Seasonal drivers of dissolved oxygen across a tidal creek–marsh interface revealed by machine learning

Abstract Dissolved oxygen (DO) is a key biogeochemical control in coastal systems, and its concentration drivers vary markedly through time space. This makes it difficult to accurately represent DO associated processes models, limiting our ability predict how these systems will respond global change. We obtained high‐frequency (5‐min) situ measurements of collected at three locations across the interface tidal creek marsh Pacific Northwest, USA. Random Forest machine learning models quantified importance categories environmental (Aquatic, Climatic, Terrestrial) variability creek–marsh interface. selected two 4‐month datasets representing Summer Winter seasonal periods test hypotheses on dominant found that Terrestrial driver—characterized by long anaerobic conditions episodic pulses after floods—was most important during Winter, whereas Aquatic over tidal, diel, lunar cycles—was Summer. explored future climate change scenarios could alter using cumulative sums driver–response framework. Our results suggest under change, Climatic may increase Summer, potentially linked changing metabolic regimes sea level, with driver increasing Winter. approach highlights useful methods for understanding spatiotemporal complexity interfaces quantifying relative distinct drivers.