The purpose of this article is to understand the effect of multi-temporalmulti-angle data on vegetation community type mapping in desert regions. Based on data from the multi-angle imaging spectroradiometer (MISR), a set of 46 multi-temporal classification experiments were carried out in the Jornada Experimental Range in New Mexico, USA. Besides multi-angle observations, bidirectional reflectance distribution function (BRDF) model parameters were also used as input data for the classifications. The experiments used two widely accepted BRDF models, the Rahman-Pinty-Verstraete (RPV) model and the Ross-thin Li-sparse reciprocal (RTnLS) model. The experiments show that multi-temporal multi-angle classifications can yield a more accurate mapping than multi-temporal nadir classifications, and multi-temporal BRDF model parameters combined with a single nadir image can provide an accuracy roughly the same as all multi-temporal multi-angle observations for the vegetation mapping. These findings opened not only a path of reducing data dimensionality for multi-temporal multi-angle classifications, but also a way of merging products of both MISR and moderate resolution imaging spectroradiometer (MODIS) to improve semi-arid vegetation mapping.