Response of Dissolved Trace Metals to Dust Storms, Sediment Resuspension, and Flash Floods in Oligotrophic Oceans

Trace metals (TM) delivered by atmospheric dust play a key role in oceanic biogeochemical cycles. However, the impact of short‐term environmental perturbations such as dust storms and sediment resuspension events on the oceanic water column is poorly constrained due to the low temporal sampling resolution and episodic nature of these events. The Gulf of Aqaba (GoA), Red Sea, is a highly accessible deep oligotrophic water body featuring exceptionally high atmospheric deposition fluxes that provide the main source of TMs to the GoA surface water. Here, we present a 2‐year time series of dissolved manganese, cobalt, nickel, copper, zinc, cadmium, and phosphate concentration profiles sampled in the GoA. The study focuses on daily time scale dust storms and episodes of sediment resuspension to quantify the immediate impact of these events on dissolved TM cycling. Counter‐intuitively, upper mixed layer TM inventories decrease with increasing aerosol loads, with the effects of aerosol‐induced TM scavenging and dissolution peaking 5–6 days after aerosol deposition. Dust storms promote intense TM scavenging, with TM inventories decreasing by up to 44%, but seldom lead to TM enrichment. Similarly, sediment resuspension and flash flood events triggered significant TM scavenging. These findings highlight the potential dual role of atmospheric deposition in the oceans as a long‐term source of dissolved TMs and a short‐term sink. The in situ observations presented here may be used to understand and quantify the global impact of abrupt environmental events on oceanic chemical compositions.

Atmospheric dust deposition is a primary source of TMs and nutrients to the open ocean surface waters (Baker et al., 2016;Jickells et al., 2016;Mahowald et al., 2018) and hence plays a key role in marine biogeochemical cycles and global climate dynamics (Jickells et al., 2005).The extent by which TMs are leached from atmospheric aerosols upon contact with seawater varies with deposition modes (dry/wet), aerosol loads, solubility, provenance, chemical composition, and particle size (Baker et al., 2020;Chance et al., 2015;Fishwick et al., 2018;Jickells et al., 2016;Shelley et al., 2018).Moreover, atmospheric aerosols can become enriched with anthropogenic sourced TMs such as lead (Pb), Ni, Cu, Zn and Cd by interacting with anthropogenic emissions released to the atmosphere (Erel et al., 2007;Nriagu & Pacyna, 1988).As the anthropogenic aerosol phases are more soluble than the natural mineral phases (Desboeufs et al., 2005;Shelley et al., 2018), they readily desorb upon interaction with seawater, enriching ocean surface waters with anthropogenic TMs (Baker & Jickells, 2017;Boyle et al., 2014;Middag et al., 2022).In addition to atmospheric deposition, benthic inputs are recognized as prominent sources of TMs to the deep ocean through the partial dissolution of resuspended sediments (Homoky et al., 2016;Millward et al., 1998) and diffusion of TMs from pore waters and sediments into overlying waters (Homoky et al., 2016;Liu et al., 2022;Plass et al., 2021).
To date, the temporal sampling resolution of oceanic TM measurements is limited by the complexity of working in remote oceanic environments.As such, much of our understanding of the biogeochemical cycling of oceanic TMs depends upon cruise transects, which provide "snapshots" of the water column composition at the moment of sampling.This, coupled with the sporadic nature of atmospheric deposition (Guieu et al., 2010;Mahowald et al., 2009;Ternon et al., 2010) and storm induced sediment resuspension (Gross et al., 1988), makes sampling across short-term perturbations very challenging and scarce (e.g., Bressac et al., 2021;Ren et al., 2011;Rijkenberg et al., 2008).Consequently, a large knowledge gap exists regarding the impact of short-term events on the oceanic water column.The impact of dust deposition on seawater TM compositions is often evaluated through controlled experiments (e.g., Desboeufs et al., 2005;Mackey et al., 2015;Roy-Barman et al., 2021).However, the complex interplay between abrupt particle perturbations and dissolved TMs in seawater cannot be fully captured or quantified based on low-resolution sampling and controlled experiments (Baker et al., 2016;Jickells et al., 2016), resulting in the need for high-resolution in situ observations in open-ocean proxy sites.

Study Site
The Gulf of Aqaba (GoA) is an elongated and deep water body, located at the northern end of the Red Sea.The GoA is connected to the main body of the Red Sea through the shallow Straits of Tiran, while the Red Sea is connected ∼2,000 km to the south to open ocean waters through the Bab-el-Mandeb Straits (Figure 1).As influxing surface waters flow northward through the Red Sea and toward the GoA, they are continuously depleted in nutrients, leading to the development of oligotrophic conditions in the GoA (Lazar et al., 2008;Reiss & Hottinger, 1984).The GoA water column configuration varies seasonally.During winter (October/November to March), a mixed layer of uniform density forms at the upper water column (Figure 2a).As the mixed layer deepens, deep nutrients are upwelled and entrained in the euphotic zone through vertical convection, supporting new primary productivity.With the onset of stratification, the confinement of nutrients above the thermocline triggers brief phytoplankton blooms, promptly exhausting surface water nutrients (Meeder et al., 2012;Zarubin et al., 2017).Throughout the summer (April to September/October), water column stratification results in low primary productivity characterized by a deep chlorophyll maximum (Figure S1 in Supporting Information S1), while deep nutrient reservoirs are mainly sustained by the continuous remineralization of sinking biological material (Lazar et al., 2008;Meeder et al., 2012).
Due to the hyper-arid regional climate (<30 mm rain year −1 ), there are no major tributaries flowing into the GoA and rainfall is limited to 1-5 events per year (Figure 3b), rendering these inputs quantitatively negligible as long-term TM sources to the GoA surface waters.However, intense rainfall may occasionally produce short-lived (hours) flash floods, which carry large sediment volumes to the GoA surface waters (Katz et al., 2015).Sediment resuspension in the GoA has an episodic, yet somewhat predictable nature, characterized by discrete events taking place annually during winter (Figure 3a), as recorded by an array of sediment traps deployed in the GoA (Chernihovsky et al., 2020;Torfstein et al., 2020).Due to its location in between the Sahara and Arabian Deserts, atmospheric deposition fluxes in the GoA are among the highest on Earth (Albani et al., 2015;Benaltabet et al., 2022;Chase et al., 2006) and are mainly delivered by short-term dust storms (Torfstein et al., 2017).Since current advection processes to the GoA are limited due to its semi-enclosed configuration (Biton & Gildor, 2011;Silverman & Gildor, 2008), dry atmospheric deposition is the main source of TMs to the GoA surface waters (Chase et al., 2011;Chen et al., 2008).
These unique conditions, combined with a water residence time of 3-8 years (Reiss & Hottinger, 1984;Silverman & Gildor, 2008), render the GoA a "natural laboratory" for studying the direct and immediate impact of shortterm events on dissolved seawater TM compositions in a well-studied, oligotrophic, open-ocean proxy environment (e.g., Benaltabet et al., 2020Benaltabet et al., , 2022;;Chase et al., 2006Chase et al., , 2011;;Chen et al., 2008).Recently, it was shown that GoA dust storms drive strong perturbations in the concentration and isotopic composition of dissolved Pb (Benaltabet et al., 2020) and Al (Benaltabet et al., 2022), with atmospheric deposition serving both as a long-term source and a short-term sink for these metals.
This study is part of the Red Sea Dust, Marine Particulates and Seawater Time Series (REDMAST) GEOTRACES process study GLpr09 (Benaltabet et al., 2020;Torfstein et al., 2020).We present a 2-year time series of water column profiles of dissolved Mn, Co, Ni, Cu, Zn, Cd, and phosphate (PO 4 ).Observations are focused on daily time scale events including dust storms, episodes of sediment resuspension and flash floods.These are evaluated in conjunction with high resolution measurements of atmospheric aerosol loads and water column sinking particle fluxes, in order to examine the direct impact of abrupt events on dissolved TM compositions.(Benaltabet et al., 2022;Chernihovsky et al., 2020;Torfstein et al., 2020).Sediment resuspension episodes during winter are reflected by peaks in particle fluxes (note the broken Y-axis).

