We elucidate the impact of bottom trawling on the biological carbon pump by focusing on changes in the water column, sea surface pCO₂ and new production that could arise from the additional release of CO₂ into the bottom water after the remineralization of resuspended sediments. It was shown that the release of sedimentary carbon through bottom trawling may extend up to 10 m or more above the seafloor (Churchill, 1989) and lead to resuspended sediments being remineralized and further advected laterally and vertically across the water column (Pusceddu et al., 2014; De Borger et al., 2021; Morys et al., 2021). Additional experimental model results from a study in the Eastern Mediterranean indicated bottom trawl-induced plumes of sediment along the outer continental shelf and upper slope promoted the supply of remineralized benthic nutrients into the euphotic zone during coastal upwelling, increasing annual net primary productivity by 15% (Dounas et al., 2007). Thus, we assumed the released sedimentary carbon as reported by (Sala et al., 2021) to be fed into the upwelling source waters that are overlaying the bottom shelf region, and subsequently, to be upwelled together with the source waters into the surface region within the individual subsystems. We also took into account the release of nutrients during the remineralization process of resuspended organic matter and their transfer into the upwelling source water masses. The extent to which source water mass concentrations of carbon and nutrients are raised by bottom trawling is hereby controlled by the volume of upwelling waters which the benthic nutrients and carbon are dispersed into. To address the effect of bottom trawling on the air-sea gas exchange, we simulated sea surface pCO₂ using the CO2SYS routine with source water masses’ DIC, Total Alkalinity (TA) and nutrient concentrations, as well as source water temperature and salinity (SWT, SWS) as input parameters ( Table 1 ). We assumed upwelled nutrients and carbon of the source water masses to be consumed at the surface following the Redfield C:N ratio of 106:16 (Rixen et al., 2023), and ultimately to be transformed into organic matter and exported into deeper water layers below the euphotic zone. With this, we follow the bottom-up principle which has been previously applied within the BUS to outline new production rates and carbon export fluxes on the basis of upwelling source water mass inventories (Eppley and Peterson, 1979; Messié et al., 2009; Waldron et al., 2009). With a constant Redfield-ratio of 106:16 (C:N), the impact of biology was estimated by subtracting the amount of nitrate-associated carbon from the source water’s initial DIC concentration, and adding the associated release of total alkalinity (TA) to the original source water TA. Additionally, we considered the surface warming effect of the upwelling source waters to account for the decrease in the CO₂ solubility of seawater and its impact on pCO₂ by using the locally measured sea surface temperatures and salinities (SST, SSS) instead of those of the source water masses. In order to quantify the effect of bottom trawling on the sink and source functionalities of the BUS, we translated the increase in pCO₂ into equivalent changes in CO₂ fluxes. Therefore, we used the offshore increase in pCO₂ through bottom trawling to add to the basin-scale average pCO₂ as estimated from ordinary kriging interpolations (see section 3.5) for calculating the annual CO₂ fluxes within both subsystems. Lastly, we shed light on changes in new production rates on the basis of nutrients and carbon that were released after bottom trawling and upwelled into the surface region, and by taking note of the efficiency at which these nutrients and carbon are being recaptured into biomass. A schematic overview of the bottom-up approach is given in Supplementary Figure 1 .

We used undisturbed sediment core samples that were collected on board RV Africana (AFR258), Meteor (M48/2, M76/2, M103/1) and Maria S. Merian (MSM17/3) (see Table 2 ) with a multicorer (Oktopus Kiel) to outline the stoichiometric carbon to nutrient ratios within the top sediment layer. The sampler was equipped with acrylic tubes of 10 cm diameter and 60 cm length, and was used to retrieve sediment cores that were sliced in 1cm intervals, kept frozen under -20°C and freeze-dried in the home laboratory, and analyzed for concentrations of total nitrogen and organic carbon with an Elemental Analyzer (Carlo Erba NA 1500).

