Multivariate methods in modelling the removal of nitrogen and phosphorus in deep-bed two-media down-flow sand filters

Abstract

Multiple linear regression (MLR) was applied to study the operation and efficiency of simultaneous denitrification with sodium acetate (NaAc) as a carbon and energy source and the chemical precipitation of phosphate with iron salts in a pilot sand filter. Different loads were simulated with the addition of primary settled wastewater (PW) containing suspended solids (SS) and biological SS from the aeration tank. SS from PW partly passed the filter. Biological SS was caught by the filter. The validation of the model showed that phosphorus concentration and chemical oxygen demand reduction were easier to predict correctly than the time of operation of the filter. The reduction of total phosphorus differed between 0.01 and 0.31 mg P/L, and the nitrate concentration in the filtrate differed between 0 and 5.2 mg N/L when comparing the experimental values with the model. Experiments showed filtrate concentrations of 0–0.12 mg PO4-P/L, 3.0–8.8 mg NO3-N/L, and 1.9–34 mg SS/L. The addition of NaAc gave a small increase in the reduction of PO4-P probably because of the assimilation of PO4-P into the denitrifying sludge. The dosage of Fe gave a small decrease in the reduction of NO3-N probably because of PO4-P limitation for denitrifying bacteria, as Fe acts as a precipitation agent upon PO4-P. HIGHLIGHTS Multiple linear regression (MLR) was applied to study processes in sand filters.; Different loadings and removal of suspended solids, organic material, and phosphorus were evaluated.; The models were found to well describe the responses of operational cases.; Suspended solids were removed more efficiently from secondary than primary settled wastewater.; The dosage of iron salt decreased the reduction of NO3-N due to PO4-P limitation.;

Keywords

• mlr
• multiple linear regression
• nutrient removal
• operational time
• sand filter
• validation