Predicting soil farming system and attributes based on soil bacterial community

Abstract

Soil of agro-ecosystems consists of complex biological interactions directly shaped by human activity, accounting for the overall soil quality. Bacterial communities are key players in most biogeochemical cycles. Here we investigated soil bacterial communities in four adjacent maize farming systems, conventional, transition, natural, and organic in order to study the effects of these practices on soil bacterial community, uncover most influential bacterial taxa and explore them as predictors of categorical and continuous soil variables (i.e. attributes). While richness and diversity were only marginally affected, we found pronounced farming system effects on bacterial community composition. The organic farming system exhibited higher bacterial community stability over two crop years. Conventional farming network exhibited the highest density and number of keystone taxa, however all keystone taxa were shared by one module only. Whilst organic farming presented the biggest and less dense network with the smallest number of keystone taxa. Blastocatellales was one of the most influential orders in organic and natural farming system networks. Bacterial taxa associated to the farming systems were visualized by a bipartite graph and used to build random forest models capable of predicting the farming system regardless of the crop year with high accuracy. The models were also able to predict soil chemical and biological attributes to a different extent. Altogether, we have shown that soil bacterial communities are deeply shaped by the underlying farming system, and can be used as a tool in sustainable farming monitoring.