The coastal environment is not managed in a way that considers the impact of multiple threats, despite being subject to threats from all realms (marine, land and atmosphere). Relationships between threats and species are often non-linear, thus linear approaches to estimating the impact of threats may be misleading. We develop a data-driven approach to model these relationships whilst limiting the effects of confounding variables when assessing cumulative impacts on temperate reef fish. We use data on water quality, commercial fishing, climate change, and indicators of recreational fishing and urbanisation to build a cumulative threat map of the Northern Region in New South Wales. Using statistical models of fish abundance we quantified associations among threats and biophysical covariates and predicted where cumulative impacts are likely to have the greatest impact on fish. We also assessed the performance of no-take zones, to protect fish from cumulative threats across two marine parks. We found that fishing has a greater impact on fish than water quality threats, and that fishing outside of no-take zones affects fish abundances inside no-take zones. Understanding the spatial influence of multiple threats is important because it enables managers to understand the complexity of management actions required to address them.