Drivers of river carbon cycle processes
Understanding the drivers of carbon cycling in rivers is needed to predict the effects of modification by stressors such as warming, land use changes and flow regulation (Bernhardt et al., 2018). Unlike terrestrial systems, carbon processing in rivers may not be synchronized with subsequent emissions of the produced GHG due to the dynamic nature and spatiotemporal variability in their physicochemical and biological characteristics(Dodds et al., 2013). Continuous measurements of the key drivers of metabolism in rivers (e.g. hydrological conditions, light, temperature, organic matter availability, and nutrient concentrations) are needed therefore to compare and pinpoint their relative importance under different stressor regimes. Additionally, because metabolism and physicochemical drivers act at multiple spatial scales, from local (riparian vegetation, channel morphology) to regional (climate, topography), and vary along the river network (Alberts et al., 2017), their combined impacts can only be examined by in-situ sensor networks and remotely-sensed data products. For example, the extra dimensionality offered by mobile robots with on-board sensors offers a potential solution to measuring spatial variation in such parameters along the course of a river.
River hydrology plays a significant role in shaping metabolism in rivers due to its control on ecosystem structure and functioning (Von Schiller et al., 2008, Hosen et al., 2019, Maavara et al., 2023). In their study based on sensor data from 222 US rivers, Bernhardt et al. (2022) found light and flow stability (and their interaction) to be key controls on primary production and respiration. Studies in temperate rivers have shown that, in addition to the obvious seasonal drivers of Gross Primary Production (GPP) (i.e. light availability, including canopy shading, and temperature), GPP’s dependence on flow should be considered in the context of river size (Hosen et al., 2019). In large rivers, GPP is maximized at low flows, but reduced in high flows due to short water residence times and high turbidity obscuring light availability (Roberts et al., 2007, Pathak et al., 2022), while there is little flow-related change on GPP fluxes in smaller rivers (Hosen et al., 2021, Maavara et al., 2023). The dependence of ecosystem respiration (ER) on flow is somewhat less straightforward; low flows may reduce benthic production due to riverbed drying, but ER can increase after flow resumes, fuelled by the accumulation of terrestrial organic matter on the dry riverbed (Acuna et al., 2005). Maavara et al. (2023) showed that ER was generally maximized in a temperate forested watershed close to median flows when water residence times allowed ample time for carbon uptake to occur, with higher flows resulting in a deeper and wider water column allowing for more DOC availability and more uptake due to larger water column volume.
Although light availability is a key driver of primary production (Savoy & Harvey, 2021), it is not straightforward to model GPP as other factors impact river autotrophs such as turbidity, vegetation, nutrient availability/stoichiometry, and shading (Behrenfeld & Falkowski, 1997). Light and temperature models can be developed for whole river networks by calibrating ML approaches from local sensor data and scaling the findings using remote sensing products such as land use/cover classifications and digital elevation models for topographical information (Segatto et al., 2021, Maavara et al., 2023). Drones with attached sensors could be utilised to build and validate such models with high spatial resolution data. Light availability can also modify temperature (Nebgen & Herrman, 2019), which has a stronger control on ER compared to GPP, indicating a possibility of higher CO2 emissions from rivers with climate warming (Demars et al., 2011). However, the impact of warming on emissions is still less predictable compared to estimates derived from metabolic theory (Battin et al., 2023). Incoming solar radiation can also mineralise DOC to inorganic forms and this must be considered alongside aquatic biological processes (Maavara et al., 2021). Indeed, recent research has shown that the magnitude of photomineralization during low flows as well as in winter often dramatically exceeds ER fluxes. Failing to consider year-round photomineralization fluxes may vastly underestimate the total magnitude of DOC uptake and CO2 production (Maavara et al., 2023).
While GPP and ER are sensitive to nutrient loading, studies have yielded equivocal results regarding the impact of nutrient loading. Some studies have reported an increase in GPP and ER due to nitrogen and/or phosphorus loading (Kominoski et al., 2018), whereas others have suggested that nutrient concentrations may be only secondary drivers due to the effects of light and food web structure (Dodds & Cole, 2007). Conversely, some studies have found evidence for reverse causality, where metabolism variations strongly control riverine nutrient dynamics (Jarvie et al., 2018, Pathak et al., 2022). The development of sensor networks, and sensor deployment technologies, designed specifically to monitor river ecosystem carbon dynamics will therefore enable the relative importance of multivariate drivers of metabolic processes and their feedback to be understood in far more detail. Moreover, bidirectional relationships between water quality and biogeochemical cycles could eventually be explained and predicted through ML using both water quality information from sensors alongside next generation sequencing data focused on microbial communities and their biogeochemical processes (Díaz-Torres et al., 2022) such as sulphur, nitrogen, phosphorus and carbon metabolism.