Summary form only given. The scaling of statistical fluctuations due to two particle-in-cell loading schemes is examined. A global loading scheme inverts the spatial distribution function over the entire system length, while a local loading scheme inverts the spatial distribution independently within each cell. The particle density is then interpolated to grid nodes, and the fluctuations are quantified. The scaling with particle number, N, is the familiar N-1/2, with the local scheme having a slightly smaller coefficient. The scaling of the fluctuation level with the number of cells, J, is J-3/2 for non-uniform distributions due to the filtering errors generated by interpolation to the mesh. The noise fluctuation level shows a minimum for a fixed number of particles, indicating that there exists an optimum choice for the number of cells