The dataset features radon-222 (222Rn), a radioactive tracer naturally present and frequently employed to assess submarine groundwater discharge (SGD). This collection is part of a study aimed at refining SGD estimations in shallow estuaries through the prediction of 222Rn variations using accessible hydroclimatic parameters [1]. The dataset includes measurements of 222Rn in water gathered recurringly from Aug. 2019 to June 2021 at half-hour intervals, at a monitoring station near the shore in Corpus Christi Bay, TX, USA (n = 10,660). Additionally, the data set encompasses continuous, accessible hydroclimatic parameters (e.g., wind speed and direction, atmospheric pressure, water temperature, tide height, creek and river discharge rate, n = 35,088). These parameters were integrated into two machine learning models - Random forest (RF) and Deep Neural Network (DNN) - aiming to interpret the variations in 222Rn and forecast during the data gap. A generalized additive model (GAM) was utilized, focusing on interpreting the variability in 222Rn inventory, particularly influenced by windspeed and direction. The tools and data presented herein afford prospects to 1) forecast 222Rn inventories in areas with significant data voids using only publicly accessible hydroclimatic parameters, and 2) refine SGD estimations affected by wind, thereby offering valuable insights for the planning of field expeditions and the development of management strategies for coastal water and solute budgets.
Keywords: Coastal groundwater; Machine learning; Radioactive tracers; Random forest; Submarine groundwater discharge (SGD).
© 2023 The Author(s).