Globally, surface water quality deterioration is an important issue exacerbated by increasing urbanisation, intensified agricultural activities and climate change. To mitigate this increase in waterway pollution there is a strong need for effective catchment management strategies. However, this is currently limited by: (1) our lack of understanding of the key processes and mechanisms driving water quality change in waterways, and (2) our inability to predict future water quality change. To address these, we need to improve our ability to understand and model water quality. This improved capacity will enable us to better identify important water quality changes under various land use, land management and climate scenarios. These improvements could thus assist waterway managers to better prioritise regions which require special management attention, as well as better assess the benefits of different management interventions and preventative actions to reduce pollution levels. Given the above, there is a vital role for predictive models for supporting catchment water quality management. There is a wide spectrum of surface water quality modelling approaches, ranging from purely black-box empirical to detailed process-oriented models; these often have contrasting strengths and weaknesses. This study aims to (1) provide a review of key considerations for designing and developing effective modelling approaches for water quality prediction; and to (2) outline reasons for choices in modelling approaches for rural and urban water quality across Victoria. A fundamental consideration is the modelling purpose which provides preliminary guidelines for other modelling decisions. The core decision in modelling water quality is the conceptualisation of key physical processes, which involves the definition of pollution sources, flow/transport pathways and other catchment features. In addition, practical considerations such as data availability and model development effort are also important. For each consideration we provide recommendations for different types of catchments, and specifically highlight distinctions between natural/rural and urban catchments. We illustrate these considerations with two modelling approaches: a) a data-driven statistical approach informed by conceptual understanding and b) a more process-oriented approach. These models have been applied in rural and urban settings respectively. Both approaches are designed to identify spatio-temporal differences in water quality, and to predict how future water quality will change. Being designed for vastly different environments (rural vs urban), these two models present separate, yet parallel approaches for modelling spatio-temporal variability in water quality. By contrasting these models, we aim to highlight strengths and weaknesses of different approaches, to share practical experiences from implementing these approaches and to identify the ways forward for both approaches. The experiences and recommendations from these modelling approaches can provide valuable r...