Abstract: Community invaded by Praxelis clematidea in karst mountainous area of Pingguo, Guangxi Province was investigated including 4 land types, i.e. abandoned cropland, young plantation, roadside and shrub. Meta analysis, analysis of covariance, linear mixed effect model and recursive partitioning were used to analyze their herb diversity and richness in relation to environmental variables. The results showed that summary effects were positive and significant, effects of groups were positive, effect of abandoned cropland was negative, and effects of other three land types were positive, and shrub had high weight in the analysis. Analysis of covariance indicated that importance value of P. clematidea, distance to driveway and soil were the main numeric factors influencing herb diversity and herb species richness, and plots invaded had higher herb diversity and richness than those uninvaded. Land types had extremely significant impacts on herb diversity but had nothing on herb species richness. Linear mixed effect model partitioned the explained variances of random effects, factor P. clematidea present or absent and residuals respectively contributed 50.9% and 49.1% in Shannon model, while 39.0% and 61.0% in richness model, groups and land types contributed zero to the random effects. P. clematidea invasion could be classified into 5 levels, and modest invasion caused higher herb diversity and richness. When P. clematidea importance value was more than 23.6%, the critical threshold, both herb diversity and richness declined.