Poplar is among the most important deciduous tree species in plantations over the world and has been used as an important model system for molecular genetics of trees. The analysis of codon usage may improve the understanding of the mechanism of codon distribution and variation in poplar and the understanding of factors shaping the codon usage patterns. Here, an EN(c) (effective number of codons)-plot method and multivariate statistical method called correspondence analysis (COA) were used to examine the codon usage of 314 genes of poplar. The results show that the main trend was highly negative correlated with the gene expression level assessed by the ''Codon Adaptation Index'' value. Moreover, there were two significant correlations between axis 1 coordinates and GC3(s) content and gene length, we infer that gene nucleotide composition and gene length also play an important role in shaping the codon usage bias in poplar. The result of relative synonymous codon usage (RSCU) analysis shows a high bias of codon usage toward the codon with A or T ending. In addition, we compared the codon preferences among poplar, Arabidopsis thaliana, Oryza sative, Homo sapiens and Escherichia coli, and poplar was found to be most similar to A.thaliana and least similar to E.coli. In this paper, 10 codons defined firstly as optimal codons through an analysis of the high-expression codon in poplar may provide some useful information for genetic engineering of poplar.