Predictive microbial-based modelling of wheat yields and grain quality across a 500km transect in Quebec

Crops yield and quality are difficult to predict using soil physico-chemical parameters. Because of their key roles in nutrient cycles, we hypothesized that there is an untapped predictive potential in the soil microbial communities. To test our hypothesis, we sampled soils across 67 wheat fields of the province of Quebec at the beginning of the growing season in May-June. We used a wide array of methods to characterize the microbial communities, their functions, and activities, including: 1) amplicon sequencing, 2) real-time PCR quantification, and 3) community-level substrate utilization. We also measured grain yield and quality at the end of the growing season, and key soil parameters at sampling. The diversity of fungi, the abundance of denitrification and nitrification genes, and the use of specific organic carbon sources were often the best predictors for wheat yield and grain quality. Using less than 15 microbial parameters, we were able to explain 55-93 % of the variation in wheat yield and grain quality across the province of Quebec. Microbial-based regression models outperformed basic soil-based models for predicting wheat quality indicators. Taken together, our results suggest that the measurement of microbial parameters early in the season could help predict accurately grain quality and quantity.