Prediction of denitrification capacity of alkalotolerant bacterial isolates from soil - an artificial neural network model

Abstract

In the past decades, the bioremediation process based on denitrification by aerobic heterotrophic bacteria was extensively studied for different engineering approaches. Besides the fact that only non-pathogenic and non-biofilm forming bacteria must be used, it is very important to isolate bacteria or a group of bacteria in nature with the capacity to remove completely nitrate without accumulation of nitrogen oxides or ammonia as intermediates. In this article, the denitrification capacity of 43 bacterial strains isolated from slightly alkaline and calcite soils along the Danube River were investigated by artificial neural network (ANN) modelling. According to the obtained results, an ANN model was developed for the prediction of denitrification capacity of bacterial soil strains based on six signification denitrification indicators: biomass and N2 gas production, nitrate and nitrite concentration as well as nitrite and ammonia formation. The ANN model showed a reasonably good predictive capability of the outputs (overall R2 for prediction was 0.958). In addition, the experimental verification of the ANN in laboratory testing indicated that the ANN could predict the denitrification capacity of soil bacteria during the denitrification process in laboratory conditions.

Keywords

• nitrogen cycle
• denitrification capacity
• denitrifying soil bacteria
• experimental verification