Anthropogenic disturbance keeps the coastal seafloor biogeochemistry in a transient state

Coastal sediments and continental shelves play a crucial role in global biogeochemistry, as they form the prime site of organic carbon burial. Bottom trawling and dredging are known to increasingly impact the coastal seafloor through relocation and homogenisation of sediments, yet little is known about the effects of such anthropogenic sediment reworking on the overall cycling of carbon and other elements within the coastal seafloor. Here, we document the transient recovery of the seafloor biogeochemistry after an in situ disturbance. Evidence from pore-water data and model simulations reveal a short-term increase in the overall carbon mineralisation rate, as well as a longer-term shift in the redox pathways of organic matter mineralisation, favouring organoclastic sulphate reduction over methane formation. This data suggests that anthropogenic sediment reworking could have a sizeable impact on the carbon cycle in cohesive sediments on continental shelves. This imprint will increase in the near future, along with the growing economic exploitation of the coastal ocean.


Solid phase analysis and water content 33
Water content was estimated from the same samples by weighing the empty centrifuge 34 tubes, the centrifuge tubes filled with wet sediment and the tubes with sediment after freeze 35 drying. Sediment porosity (volume of pore water per total volume of sediment) was 36 determined from water content and solid phase density measurements, accounting for the salt 37 content of the pore water. The solid phase density was determined by adding a known mass of 38 grinded, freeze-dried sediment to a 100 ml graduated cylinder filled with water and recording 39 the volume displacement. ThermoScientific Element 2) after 100x dilution with Milli-Q water. Indium (2.5 ppb) 51 containing 2% HNO3 was injected simultaneously with the samples as an internal standard. 52

Diagenetic model formulation 53
The applied reactive transport model is an adaptation of the early diagenetic model used by 54 van de Velde et al. 2 The mass balance equations are representative for a cohesive marine 55 sediment without advection and bioturbation 3,4 56 where i C represents the concentration of a solute in the pore water, i S is the concentration of 58 a solid component, z is the depth into the sediment and  the porosity. Rk represents the 59 reaction rate of the k-th reaction in the reaction set and i ,kdenotes the stoichiometric 60 The solution procedure has been described previously 2,8 , but in short, the open-source 75 programming language R was used to implement a numerical solution procedure for the 76 partial differential equations (Eq. Biogeochemical reaction set 82 The aim of the model simulations was to reproduce the disturbance event between May 83 and June 2014, in order to estimate how organic matter mineralisation pathways evolve 84 through time. The reaction set included (n=14) was tailored to this (Table S2) were not included, as these process are generally slow 11 and will be of minor importance for 103 the purpose of the model. In a similar fashion, DIR releases ferrous iron (Fe 2+ ), which can (i) 104 adsorb onto solid phase particles 16,18 , (ii) become reoxidised by oxygen, or (iii) precipitate as 105 iron sulphide. Sulphate reduction produces free sulphide, which can be (i) reoxidised by 106 oxygen (ii) reoxidised by iron (oxyhydr)oxides or (iii) precipitate as iron sulphide. 107 Methanogenesis produces methane (CH4) which can be oxidised by oxygen, iron oxides or 108 sulphate (Table S2). 109 The adsorption of Fe 2+ , NH4 + , Mn 2+ are included as a reversible, linear adsorption process 110 (Table S2), where the concentration of the adsorbed species is in equilibrium at all times with 111 the surrounding pore water, e.g.,  (Table S3). The pH of the pore water is a controlling factor in the 118 precipitation of FeS, but it is not explicitly modelled, in order to reduce model complexity. 119 Instead, a constant depth profile with pH = 7.5 was adopted.  (Table S4). Upper boundary conditions for the solutes were set to a fixed 124 concentration, based on the values measured in this and previous field campaigns 7 (Table S5). 125 Upper boundary conditions for the solids were all set at a zero input flux, assuming that there 126 was no input of iron and manganese oxides during the period of recovery (which is a 127 reasonable assumption, considering the absence of iron and manganese oxides during the 128 preceding months 7 ). Bottom boundary conditions for all state variables were set at zero 129 gradients. 130 Two types of model simulation were carried: (1) a steady state simulation describing the 131 steady situation before the disturbance event and (2) transient simulation that describe the 132 development of the sediment chemistry after the disturbance event. Table S4 shows the 133 overview of all parameters as used in the steady-state simulation (the month May). For the 134 dynamic model simulation after the disturbance event (months June, July, September) one 135 needs to consider that in reality, adsorption is not instantaneous. It is a dynamic process and 136 not much is known about the dynamics of reversible adsorption in natural systems. Therefore, 137 there is a lot of uncertainty on the rates of adsorption. Rather than modelling adsorption 138 dynamically (which immensely increases model calculation time) we have reduced the 139 equilibrium constants of ammonium, manganese and iron, which partially accounts for the 140 dynamics of adsorption (slower adsorption leads to less adsorbed species vs dissolved species, 141 which is essentially equivalent to a lower saturation constant). The steady-state simulation 142 was run with the K-values given in Table S4, while for the dynamic simulation run, K-values 143 were set at 1.75 cm 3 g -1 , 5 cm 3 g -1 and 10 cm 3 g -1 for adsorbed ammonium, manganese and 144 iron, respectively. 145

146
In a first step, the model was allowed to reach a steady state with (i) no iron and 147 manganese oxides (ii) no fast degradable organic carbon (which is only brought in during the 148 disturbance event) and (iii) a fixed concentration depth profile of (slow degradable and 149 refractory) organic carbon of 2%. The fraction of slow degradable vs refractory organic 150 matter was tuned to the DIC and SO4 2profiles. The adsorption coefficient and the C:N ratio 151 of organic matter were tuned to the NH4 + profile. Resulting model profiles are represented in 152  Table A5. The concentrations in the 5 cm layer in June 218 were consistently higher than in May, for POC (~0.7%), PN (~0.11%), extractable Fe (~110 219 µmol g -1 ) and extractable Mn (~5 µmol g -1 ). In contrast, PIC was constant for all layers and 220 for all months ( Figure S9). However, the standard deviations are relatively big when 221 compared to these differences. To check whether there was a statistically significant 222 difference between the different months, a standard two-way ANOVA with replicate was 223 performed with month and depth interval as factors (significance level p = 0.05). The only 224 significant difference was found for PIC (p = 0.03, Similarly, trace metal concentrations (As, Cd, Cr, Cu, Pb, Ni and Zn; aqua regia 228 extractable) is given in Figure S10 (mean values and standard deviations in Table S6). If the 229 layer was a deposition of dredged sediment originating from navigation channels and/or 230 harbours, one would expect that the solid phase of the sediment shows similar characteristics 231 as the sediment in the harbour zones. Therefore, the trace metal content of the dredged 232 material from the harbours in the Belgian Coastal Zone (dashed and filled line in Figure S10  233 and second and third column Table S7) are given as a reference for the measured 234 concentrations. As for the major element concentrations, a two-way ANOVA was used to 235 evaluate the apparent differences in trace metal content. Only lead and arsenic gave a 236 significant result (Table S8). This shows as well that the high value in June is most likely not 237 because of different sediment characteristics.  Table S1

320
The associated kinetic expressions are listed in

345
*for September only 1 core was measured for As, and only 1 measurement was within the 10 -15 cm range, thus no standard deviation could be given.