Figure 1 shows that marine isoprene emissions in the MarOrg simulation resulted in a change in surface iso- prene concentrations from the simulation with default treatments over marine regions. The strong summertime solar radiation and high [Chl-a] levels in northern Atlantic and Pacific Oceans contributed to the high emission rates in those regions, while moderate ocean biological activity throughout the year over much of the Southern Ocean also resulted in high marine isoprene emission rates. Over remote marine regions, marine isoprene emissions increased surface isoprene concentrations by 5 - 10 ppt (see Figure 1). This increase was within the range of most isoprene concentrations (<1 to 300 ppt, typically < 50 ppt) observed over remote marine regions [3] . The large emission rates occurring near coastal regions (and some inland lakes) were due to the combination of isolated regions of very high [Chl-a] and coarse (4˚ × 5˚) grid resolution. The greatest increases in surface isoprene concentrations were predicted over the northern Atlantic/Pacific and Southern Oceans where marine isoprene

emission rates are high. These areas also had low oxidant levels and therefore long lifetimes of isoprene emitted from marine sources and transported from terrestrial regions. Over terrestrial areas such as the rainforests of South America, Southeast Asia, and Africa, a mixture of positive and negative changes in the surface isoprene concentrations is evident. Such changes were likely due to the fact that terrestrial emissions were computed online and were sensitive to the slight variability in meteorology occurring due to model reinitializations. Although the changes in surface isoprene concentrations over terrestrial areas were of similar magnitude as those of marine regions, they were much smaller when calculated in terms of percentage changes (typically < 5%).

Predictions of other atmospheric chemistry-relevant gases in remote marine regions were also affected by emissions of marine isoprene. The surface concentrations of hydroxyl radical (OH) decreased by 10⁻³ ppt or more in parts of the northern Pacific and Southern Oceans(see Figure 1) because of its reaction with the additional isoprene from marine sources. Small changes in surface O₃ concentration due to marine isoprene emissions(both increases and decreases of over 0.1 ppb) were also predicted; oceanic regions immediately downwind of anthropogenic pollution centers (China, U.S., etc.) experienced increased concentrations while remote marine regions typically have decreased concentrations (Figure 1). Decreases in predicted surface O₃ concentrations over most of the eastern North Pacific Ocean were consistent with Gantt et al. [16] , who found that the emission of isoprene and monoterpenes from marine sources caused a small (−0.1%) decrease in CMAQ-predicted O₃ mixing ratios off the Pacific coast of the U.S.

Although coincident measurements of isoprene emissions and concentrations over remote marine regions are very limited [3] , the May 2001 simulation period allowed for model evaluation of predicted isoprene emissions/ concentrations in the northwest Pacific Ocean [38] . Compared to the Matsunaga et al. [38] observations, the model overpredicted marine isoprene emissions (especially during the daytime) and underpredicted surface concentrations (Figure 2). Concurrent over predictions of emissions and under predictions of surface concentra- tions suggested that the simulated isoprene lifetime by GU-WRF/Chem over the marine boundary layer was too short, potentially due to the coarse grid resolution. The low temporal resolution of the observed emissions in Matsunaga et al. [38] made a more quantitative analysis of the diurnal cycle of emissions difficult, but the few observations available were inconsistent with the large increase in daytime emissions predicted by the model. The very low surface isoprene concentrations predicted by the Default simulation do not support the suggestion from Matsunaga et al. [38] that terrestrial isoprene sources influence the surface concentrations in this region;

although we cannot definitely rule out the possibility of terrestrial isoprene sources affecting surface concentrations because of the use of a coarse model grid resolution in our simulations, proximity to the Japanese coast (~500 km) and the short isoprene lifetime predicted by GU-WRF/Chem in the marine boundary layer. The chemical mechanism used in GU-WRF/Chem, CB05-GE, is part of series developed for urban photochemistry [41] that may have difficulty representing remote marine conditions despite recent updates [38] . Nonetheless, according to Figure 2, the inclusion of marine isoprene emissions in the MarOrg simulation brought the predicted concentrations closer to observations.

