A Quantile-based Matching (QM) method has been widely used to correct the biases in global and regional climate model outputs. The basic idea of QM is to adjust the Cumulative Distribution Function (CDF) of model for the projection period on the basis of the difference between the model and observation CDFs for the training period. Therefore, the CDF of observation on training period plays an important role in quantile-based matching. Also, ensembles are highly correlated because ensemble forecasts generated from a combination of randomly perturbed initial conditions and different convective schemes in numerical weather model. We discuss the dependence of the bias correction results obtained from Qunatile-based Matching when there is correlation between ensembles and the variance of observation is larger than that of model. A simulation study is employed to understand the relation and distributional characteristics of observation and model when applying Quantile-based Matching method.