Salinization of shallow coastal aquifers is particularly critical for ecosystems and agricultural activities. Management of such aquifers is an open challenge, because predictive models, on which science-based decisions are to be made, often fail to capture the complexity of relevant natural and anthropogenic processes. Complicating matters further is the sparsity of hydrologic and geochemical data that are required to parameterize spatially distributed models of flow and transport. These limitations often undermine the veracity of modeling predictions and raise the question of their utility. As an alternative, we employ data-driven statistical approaches to investigate the underlying mechanisms of groundwater salinization in low coastal plains. A time-series analysis and auto-regressive moving average models allow us to establish dynamic relations between key hydrogeological variables of interest. The approach is applied to the data collected at the phreatic coastal aquifer of Ravenna, Italy. We show that, even in the absence of long time series, this approach succeeds in capturing the behavior of this complex system, and provides the basis for making predictions and decisions.