An ANN application for water quality forecasting

Abstract

Rapid urban and coastal developments often witness deterioration of regional seawater quality. As part of the management process, it is important to assess the baseline characteristics of the marine environment so that sustainable development can be pursued. In this study, artificial neural networks (ANNs) were used to predict and forecast quantitative characteristics of water bodies. The true power and advantage of this method lie in its ability to (1) represent both linear and non-linear relationships and (2) learn these relationships directly from the data being modeled. The study focuses on Singapore coastal waters. The ANN model is built for quick assessment and forecasting of selected water quality variables at any location in the domain of interest. Respective variables measured at other locations serve as the input parameters. The variables of interest are salinity, temperature, dissolved oxygen, and chlorophyll-a. A time lag up to 2Δt appeared to suffice to yield good simulation results. To validate the performance of the trained ANN, it was applied to an unseen data set from a station in the region. The results show the ANN’s great potential to simulate water quality variables. Simulation accuracy, measured in the Nash–Sutcliffe coefficient of efficiency (R²), ranged from 0.8 to 0.9 for the training and overfitting test data. Thus, a trained ANN model may potentially provide simulated values for desired locations at which measured data are unavailable yet required for water quality models.

Introduction

Seawater is a primary natural resource for many coastal development sectors. It is also one of the most sensitive and vulnerable resources, because it is negatively impacted by a variety of anthropogenic activities. Deterioration of coastal water quality has triggered the initiation of serious management efforts in many countries. Tourism, recreation, and fishing require at least an acceptable level of seawater quality. Singapore, an island state, has stringent enforcement and continuously upgraded waste handling facilities to guarantee seawater of a certain quality.

Most acceptable ecological and social decisions are difficult to make without careful modeling, prediction/forecasting, and analysis of seawater quality for typical development scenarios. Water quality prediction enables a manager to choose an option that satisfies a large number of identified conditions. For instance, water quality variables, such as salinity, temperature, nutrients, dissolved oxygen (DO), and chlorophyll-a (Chl-a), in coastal water describe a complex process governed by a considerable number of hydrologic, hydrodynamic, and ecological controls that operate at a wide range of spatiotemporal scales. Sources of the admixtures often cannot be clearly identified, and the locally influenced complex mass exchange between the variables may not be known. Due to the correlations and interactions between water quality variables, it is interesting to investigate whether a domain-specific mechanism governing observed patterns exists to prove the predictability of these variables. The identification of such forecast models is particularly useful for ecologists and environmentalists, since they will be able to predict seawater pollution levels and take necessary precaution measures in advance.

Classical process-based modeling approaches can provide good estimations of water quality variables, but they usually are too general to be applied directly without a lengthy data calibration process. They often require approximations of various processes, and these approximations may overlook some important factors affecting the processes in seawater. A process-based model requires a lot of input data and model parameters that are often unknown, while data-driven techniques provide an effective alternative to conventional process-based modeling. Models developed by data-driven techniques are computationally very fast and require fewer input parameters than process-based models. Data-driven modeling techniques have gained popularity in the last 20 years. The scientific and engineering communities have acquired already extensive experience in the development and usage of data-driven techniques. ANNs are, however, still not widely used tools in the fields of water quality prediction and forecasting. ANNs are able to approximate accurately complicated non-linear input–output relationships. Like their physics-based numerical model counterparts, ANNs require training or calibration. After training, each application of the trained ANN is an estimation of a simple algebraic expression with known coefficients and is executed practically instantaneously. The ANN technique is flexible enough to accommodate additional constraints that may arise in the application.

This paper demonstrates the application of ANNs to model the values of selected seawater quality variables, having the dynamic and complex processes hidden in the monitored data itself. The ANN model can reveal hidden relationships in the historical data, thus facilitating the prediction and forecasting of seawater quality. The steps followed in the development of such models include the choice of performance criteria, division and pre-processing of the available data, determination of appropriate model inputs and network architecture, optimization of the connection weights (training), and model validation. In this paper, a study of ANN modeling to predict and forecast temperature, salinity, DO, and Chl-a in Singapore coastal waters is presented. These water quality parameters were measured weekly at various locations. These models could be used as a prediction tool, which complements the process-based model and ongoing field monitoring program in the region. The results of the ANN prediction and forecast model in the East Johor Strait are discussed in this paper.