It is important to know the comprehensive microbial communities of fecal pollution sources and receiving water bodies for microbial source tracking. Pyrosequencing targeting the V1-V3 hypervariable regions of the 16S rRNA gene was used to investigate the characteristics of bacterial and Bacteroidales communities in major fecal sources and river waters. Diversity analysis indicated that cow feces had the highest diversities in the bacterial and Bacteroidales group followed by the pig sample, with human feces having the lowest value. The Bacteroidales, one of the potential fecal indicators, totally dominated in the fecal samples accounting for 31%-52% of bacterial sequences, but much less (0.6%) in the river water. Clustering and Venn diagram analyses showed that the human sample had a greater similarity to the pig sample in the bacterial and Bacteroidales communities than to samples from other hosts. Traditional fecal indicators, i.e., Escherichia coli, were detected in the human and river water samples at very low rates and Clostridium perfringens and enterococci were not detected in any samples. Besides the Bacteroidales group, some microorganisms detected in the specific hosts, i.e., Parasutterella excrementihominis, Veillonella sp., Dialister invisus, Megamonas funiformis, and Ruminococcus lactaris for the human and Lactobacillus amylovorus and Atopostipes sp. for the pig, could be used as potential host-specific fecal indicators. These microorganisms could be used as multiple fecal indicators that are not dependent on the absence or presence of a single indicator. Monitoring for multiple indicators that are highly abundant and host-specific would greatly enhance the effectiveness of fecal pollution source tracking.