Advances in Earth Science ›› 2011, Vol. 26 ›› Issue (8): 837-847. doi: 10.11867/j.issn.1001-8166.2011.08.0837

• Articles • Previous Articles     Next Articles

A Summary of Methods for Statistical Downscaling of Meteorological Data

Liu Yonghe 1, Guo Weidong 2, Feng Jinming 3, Zhang Kexin 4   

  1. 1.Institute of Resources and Environment, Henan Polytechnic University, Jiaozuo454000,China;2.ICGCR, School of Atmospheric Sciences, Nanjing University, Nanjing210093,China;3.Key Laboratory of Regional ClimateEnvironment Research for Temperate East Asia, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing100029,China;4.Linyi Meteorological Bureau,  Linyi276004,China
  • Received:2011-02-14 Revised:2011-04-14 Online:2011-08-10 Published:2011-08-10

Liu Yonghe, Guo Weidong, Feng Jinming, Zhang Kexin. A Summary of Methods for Statistical Downscaling of Meteorological Data[J]. Advances in Earth Science, 2011, 26(8): 837-847.

As one of the means for bridging the gap of the data between low resolution data obtained from weather models and that needed in basin scale, statistical downscaling become a important field for study. The approaches for statistical downscaling are abundant and can be roughly divided into three groups: Transfer functions, weather-typing approaches and stochastic weather generators. The transfer functions may be linear methods, such as multivariate linear regression, canonic correlation, and Singular Value Decomposition, or nonlinear methods like artificial neural networks and support vector machine. Weather generators are designed initially for generating the missing weather data, but in downscaling applications it can be used as the output backend of other statistical downscaling techniques. The weather-typing approaches can be regarded as some variant of weather generators combining with some transfer functions or classifications. Thus, no strict boundaries exist among the three groups. The statistical downscaling problems have some attributes involved with temporal downscaling or spatial downscaling, stochastic downscaling or deterministic downscaling, temporal self-correlation and spatial correlation, point-site oriented or grids oriented. The differences of downscaling performance between all techniques are mainly related to these attributes. In recent years, the analog method, weather classifying, hidden Markov modeling, generalized linear models, Poisson clustered point process and the multiplicative cascade process based on multifractal theories are developed and used for statistical downscaling application, and many new nonlinear methods such as generalized additional models and physical-statistical methods are arising. Meanwhile, there is also some widely used model software available. Among all new emerging techniques, the nonlinear methods, stochastic simulation techniques of climate scenario, downscaling models for shortterm weather prediction and the statistical downscaling methods combining physical mechanism may become the main trend of study in the future.

No related articles found!
Full text