Yields and soil properties
The rice yield of each treatment showed a trend of GMCF > CF > GM > NF (Fig. 1). Compared with NF treatment, the rice yields significantly increased in GMCF, CF and GM treatments by 161.2%, 130.6% and 77.2%, respectively. Hunan province had the highest yield, the average yield was 10950 kg ha− 1, followed by Henan, Jiangxi and Fujian provinces (average yields of 9483, 9090 and 7904 kg ha− 1, respectively) (Fig. 1). Rice yields significantly increased in GMCF treatment compared with CF treatment in Jiangxi province. The rice yields of GM, CF and GMCF treatments were significantly higher than NF treatment in Henan province.
Soil chemical properties showed various trends under different treatments at the four sites (Table 4). In Hunan province, soil pH was increased in CF and GMCF, compared with NF and GM treatments, and SOM was increased by GM, compared with other treatments. In Jiangxi province, GM treatment significantly increased NO3−-N, compared with other treatments. In Henan province, SOM and TN were increased in GMCF treatment, but pH was reduced compared with CF treatment. In Fujian province, NH4+-N was increased in CF and GMCF, compared with other treatments. AP was increased by CF and GMCF, compared with NF and GM treatments in Hunan, Jiangxi and Henan provinces (Table 4).
The functional genes involved in N cycling
The copy numbers of nifH gene had no significant difference among treatments in Hunan and Fujian provinces. In Jiangxi province, the copies of nifH gene were increased significantly in GMCF, compared with NF treatment. The copies of nifH gene were increased significantly in GMCF, compared with CF and NF treatments in Henan province (Fig. 2A). The abundance of nifH gene varied in different sites (Fig. 2B), Henan province had the highest copy numbers of nifH gene, the average copy number was 3.70×1010 copies per gram dry soil, followed by Hunan, Fujian and Jiangxi provinces (average numbers of 2.26×1010, 9.19×109 and 7.86×109 copies per gram dry soil, respectively).
The copy numbers of AOA-amoA and AOB-amoA genes had similar trends under different treatments at the four sites (Fig. 3). The copy numbers of AOA-amoA gene increased in CF treatment, compared with NF and GM treatments in Hunan province, and no significant difference was found among treatments of other sites (Fig. 3A). The AOA-amoA abundances were highest in Henan province and lowest in Jiangxi province; no significant difference was found among sites (Fig. 3B). The copy numbers of AOB-amoA gene were increased by CF treatments in Jiangxi province and decreased by GMCF in Fujian province, compared with other treatments (Fig. 3C). AOB-amoA gene abundance had no significant difference among sites (Fig. 3D).
The ratios of AOA-amoA to AOB-amoA gene ranged from 2.70 to 63.25, indicating that AOA-amoA was more abundant than AOB-amoA in the studied paddy soils. AOA-amoA to AOB-amoA gene ratios were increased by GMCF treatment in Hunan, Jiangxi and Fujian provinces (Fig. 4). The average value of AOA-amoA to AOB-amoA gene ratios varied among sites. Fujian had the highest AOA-amoA to AOB-amoA ratios (22.33), followed by Hunan (18.99), Jiangxi (7.60) and Henan (4.93) provinces (Fig. 4).
The nirK and nosZ genes were selected to reflect the denitrification process. The copy numbers of nirK gene increased in GMCF treatment in Hunan province and increased in CF treatment in Henan and Fujian provinces, compared with NF (Fig. 5A). At the four sites, the nirK gene was more abundant in Henan and Fujian provinces than that in Hunan and Jiangxi provinces (Fig. 5B). The copy numbers of nosZ gene had no difference in Hunan, Henan and Fujian provinces. In Jiangxi province, the application of green manure increased the nosZ gene copies (Fig. 5C). The highest average copy number of nosZ gene among the four sites was found in Henan province, i.e., 7.81×107 copies per gram dry soil, and was significantly higher than that in Jiangxi and Fujian provinces (2.97×105 and 1.34×107 copies per gram dry soil, respectively) (Fig. 5D).
The ratios of nosZ to nirK gene copies ranged from 1.14 to 673.85, varied a lot among the treatments and sites (Fig. 6). In all the four provinces, the nosZ to nirK ratios were decreased in GMCF treatment, compared with NF. The average value of nosZ to nirK ratio in Hunan province was 503.75, significantly higher than other sites. The average values in Jiangxi, Henan and Fujian provinces were 61.48, 16.08 and 3.47, respectively.
Correlations among the functional genes involved in N cycling and soil properties
Pearson’s correlation coefficients were calculated to explore the relationships among the five functional genes involved in N cycling and the correlations between functional genes and soil properties (Fig. 7). The nirK and nosZ genes in denitrification were positively related to nifH gene and AOA-amoA and AOB-amoA genes, indicating that denitrification was closely related to N-fixation and nitrification process. The archaeal and bacterial amoA genes had significantly positive correlations with each other, and the same as nirK and nosZ genes (Fig. 7).
The nifH and nosZ genes were negatively correlated with soil NH4+-N content, and positively correlated to AK content. The nirK gene in denitrification had significantly negative correlations with soil SOM, TN, NH4+-N, NO3−-N and AP content (Fig. 7).
Relative influences of environmental factors on rice yield
The relative influences of the functional genes and soil chemical properties on rice yield were evaluated based on the aggregated boosted trees analysis (ABT) (Fig. 8). The nirK gene in denitrification was the most crucial factor influencing rice yield, accounting for 63.5% of the total influence. The nifH gene accounted for 14.1%, and the other three genes together contributed only 22.4% to rice yield. SOM had the most important factor contributing to rice yield among soil chemical properties, with a contribution rate of 41.0%, followed by TN (26.1%). The contribution of NH4+-N, AP, NO3−-N, AK and pH to rice yield ranged from 4.9–7.7%.