This document summarizes an approach for uncertainty-aware multidimensional projection of ensemble data. The key contributions are a new dissimilarity measure between ensemble data objects based on both mean distance and distribution distance, and an enhanced Laplacian-based projection scheme. The approach first estimates probability distributions for each ensemble data object using kernel density estimation. It then projects a subset of control points using MDS and interpolates the remaining points based on nearest neighbors. Visualization widgets allow exploration of projection results and uncertainty quantification. The approach is demonstrated on synthetic, NBA player statistic, and weather simulation datasets.