Seawater Sampling
Seawater profiles were sampled between January 2017 and November 2018 (Figure 2) at the GoA time series station (Station A, water depth ∼700 m, 29°28′N 34°55′E; Figure 1).Deep seawater profiles were sampled in tandem with CTD (SeaBird) profiles, from the RV Sam Rothberg and shallow profiles (up to 220 m) were sampled from a fiberglass speedboat.Seawater samples were collected using acid-cleaned Teflon coated GO-Flo bottles (General Oceanics) from which metal parts were replaced with non-metallic strings.The GO-Flo bottles were mounted on a Vectran rope and tripped with Teflon coated messengers, following GEOTRACES sampling protocols (Cutter et al., 2017).Sampling of post-flood surface seawater was done using acid cleaned HDPE bottles.Sampling bottles were transported within ∼1-2 hr of collection to the Interuniversity Institute (IUI) for Marine Sciences on shore clean laboratory (Figure 1), where the seawater was filtered with Tygon (Masterflex) tubing through a 0.2 μm pore size acid cleaned cartridge filter (Pall AcroPak 1000, Supor Membrane).Samples for dissolved Mn, Co, Ni, Cu, Zn, and Cd concentrations were filtered into acid cleaned LDPE cubitainers (10 L, Fisher Scientific) or bottles (1L, Nalgene) and acidified for storage with 6M HCl (final concentration 0.024 M HCl) to pH <1.8.Samples for PO 4 concentrations were either aliquoted from the TM samples or filtered separately and stored in polypropylene (Corning) tubes and acidified with 1 M HCl to a final concentration of 0.007 M HCl.Dissolved PO 4 replicates analyzed in both storage and acidification procedures yielded similar results (Figure S2 in Supporting Information S1).Sample processing was carried out at the IUI clean laboratory (class 1,000 with class 100 workstations).All acids and reagents used in this study were either in-house double-distilled (Analab Evapoclean®) or commercially purchased ultra-pure solutions.The acid cleaning procedure for sample cubitainers included a rinse in diluted detergent, soak in 4 M HNO 3 and 4M HCl for 48 hr each, and a final soak in 0.1 M HCl (>72 hr).Sample LDPE bottles were rinsed with a diluted detergent, sequentially filled with 4 M HNO 3 and 4 M HCl and heated at 80°C for 72 hr with each acid.All sample containers were rinsed three times with MQ water (18.2MΩ cm, Millipore) between each cleaning step.TM concentrations in acidified MQ water (0.024 M HCl) stored for 1 month in LDPE cubitainers and bottles were similar to fresh acidified MQ water, suggesting a negligible contamination from sample containers during storage (Figure S3 in Supporting Information S1).

Analyses of Dissolved Mn, Co, Ni, Cu, Zn, Cd, and PO 4 in Seawater
Pre-concentration and calibration protocols were modified after Sohrin et al. (2008) and Biller and Bruland (2012).Samples for Mn, Co, Ni, Zn, Cu, and Cd concentrations, along with Pb (Benaltabet et al., 2020) and Al (Benaltabet et al., 2022) were UV irradiated in FEP bottles (Nalgene).Prior to pre-concentration, the samples pH was adjusted to 6.1 ± 0.2 with an ammonium acetate (NH 4 Ac) buffer to a final concentration of 0.01 M. Columns (Biorad) loaded with the NOBIAS chelate PA-1 resin (HITACHI high technologies) were used for sample pre-concentration and buffer purification.Prior to sample loading, columns were sequentially conditioned with 1M HNO 3 and 0.05 M NH 4 Ac.After sample loading and before elution, 0.05 M NH 4 Ac was used for salt matrix removal.Triple distilled 1 M HNO 3 was used as an elution acid and the concentration factor was determined gravimetrically.The collection of the eluant was performed in two ways: larger volume samples (∼230-250 ml) were eluted into acid cleaned vials (PP, Sarstedt) and measured directly, while lower volume samples (∼100-130 ml) were eluted to PFA vials (Savillex), evaporated to a droplet and re-dissolved with 500 μl triple distilled 0.7 M HNO 3 .Replicates processed both ways yielded similar results (Figure S4 in Supporting Information S1).An external calibration curve was established with TM purified seawater (pretreated three times with the NOBIAS chelate PA-1 resin) spiked with varying amount of a gravimetric multi elemental stock standard (ESI).This curve was also used to assess recovery (Table S1 in Supporting Information S1) by comparing to standards of similar concentrations made up in HNO 3 .The procedural blank for each new batch of NH 4 Ac buffer solution was determined by pre-concentrating 4 splits of a solution containing the amount of reagents added to each sample (HCl and NH 4 Ac) mixed with TM purified seawater (Table S1 in Supporting Information S1).The mean signal of the splits (n = 4) was subtracted from the signals of the samples processed with the respective NH 4 Ac batch.On average, each batch of NH 4 Ac buffer was used for the processing of ∼30 samples.Samples were processed in duplicates or triplicates and the uncertainty was estimated as 1 standard deviation (σ) from the mean of the replicates.Data were acquired in five separate sessions of processing and analyses, which varied in recovery, procedural blanks and detection limits (defined as 3σ of the procedural blank).These are reported in Table S1 of Supporting Information S1 as the average and 1σ values from the five sessions.An in-house seawater standard was processed in each session to follow and correct for any methodological drifts.TM concentrations were analyzed using an Agilent 7500cx Magnetic Sector ICP-MS at the Institute of Earth Sciences, Hebrew University of Jerusalem.Instrumental background values and 10.1029/2023GB007858 6 of 24 drifts were monitored by measuring an HNO 3 wash solution and a multielemental standard solution, respectively, every 6-8 samples.Accuracy was validated with CASS-6 and NASS-7 Certified Reference Materials (Table S1 in Supporting Information S1).Good agreement was achieved for all TMs in both standards except Cu in CASS-6, which was higher than the certified value.The higher Cu concentrations could potentially stem from variations in UV irradiation procedures, shown to increase dissolved Cu concentrations (Biller & Bruland, 2012;Yang et al., 2018).Dissolved TM profiles correspond to previously reported measurements in Station A (Chase et al., 2011) in both concentration range and vertical distribution (Figure S5 in Supporting Information S1).
Dissolved PO 4 concentrations were measured at the National Monitoring Program (NMP) laboratory, using a Flow Injection Quik-Chem 8500 Auto-analyzer (LACHAT Instruments) following a protocol modified after Grasshoff et al. (2009).A calibration curve was established by gravimetrically diluting a concentrated commercial standard solution (MERCK) with low nutrient-filtered seawater acidified to the same HCl concentrations as the samples.Blanks and detection limits (defined as 3σ of the blank), estimated by analyzing acidified MQ water, were 0.002 ± 0.003 and 0.009 μmol kg −1 , respectively.Uncertainty for selected PO 4 samples (n = 52) was calculated as 1σ of duplicate analysis.The uncertainty for the rest of the samples (n = 95) was estimated as the mean σ of the same run.To validate accuracy, the NMP laboratory participated in an inter-laboratory study organized by the QUASIMEME program (e.g., Smith & Wells, 2007), which compared nutrient concentrations in seawater.Dissolved PO 4 concentrations measured by the NMP laboratory were within the uncertainty of the inter-calibration mean value in all the five reference samples analyzed (Figure S6 in Supporting Information S1).

