Abstract
While air heat waves often grab headlines, riverine heat waves have gone quietly unnoticed because rivers are commonly perceived as cool refuges. Analysis of riverine heat waves has been hindered by fragmented datasets, despite a proliferation in water-temperature monitoring with sensors and satellites. Here, we analyze riverine heat wave events by training one single deep learning (long short-term memory) model and reconstructing consistent and continuous daily water temperatures (WT) in 1471 sites in the Contiguous United States (1980-2022). We show that riverine heat waves occur at about half the frequency (2.3 versus 4.6 events/year), a third intensity (2.6 versus 7.7 °C/event), but almost double the duration (7.2 versus 4.0 d/event) of air heat waves. Riverine heat wave events have increased at double to quadruple rates of air heat wave events, amounting to an additional 1.8 events/year in frequency, 0.43 °C/event in intensity, 3.4 d/event in duration, and 7 to 15 additional thermal stress days for aquatic ecosystems in 2022 compared to 1980. Rising riverine heat waves have outpaced those of air heat waves in 65 to 76% of the sites, particularly in regions experiencing accelerated warming (e.g., the Rockies). Riverine heat wave trends are driven predominantly by climate-induced changes such as warming and dwindling snowpacks and water flow. Human activities do play important roles: large dams elongate, whereas agriculture reduces heat waves. These results highlight anthropogenic climate change as the primary external driver, whereas human-induced structural changes as the secondary internal modulators of river response to heat disturbance. The widespread rise of riverine heat waves threatens aquatic ecosystems and water-energy-food security, underscoring the need for their global characterization and risk assessment.