We used published global benthic CO₂ efflux estimates of bottom trawling and dredging fishing practices (Sala et al., 2021). The benthic CO₂ efflux was derived from the amount of carbon stored within the top layer of the sediment as based on recently published global carbon stocks (Atwood et al., 2020), and the labile fraction of carbon that is remineralized and released at the sediment-water interface after the occurrence of bottom trawling. The latter is inferred from the sediment type and trawl activity as estimated from automatic identification systems (AIS) data that was used for tracking the fishing trawlers’ distance, speed, and the applied gear type with its corresponding penetration depth. The sediment type was thereby used as a proxy for the labile carbon fraction due to its impact on organic matter preservation and remineralization. For each sediment type (coarse, sandy, fine and biogenic), the proportion of labile organic carbon was assigned using literature values, amounting e.g., to 0.04 for sandy sediments and to 0.7 for muds, silts or biogenic sediments. The CO₂ efflux was then modelled using an average first-order reaction constant k for the degradation process as a function of oceanic region (Atlantic = 1.00).

Hereby, several studies outlined the underlying first-order reaction constant k as applied by Sala et al. (2021) to be overestimated, since it based on a reactivity value for highly reactive, fresh organic carbon that has recently been transferred to the sediment surface, and applied to a bulk of less reactive compounds within deeper sediment layers.

To shed light on the applicability of k by Sala et al. (2021) for the BUS, we took into account the degradation index (DI) inferred from amino acids of sediment samples collected during cruise MSM17/3. According to Dauwe et al. (1999), the DI derived from amino acids of particulate matter samples can be directly linked to the degradation rate, allowing to assess the quality of organic matter and first-order reaction constants. We estimated an average DI of 0.12 (standard deviation ±0.1) for surface sediment samples collected at water depths between 200 – 1000 m (in line with the bottom trawled area), with 0.36 as the highest value. Given the linear correlation between DI and k, these values imply the reactivity constant k to fall within a spectrum of 0.1 – 1 (Dauwe et al., 1999), where k=1 is used by Sala et al. (2021) to infer the benthic CO₂ efflux of the BUS. Hereby, higher values are indicative for well-preserved, fresh organic matter and lower values implying less well-preserved organic matter and faster degradation process.

We used data on upwelling source water mass compositions within the NBUS and SBUS taken from Siddiqui et al. (2023). This dataset is based on water samples collected with multiple CTD/Rosette systems during various cruises to the BUS which we listed in Supplementary Table 1 , including data from the Global Ocean Data Analysis Project version 2.2020 (GLODAPv2_2020). Water samples were analyzed for temperature, salinity, oxygen, DIC, total alkalinity (TA) and nutrients (phosphate P, nitrate N) following the methods described in Emeis et al. (2018) and Flohr et al. (2014). The upwelling waters, SACW and ESACW, were defined using potential temperatures (theta, θ ) and definitions by Mohrholz et al. (2014) following Equations 1, 2, respectively:

Additional samples for suspended matter were retrieved with the multiple CTD/Rosette system during Meteor cruise M153, Sonne cruise SO283 and SO285. The filtration volume of sea water on pre-combusted (450°C) and tarred glass fibre filters (WHATMAN GF/F, ~0.7 μm, 47 mm diameter) varied between 5 and 30 L. The filtration was stopped when filters were satisfactorily covered. After filtration, the samples were rinsed with deionised water to remove sea salt and subsequently dried in the ship’s dry oven at 40 °C for 48 hours prior to analysis of total nitrogen and organic carbon.

The analysis and quantification of air-sea CO₂ fluxes was based on data taken from Siddiqui et al. (2023), comprising of continuous underway measurements collected between 2008 and 2019 according to methods described in Emeis et al. (2018), and additional quality-controlled measurements from the Surface Ocean CO₂ Atlas (SOCAT) v2020 (Bakker et al., 2016). Data on the sea surface CO₂ partial pressure (pCO₂) were normalized to the reference year 2020 (Takahashi et al., 2009, 2014) and spatially interpolated on a 0.1° x 0.1° grid using ordinary kriging performed with the R automap package (Hiemstra et al., 2009). Annual variogram models applied during the ordinary kriging procedure, together with autocorrelation length scales that were used to correct for spatial autocorrelation following Landschützer et al. (2014), can be found in Siddiqui et al. (2023).