Suspended Atmospheric Particles and Rain Time Series
Concentrations of total suspended atmospheric particles (TSaP) were retrieved from the Israel Ministry of Environmental Protection meteorological monitoring stations (www.svivaaqm.net).The mass concentration of airborne suspended particulate matter was measured using a Thermo Scientific FH62-C14 continuous particulate by the use of beta attenuation (see details in Torfstein and Kienast (2018)).Data presented here (Figure 3b) is an average of 3 sampling stations (Eilat1, Eilat3, and Eilat4; Figure 1) with each station providing continuous data at 6 hr intervals in order to smooth out any sporadic shifts in TSaP values.Dust storm events were defined as intervals during which TSaP levels were at least double than the annual average value of 91 μg m −3 (i.e., >182 μg m −3 ).The time span for these intervals typically ranged between a few hours and 3 days.Air mass back trajectories (Figure S7 in Supporting Information S1) were calculated using the National Oceanic and Atmospheric Administration Hybrid Single-Particle Lagrangian Integrated Trajectories (HYSPLIT) program (Stein et al., 2016), using the GDAS meteorological data set for a period of 72 hr with an isentropic vertical motion calculation method for a target height of 100 m above ground level.Precipitation data were retrieved from the Israel Meteorological Service (https://ims.data.gov.il)meteorological stations (Figure 1c) equipped with an automated tipping bucket rain gauge that measures the amount of rain at a 10-min resolution.

Sinking Particle Flux Time Series
The flux of sinking water column particles was sampled using an automated McLane PARFLUX-II (aperture area 0.5 m 2 ) time series sediment trap mounted on a bottom-tethered mooring stationed near Station A (29°29′N 34°56′E; Figure 1) at a water depth of ∼610 m (data set previously published in Benaltabet et al. (2022), see full technical details in Chernihovsky et al. (2020) and Torfstein et al. (2020)).The trap, deployed at a depth of 410 m, collected sinking particles in 21 rotating bottom cups filled with a saturated NaCl brine poisoned with 150 mg L −1 HgCl 2 .The cups rotated every 12-108 hr and the mooring was retrieved and redeployed every 2-14 weeks.Following the retrieval of the mooring, the bottom cups were transferred to the clean lab where they were refrigerated overnight to allow the settling of suspended particles.Then, the samples were wet-sieved (1 mm), rinsed three times with deionized water, freeze-dried, and weighted to determine bulk sinking flux.

Calculation of Weighted Average TM Concentrations and Change Rates
In order to demonstrate daily time scale shifts in TM concentrations, the mixed layer and deep-water TM weighted average concentrations were calculated using Equation 1: where the integral represents the inventory (nmol m −2 ) of a given TM over the shallowest (Z shallow ) and deepest (Z deep ) samples in the mixed layer or deep waters.The integral is divided by the difference in the measured depths to produce the weighted average TM concentration in the mixed layer (TM ML ) or deep waters (TM DW ).The mixed layer depth (MLD) is defined as the depth at which the density anomaly value (σ t ) is >0.03 kg m −3 than the σ t value measured at 10 m depth (de Boyer Montégut et al., 2004).The deep waters are defined as deeper than the MLD, with the lower bound defined as the deepest depth sampled.For each weighted average concentration, Equation 1 was calculated 10,000 times using a Monte Carlo simulation, with each calculation yielding a random value that is within the range of ±1σ of the replicate extractions.Final weighted average concentrations were calculated as the average value (n = 10,000) obtained from each simulation and TM ML and TM DW uncertainties were estimated as 2σ from the simulation average.
The change in weighted average TM concentration in response to perturbative events (i.e., dust storms and sediment resuspension episodes) was quantified by the difference in TM ML or TM DW between profiles measured before and after events (Equation 2): where i pertains either to the mixed layer (ΔTM ML ) or the deep waters (ΔTM DW ) and TM i post and TM i pre denotes the weighted average concentration (Equation 1) of a given TM measured after and before events, respectively.The statistical significance of ΔTM i was determined on the basis of an unpaired t-test with a 95% confidence interval (p < 0.05, Equation 3).
where 2σ represents the uncertainty yielded by the Monte Carlo simulation and n is the number of samples used for the calculation of each weighted average concentration (Equation 1).
The observed shifts in mixed layer TM concentrations following dust storms (ΔTM ML ) were converted to TM change rates (ΔTM rate ) using Equation 4: where Δt represents the time elapsed between the sampling of the pre and post dust storm profiles (Table 1).
Theoretical TM dissolution rates (DISS rate ) from incoming dust storms were estimated from the total TSaP measured between profiles bracketing dust storms and aerosol properties (Equation 5, Benaltabet et al., 2022): where TSaP total denotes the sum of the measured TSaP between two profiles bracketing a dust storm (Table 1).
V d is the airborne particle settling velocity of 0.7 cm s −1 calculated for GoA mineral dust at a particle diameter of 0.1-5 μm (Chen et al., 2007).D is the bulk long-term average elemental concentration in GoA aerosols (Table S2 in Supporting Information S1, Chen et al., 2006;Torfstein et al., 2017).S represents the elemental fractional solubility of GoA aerosols in seawater following a 10-30 min leach (Chen et al., 2006;Mackey et al., 2015) (Table S2 in Supporting Information S1).t TSaP represents the interval of each TSaP measurement (which was constant at 6 hr), MLD is the measured mixed layer depth (Table 1), and ρ SW is the average density of GoA seawater (1,028 kg m −3 ).We note that the constant values for V d , D, and S, which contain large uncertainties and variability (Table S2 in Supporting Information S1), might not be ideal for the time scales of this study and for representing individual dust storms.Applying the lower and upper limits of these parameters yields DISS rate values that are up to an order of magnitude lower and higher, respectively.However, even if the higher bound values of DISS rate are considered, they are still 1-2 order of magnitude lower than measured ΔTM rate of Ni, Cu, Zn, and PO 4 .Hence, specific dissolution rates are not explicitly discussed in Sections 4.3-4.5 and are only used to present general trends.

Oceanographic Settings
Twenty-six profiles comprised of 171 seawater samples were sampled between January 2017 and November 2018 at Station A (Data Set S1) (Benaltabet et al., 2023) across two annual cycles of water column deep mixing and stratification (Figure 2a).In addition, CTD and PO 4 profiles were sampled monthly at Station A by the NMP (Shaked & Genin, 2019).In 2017, the maximum MLD of 520 m was reached in February and the onset of stratified conditions in April was maintained until November.In 2018, the maximal MLD (274 m) was reached in January.Subsequently, stratified conditions were resumed in April 2018, until October 2018 (Figure 2a).

Dust Storms Patterns, Magnitude, and Provenance
Sampling efforts focused on five dust storm events (Figures 4 and 5, Table 1), which were supplemented by additional background profiles sampled during low TSaP periods in March, September, November, and December 2017 and February and August 2018 (Figure 3).Dust storm names, dates and estimated back trajectory-based air mass origin, along with the seawater profiles bracketing them, are presented in Table 1 and a detailed description of each storm is given in Benaltabet et al. (2022).

