Time series analysis of water pollution data

Abstract

Aim of this paper is to apply time series analysis methods to investigate water pollution in Ukraine. We are using datasets provided by the Ukrainian government (State Water Resources Agency of Ukraine) which contain information about biochemical oxygen demand (BOD), ammonium ions concentration and phosphate ions concentration in river water. Values of concentration are measured at eight consequent water stations. The original data contest was aimed to predict concentration level at one of the stations using values at the other ones. Time series values have got periods of missing data. First of all, these missing parts need to be calculated. Spline functions have been used to solve this task. In the paper mathematical models describing behavior of concentration and BOD level time series at the target station are built. Also, models aimed to describe dependence of concentration at the target station on levels at the other stations are constructed and investigated. Influence of pollution values at intermediate stations on levels at the target stations are investigated with statistical tests. The proposed technique can be used to investigate pollution at plants in other domains of industry and to handle missing values in various time series.