In this paper we use density estimation to investigate the particle size distribution in surface water and to extract the components of density functions obtained. This distribution helps to investigate quality of water with respect to its physical characterization. The analysis is conducted using the two-step density estimation method. The first step of the method is based on parametric density estimation and uses a modification of the EM algorithm to a normal mixture model. In the second step we apply the generalized kernel density estimator, being the convex combination of the well-known kernel density estimators.