Dissolved Trace Metal Distributions
During the 2-year time series of the study, dissolved Mn concentrations ranged between 0.2 and 4.1 nmol kg −1 (Figure 2d).In the upper mixed layer, high Mn concentrations (Mn ML = 1.0-4.1 nmol kg −1 ) shifted sharply to low concentrations in the deep waters (Mn DW = 0.2-1.2nmol kg −1 ).While Mn DW was relatively constant, Mn ML varied seasonally, with the lowest concentrations measured during 31.1.17, 6.3.2017, and8.3.2018 (Figures 4 and5).The highest Mn concentrations were measured during 5.8.18 when the water column was fully stratified.
Dissolved Zn concentrations were typically between 1.0 and 2.8 nmol kg −1 (Figure 2h), with surface concentrations of up to 6.3 nmol kg −1 recorded during summer stratification and concentrations lower than 1.0 nmol kg −1 during 8.3.18(Figure 5m).
Dissolved Cd concentrations increased with depth, featuring Cd ML concentrations between 5.7 and 14.7 pmol kg −1 compared to Cd DW values of 10.1-26.3pmol kg −1 .Deep Cd concentrations of up to 29.0 pmol kg −1 were observed between January and November 2018, while concentrations as low as 3.5 pmol kg −1 were measured across the mixed layer during 31.1.2017,6.3.2017, and 13.12.2017(Figure 2i).Similarly, PO 4 concentrations were depleted across the mixed layer (Figure 2c), with PO 4ML < 0.09 μmol kg −1 .PO 4 enrichment was observed at depth, with PO 4DW concentrations of 0.09-0.31μmol kg −1 .

Long-Term Trace Metal Distributions
The strong seasonal cycle of the GoA water column is reflected by the distribution patterns of several TMs.The distribution of dissolved Mn is controlled by redox cycling through Mn partitioning into insoluble Mn oxides and bacteria-mediated photoreduction into soluble Mn(II) (Sunda & Huntsman, 1988, 1994;Wu et al., 2014).
In the GoA, the resemblance between dissolved Mn concentrations and seawater density (Figure 2d), suggests that the vertical and temporal distribution of dissolved Mn is primarily dictated by the physical properties of the water column and photochemical processes.As such, the active vertical mixing of seawater in the mixed layer, which continuously exposes seawater in the mixed layer to sunlight, enhances bacteria-mediated photoreduction rates and inhibits the formation of insoluble Mn oxides (Sunda & Huntsman, 1988).This, along with the local high irradiance levels (Shaked, 2008) and high atmospheric deposition fluxes (Benaltabet et al., 2022;Chase et al., 2006), sustains the high dissolved Mn concentrations through the mixed layer.A similar mechanism was previously proposed to support the redox cycle of Fe(II) in the GoA (Shaked, 2008).In the deep waters, the lack of sunlight results in an accelerated formation of Mn oxides (Sunda & Huntsman, 1988, 1994), which sharply drives down dissolved Mn concentrations (Figure 2d).
Dissolved Co presents a scavenged-type distribution, depicting surface enrichment due to leaching from atmospheric particles and decreasing concentrations with depth brought on by scavenging induced by sinking particles (Dulaquais et al., 2017;Hawco et al., 2018).Co scavenging is known to be associated with Mn oxides, as Co is microbially co-oxidized with Mn (Chmiel et al., 2022;Hawco et al., 2018;Moffett & Ho, 1996).Indeed, Co and Mn distributions are similar (Figure 2e) and display a strong covariance (Figure S8 and Table S3 in Supporting Information S1), suggesting that on long time scales, Mn oxides are a prominent scavenger of Co in the GoA.
Co, Ni, Cu, and Zn are considered to play a biological role and are often present in low concentrations in open ocean surface waters (e.g., Boyle et al., 1976;Bruland, 1980).However, in the GoA, phytoplankton are mostly limited by macronutrients (i.e., PO 4 and NO 3 ) (Lazar et al., 2008;Mackey et al., 2007), which coupled with the high atmospheric inputs, prevents these metals from limiting primary productivity and being depleted at the surface (Chase et al., 2011).Nevertheless, Ni concentrations in the GoA (Figure 2f) are considerably lower compared to other oceans, which rarely feature dissolved Ni concentrations below 1.7 nmol kg −1 (Middag et al., 2020).It is possible that the high Fe concentrations in the GoA (Chase et al., 2006;Shaked, 2008) facilitate the enhanced biological uptake of Ni during nitrogen fixation (Middag et al., 2020;Schlosser et al., 2014).Indeed, high nitrogen fixation rates were measured in the GoA during winter mixing in 2017 and 2018 (Landou et al., 2023), corresponding to the observed low Ni concentrations.
Dissolved Cd and PO 4 both display nutrient-type profiles, depicting surface depletion and enrichment at depth (Figure 2) as they are both taken up by phytoplankton in the euphotic zone and remineralize at depth.On a global scale, Cd and PO 4 cycles are known to be linked (de Souza et al., 2022;Middag et al., 2018).In the GoA, Cd, and PO 4 generally co-vary (Figure S8 in Supporting Information S1), with R 2 between 0.59 and 0.95 (Table S3 in Supporting Information S1), except for 24.1.2017,where R 2 = 0.01 (see Section 4.6).

Controls of Atmospheric Deposition Over Trace Metal Inventories
To evaluate the extent of atmospheric deposition controls over TM ML concentrations, TM ML in each profile is compared to integrated TSaP loads over varying time intervals prior to seawater sampling (24-502 hr).Then, the square of the correlation coefficient (R 2 ) between integrated TSaP loads and TM ML over the entire profile data set is calculated for each TSaP integration interval (Figure 6).The resulting R 2 coefficients calculated for intervals of 24-502 hr of TSaP integration range between 0.03 for Cd ML and 0.15 for Mn ML .However, these correlations include both background profiles and the profiles sampled following perturbative events (i.e., up to 3 days after dust storms, 16 days after sediment resuspension events and 12 days after flash floods; Table S4 in Supporting Information S1).When the profiles recently affected by perturbative events, which abruptly shift TM inventories (see Sections 4.3, 4.6, and 4.7), are filtered out, a systematic pattern of R 2 variability is revealed (Figure 6a).The highest correlations for Co ML , Ni ML , Cu ML , and Cd ML are achieved for TSaP integration periods of 120-144 hr, with R 2 values of 0.56, 0.55, 0.37, and 0.43, respectively.For Zn ML , the highest correlations to TSaP loads are observed for integration periods of 168-216 hr, with R 2 = 0.36.We note that the smaller number of dissolved Zn measurements relative to other TMs may be the cause to the shift in peak correlation to higher TSaP integration periods.
The correlations between TSaP and TM ML imply that atmospheric deposition imposes some control over the dissolved inventories of Co, Zn, Ni, Cu, and Cd.These controls, which include the competing processes of scavenging and dissolution associated with atmospheric aerosols, reach their maximal effect 5-6 days (7-9 days for Zn) following aerosol deposition.Hence, the maximal shift of TM ML signals associated with atmospheric deposition will probably occur 5-6 days after deposition.Thereafter, the effects of scavenging will gradually fade as dissolution gradually shifts TM concentrations back to their initial baseline values.The exclusion of profiles recently affected by dust storms, sediment resuspension events and flash floods and the resulting higher TSaP-TM ML correlations shows that the profiles associated with particle perturbations are separated from the general TSaP-TM ML steady-state trend (Figures 6d and 6f-6j).Hence, the dissolved background inventories of mixed  S4 in Supporting Information S1) display higher correlations to TSaP relative to the entire set of profiles (empty symbols, Table S4 in Supporting Information S1).The highest correlations for Co ML , Ni ML , Cu ML , Cd ML and Al ML were achieved for TSaP integration periods of 120-144 hr, whereas Mn ML and PO 4ML R   layer Co, Ni, Cu, Zn, and Cd, are transiently shifted away from steady state by the abrupt increase in seawater particle loads induced by these perturbations (see Sections 4.3,4.6,and 4.7).
By contrast, excluding profiles sampled after perturbative events did not result in higher TSaP-Mn ML correlations, as these remain rather poor with maximal R 2 = 0.17 (Figure 6k).Hence, atmospheric controls over long-term background Mn ML are probably masked by redox processes involving the formation and photoreduction of Mn oxides (Sunda & Huntsman, 1988, 1994;Van Hulten et al., 2017).Similarly, correlations of PO 4ML to integrated TSaP are generally low (R 2 ≤ 0.29) and the correlation pattern is not shifted by the exclusion of perturbative events (Figures 6a and 6e), demonstrating the strong biological controls exerted on PO 4ML relative to atmospheric deposition.
Similar observations were previously reported for dissolved Al (Benaltabet et al., 2022) and Pb (Benaltabet et al., 2020) following the same approach (Figures 6a, 6c, and 6d).However, the dissolved Pb isotopic signature associated with atmospheric deposition was positively correlated with increasing aerosol loads (Figure 6b) and the peak in Pb isotopic shift was reached ∼3.5 days after aerosol deposition (Figure 6a), representing the different rates of rapid anthropogenic Pb desorption and slower scavenging from/onto atmospheric particles.
The combined observations highlight the specific nature and temporal dynamics of various metals with respect to dissolution and scavenging processes imposed by atmospheric deposition and further place new constraints on the manner and time scales at which atmospheric TM inputs are delivered to the oceans.We note however, that the background atmospheric deposition fluxes in the GoA are considerably higher than most ocean basins (Benaltabet et al., 2022;Mahowald et al., 2005;Measures & Vink, 2000).Thus, the general insights presented here most likely pertain to oligotrophic regions of similar atmospheric deposition fluxes, such as the Eastern Tropical Atlantic (Anderson et al., 2016;Mahowald et al., 2005;Measures et al., 2015).

