This paper studies, in a survey sampling framework with unequal probability sampling designs, three nonparametric kernel estimators for the mean curve in presence of discretized trajectories with missing values. Their pointwise variances are approximated thanks to linearization techniques.