Abstract: Understanding the interspecific relationships in harsh environments can reveal how plant species adapt and assemble into communities under disadvantageous conditions. In this study, we used variance tests to detect species’ associations, and used chi-squared tests, the Jaccard index, and Spearman’s rank correlation analyses to analyze interspecies associations and correlations for 18 dominant species from Savanna communities in Yuanmou, a dry-hot valley of the Jinsha River, upstream of the Yangtze River in southwest China. The results showed that: 1) There were significant negative correlations among Savanna grassland plants in Yuanmou. These negative correlations were indicative of strong resource competition or difficult ecological relationships among dominant species; 2) The 18 dominant species were weakly inter-associated, and had a low frequency of common occurrence in plots. Spearman’s rank correlation analyses of the 18 dominant species revealed positive interspecific correlations, with 31 significant or very significant species pairs and 18 very significant species pairs (20.3% and 11.8%, respectively, of a total of 153 species pairs); 3) The 18 dominant species were divided into four specific ecological groups. Among these species, Cymbopogon goeringii, Eulaliopsis binata, Bothriochloa pertusa and widespread species Heteropogon contortus were key band species bundled with other species to form communities in the dry-hot ecosystem in Yuanmou valley. These results suggested that Savanna grassland plants have strong resource competition and/or difficult ecological relationships. The main species showed insolated distribution patterns and the communities were easily disturbed. In practice, species like H. contortus should be prioritized as key species for vegetation restoration, and functional communities should be protected to preserve groups of species that are adapted to the harsh conditions of the hot-dry ecosystem.

Key words: drought, adverse environment, community structure, ecological