Daily Time Scale Scavenging and Dissolution Rates in Response to Dust Storms
Daily time scale shifts in TM weighted average concentrations with respect to dust storms are presented in Figures 7 and 8.The oligotrophic conditions in the GoA generally allow excluding biological factors from being significant in driving daily time scale variations in TM ML concentrations.Indeed, a lack of any correlation between surface seawater chlorophyll-α concentrations and atmospheric loads (Torfstein & Kienast, 2018) suggests that dust storms are less likely to trigger phytoplankton blooms that would consume TMs in the euphotic zone.In addition, seasonal-scale advection rates to the northern GoA (Biton & Gildor, 2011) are likely to play a limited role in modulating TM ML concentrations on the time scale of days.Thus, we consider TM dissolution and desorption from atmospheric aerosols into seawater and TM scavenging onto atmospheric particles to be the main factors in controlling shifts in TM ML concentrations following dust storms (ΔTM ML ; Equation 2), as previously shown for Pb and Al (Benaltabet et al., 2020(Benaltabet et al., , 2022)).Hence, a significant increase in TM ML following dust storms (ΔTM ML > 0, unpaired t-test, p < 0.05; Equation 3) will suggest that dissolution rates outcompete scavenging rates and a decrease in TM ML (ΔTM ML < 0) will point to higher scavenging rates.A statistically insignificant change (unpaired t-test, p > 0.05; Equation 3) in TM ML could either stem from similar scavenging and dissolution rates or from undetectable variations in ΔTM ML.
Figure 9 depicts the relationship between TM scavenging and dissolution rates by comparing post dust storms TM change rates (Equation 4) to theoretical dissolution rates (Equation 5).These rates feature high variability between different metals and dust storms.The only cases where dust storms resulted in a significant enrichment in TM ML (ΔTM rate > 0) are for Ni following DS-M18 (Figure 9c) and for Cu following DS-J18 and M18 (Figure 9d).Concomitantly, these storms also brought to a significant decrease in dissolved Mn, Co, and Zn (and  4) and theoretical trace metal dissolution rates (DISS rate , Equation 5) in response to dust storms.Statistically significant and non-significant shifts following dust storms are depicted by full and empty circles, respectively.Statistically significant and non-significant shifts following low dust intervals (DS-O18-2, N18) are represented by full and empty squares, respectively.Negative ΔTM rate values indicate higher dust storm induced scavenging rates with respect to dissolution rates while positive ΔTM rate values indicate higher dissolution rates.Error bars represent ΔTM rate uncertainty propagated after TM ML 2σ calculated from the mean of a Monte Carlo simulation (n = 10,000).(h) Relatively high correlations between ΔAl rate (Benaltabet et al., 2022) and ΔMn rate suggest that the post dust storm Mn scavenging mechanism is similar to that of Al.
Interestingly, DS-O18-2, which was previously recognized as inconsequential due to its short duration (only 6-12 hr) and relatively low magnitude (Benaltabet et al., 2020(Benaltabet et al., , 2022)), was followed by significant decreases in ΔNi rate , ΔCu rate , and ΔZn rate (Figures 9c-9e).Similarly, a significant increase in ΔZn rate was observed for the low TSaP interval during N18 (Figure 9e).Hence, it is possible that the shift in ΔTM rate measured after DS-O18-2 and N-18 represents a delayed response of some TMs to earlier, higher magnitude events, that is, DS-O18-1 and DS-O18-3, respectively (Figure 8c).However, this delayed effect is not seen following DS-J18 as TM ML concentrations between the profiles sampled 2 and 5 days afterward, on 7 and 10.1.2018,respectively, are statistically similar (Figure 8a).
Although the studied dust storms were previously shown to substantially shift Al and Pb inventories (Benaltabet et al., 2020(Benaltabet et al., , 2022)), it seems that these storms often result in insignificant shifts in the TMs reported here (Figure 9).Still, in the cases where significant shifts are observed, dust storms prevalently lead to higher TM scavenging rates relative to dissolution rates, resulting in a net decrease in TM inventories (Figure 9).When TM enrichment following dust storms is observed, it is associated with anthropogenic sourced metals, that is, Ni, Cu, Zn, and Pb (Benaltabet et al., 2020), shown to be enriched in GoA aerosols (Chen et al., 2008;Chien et al., 2019;Torfstein et al., 2017).
Considering the entire October-November 2018 sampling session, significant shifts in deep-water TM concentrations (ΔTM DW ) were also observed between 18.10.2018and 7.11.2018for all metals but Cd (Figures 5p-5u).This suggests that the impact of dust storms is not limited to the upper mixed layer, as aerosol particles settling throughout the water column also hold the potential of modulating deep TM concentrations to a depth of at least 660 m.

Factors Controlling Scavenging and Dissolution Rates
The general inconsistency and variability in the response of TM ML concentrations to dust storms imply that other factors controlling scavenging and dissolution rates might be at play.Aerosol TM concentrations and solubility, mineralogy and particle size are all known to vary with aerosol provenance (Baker et al., 2020;Chance et al., 2015;Fishwick et al., 2018;Shelley et al., 2018).Hence, it is possible that the different post-storm TM scavenging and dissolution rates during different events are controlled by the different provenance of particles (Figure S7 in Supporting Information S1).
In addition, dust storm duration may also exert controls over TM scavenging and dissolution rates.The March 2018 sampling session, bracketing DS-M18, ranged over an extended period of high TSaP loads (Figure 8b).
Here, Co ML and Zn ML presented high scavenging rates (Figures 9b and 9e) and decreased continuously between 26, 28, and 29.3.2018(Figures 8b, 9b, and 9e), while Ni ML and Cu ML gradually increased as a result of aerosol leaching (Figures 8b, 9c, and 9d).Therefore, it seems that regardless of aerosol provenance, a continuum of high aerosol deposition will lead to a progressive and gradual change in TM ML .Contrarily, discrete storms such as DS-J18, will result in a sharp response of dissolved TMs (Figures 8a and 9).
Compared to dry deposition, wet deposition is known to increase the solubility of aerosol TMs and hence the soluble TM flux from atmospheric deposition (Chance et al., 2015;Jickells et al., 2016).Yet, the expected increase in ΔTM rate after the rain event that followed DS-O18-1 was displayed only by Al (Benaltabet et al., 2022) and to a lesser extent by Ni (Figure 9c; statistically significant at p < 0.1).This suggests that wet deposition may deliver higher fluxes of lattice-bound metals.Conversely, the flux of soluble anthropogenic TMs such as Cu, Zn, and Pb (Benaltabet et al., 2020) is not necessarily increased by wet deposition, as it makes little difference whether these soluble metals are delivered by dry deposition and dissolve directly in seawater or delivered already in the dissolved phase by rainwater.
The response of dissolved TMs to dust storms is conventionally evaluated through controlled aerosol leaching experiments, which often predict TM enrichment from dust (e.g., Mackey et al., 2015;Mendez et al., 2010;Thuróczy et al., 2010).However, these predictions are rarely replicated by the in situ results presented here, as net scavenging induced by dry deposition dust storms is much more prevalent than dissolution (Figure 9).Hence, it is possible that the high particle/solvent ratio often employed in aerosol leaching experiments (Mackey et al., 2015;Roy-Barman et al., 2021) actually over estimates the magnitude of dust storms, at least in comparison to the storms discussed here.Hence, extreme dust storms could potentially lead to higher dissolution rates and to a net addition of TMs.

