Identification of primary effecters of N2O emissions from full-scale biological nitrogen removal systems using random forest approach

Emerging investigator series: modeling of wastewater treatment bioprocesses: current development and future opportunities

“…116 In the wastewater treatment field, the random forest algorithm has been implemented for activated sludge processes, anaerobic digesters, membrane bioreactors, and anammox processes. [117][118][119][120] The main use of random forestbased models includes the prediction of system performance, fault finding, big data handling, model comparisons, and exploration of datasets with applicable reservations and constraints. 121 Although random forest-based models, similar to other data-driven models, are not able to integrate biological principles, these models allow for the identification of key features and conditions that are most influential on the process.…”

“…Song et al implemented this modeling approach with wastewater treatment inputs as multivariate datasets to predict N 2 O emission from the aerated zones of activated sludge processes. 118 Based on the model inference, they identified inorganic carbon concentration and specific ammonia oxidation activity as two of the dominant factors that determined treatment performance. 118 The model was further used to identify the different mechanisms of N 2 O generation in oxic and anoxic environments and demonstrated the key role of N 2 O in those zones in promoting niche-specific biochemical reactions.…”

“…118 Based on the model inference, they identified inorganic carbon concentration and specific ammonia oxidation activity as two of the dominant factors that determined treatment performance. 118 The model was further used to identify the different mechanisms of N 2 O generation in oxic and anoxic environments and demonstrated the key role of N 2 O in those zones in promoting niche-specific biochemical reactions. 122 Random forest can be combined with other algorithms such as principal component analysis (PCA) and neural networks to improve the prediction of effluent quality.…”

“…118 The model was further used to identify the different mechanisms of N 2 O generation in oxic and anoxic environments and demonstrated the key role of N 2 O in those zones in promoting niche-specific biochemical reactions. 122 Random forest can be combined with other algorithms such as principal component analysis (PCA) and neural networks to improve the prediction of effluent quality. Preprocessing of data using PCA could enhance the robustness of random forest-based models, leading to a more accurate prediction of membrane flux in membrane bioreactors as compared with neural networks.…”

Linking seasonal N2O emissions and nitrification failures to microbial dynamics in a SBR wastewater treatment plant

“…However, these factors are not exclusive and could only partly explain emission patterns assessed in long-term monitoring campaigns ( Vasilaki et al., 2019 ). Statistical regression algorithms and mechanistic process modeling based on the activated sludge modeling framework have been applied with limited success to model N 2 O emissions from WWTP ( Ni and Yuan, 2015 ; Song et al., 2020 ; Vasilaki et al., 2018 ). Thus, to better understand the N 2 O emissions from WWTP and identify relevant mechanisms, new aspects may have to be taken into account.…”