Large-scale molecular dynamics simulations are used to simulate a layer of nanoparticles floating on the surface of a liquid. Both a low viscosity liquid, represented by Lennard-Jones monomers, and a high viscosity liquid, represented by linear homopolymers, are studied. The organization and diffusion of the nanoparticles are analyzed as the nanoparticle density and the contact angle between the nanoparticles and liquid are varied. When the interaction between the nanoparticles and liquid is reduced the contact angle increases and the nanoparticles ride higher on the liquid surface, which enables them to diffuse faster. In this case the short-range order is also reduced as seen in the pair correlation function. For the polymeric liquids, the out-of-layer fluctuation is suppressed and the short-range order is slightly enhanced. However, the diffusion becomes much slower and the mean square displacement even shows sub-linear time dependence at large times. The relation between diffusion coefficient and viscosity is found to deviate from that in bulk diffusion. Results are compared to simulations of the identical nanoparticles in 2-dimensions.