Weekly sampling was conducted at Tudor Hill, Bermuda from November 2018 to March 2020. The sampling tower platform is approximately 30 meters above sea level with coordinates of 32.265°N, 64.879°W.

The following collection and chemical processing methods are reproduced from related dataset https://www.bco-dmo.org/dataset/906770 (doi: 10.26008/1912/bco-dmo.906770.1) and Sedwick et al. (2023).

Composite samples of bulk aerosol were collected atop the 23-meter height sampling tower at Tudor Hill, Bermuda, on an approximately weekly basis from November 2018 through March 2020, bracketing the four BAIT project cruises. A high-volume (~700 liters per minute) air sampler was used to collect aerosols on acid-cleaned Whatman-41 cellulose filters (8"x10", 20-micrometer nominal pore size), which collect particles as small as 1 micrometer (µm) with >90% efficiency (Stafford and Ettinger, 1972). The coastal Tudor Hill site faces into the prevailing southwesterly winds, so aerosols were only collected during winds >1 meter per second (m s-1) from the 210°-315° sector in order to avoid local sources. Sample filters were stored in zip-lock polyethylene bags in a vacuum desiccator at room temperature. For analysis of total aerosol iron, 1/16th portions of the aerosol sample filters were digested with a mixture of ultrapure concentrated nitric and hydrofluoric acids and hydrogen peroxide (Fisher Optima) in Teflon vessels (Morton et al., 2013), using a microwave heating system (CEM MARS 6), then evaporated on a hot plate and diluted to volume with 2% (v/v) ultrapure nitric acid. Replicate 1/16 portions of the aerosol filters were also subjected to a flow-through leaching procedure modified from Buck et al. (2006). Briefly, the aerosol filter portions were placed atop an acid-washed 0.2 µm pore polycarbonate membrane filter loaded into a perfluoroalkoxy (PFA) resin filtration tower (Savillex), leached with 250 milliliters (mL) of high-purity deionized water (DIW, Barnstead Nanopure, >18.2 MΩ-cm resistivity), and the leachate acidified to 0.4% (v/v) with 6 M ultrapure hydrochloric acid (Fisher Optima) for analysis of "DIW-soluble aerosol iron". The same filter portions were then subjected to a batch leaching procedure modified from Kadko et al. (2019) using 25% acetic acid (HOAc, Fisher Optima) and 0.02 M hydroxylamine hydrochloride (Sigma) at 90 degrees Celsius (°C), and the supernatant leachate was evaporated and then diluted in 2% ultrapure nitric acid (Fisher Optima) for analysis of "HOAc-soluble aerosol iron". Field blanks for the aerosols (an acid-cleaned filter) were deployed on the Tudor Hill tower and processed in the same manner as samples, but without operating the aerosol sampler pump.

Iron (Fe) isotope Analysis:
Subsamples of the total digested, DIW-leached, and HOAc leached aerosol samples were taken for Fe isotope analysis at the University of South Florida. Samples were then processed for dissolved Fe isotope and concentration analysis at the University of South Florida following Sieber et al. (2019), modified from Conway et al. (2013) and Conway et al. (2019). For all samples a 57Fe-58Fe double spike was added prior to chemical processing, in an approximate 1:2 sample:spike ratio based on expected concentrations from other analysis. Total digest aliquots were then evaporated to dryness, while leached samples were put through a batch extraction using Nobias PA-1 chelating resin. All samples were refluxed with a HNO3-H2O2 solution to digest organics, followed by purification by anion-exchange chromatography using AG-MP1 resin. Isotope analyses were then performed on a Thermo Neptune Plus MC-ICPMS in the Tampa Bay Plasma Facility at the University of South Florida using the double spike technique via a ~100 microliters per minute (uL min-1) PFA nebulizer and Apex Ω introduction introduction system, Pt Jet Sampler cone and an Al X skimmer cone.

Fe stable isotope ratios are expressed in delta notation (δ56Fe) relative to the IRMM-014 standard. A secondary Fe standard, NIST-3126, was analyzed over 44 sessions to provide an estimate of long-term instrumental precision. We obtain a value of +0.36 ± 0.05‰ (2SD, n = 524; runs = 37), in agreement with consensus values (Hunt et al. 2022; Conway et al., 2013). Therefore, we consider a 2SD uncertainty of 0.05‰ as an estimate of analytical precision, and have applied it to all samples, except for low concentration samples where the larger internal error is considered a more conservative estimate of uncertainty. The isotope method described above yields a total procedural blank of ~0.4 nanograms (ng) Fe (n = 5) per sample, which corresponds to less than 0.5% of sample measured here. Aerosol compositions have not been corrected for blank contribution.