Rapid assessment of surface water quality using statistical multivariate analysis approach: Oder River system case study

Abstract

Many physicochemical and biological monitored parameters must be taken into consideration to fully evaluate the surface water environmental condition. However, there are situations where a simple and rapid assessment of the poor water quality situations is critically important. This work presents a universal methodology for monitoring of many parameters simultaneously and early detection out-of-control samples in a real-time mode. The approach uses multivariate statistical quality control chart based on Principal Component Analysis (PCA) model with two well-known measures of abnormal behaviour in a process or system: Hoteling's T2 statistics and Q-statistic. The proposed TQ_PCA quality index provides on-line assessment of the water sample quality, with no specific knowledge and assumptions about control limits for monitored parameters required. A water sample is assessed through the simple control chart using the PCA model established for training/reference samples. The power of the proposed index has been tested using water quality data from the Oder River, including the time of the largest ecological disaster in recent European river history. The proposed index showed excellent analysis performance for physicochemical water quality dataset from Polish stations and physicochemical and biological water quality dataset from German/Frankfurt station, confirming earlier reports. There were consecutive number of alarms reported by the statistical index, a month prior to the disaster when there were no evident changes in the individual parameters. The method presented in this study demonstrated capability of assessment of the major water quality parameters, whose changes preempt the uncommon event. The presented TQ_PCA index could be easily extended to any research involving a large dataset of monitoring parameters from any industrial chemical process.