Advances in projection of climate change impacts using supervised nonlinear dimensionality reduction techniques
详细信息 查看全文
- 作者：Ali Sarhadi ; Donald H. Burn ; Ge Yang ; Ali Ghodsi
- 关键词：Dimensionality reduction ; Statistical downscaling ; Supervised Principal Component Analysis (S ; PCA) ; Climate change ; Supervised learning
- 刊名：Climate Dynamics
- 出版年：2017
- 出版时间：February 2017
- 年：2017
- 卷：48
- 期：3-4
- 页码：1329-1351
- 全文大小：
- 刊物类别：Earth and Environmental Science
- 刊物主题：Geophysics/Geodesy; Climatology; Oceanography;
- 出版者：Springer Berlin Heidelberg
- ISSN：1432-0894
- 卷排序：48
文摘
One of the main challenges in climate change studies is accurate projection of the global warming impacts on the probabilistic behaviour of hydro-climate processes. Due to the complexity of climate-associated processes, identification of predictor variables from high dimensional atmospheric variables is considered a key factor for improvement of climate change projections in statistical downscaling approaches. For this purpose, the present paper adopts a new approach of supervised dimensionality reduction, which is called “Supervised Principal Component Analysis (Supervised PCA)” to regression-based statistical downscaling. This method is a generalization of PCA, extracting a sequence of principal components of atmospheric variables, which have maximal dependence on the response hydro-climate variable. To capture the nonlinear variability between hydro-climatic response variables and projectors, a kernelized version of Supervised PCA is also applied for nonlinear dimensionality reduction. The effectiveness of the Supervised PCA methods in comparison with some state-of-the-art algorithms for dimensionality reduction is evaluated in relation to the statistical downscaling process of precipitation in a specific site using two soft computing nonlinear machine learning methods, Support Vector Regression and Relevance Vector Machine. The results demonstrate a significant improvement over Supervised PCA methods in terms of performance accuracy.