Episodic flooding causes sudden deoxygenation shocks in human-dominated rivers

Fig. 1 | Nationwide multi-year mean of surface water dissolved oxygen (DO) and its percent saturation (DO%sat) in Chinese rivers, and relationships between river surface water discharge (Q), DO, and DO%sat. A–B, mean DO concentrations and DO%sat in 1156 rivers from the daily mean measurements in China from November 8, 2020, to December 31, 2023; C–D, pearson correlation coefficients (r) of linear fits between river Q and DO, and between Q and DO%sat in the 1156 rivers in China. Results with p > 0.05 are shown as white circles. Stacked percentage histograms in A–D show different categories of DO and DO%sat, and Pearson r of linear fits. E–F, box plots showing different categories of Pearson r between Q and DO and DO%sat in rivers with different Strahler orders, and G–H, the relationships between the above mentioned Pearson r of linear fits and agricultural and urban land use (%). Each box shown in panels E-F represents the interquartile range, the horizontal line is the median and the whiskers are 1.5× the interquartile range.
Fig. 2 | Relationships between oxygen metrics (DO, DO%sat) and water quality parameters. A–B, Pearson correlation coefficients (r) of linear fits between river DO and ammonium (NH4+-N), and between DO%sat and NH4+-N in the 1156 rivers in China. White circles show non-significant results (p > 0.05). Inserted are examples of site-specific relationships (colored dots map to locations). C–D, corresponding relationships between river DO and chemical oxygen demand (COD), and between DO%sat and COD in the 1156 rivers in China. Relationships between DO and COD, and between DO%sat and COD, with example sites shown.
Fig. 3 | DO and DO%sat variability between the flood and non-flood periods and the occurrence of hypoxia and anoxia. A–B, ratios of DO and DO%sat between the flood period (discharge ranked above the 95th percentile threshold) and the non-flood period, i.e., DOflood:DOnon-flood and DO%sat flood:DO%sat non-flood calculated from nationwide 1156 rivers monitored daily from November 8, 2020, to December 31, 2023. C–D, boxplots showing different categories of DOflood:DOnon-flood and DO%sat flood:DO%sat non-flood in rivers with different Strahler orders. E–G, locations of sites with at least one occurrence of hypoxia (DO ≤ 2 mg L-1) or anoxia (DO ≤ 0.5 mg L-1) during flood events, and these sites were categorized into different rivers with different Strahler orders. Each box shown in panels C–D and F–G represents the interquartile range, the horizontal line is the median and the whiskers are 1.5× the interquartile range. H–K, relationships between DOflood:DOnon-flood and NH4+-Nflood: NH4+-Nnon-flood, and between DOflood:DOnon-flood and agricultural and urban land use (within 30 km radius of each site); Relationships between DO%sat flood:DO%sat non-flood and NH4+-Nflood: NH4+-Nnon-flood, and between DO%sat flood:DO%sat non-flood and agricultural and urban land use. These relationships were categorized for rivers with different Strahler orders.
Fig. 4 | Agricultural and urban land use (%) and mean ammonium concentration (NH4+-Nflood) for the rivers that experienced hypoxic (≤ 2 mg L–1) and anoxic (≤ 0.5 mg L–1) events during floods. A–B, agricultural and urban land use (in %, within 30 km radius of each site) for the rivers that experienced hypoxic (A) and anoxic (B) events during floods (discharge ranked above the 95th percentile threshold). C–D, the mean concentration of NH4+-Nflood (mg L–1) for the rivers that experienced hypoxic and anoxic events during floods. Similar NH4+-Nflood concentrations were observed between hypoxic and anoxic sites during flood events. The results have been categorized into rivers with different Strahler orders.
Fig. 5 | Post-flood recovery rates of DO and DO%sat. A–B, the post-flood recovery rates of DO and DO%sat were calculated as the linear regression slopes of these parameters at each monitoring site, measured at 1, 3, 5, 7, and 9 days following flood events. C–D, relationships between agricultural and urban land use (within 30 km radius of each site) and the post-flood recovery rates of DO and DO%sat. These relationships were categorized for rivers with different Strahler orders.
Extended Data Fig. 1 | Spatial coherence tests were performed on all 1156 rivers studied. The results show that at 116 river sites, DO and DO%sat exhibited spatial coherence, with upstream values predicting those obtained downstream (A–B). White dots indicate sites that have no direct connection to other sites. The color of each data point represents the percentage contribution of DO and DO%sat variability at each site to its downstream site. The sites have been classified on the basis of different stream Strahler orders.
Extended Data Fig. 2 | The frequency histograms of surface water discharge (Q) in three randomly selected rivers and its relationships with dissolved oxygen (DO) concentration and percent saturation (DO%sat). A, location of three randomly selected sites (the Hai River at Yuancunji, the Huai River at Tuanjiezha, and the Yangtze River at Chenjiadun near the Datong hydrological gauging station). B–D, normal distribution of log-transformed Q frequency at each site (Yuancunji, Tuanjiezha, and Chenjiadun). E–F, relationships between linear, power and logarithmic fits between river Q and DO, and between Q and DO%sat of the Yangtze River at Chenjiadun (near Datong).
Extended Data Fig. 3 | Land use in China in 2020. Land use practices, including farmland, forest, grassland, waterbody, city, and bare soil in different river basins in 2020, and highly developed landscapes were typically found in the East China Plain.
Extended Data Fig. 4 | Thermal and turbidity responses to flood events. A–B, ratios of water temperature (WT) and turbidity during flood (>95th percentile discharge) versus pre-flood (4–9 days prior) periods (WTflood:WTpre-flood; Turbidityflood:Turbiditypre-flood). C–D, corresponding post-flood (4–9 days after) recovery ratios, from daily monitoring (2020–2023) of 1156 rivers.
Extended Data Fig. 5 | Pre-flood oxygen conditions and post-flood recovery patterns. A–B, ratios of flood-period to pre-flood (4–9 days prior) values for DO and DO%sat, i.e., DOflood:DOpre-flood and DO%sat flood:DO%sat pre-flood. C–D, corresponding post-flood (4–9 days after) ratios for DO and DO%sat, i.e., DOflood:DOpost-flood and DO%sat flood: DO%sat post-flood. All data from daily monitoring of 1156 rivers nationwide from November 8, 2020, to December 31, 2023.
Extended Data Fig. 6 | Land use effects at multiple spatial scales. A, relationships between DOflood:DOnon-flood and agricultural and urban land cover within 5, 10, and 20 km buffers. B, corresponding relationships for DO%sat flood:DO%sat non-flood ratios.
Extended Data Fig. 7 | Ratios of the frequency of hypoxic (≤ 2 mg L–1) or anoxic (≤ 0.5 mg L–1) events during flood relative to non-flood periods. A, ratio of the frequency of hypoxic events during flood (discharge ranked above the 95th percentile threshold) relative to non-flood periods, i.e., hypoxiaflood:hypoxianon-flood. B, ratio of the frequency of anoxic events during flood relative to non-flood periods, i.e., anoxiaflood:anoxianon-flood. Note that sites that did not experience hypoxic and anoxic events during either flood or non-flood periods were not included here.
Extended Data Fig. 8 | The contribution of flood-induced water temperature changes to the observed DO decline. Using both the actual temperature change (WTbefore – WTflood) and the established physical relationship between water temperature and DO concentration, we estimated the temperature-induced oxygen loss (in mg DO L–1). The percentage contribution was then calculated as the ratio of DO decline caused by flood-induced temperature changes to the total observed DO decline (DObefore – DOflood), multiplied by 100%.