Over the past few decades, the world has witnessed an unprecedented trend of urbanization. This trend has intensified many globally recognized environmental challenges, particularly greenhouse gas (GHG) emissions from substantial human production and consumption activities in urban areas. While much research focuses on anthropogenic GHG emissions resulting from socioeconomic activities^1,2, there is comparatively less emphasis on nature-based mitigation pathways, like urban green-blue infrastructure, largely because our understanding of emissions and their controls remains underdeveloped.

Within cities, rivers serve as integral components of urban green-blue infrastructure, offering valuable socioeconomic and ecological benefits to urban residents and wildlife³. However, this beneficial role is often compromised by degraded physical, chemical, and biological conditions, which, in turn, disrupt elemental cycling and alter GHG emissions within urban rivers. Despite mounting evidence that highlights widespread alterations in the rates and composition of GHG emissions from urban rivers^4,5, characterizing the global significance and understanding the controls of these altered emissions remain challenging due to the lack of a comprehensive quantitative research framework.

In a new study published in Nature Sustainability, we innovatively developed machine learning-based predictive models by integrating GHG measurements from urban rivers and multi-source geospatial data. This approach enabled us to establish a quantitative assessment framework for understanding the global controls, patterns, and magnitudes of GHG emissions from urban rivers. Our research began by compiling a comprehensive dataset of GHG measurement from urban rivers through a systematic review of published scientific literature. We then compared the concentrations and fluxes of GHGs, along with associated water physiochemical properties, with those of non-urban rivers in the Global River Methane Database⁶ (GriMeDB). Our findings revealed striking differences: median CH₄ and N₂O concentrations and fluxes in urban rivers are 1.3 to 5.0 times higher than those in non-urban rivers from GriMeDB. However, contrasting trends in CO₂ concentration and flux highlighted larger uncertainties and the challenge of definitively distinguishing between urban and non-urban rivers in terms of CO₂ emissions.

To gain deeper insights into the broad-scale controls of urban GHGs, we conducted a rigorous standardized linear regression analysis examining the relationships between GHG concentrations and fluxes and various geospatial variables. Our global analysis revealed contrasting trends: as drainage area increases, GHG concentrations and fluxes in urban rivers significantly decrease. Moreover, we find positive correlations between population, population density, and GDP with concentrations and fluxes of all three GHGs. Conversely, concentrations and fluxes of CH₄ and N₂O exhibit a significant decrease with increasing GDP per capita. Furthermore, we observe that concentrations and fluxes of all three GHGs are negatively correlated with catchment elevation. Interestingly, the response of carbon-based (CO₂ and CH₄) and nitrogen-based (N₂O) GHGs in urban rivers to terrestrial biospheric variables displays distinct patterns.

We then developed robust random forest (RF) predictive models (R² ≥ 0.60) for urban river GHG concentrations and fluxes modelling. Based on the global Morphological Urban Areas⁷ (MUAs) and Global Reach-Level A Priori Discharge Estimates for SWOT⁸ (GRADES), we delineated the global urban river boundaries and further generated spatially explicit urban river GHG concentrations and fluxes globally through combining RF models and reach-level predictive variables. Our analysis unveiled that urban river concentrations and fluxes of all three GHGs are significantly higher in tropical regions, with larger disparities observed in CH₄ and CO₂ between tropical and temperate zones compared to N₂O. In addition, we observed nonlinear trends (inverted U-shaped relationship) in the concentrations and fluxes of all three GHGs, aligning with the Environmental Kuznets Curves theory. Specifically, the emissions varied with national income levels with the highest emissions happening in lower-middle income countries where river pollution control is deficient.

We estimated that global annual GHG emissions from urban rivers amount to 78.1 Tg CO₂-equivalent (1.1 Tg CH₄, 42.3 Tg CO₂, and 0.02 Tg N₂O, respectively). Notably, urban rivers in upper-middle income countries and Asia collectively contribute 41.7% and 66.1% of the total emissions, respectively, with river surface areas encompassing over 50% of the total area. Moreover, predicted GHG emissions from urban rivers are nearly double those from non-urban rivers⁹ (~ 815 versus 414 mmol CO₂-eq m⁻² d⁻¹), and are comparable to scope-1 urban emissions¹⁰ (1,058 mmol CO₂-eq m⁻² d⁻¹) in intensity. It's worth noting that urban rivers exhibit a higher proportion of radiative forcing from CH₄ and N₂O, while CO₂ emissions are reduced compared to global rivers in general. Given the backdrop of climate change and escalating urbanization, our analytical findings serve as a factual basis and offer scientific guidance for policymakers and managers engaged in urban GHG emission reduction and water environment pollution control efforts.

 Article reference:

 Xu, W. H. et al. Globally elevated greenhouse gas emissions from polluted urban rivers. Nat Sustain (2024). https://doi.org/10.1038/s41893-024-01358-y

1. Creutzig, F. et al. Global typology of urban energy use and potentials for an urbanization mitigation wedge. Natl. Acad. Sci. U. S. A. 112, 6283–6288 (2015).
2. Ramaswami, A. et al. Carbon analytics for net-zero emissions sustainable cities. Nat. Sustain. 4, 460–463 (2021).
3. Perini, K., Sabbion, P. Urban sustainability and river restoration: Green and blue infrastructure. John Wiley & Sons (2016).
4. Wang, G. Q. et al. Intense methane ebullition from urban inland waters and its significant contribution to greenhouse gas emissions. Water Res. 189, 116654 (2021).
5. Wang, G. Q. et al. Unexpected low CO₂ emission from highly disturbed urban inland waters. Water Res. 235, 116689 (2023).
6. Stanley, E. H. et al. GriMeDB: The global river database of methane concentrations and fluxes. Earth Syst. Sci. Data 15, 2879–2926 (2022).
7. Taubenböck, H. et al. A new ranking of the world’s largest cities—Do administrative units obscure morphological realities? Remote Sens. Environ. 232, 111353 (2019).
8. Lin, P. R. et al. Global reconstruction of naturalized river flows at 2.94 million reaches. Water Resour. Res. 55, 6499–6516 (2019).
9. Lauerwald, R. et al. Inland water greenhouse gas budgets for RECCAP2: 2. Regionalization and homogenization of estimates. Glob. Biogeochem. Cycle 37, e2022GB007658 (2023).
10. Huo, D. et al. Carbon Monitor Cities near-real-time daily estimates of CO₂ emissions from 1500 cities worldwide. Sci. Data 9, 533 (2022).