Abstract: Quantitative classification (TWINSPAN)and ordination (DCA) methods were used to study the gobi deserta in Anxi, and the stepwise vegression method was used to decide the interrelatire equations among the communities and the environmental factors. The results indicated that the gobi deserta could be divided into 6 types of plant communities in this region. These types were a closely related to the soil contents of total salt ( S), C1- and SO2-4 and the annual precipitation. They were distributed according to the environmental gradients from low to high salt content and from arid to extreme arid. The position of 60 sample plots in DCA ordinational first axis ( Xl ) significantly correlated with the environmental factors, viz. the soil contents of salt, C1- and SO2-4. And that in the second axis of the DCA ordination (X2) correlated well with the annual precipitation. Their relationships are expressed in the following equations: 1/Xl = 2.199 2 + 0.022 5/S – 0.018 8/Cl –0.0567/S0 (R = 0.797,P < 0.01);X2 = –82.8561 + 117.921 9 logR (R = 0.817, P <0.01).

Key words: Deserta, Multivariate analysis, Stepwise regression