Multivariate statistical analysis was applied to investigate the dependencies and underlying patterns between N2O emissions and online operational variables (dissolved oxygen and nitrogen component concentrations, temperature and influent flow-rate) during biological nitrogen removal from wastewater. The system under study was a full-scale reactor, for which hourly sensor data were available. The 15-month long monitoring campaign was divided into 10 sub-periods based on the profile of N2O emissions, using Binary Segmentation. The dependencies between operating variables and N2O emissions fluctuated according to Spearman's rank correlation. The correlation between N2O emissions and nitrite concentrations ranged between 0.51 and 0.78. Correlation >0.7 between N2O emissions and nitrate concentrations was observed at sub-periods with average temperature lower than 12 °C. Hierarchical k-means clustering and principal component analysis linked N2O emission peaks with precipitation events and ammonium concentrations higher than 2 mg/L, especially in sub-periods characterized by low N2O fluxes. Additionally, the highest ranges of measured N2O fluxes belonged to clusters corresponding with NO3-N concentration less than 1 mg/L in the upstream plug-flow reactor (middle of oxic zone), indicating slow nitrification rates. The results showed that the range of N2O emissions partially depends on the prior behavior of the system. The principal component analysis validated the findings from the clustering analysis and showed that ammonium, nitrate, nitrite and temperature explained a considerable percentage of the variance in the system for the majority of the sub-periods. The applied statistical methods, linked the different ranges of emissions with the system variables, provided insights on the effect of operating conditions on N2O emissions in each sub-period and can be integrated into N2O emissions data processing at wastewater treatment plants.
- Hierarchical k-means clustering
- Long-term monitoring campaign
- NO emissions
- Principal component analysis