Carbon flux rates (FCO₂) were determined using the partial pressure at the sea surface (pCO_2,sw) and of the atmosphere (pCO_2,at) following Equation 3:

with K₀ as the solubility coefficient of CO₂ (Weiss, 1974) and k as the gas transfer velocity of CO₂ (Wanninkhof, 2014), calculated using Equation 4:

with Sc as the Schmidt number of CO₂ in seawater, 660 as Sc at 20°C water temperature, and u referring to wind speed (m s^-1) at 10 m above sea surface. Sc was determined using shipboard data on wind speed, sea surface temperature (SST) and salinity (SSS) that were spatially interpolated using the ordinary kriging procedures as outlined for pCO₂. The annual mean CO₂ exchange coefficients, sea surface pCO₂ and flux rates for the northern and southern Benguela Upwelling System that we used in this study are presented in Table 1 .

To estimate the impact of bottom trawling on the biological productivity, we applied the bottom-up approach to calculate new production rates for each subsystem within the BUS based on respective upwelling volumes and benthic nitrate concentrations released into the upwelling source waters after bottom trawling (NBUS: 1.39 µmol N kg^-1, SBUS: 0.47 µmol N kg^-1). Quantifying this amount of N available within the euphotic zone and translating it into equivalent grams of C using the Redfield-ratio (6.6) resulted in new production rates of 3.14 and 0.47 Tg C year^-1 for the NBUS and SBUS, respectively. Compared to new production rates based on average DIC and N concentrations of the source water masses (NBUS: 59.83 Tg C year^-1, SBUS: 21.14 Tg C year^-1), the release of N in the aftermath of bottom trawling could increase new production by 5.25% in the NBUS and 2.22% in the SBUS.

However, compared to the corresponding benthic CO₂ efflux of 4.35 and 0.64 Tg C year^-1 in the north and south, only 2/3 of the benthic CO₂ emissions could be recaptured and assimilated into organic matter by benthic nutrients as a result of the difference in C:N ratios between the sediment ( Table 2 ) and surface biomass (Redfield, 6.6). Although this could support the overall productivity of the system, the remaining 1/3 of the benthic CO₂ emissions could lead to an ongoing enrichment of DIC and rise of pCO₂ in the surface region. To elaborate the latter, we estimated the impact of enhanced DIC and nutrient concentrations through bottom trawling on the air-sea gas exchange through sea surface pCO₂ simulations using CO2SYS. We therefore compared our results with the measured and simulated pCO₂ of the NBUS and SBUS coastal and offshore region as presented in Siddiqui et al. (2023).

According to Siddiqui et al. (2023), highest pCO₂ prevails in the coastal region ( Figures 3A, C ) where the upwelling of carbon-rich waters fosters the outgassing of CO₂, as also indicated by elevated pCO₂ measurements. Hereby, the average modelled pCO₂ in the upwelled water lies within the upper range of the nearshore measured pCO₂ (NBUS:294 – 1012 µatm, SBUS: 334 – 610 µatm). This implies biologically-mediated nutrient consumption and fixation of DIC into biomass to occur simultaneously with coastal upwelling, as also indicated by increased chlorophyll concentrations that have been observed by satellites along a narrow belt across the coastal region within the BUS (Weeks et al., 2006; Demarcq et al., 2007; Lamont et al., 2019). In addition, pCO₂ measurements were taken during different times of the year, covering (non-) upwelling seasons and variable intensities of solar radiation that could have led to variabilities in the warming of upwelling waters and its effect on sea water CO₂ solubility. The latter is addressed by considering the effect of the biologically-mediated nutrient consumption on pCO₂ within the offshore region, while also accounting for the warming of upwelled waters in a subsequent stage. This caused a decrease in pCO₂ with respect to the coast due to degassing and the biologically-mediated uptake of CO₂ within the offshore flowing waters ( Figures 3B, D ). Hereby, the simulated pCO₂ falls into the spectrum of measured estimates in cases when the surface warming effect is being accounted for. Depending on its strength, pCO₂ can be drawn below or above the atmospheric pCO₂, which, in the end, controls the regional CO₂ sink and source functionality of the two subsystems (Siddiqui et al., 2023).

