AbstractIn order to explore the correlations between desert plant community types and soil variables at the southern edge of Gurbantunggut Desert, we established 48 quadrats of 20 m × 10 m to study the vegetation by the randomized systematic method in Qitai County of Xinjiang in 2005. After vegetation survey, we selected 20 plots with obviously different physiognomy and sampled soils at 0, 10, 30, 50 and 80 cm depths, respectively. Plants and soil characters were analyzed simultaneously and the results were correlated. A hier-archical cluster and principle component analysis (PCA) were performed based on importance value of plant species and soil variables, respectively. Four types of communities were recognized and described in details, i.e. Phragmites australis + Achnatherum splendens community, Haloxylon ammodendron+Reaumuria son-garica+Nitraria sibirica community, Haloxylon ammodendron+Ceratoides ewersmanniana+ Seriphidium terrae-albae community, and Ceratoides ewersmannian+Alhagi sparsifolia community. The results of prin-ciple component analysis showed that there was a specific relationship between soil variables and plant community type. Three principal components, namely pH, organic matter content, and soil salinity, could ex-plain 85.65% of the differences among the soils of different plant communities.