Water Quality Measurements in the Brantas, Indonesia: Constructing Different Perspectives with Principal Component Analysis

Abstract

The water quality of the Brantas river in Indonesia is of concern to several agencies on East Java. These agencies all measure its water quality in their own way in terms of locations, rhythms and parameters. The goal of this thesis is to find out if these agencies measure the same and if not, how these measurements differ. From these measurements, perspectives are constructed for each agency with the use of Principal Component Analysis. The agencies investigated are the Dinas Lingkungan Hidup Jawa Timur (DLH Jatim), Belai Besar Wilayah Sungai Brantas (BBWS) and Perum Jasa Tirta I (PJT). As an addition to the PCA, a neural network model is constructed and trained to recognize the measurement agency of a datapoint from the measurement values. It was found that the all three agencies recognized oxygen as a dominant driver in water quality processes. Secondary processes were mostly driven by rainfall, but the effect of this was seen differently by the agencies. DLH Jatim distinguishes surface waste runoff separately from rainfall, while Perum Jasa Tirta I will see them as inherently connected. BBWS will not recognize the surface waste runoff process as a significant factor in the water quality. These differences found in the representation of core processes in the Brantas outline how different agencies can have a different perspective on water quality. This was further underlined by the conclusions from the neural network analysis. Here it was found that the author could be recognized from the measurement values alone on 88% of agency data.