Graphical Evaluation for Catchment Water Quality Modeling: An Overview of Methods With Applications

Graphical methods are complementary tools to conventional statistical Performance Measures (PMs) for model evaluation. Despite their recognized value, a comprehensive understanding of how graphical methods can be applied and interpreted for water quality modeling remains lacking. In this overview paper, we describe typical usage of graphical methods, aspects of model performance they reveal, and the interpretation of visualization methods based on four categories, which include (1) temporal trend and variation in water quality data, (2) distribution of model outputs and parameters, (3) relationships between variables, and (4) spatial variation in water quality responses and performance. In addition, we highlight the capability of different graphical methods in terms of visually assessing performance in model fitting, addressing model limitations and predictive uncertainties. We also underscore their potential in uncovering key patterns reflecting relevant processes to inform whether the model is sensible in getting results that are consistent with relevant characteristics of observations. Graphical methods help in establishing a level of confidence with respect to the quality of the modeling process and especially its model evaluation exercises. Recommendations on the selection of methods, qualitative interpretation of model performance, and future research on graphical model evaluation and improvement are discussed. Two example studies featuring distinct water quality problem contexts and model structures are presented. They show how different graphical methods can be used to enhance assessment of model performance and sensibility, thereby identifying areas for model improvement. This article is categorized under: Science of Water > Water Quality