Prediction and Assessment of Stream Water Quality by Considering Multiple Water Quality Parameters

Geostatistical studies entail identifying the most appropriate model to describe the observed data so that it can be used to accurately predict responses across a range of possible locations. The purpose of such a model is to depict the link between the response variables and the predictors while taking into account uncertainties in space and time. We propose a novel approach to model such data via a multivariate spatio-temporal additive model derived through considering a multivariate normal approximation. To demonstrate how the proposed approach works, we use numerous water quality parameters to model and predict the water quality of a stream network. To re ect the spatial variability of the stream network, we employed hydrologic distances in the model, which allowed certain properties of streams and rivers, such as stream ow connectivity, to be effectively described. It was observed that the proposed multivariate model produces accurate predictions at un-sampled locations compared to its univariate counterparts. Accordingly, this study reveals that the proposed multivariate modelling approach is a viable alternative for modelling complicated data such as the data found in water quality monitoring.