용현황 표 - 등록번호, 청구기호, 권별정보, 자료실, 이용여부로 구성 되어있습니다.
등록번호 청구기호 권별정보 자료실 이용여부
0003178397 551.630285 -25-3 v.3 서울관 인문자연과학자료실(314호) 이용가능
0003178398 551.630285 -25-3 v.3 서울관 인문자연과학자료실(314호) 이용가능

추천서가 (다양한 추천 자료를 만나보세요)

권호기사보기

권호기사 목록 테이블로 기사명, 저자명, 페이지, 원문, 기사목차 순으로 되어있습니다.
기사명 저자명 페이지 원문 기사목차

권호

기사명 원문보기 기사목차
닫기