In this study, we investigate the feasibility of using a blending scheme to combine global coarse-resolution analysis with a high-resolution limited-area model. By regularly introducing large-scale information, we intend to correct accumulated forecast error in a high-resolution model due to continuous data assimilation cycles. The impact of incorporating the blending scheme into three-dimensional variational (3DVAR) and local ensemble transform Kalman filter (LETKF) data assimilation systems are evaluated using a frontal rainband case. In this study, we aim to answer three questions: (1) Can including larger-scale information correct accumulated forecast error and improve model quantitative precipitation forecast (QPF) skill? (2) What is the optimal cut-off length scale (CLS) for the two systems? (3) Which prognostic variables are influential if we want to optimize model QPF skill by applying the blending technique on it? Results from 24 cycles with an hourly update reveal that incorporating the blending scheme successfully mitigates accumulated forecast error and improves model QPF skill for both systems. For this case, 600 km is the optimal CLS that most enhances model QPF. Also, blending large-scale water vapor to correct moisture field plays a key role in this case. Furthermore, the blending scheme imposes a larger impact in 3DVAR than that in LETKF. The reason is that the large-scale information is directly blended into the 3DVAR background which contains more convective-scale features while it is used to recenter the smoother LETKF ensemble mean. This results in smoothed blending analysis and could further aggravate model spin-up time in high-resolution forecast.