16S rRNA-based Analysis of Microbial Community Changes in Different Floc Size Aggregates in a Full-scale Landfill Leachate Treatment Plant During Nitrogen Removal Process Upgrading

This study utilized 16S rRNA gene analysis to investigate the microbial community dynamics in different floc size aggregates during the nitrogen removal process upgrade in a full-scale Landfill Leachate Treatment Plant (LLTP). The transition from conventional nitrification-denitrification to Partial Nitrification-Denitrification significantly altered microbial community diversity and structure. Neutral Community Model and null model analyses indicated that environmental changes due to nitrogen removal process upgrades and floc size variations shifted the balance between stochastic and deterministic processes. Principal Component Analysis, Partial Least Squares Discriminant Analysis, and non-metric multidimensional scaling emphasized significant segregation of microbial community structures across different floc size aggregates and nitrogen removal processes. Monitoring community information from different floc sizes during the process upgrade revealed that small flocs, with their homogeneity and high mass transfer efficiency, enriched dominant Ammonia-Oxidizing Bacteria while suppressing dominant Nitrite-Oxidizing Bacteria. LDA Effect Size analysis showed that key microbial taxa such as Defluviicoccus, Thauera, Nitrosomonas, Truepera, and Lentimicrobium exhibited significant abundance differences across different floc sizes and denitrification stages, serving as biomarkers for process conversion. Additionally, based on species abundance and environmental factors, a DeepInsight deep learning model combined with convolutional neural networks was developed to predict the plant's denitrification efficiency. This study not only unveiled the evolutionary characteristics and driving mechanisms of bacterial communities in different floc size aggregates during process upgrades but also provided sensitive microbial indicators for identifying process conversion and established a deep learning model to predict denitrification efficiency. These findings offer critical theoretical and practical guidance for optimizing LLTP denitrification processes and address research gaps in process upgrading from a microbial community perspective based on aggregate floc size.