
For assessing future climate change impact, bias correction of GCMs is necessary to reduce the gaps between modeling results and reality. Many bias correction methods have been developed for implementing GCMs in hydrologic modeling. Recently, joint variable bias correction is introduced and this method is not fully reviewed before, in this study we assessed joint variable bias correction method comparing to CDF mapping, one of most common bias correction methods, as a univariate bias correction. Precipitation and evapotranspiration estimated by PenmanMonteith method from NLDAS2 (the forcing data for Phase 2 of the North American Land Data Assimilation System) were used as references and nine GCMs of CMIP5 (Coupled Model Intercomparison Phase 5) were bias corrected with CDF mapping (univariate bias correction) and joint variable bias correction. Joint variable bias correction was conducted with four different ways: (1) matching probability of wetday of simulation data (GCMs) to probability of wetday of reference data (NLDAS2) in probability space, (2) matching probability of wetday in physical space, (3) without consideration of matching probability of wetday in probability space, and (4) without consideration of matching probability of wetday in physical space. Matching probability of wetday and bias correction in probability space reduce biases of precipitation and evapotranspiration the most among four different ways, however CDF mapping reduced biases better than joint variable bias correction. In addition, CDF mapping reproduced joint CDF of precipitation and evapotranspiration better than joint variable bias correction did. 