In order to account for the impact of bottom trawling, we repeated the simulations of pCO₂ for the coast and offshore regions by adding the bottom trawl-induced efflux of DIC and nutrients as previously outlined to the given source water concentrations. The effect of bottom trawling thereby led to an additional increase in pCO₂ at the coast of +100 and +20 µatm in the NBUS and SBUS, respectively, that gradually decreased towards offshore, where pCO₂ was merely raised by 5 µatm in the north and 3 µatm in the south.

Given the increase in pCO₂ through bottom trawling by 5 µatm in the NBUS, the corresponding annual CO₂ flux would increase by ~1 Tg C year^-1, resulting in 17.13 (-2.64 – 77.25) Tg C year^-1 that would be emitted into the atmosphere. In the SBUS, the bottom trawl-induced rise in pCO₂ would offset the annual CO₂ flux by 0.3 Tg C year^-1, leading to a comparatively lower CO₂ uptake of -2.79 (-4.55 – 4.25) Tg C year^-1. Thus, an additional release of sedimentary carbon into upwelling source waters could on average reduce the SBUS’s sink functionality of atmospheric carbon by ~10%, and increase the rate of outgassing by 6.5% in the NBUS.

These values are in the order of CO₂ emissions by land use and land cover changes as ascribed to the AFOLU (agriculture, forestry, and other land use) sector for Namibia, which takes into account carbon stocks of coastal reservoirs like mangrove forests, tidal marshes, and seagrass meadows (“blue carbon”) (Rixen et al., 2023). The bottom trawl-induced release of CO₂ into the atmosphere within the BUS (~1Tg C year^-1) corresponds to ~3% of the CO₂ currently stored by AFOLU (-30.6 Tg C year^-1) (Rixen et al., 2023), showing the scale to which bottom trawling could potentially alter CO₂ emissions in coastal settings.

Hence, the efficiency of the biological carbon pump in sequestering and storing atmospheric CO₂ is likely affected by bottom trawling: On the one side, such fishing practices impair the sink function of the BUS by reactivating ~5 Tg C year^-1 previously stored in the sediment of which 2/3 may be transformed back into organic matter. On the other side, the trawl-induced release of sedimentary CO₂ into upwelling source water masses can potentially foster an additional outgassing of ~1 Tg C year^-1 into the atmosphere. Hereby, our estimations are representative for an upper limit, assuming the entirety of the benthic CO₂ efflux and associated nutrients to be upwelled into the surface region and assimilated into organic matter, while neglecting any processes hampering the biologically-mediated CO₂ uptake, such as light and iron limitations or water column denitrification. The rise in pCO₂ through bottom trawling is thereby related to the efficiency at which benthic nutrients and DIC are recaptured and assimilated into organic matter after being released from the sediment and upwelled into the surface region. Thus, using a constant stoichiometric carbon to nutrient ratio further disregards any variability in nutrient utilization and remineralization, which, in the end, is a pivotal factor that constitutes the recapture efficiency of the bottom trawl-induced CO₂ efflux. Additionally, the amount of the benthic CO₂ efflux and associated nutrients as elucidated for the BUS are subject to uncertainty because underlying calculations of the labile carbon fraction and remineralization were based on basin-scale average values that curtailed spatial variabilities within the BUS (Sala et al., 2021; Epstein et al., 2022). In this regard, the remineralization potential is related to oxygen availability, with oxygen-rich environments likely promoting organic matter degradation and oxygen deficiencies limiting remineralization (Sciberras et al., 2016; Bradshaw et al., 2021). Hereby, a potential expansion of Oxygen Minimum Zones (OMZs) could add further pressure on benthic ecosystems under the influence of bottom trawling by impeding the recovery potential and protection of benthic flora and fauna, which would otherwise have a stabilizing effect on the sediment and increase its resistance to physical disruptions (Roberts et al., 2017). Hence, in addition to monitoring of hydrographic changes and benthic ecosystem responses, higher spatially resolved data is needed to account for heterogeneity in sedimentary carbon characteristics and to refine trawl-induced benthic carbon and nutrient effluxes within the BUS.