학술발표
연구를 나누고, 예측의 지평을 넓히다
학술발표를 통해 나누는 KIAPS의 연구성과를 소개합니다.
지식과 아이디어의 교류로 수치예보의 새로운 가능성을 모색합니다.
[세미나] 전문가 초청 세미나_손병주 교수
- 작성자
- 마스터관리자
- 작성일
- 조회수
- 88
- 일시: 2024년 10월 4일(금) 15:00~
- 장소: 사업단 7층 세미나실
- 연사: 손병주 교수(서울대)
- 제목: Satellite-Estimated Microwave Emissivity and Emission Temperature for DA over the Arctic Sea Ice Region: ANN-based Algorithm
- Abstract
Over the Artic Ocean, the data assimilation, which is one of most important techniques for weather forecasting, largely relies on satellite measurements because of the lack of conventional measurements. Microwave radiometer measurements are especially important because of the less sensitive nature of microwave to clouds. However, surface-sensitive channel measurements, that can provide atmospheric thermal information for the surface - lower troposphere layers, are much limited because of the poorly known surface emissivity and emitting layer temperature of the snow/ice. Here, based on simulations from the developed ice growth model, a retrieval method for directly measuring the necessary surface information from microwave channel measurements. In doing so, an artificial neural network (ANN) technique based on deep learning was introduced, and the non-linear relationship between satellitemeasured brightness temperatures and simulated thermal status of the snow and ice was learned. The ANN model was mapped and verified using the 10-fold cross-validation technique. Beside restoring surface emissivity and emission temperature using the training data, we indirectly validate the retrieval method by comparing simulated brightness temperatures against observations, in which retrieved emissivity and emission temperature were used as inputs for simulations. The comparison results at ATMS surface-sensitive sounding channels indicate an excellent agreement, promising that the surface emissivity and emission emperature can be directly obtained from ATMS measurements for the data assimilation when ATMS data are ingested to the assimilation system for forecasting.
- 장소: 사업단 7층 세미나실
- 연사: 손병주 교수(서울대)
- 제목: Satellite-Estimated Microwave Emissivity and Emission Temperature for DA over the Arctic Sea Ice Region: ANN-based Algorithm
- Abstract
Over the Artic Ocean, the data assimilation, which is one of most important techniques for weather forecasting, largely relies on satellite measurements because of the lack of conventional measurements. Microwave radiometer measurements are especially important because of the less sensitive nature of microwave to clouds. However, surface-sensitive channel measurements, that can provide atmospheric thermal information for the surface - lower troposphere layers, are much limited because of the poorly known surface emissivity and emitting layer temperature of the snow/ice. Here, based on simulations from the developed ice growth model, a retrieval method for directly measuring the necessary surface information from microwave channel measurements. In doing so, an artificial neural network (ANN) technique based on deep learning was introduced, and the non-linear relationship between satellitemeasured brightness temperatures and simulated thermal status of the snow and ice was learned. The ANN model was mapped and verified using the 10-fold cross-validation technique. Beside restoring surface emissivity and emission temperature using the training data, we indirectly validate the retrieval method by comparing simulated brightness temperatures against observations, in which retrieved emissivity and emission temperature were used as inputs for simulations. The comparison results at ATMS surface-sensitive sounding channels indicate an excellent agreement, promising that the surface emissivity and emission emperature can be directly obtained from ATMS measurements for the data assimilation when ATMS data are ingested to the assimilation system for forecasting.