Spectral images contain a large volume of data and can be efficiently compressed by low dimensional linear models. However, there is a trade-off between the accuracy of spectral and colorimetric representation. When a spectral image is reproduced by a low-dimensional linear model, spectral error and color difference are contrary to each other and minimizing the colour error is by no means equivalent to minimizing the spectral error. Although one aim of a spectral-image file format is to preserve and represent the spectral information, most users are likely to reproduce a spectral image on a trichromatic image-reproduction system and therefore it is important that the spectral information is not preserved at the expense of colorimetric accuracy. In this study a method for spectral encoding that provides an efficient representation of the spectral information whilst perfectly preserving the colorimetric information is analysed. The lossy compression technique that is considered in this work is based on a low-dimensional linear model of spectral reflectance, with the first three basis functions represent color information and the additional basis functions are metameric blacks which preserve spectral information.