Fourier analysis as a forecasting tool for absorbance time series received by UV-Vis probes installed on urban sewer systems

The objective of this work was to apply the Discrete Fourier Transform DFT to time series of waste water UV-Vis absorbance spectra for forecasting purposes. Three time series for three different study sites were used: i) Salitre WWTP in Bogotá, ii) Gibraltar pumping station in Bogotá, iii) San Fernando WWTP in Itagüí (south part of Medellín). Each of these time series had equal number of samples (1051), with an acquisition period of one sample per minute during approximately 18 hours. A specific methodology for forecasting UV-Vis time series from DFT is proposed, considering site specific behaviours of water quality and target pollutants to be monitored. Results obtained show that: (i) for all the time series studied, forecasting of UV part of the spectrum is more feasible than the visible part; (ii) forecasting on visible spectrum (near UV) is more feasible for San Fernando WWTP than for the other study sites; (iii) the recommended number of components varies as a function of the forecasting period; (iv) minimum forecasting errors and maximum forecasting periods vary according to time series and target pollutants.