Declining dissolved oxygen levels in the world’s rivers due to climate change

Dissolved oxygen is expected to decline in many of the world’s rivers due to increasing water temperatures under climate change. This may cause significant adverse effects for freshwater ecosystems, such as mass mortality events and fish kills. However, previous studies related to the effects of climate change on dissolved oxygen are mostly carried out at local or regional scales. In our study, we perform the first global-scale analysis of dissolved oxygen concentrations under climate change for both historic (1980-2019) and future periods (2020-2100) (Graham et al., 2025). This involves the development of a hybrid process-based and machine learning model framework of dissolved oxygen concentration at the global-scale. The model framework includes the process-based DynQual surface water quality model and a random forest machine learning model for error correction, trained on roughly 2.6 million observations of dissolved oxygen concentrations.

The hybrid approach shows a significantly improved performance in simulating dissolved oxygen concentrations compared to the process-based model alone. For instance, there is on average a 43% reduction in the normalised Root-Mean-Squared-Error (nRMSE) when applying residual error correction with machine learning. Additionally, the hybrid DynQual_Random Forest model was able to better capture the impacts of extremes compared to the standalone process-based model. We applied the hybrid model globally at 5arcmin (approximately 10km) spatial and daily resolution for the periods 1980-2019 and 2020-2100. Our results show significant decreasing trends in dissolved oxygen concentration for the majority of rivers worldwide, which leads to on average 8.8 ± 2.3 more hypoxia days (with DO < 3 mg l^-1) per decade globally over the period 2020-2100. This study highlights the strengths of a hybrid process-based and machine learning modelling framework to capture water quality responses at high spatial and temporal resolution as well as during hydro-climatic extremes. It shows that increasing water temperatures and increasing biochemical oxygen demand (BOD) are likely the key drivers of decreasing dissolved oxygen concentrations under climate change. Furthermore, our study emphasises the importance of dissolved oxygen as a key driver of freshwater ecosystem health in the coming decades.

References

Graham, D.J., Bierkens, M.F.P., Jones, E.R. et al. Climate change drives low dissolved oxygen and increased hypoxia rates in rivers worldwide. Nat. Clim. Chang. 15, 1348–1354 (2025). https://doi.org/10.1038/s41558-025-02483-y