Serious obstacles exist for the continuous cropping of watermelon (Citrullus lanatus). It has previously been shown that following application of living Ceratobasidum stevensii B6 mycelia, the quantity of Fusarium oxysporum (FO) in continuously cropped soil at the harvest stage decreased by 29.9%, while the total yield and soluble sugar content of the fruit increased by 103.8% and 35.1% compared with the control. Here we further discuss the mechanisms by which B6 may alleviate watermelon cropping problems. We developed a real-time polymerase chain reaction (PCR) assay to detect and quantify B6 in soil, and monitored the dynamics of the FO population in the soil. We also evaluated the effects of different fermentation components on the soil microflora, using Denatured Gradient Gel Electrophoresis (DGGE). B6-specific DNA primers were designed based on the internal transcribed spacer (ITS) sequence. Amplification of B6 DNA using Cf1/Cr1 primers yielded a single 371 bp-long product; no product was obtained in the other ten fungal species tested. The detection limit of the system was 100 fg/μL of B6 genomic DNA. To replicate the watermelon cropping system as closely as possible, a surface layer (5-20 cm depth) of an Ultisol was sampled from a 3-year old continuously cropped upland watermelon site, at the Red Soil Experimental Station, Chinese Academy of Agricultural Sciences, Yongzhou, Hunan Province (N26°45', E111°53'). Pot experiments were conducted in the botanical garden of Nanjing Normal University; these were a control (A), an inactivated B6 liquid broth (B), an activated B6 liquid broth (C), an inactivated B6 solid fermentation (D), and an activated B6 solid fermentation (E). When the solid fermentation of B6 was inoculated into soil, the level of genomic DNA reached its highest at 7.4 log (pg/g dry soil) after one week, but decreased below the detection limit after 5 weeks. When liquid broth of B6 was inoculated into soil, the level of genomic DNA decreased immediately, and was below the detection limit after 4 weeks. B6 was not detected in other treatments, meaning that B6 was low in the natural environment. The population of FO was effectively controlled by treatments C and E during the first 4 weeks (FO was about 5×10³ CFU/gDM), and then from the fifth week FO increased rapidly. DGGE fingerprinting of bacterial cluster analysis showed that treatments A and B clustered in one class (with 57.5% similarity), and treatments C, D, and E clustered in another class (with 55.2% similarity). DGGE fingerprinting of fungal cluster analysis showed that treatments A, B, and C clustered in one class (with 66.0% similarity). Treatments D and E exhibited significant differences. Except for treatments C and E, there was little difference between the bacterial or fungal Shannon diversity indices, or evenness between the treatments. Our results suggest that B6 survived in the soil for about one month under these experimental conditions, and that the strain tested was environmentally friendly and interfered little with the soil microenvironment.