Mn Scavenging Mechanisms
By excluding the low aerosol loads of DS-O18-2 and N18 (and DS-A17, see further), ΔMn rate and DISS rate form a linear correlation (R 2 = 0.99, Figure 9a).Similar trends were observed for dissolved Al (Benaltabet et al., 2022), indicated by the correlation between ΔMn rate and ΔAl rate (R 2 = 0.69, Figure 9h).Hence, it is possible that the post dust storm Mn scavenging mechanism is similar to that of Al (detailed in Benaltabet et al. (2022)), and is associated with pre-formed aerosol Mn oxides carried by dust storms (Torfstein et al., 2017).Accordingly, low particle flux dust storms delivered during times of a deep mixed layer (i.e., low TSaP total /MLD ratios; Equation 5), will result in low particle loads and hence low particulate Mn oxide concentrations in the mixed layer.These conditions will favor the scavenging of dissolved Mn onto aerosol Mn oxides over the photoreductive dissolution of these oxides (Figure 9a).As dust storms intensify and the MLD shallows (i.e., high TSaP total /MLD ratios; Equa tion 5), the higher particulate Mn oxide concentrations in the mixed layer will result in enhanced photoreduction and higher Mn DISS rate values.However, the positive change rate (ΔMn rate > 0) expected from the ΔMn rate -DISS rate linear regression for DS-A17 is not observed (Figure 9a).Hence, the dissolution of Mn during high dust loads might be limited by other factors controlling photoreduction rates such as irradiance levels, photoreducing bacteria abundances, and oxygen concentrations (Sunda & Huntsman, 1988, 1994;Van Hulten et al., 2017).The interplay between these factors can probably support Mn dissolution until an upper limit of particulate Mn oxide concentrations is reached.When this concentration threshold is abruptly crossed during high dust loads and shallow mixing depths, the photoreductive dissolution of Mn oxides will probably be inhibited.

Scavenging Onto Resuspended Sediments
In January 2017, four deep profiles (supplemented by an additional CTD profile) recorded the deepening of the MLD from 234 m on 4.1.2017to 390 m on 31.1.2017(Figures 4a and 10) and the concomitant surge in sinking water column particle flux during 15.1.2017(Benaltabet et al., 2022;Figures 3b and 10).These profiles present high and significant shifts in ΔTM ML and ΔTM DW in all elements (except Zn that was not measured) (Figure 10).
With the deepening of the mixed layer, deep waters are actively upwelled and entrained in the upper mixed layer, resulting in a possible shift in TM ML concentrations.To quantify the controls of mixed layer deepening over the measured variations in TM ML concentrations, the expected TM ML from vertical mixing and entrainment of upwelled deep waters (TM ML mix ) was calculated for each profile in January 2017 (except 4.1.2017)using Equation 6: where MLD pre and MLD post are the measured MLDs of two consecutive profiles and the difference between them (MLD post − MLD pre ) is a measure of the entrained deep waters.TM ML pre is the mixed layer weighted average concentration (Equation 1) of a given TM in the preceding profile, and TM ent is the TM weighted average concentration of the deep water entrained in the upper mixed layer.
In addition to mixing with deep waters, TM ML may also decrease as a result of micro-nutrient phytoplankton (co)uptake facilitated by the availability of entrained deep macronutrients.The TM inventory taken up by phytoplankton (TM ML uptake ) can be estimated from the difference between the measured PO 4ML and expected PO 4ML mix in each profile, assumed to represent the biological utilization of entrained deep PO 4 .The difference in PO 4 is then multiplied by the average TM/PO 4 regression slope ((TM/PO 4 ) reg ) of the North Atlantic, North Pacific, and Southern Oceans (Twining & Baines, 2013 and references therein), which reflects the TM/PO 4 stoichiometry in phytoplankton, as described in Equation 7: ) reg (7) Next, the expected TM ML from mixing and biological uptake (TM ML exp ) can be estimated by subtracting TM ML uptake from TM ML mix (Equation 8).Accordingly, any positive or negative deviation from TM ML exp could indicate resuspended particles induced dissolution or scavenging, respectively (note that PO 4ML exp is based on PO 4ML mix alone).
Afterwards, during 18.1.2017and 24.1.2017,the MLD rapidly deepened by 93 and 109 m, respectively.Concomitantly, a significant peak in sinking particle flux, attributed to a yearly-recurring sediment resuspension event (Chernihovsky et al., 2020;Torfstein et al., 2020), was registered on 15.1.2017(Figure 10).On 24.1.2017,measured PO 4ML is significantly lower than PO 4ML exp since any entrained deep PO 4 was rapidly utilized by primary productivity in the upper water column (Meeder et al., 2012;Zarubin et al., 2017).Still, statistically similar values are attained for TM ML and TM ML exp on 24.1.2017for all metals apart from Mn ML and Cd ML , which decreased and increased, respectively, due to the entrainment of lower Mn (Figure 4b) and higher Cd deep waters (Figure 4f).
In contrast to the upper water column, all deep-water TM concentrations registered a significant decrease of 20%-31% in TM DW between 10.1.2017and 24.1.2017(Figure 10).In addition, the Cd-PO 4 correlation dropped to a value of R 2 = 0.01, compared to 0.72 on 10.1.2017(Table S3 in Supporting Information S1), invoking a sink for TMs but not PO 4 .This deep-water sink is most likely related to scavenging onto resuspended sediments (Benaltabet et al., 2022;Figure 11) rich in Fe-Mn oxides (Steiner et al., 2019).In addition, it is possible that the resuspension of sediments led to the release of sedimentary pore water Fe and Mn (80 and 100 μmol kg −1 , respectively; Blonder et al., 2017;Steiner et al., 2019).Upon contact with overlaying oxygenated waters, the excess dissolved Fe and Mn triggered the rapid formation of Fe-Mn oxides and subsequent scavenging of dissolved TMs (Hawco et al., 2018;Murray, 1975;Tonkin et al., 2004).Afterwards, although the MLD only deepened by 29 m between 24.1.2017and 31.1.2017,large variations in TM inventories were registered across the entire water column (Figure 10).In the deep waters, all TM concentrations apart from Cd DW continued to decline by 16%-36% due to scavenging.In addition, during 31.1.17,significant shifts in TM inventories were registered in the upper mixed layer as well, with Mn ML , Co ML , Ni ML , Cu ML , and Cd ML dropping by 26%, 32%, 41%, 28%, and 58%, respectively (Figure 10).Furthermore, during 31.1.2017,significantly large deviations from TM ML exp are observed for all TMs, suggesting that in addition to inducing TM scavenging in the deep waters, resuspended sediments and/or Fe-Mn oxides, also promoted TM scavenging in the upper mixed layer via active upwelling of deep particles (Benaltabet et al., 2022;Torfstein et al., 2020).As such, scavenging processes are responsible for 73%, 87%, 92%, 93%, and 88% of the decrease in Mn ML , Co ML , Ni ML , Cu ML , and Cd ML , respectively, measured between 24.1.2017and 31.1.2017.Interestingly, the effects of scavenging in the upper mixed layer are only evident during 31.1.2017,∼2 weeks after the peak in sinking particle flux (Figure 10).This probably reflects a methodological bias related to the difference between the relatively coarser and faster sinking particulate matter collected in the sediment traps (Buesseler et al., 1992) compared with the relatively fine grain resuspended particles, which are an efficient scavenger of TMs (Jeandel et al., 2015;Lam et al., 2015).The latter remained suspended in the water column for ∼2 weeks after the resuspension event, resulting in a measurable effect on TM inventories even after this time period.

