We present a non-parametric method for compressing multichannel audio data for reproduction through Wave Field Synthesis. The method consists of applying a two-dimensional filterbank to the input mul- tichannel signal, in both time and channel dimensions, and coding the two-dimensional spectra using a spatio-temporal frequency masking model. The coded spectral data is organized into a bitstream together with side information containing scale factors and Huffman codebook information. We demonstrate how this coding method can be applied to any smooth distribution of loudspeakers in space, while obtaining a stable bitrate that is 15% lower compared to coding each channel independently.