Predicting organic matter and nutrient removal efficiency in a constructed wetland using data mining

Vertical and horizontal flow constructed wetlands (VF and HF CWs) are most widely used in the treatment of domestic sewage, storm runoff and selected industrial wastewaters. Current research focuses on the search for new applications, alternative substrates and developing mathematical and statistical methods to model the process of removing pollutants in CWs. This is important from a scientific and practical point of view. The performance of VF CWs with two types of substrate (gravel and Certyd - a waste material) was investigated over a two-year period. The following parameters were analyzed: biological oxygen demand (BOD 5 ), chemical oxygen demand (COD), ammonia nitrogen (N-NH 4 ), total nitrogen (N-Total), total phosphorus (P-Total) and dissolved oxygen (DO). At the same time, both the temperature of the inlet sewage and the air temperature were monitored. A support vector machine with a Gaussian kernel function was used to model the treatment efficiency. To select the model parameters, a 10-fold cross-validation was applied. The models applied were characterized by a very good fit. The expected average efficiency of organic removal was 88.3% (BOD 5 ) and 80.8% (COD) in the bed filled with Certyd, and 83.8% (BOD 5 ) and 76.9% (COD) in the bed with a classic filling. The predicted average efficiency of N-NH 4 and N-Total removal in CWs filled with Certyd and gravel were 80.4 and 74.2%, and 59.6% and 57.0% respectively.