Flood Induced Scavenging
Compared to the profiles measured on 14.2.201814.2. and 26.3.201814.2. , TM concentrations on 8.3.2018 are markedly lower, and are among the lowest concentrations registered throughout the study (Figures 5i-5o).Based on our continued monitoring of the water column and considering the short time window of this drop in TM concentrations, it is highly likely that this was triggered by particle scavenging (in contrast to the admixing of a new, TM-depleted water mass).However, no major dust storms (Figure 3b) or sediment resuspension events (Figure 3a) occurred in the days prior to 8.3.2018,entailing a different source of particles.Indeed, the rain events taking place on 22-24.2.2018 resulted in major flash floods (Figure 3b) in two ephemeral streams: the Kinnet Canal and Shahamon Stream (Figure 1).GoA flash floods are known to carry high sediment loads that can be transported along surface waters over large distances for long periods of time (Katz et al., 2015;Lamb & Mohrig, 2009).Hence, while the flood waters are characterized by significant enrichment of dissolved TMs (Table S5 in Supporting Information S1), we suggest that the fine fraction of the fluvial particulate matter was retained in suspension for at least 2 weeks and led to the scavenging removal of dissolved TMs in the open-water Station A (Figures 5i-5o), some ∼7-10 km away from the estuaries (Figure 1).This is further supported by low salinity measured in the upper 80 m in Station A during 13.3.2018(Figure 5h), which may correspond to lower salinity flood waters (Katz et al., 2015).

Summary
Dissolved Mn, Co, Ni, Cu, Zn, Cd, and PO 4 concentration profiles are reported in a deep, oligotrophic, open ocean proxy site (the Gulf of Aqaba, northern Red Sea).The results are discussed in the context of daily resolution measurements of atmospheric aerosol loads and water column sinking particle fluxes.Seawater sampling focused on daily time scale dust storms, episodes of sediment resuspension and flash floods, to quantitatively estimate the direct in situ long-and short-term impact of these perturbative events over dissolved trace metal inventories.
Upper mixed layer Co, Ni, Cu, Zn and Cd inventories are negatively correlated with increasing aerosol loads due to scavenging induced by atmospheric particles.The highest correlations are achieved when considering aerosol loads integrated over 120-144 hr, suggesting that the peak aerosol-derived signal offset in these TM in seawater is reached ∼5-6 days after aerosol deposition.In contrast, dissolved Mn is not affected by long-term varying aerosol loads because of the stronger controls exerted by surface water redox processes.
On shorter time scales, scavenging induced by dust storms abruptly drives down Mn, Co, Ni, Cu, Zn, and Cd mixed layer concentrations by up to 16%, 15%, 6%, 12%, 44%, and 42%, respectively (Figure 11).Although less prevalent than scavenging, the partial dissolution of settling dust storm particles (most likely of an anthropogenic origin) increases Ni, Cu, and Zn mixed layer concentrations by up to 5%, 5%, and 71%, respectively (Figure 11).No enrichment of Mn, Co, Cd, and PO 4 was registered following dust storms.These results demonstrate that atmospheric deposition, which serves as a long-term source of TM to the oceans, also acts as a short-term sink.
A rapid deepening of the mixed layer coupled with a major sediment resuspension event resulted in a decrease of 16%-36% in deep water trace metal concentrations due to scavenging induced by resuspended particles (Figure 11).Moreover, a drop in trace metal concentrations of 26%-59% was registered in the upper mixed layer ∼2 weeks after the sediment resuspension event, due to scavenging onto fine resuspended particles that were actively upwelled to the surface waters (Figure 11).
Trace metal scavenging was identified in association with a flash flood, suggesting that fine flood particulate matter can be transported over long distances and result in the removal of TM in open surface waters (Figure 11).
The unprecedented temporal sampling resolution implemented here provides first order quantitative estimates of the in situ impact of dust storms and sediment resuspension events on dissolved seawater TM.By employing site-specific adjustments, the scavenging and dissolution rates generated here may be implemented in large-scale open ocean studies to help interpret trace metal distributions in the context of interaction with terrigenous particulate matter.Rivlin and Yonathan Shaked of the NMP for assisting with nutrient measurements at the IUI.We further wish to thank Barak Yarden, the IUI marine crew, and Guilhem Banc-Prandi for their assistance in seawater sampling.We are grateful for the constructive comments of two anonymous reviewers, who helped improve this manuscript significantly.This work is part of the GEOTRACES REDMAST process study (Glpr09) and was funded by the Israel Science Foundation Grant 834/19 to AT.Additional support to TB was provided by the Bester and Pfeifer scholarships, and the HUJI Advanced School for Environmental Studies scholarship.

Figure 1 .
Figure 1.Location map.(a) The Red Sea and surrounding Saharan and Arabian Deserts.(b) The Gulf of Aqaba at the northern Red Sea.(c) Locations of the time series study site at Station A (∼700 m depth), sediment trap mooring, on-shore research station (IUI), total suspended atmospheric particles monitoring stations (TSaP), and meteorological station rain gauge (MS).The Shahamon Stream and Kinnet Canal and the corresponding seawater sampling locations are represented by broken red lines and triangles, respectively.Aerial footage from Google Earth. Figure adapted from Benaltabet et al. (2022).

Figure 3 .
Figure 3. (a) Sinking particle flux (g m −2 day −1 ) at Gulf of Aqaba during 2017-2018.The fluxes were measured at a daily resolution by a sediment trap deployed near Station A (Benaltabet et al., 2022; Chernihovsky et al., 2020; Torfstein et al., 2020).Sediment resuspension episodes during winter are reflected by peaks in particle fluxes (note the broken Y-axis).(b) Total suspended atmospheric particles (TSaP, black line, μg m −3 ) time series (6-hr resolution) as measured by the Israel Ministry of Environmental Protection monitoring stations.Dust storms are depicted by peaks in TSaP.Blue vertical bars represent precipitation events (Israel Meteorological Service).Subsequent flash flood events are represented by black stars.Thin gray vertical bars denote the timing of seawater sampling cruises and pale blue bars outlined by broken lines represent sediment resuspension events and dust storms bracketed by seawater profile sampling.
Figure 3. (a) Sinking particle flux (g m −2 day −1 ) at Gulf of Aqaba during 2017-2018.The fluxes were measured at a daily resolution by a sediment trap deployed near Station A (Benaltabet et al., 2022; Chernihovsky et al., 2020; Torfstein et al., 2020).Sediment resuspension episodes during winter are reflected by peaks in particle fluxes (note the broken Y-axis).(b) Total suspended atmospheric particles (TSaP, black line, μg m −3 ) time series (6-hr resolution) as measured by the Israel Ministry of Environmental Protection monitoring stations.Dust storms are depicted by peaks in TSaP.Blue vertical bars represent precipitation events (Israel Meteorological Service).Subsequent flash flood events are represented by black stars.Thin gray vertical bars denote the timing of seawater sampling cruises and pale blue bars outlined by broken lines represent sediment resuspension events and dust storms bracketed by seawater profile sampling.

