Thông tin chung
  English
  Đề tài NC khoa học
  Bài báo, báo cáo khoa học
  Hướng dẫn Sau đại học
  Sách và giáo trình
  Các học phần và môn giảng dạy
  Giải thưởng khoa học, Phát minh, sáng chế
  Khen thưởng
  Thông tin khác
  Tài liệu tham khảo
  Hiệu chỉnh
 
Số người truy cập: 107,045,312

 Combining Geophysical Variables for Maximizing Temporal Correlation Without Reference Data
Tác giả hoặc Nhóm tác giả: S. Kim, H. T. Pham, Yi Y. Liu, A. Sharma and L. Marshall
Nơi đăng: The 23rd International Congress on Modelling and Simulation (MODSIM2019); Số: 2019;Từ->đến trang: 843;Năm: 2019
Lĩnh vực: Khoa học công nghệ; Loại: Báo cáo; Thể loại: Quốc tế
TÓM TẮT
Geophysical datasets are important for understanding the principles of nature. Ground-based measurement is a classic way to obtain geophysical information, but is generally limited in space and time. Satellite and model-derived geophysical estimates can be an appealing alternative to ground measurement due to improved spatiotemporal availability. However, these estimates generally have their own advantages and weaknesses, and there is still room for improvements in their performance. To achieve this, many efforts have been made through de-noising, filtering and merging. Here we focus on merging existing satellite- and model-derived datasets for improved measurement of geophysical variables. Such satellite- and model-derived variables have shown complementarity in their performance in terms of statistical metrics such as bias, root mean square error, and correlation coefficients. This complementary behaviour results from their different skills under different physical/ climatological retrieval conditions. For remote sensing data, the ability to capture temporal variability has been regarded as important in many applications. A linear combination of data sets is a simple but effective way to take the strengths of the original products to achieve better performance. However, a limitation is that calculating optimal weight assigned to each original product during the linear combination needs a truth reference that is often not available and thus hinders their practical applications. The extended triple collocation (ETC) approach does not have this limitation and it can provide error variances and Pearson correlations of the parent products against (hidden) truth without any reference data. The ETC approach has been widely used for uncertainty estimation of environmental variables (e.g. soil moisture, rainfall), but it has not been directly used for data merging, especially for maximizing correlation. The aim of this study is, therefore, to develop a merging approach which can take advantages of both a linear combination and ETC approach. That is, we derive a linear combination method maximising correlation using data-truth correlations derived from ETC. For this, we first present the theoretical background of the proposed method and then verify it through synthetic experiments. In addition to this, we also perform the ETC-based combination scheme using various satellite- and model-derived soil moisture data. Finally, the combination results are compared to ground-based measurements.
ABSTRACT
Geophysical datasets are important for understanding the principles of nature. Ground-based measurement is a classic way to obtain geophysical information, but is generally limited in space and time. Satellite and model-derived geophysical estimates can be an appealing alternative to ground measurement due to improved spatiotemporal availability. However, these estimates generally have their own advantages and weaknesses, and there is still room for improvements in their performance. To achieve this, many efforts have been made through de-noising, filtering and merging. Here we focus on merging existing satellite- and model-derived datasets for improved measurement of geophysical variables. Such satellite- and model-derived variables have shown complementarity in their performance in terms of statistical metrics such as bias, root mean square error, and correlation coefficients. This complementary behaviour results from their different skills under different physical/ climatological retrieval conditions. For remote sensing data, the ability to capture temporal variability has been regarded as important in many applications. A linear combination of data sets is a simple but effective way to take the strengths of the original products to achieve better performance. However, a limitation is that calculating optimal weight assigned to each original product during the linear combination needs a truth reference that is often not available and thus hinders their practical applications. The extended triple collocation (ETC) approach does not have this limitation and it can provide error variances and Pearson correlations of the parent products against (hidden) truth without any reference data. The ETC approach has been widely used for uncertainty estimation of environmental variables (e.g. soil moisture, rainfall), but it has not been directly used for data merging, especially for maximizing correlation. The aim of this study is, therefore, to develop a merging approach which can take advantages of both a linear combination and ETC approach. That is, we derive a linear combination method maximising correlation using data-truth correlations derived from ETC. For this, we first present the theoretical background of the proposed method and then verify it through synthetic experiments. In addition to this, we also perform the ETC-based combination scheme using various satellite- and model-derived soil moisture data. Finally, the combination results are compared to ground-based measurements.
© Đại học Đà Nẵng
 
 
Địa chỉ: 41 Lê Duẩn Thành phố Đà Nẵng
Điện thoại: (84) 0236 3822 041 ; Email: dhdn@ac.udn.vn