Nutritional and environmental effects on biological nitrogen fixation in soybean: A meta-analysis

Biological N2 fixation (BNF) and mineral soil absorption are complementary N sources for soybean growth. BNF is sensitive to nutrient and environmental conditions, though a comprehensive synthesis of the studies underlying these conclusions is currently lacking. Our objective was to conduct a meta-analysis of nutrient and environmental effects on BFN. Articles reporting manipulative treatments and BNF estimations were compiled, and data and metadata were extracted. N fertilization reduced BNF ∼70% relative to the unfertilized control in greenhouse experiments and ∼44% in field experiments. This effect was higher for vegetative than for reproductive stage applications. Fertilization with other nutrients stimulated BNF relative to the unfertilized treatments. Water stress reduced BNF 40% relative to the unstressed control. The negative impact of water stress was larger when stress was applied at vegetative (-70%) compared to reproductive (-30%) stages. Flooding reduced BNF by 40% relative to the non-flooded control with highest effect when applied during vegetative stage (-82%). Increased temperature reduced BNF nitrogenase activity. Carbon dioxide enrichment has a stimulatory effect on BNF. With the exception of N fertilization, soybean and Rhizobium fitness showed a high positive correlation across nutrients and environmental factors. This work summarizes for the first time the impact and relative response of BNF to different nutrient and environmental factors. Results showed an initial critical stage (i.e. vegetative stage) for BNF that is common to most environmental factors negatively affecting BNF. This suggests that common strategies might exist to increase BNF that are independent of the specific identity of the stressor. Also, the positive correlation between host and Rhizobium fitness suggest that strategies to increase BNF potential are the same that those required to increase crop performance. We observed publication bias for some variables, suggesting that effect sizes might be inflated compared to true effect sizes. The only way to solve this problem is to find venues for publication of no significant results when sound experimental approaches and reasonable statistical power can be proved.