Figure 4 .
Figure 4. Dissolved trace metal water column profiles associated with sediment resuspension and dust storms sampled in Station A during 2017.Upper panels depict (a) density, (b-f) dissolved trace metal and (g) PO 4 concentrations sampled before and after the sediment resuspension event in January 2017.Note that the CTD profile sampled on 18.1.17in panel (a) was not supplemented with trace metal measurements.Lower panels represent (h-l) trace metal and (m) PO 4 concentrations sampled before and after dust storms occurring in April and May 2017 (DS-A17 and DS-M17).Horizontal error bars represent 1σ from the mean of duplicate-triplicate extractions.

Figure 5 .
Figure 5. Dissolved trace metal water column profiles associated with flash floods and dust storms sampled in Station A during 2018.Upper panels represent (a-f) dissolved trace metal and (g) PO 4 concentration profiles sampled before and after a dust storm occurring in January 2018 (DS-J18).Middle panels depict (h) salinity, (i-n) dissolved trace metal and (o) PO 4 concentrations sampled before and after the flash flood and dust storms occurring on 24.2.2018 and March 2018 (DS-M18), respectively.Lower panels represent (p-u) trace metal and (v) PO 4 concentration profiles sampled before and after dust storms occurring in October-November 2018 (DS-O-1, DS-O-2, DS-O-3, and N18).Horizontal error bars represent 1σ from the mean of duplicate-triplicate extractions.

Figure 6 .
Figure 6.(a) Correlation values (R 2 ) between dissolved trace metal mixed layer concentrations (TM ML ) and integrated TSaP loads at varying time intervals prior to seawater sampling.Results from previously published Al concentrations(Benaltabet et al., 2022), and Pb concentrations and isotopes(Benaltabet et al., 2020) are from the same data set.Background profiles (full symbols, TableS4in Supporting Information S1) display higher correlations to TSaP relative to the entire set of profiles (empty symbols, TableS4in Supporting Information S1).The highest correlations for Co ML , Ni ML , Cu ML , Cd ML and Al ML were achieved for TSaP integration periods of 120-144 hr, whereas Mn ML and PO 4ML R 2 values remained poor.(b-k) TM ML versus integrated TSaP.The displayed TSaP integration interval was chosen as the interval yielding the highest correlation to TM ML , as depicted in panel (a) by broken line squares.Lower TM ML concentrations in background profiles (full symbols) are correlated with increased aerosol loads.Profiles sampled after dust storms, events of sediment resuspension, and flash floods (empty symbols) are separated from the generally negative trends of background TM ML .Vertical error bars represent TM ML uncertainty calculated as 2σ from the mean of a Monte Carlo simulation (n = 10,000).
Figure 6.(a) Correlation values (R 2 ) between dissolved trace metal mixed layer concentrations (TM ML ) and integrated TSaP loads at varying time intervals prior to seawater sampling.Results from previously published Al concentrations(Benaltabet et al., 2022), and Pb concentrations and isotopes(Benaltabet et al., 2020) are from the same data set.Background profiles (full symbols, TableS4in Supporting Information S1) display higher correlations to TSaP relative to the entire set of profiles (empty symbols, TableS4in Supporting Information S1).The highest correlations for Co ML , Ni ML , Cu ML , Cd ML and Al ML were achieved for TSaP integration periods of 120-144 hr, whereas Mn ML and PO 4ML R 2 values remained poor.(b-k) TM ML versus integrated TSaP.The displayed TSaP integration interval was chosen as the interval yielding the highest correlation to TM ML , as depicted in panel (a) by broken line squares.Lower TM ML concentrations in background profiles (full symbols) are correlated with increased aerosol loads.Profiles sampled after dust storms, events of sediment resuspension, and flash floods (empty symbols) are separated from the generally negative trends of background TM ML .Vertical error bars represent TM ML uncertainty calculated as 2σ from the mean of a Monte Carlo simulation (n = 10,000).

Figure 9 .
Figure 9. (a-f) Mixed layer weighted average trace metal change rates (ΔTM rate , Equation4) and theoretical trace metal dissolution rates (DISS rate , Equation5) in response to dust storms.Statistically significant and non-significant shifts following dust storms are depicted by full and empty circles, respectively.Statistically significant and non-significant shifts following low dust intervals (DS-O18-2, N18) are represented by full and empty squares, respectively.Negative ΔTM rate values indicate higher dust storm induced scavenging rates with respect to dissolution rates while positive ΔTM rate values indicate higher dissolution rates.Error bars represent ΔTM rate uncertainty propagated after TM ML 2σ calculated from the mean of a Monte Carlo simulation (n = 10,000).(h) Relatively high correlations between ΔAl rate(Benaltabet et al., 2022) and ΔMn rate suggest that the post dust storm Mn scavenging mechanism is similar to that of Al.

Figure 10 .
Figure 10.Impact of a sediment resuspension event (January 2017) on dissolved trace metal concentrations.Overlain on sinking particle fluxes (bottom panel) is the mixed layer depth (stars), which sharply increases between 10.1.2017and 18.1.2017.Presented are measured mixed layer (TM ML , full symbols) and deep-water (TM DW , crossed symbols) trace metal concentrations and expected mixed layer trace metal concentrations from vertical mixing and biological uptake (TM ML exp , empty symbols, broken line, Equation7).Note the shifts between TM ML and TM ML exp during 31.1.2017,representing the scavenging of trace metals onto resuspended particles in the upper mixed layer of the water column.

Figure 11 .
Figure11.Conceptual model summarizing the response of dissolved trace metals (TM) to perturbative events.Pb and Al dynamics were described byBenaltabet et al. (2020Benaltabet et al. ( , 2022)).Solid red arrows represent scavenging of dissolved TMs onto particles and broken blue arrows represent the dissolution of TMs from particles to surrounding waters.Dust storm-related settling aerosols induce substantial dissolved TM scavenging of all studied metals, both across the mixed layer and at depth.However, dust storms seldom lead to TM enrichment, which is associated with anthropogenic-sourced metals.Episodes of sediment resuspension promote scavenging of TMs at depth and across the upper water column, as small resuspended particles are upwelled to surface waters across the mixed layer.Flash floods bring to significant enrichment in dissolved TMs at the coast.Concomitantly, flood particles carried along surface waters promote the scavenging removal of TMs at open waters.
Dates and Names of the Dust Storms Bracketed by the Sampling of Seawater Profiles (Benaltabet et al., 2022)sedimenttrap samples were collected proximal to Station A at a daily timescale resolution(Benaltabet et al., 2022).The long-term average particle flux at 410 m depth was Note.Also presented are the values for the parameters used in Equation5: the mixed layer depth (MLD), total TSaP (TSaP total ), and time (Δt) between profiles bracketing dust storms.See also Figures7 and 8for depiction of TSaP loads with respect to seawater sampling.aDS-J18 is defined as the interval bracketed by the profile samples during 04.01.2017 and 07.01.2017 although an additional profile was sampled during 10.01.2017.bDS-M18 is defined as the interval bracketed by the profile samples during 26.03.2018 and 29.03.2018although an additional profile was sampled during 28.03.2017.cDS-O18-1was followed by a precipitation event delivering 5.6 mm of rain over ∼1 hr.dThe N-18 interval did not feature a dust storm and represents low (background) TSaP levels.Table1