Premium
Addressing Spatial Dependence Bias in Climate Model Simulations—An Independent Component Analysis Approach
Author(s) -
Nahar Jannatun,
Johnson Fiona,
Sharma Ashish
Publication year - 2018
Publication title -
water resources research
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.863
H-Index - 217
eISSN - 1944-7973
pISSN - 0043-1397
DOI - 10.1002/2017wr021293
Subject(s) - univariate , grid , climate model , independent component analysis , series (stratigraphy) , representation (politics) , computer science , statistics , precipitation , spatial dependence , general circulation model , data set , climatology , climate change , econometrics , mathematics , multivariate statistics , meteorology , artificial intelligence , geology , geography , paleontology , oceanography , geometry , politics , political science , law
Conventional bias correction is usually applied on a grid‐by‐grid basis, meaning that the resulting corrections cannot address biases in the spatial distribution of climate variables. To solve this problem, a two‐step bias correction method is proposed here to correct time series at multiple locations conjointly. The first step transforms the data to a set of statistically independent univariate time series, using a technique known as independent component analysis (ICA). The mutually independent signals can then be bias corrected as univariate time series and back‐transformed to improve the representation of spatial dependence in the data. The spatially corrected data are then bias corrected at the grid scale in the second step. The method has been applied to two CMIP5 General Circulation Model simulations for six different climate regions of Australia for two climate variables—temperature and precipitation. The results demonstrate that the ICA‐based technique leads to considerable improvements in temperature simulations with more modest improvements in precipitation. Overall, the method results in current climate simulations that have greater equivalency in space and time with observational data.