Globally, the implementation of marine isoprene and POA emissions resulted in increased surface PM_2.5 organic aerosol concentrations (> 0.5 µg∙m⁻³) over biologically-active remote marine regions in the northern

Atlantic/Pacific and Southern Oceans (Figure 3). While the predicted increases in organic aerosol concentration occurred year-round in the Southern Ocean due to the persistent strong winds and moderate [Chl-a], in- creases in northern Atlantic and Pacific were the highest in the spring and lowest in the winter months. Surface organic aerosol concentrations in tropical oceanic regions were largely unchanged between the Default and Mar Org simulations due to low biological activity and sea spray aerosol emissions relative to mid- and high-latitude oceanic regions. Comparison of the predicted concentrations with observations in clean marine conditions com- piled from the years 1973 to 2009 [9] indicated a latitudinal model bias, with over predictions in mid- and high- latitude regions and underpredictions in the tropics. Although only a few sites have measurements throughout the year (e.g., Mace Head, Amsterdam Island), comparison with model predictions in Figure 3 suggests that the MarOrg simulation did not replicate the observed seasonal cycle of concentrations. This result is consistent with Gantt et al. [42] who found that none of the existing sea spray aerosol-based marine POA emission parameterizations are able to replicate the observed seasonality of organic aerosol concentrations.

Comparison of GU-WRF/Chem and aerosol mass spectrometer observations from Mace Head, Ireland [39] for the year 2009 revealed that the inclusion of marine isoprene and POA emissions in the MarOrg simulation improved the prediction of surface submicron organic aerosol mass concentrations (see Figure 4). Without marine isoprene and POA emissions, the Default simulation frequently predicted surface organic aerosol concentrations

at Mace Head < 0.025 µg∙m⁻³ during clean marine conditions (defined as periods having aerosol number concentration < 700 cm⁻³, onshore wind direction, and black carbon concentrations < 0.05 µg∙m⁻³) while the observed concentrations typically ranged between 0.05 and1 µg∙m⁻³ [39] . Organic aerosol concentrations rarely fell below 0.05 µg∙m⁻³ in the MarOrg simulation. Model-measurement discrepancies persisted in the MarOrg simulation, especially during the marine organic aerosol plume period (16 August 2009) observed by Ovadnevaite et al. [40] . During this plume event, the MarOrg simulation predicted organic aerosol concentrations (0.2 µg∙m⁻³) that were an order of magnitude lower than the observed maximum concentrations (3.8 µg∙m⁻³). The change of annual average submicron organic (secondary and primary) aerosol concentrations at Mace Head between the Default and MarOrg simulations was 0.15 µg∙m⁻³, a value similar to the observed annual concentrations (0.25 µg∙m⁻³) at Mace Head during clean marine conditions.

During the marine organic aerosol plume (16 August 2009) observed by Ovadnevaite et al. [40] (see Figure 5), the MarOrg simulation had a slightly higher aerosol number concentration than the Default simulation in the size range over which both simulations had the largest underpredictions (0.05 to 0.5 µm in diameter) relative to observations. Much of this difference is attributable to the increased organic aerosol number concentrations for diameters > 0.05 µm. In other clean marine periods where the MarOrg and Default simulations predicted substantially different submicron organic aerosol mass concentrations, the increase in the aerosol number concentration for the MarOrg simulation in the 0.05 to 0.5 µm diametersize range is more evident. These model results are consistent with the observations of multiple marine organic aerosol plumes from Ovadnevaite et al. [40] , who found that periods with the highest mass fraction of organics in submicron aerosols also had aerosol number size distributions peaking between 0.05 and 0.2 µm in diameter. It’s clear, however, that the MarOrg simulation did not accurately simulate this marine organic aerosol plume both in terms of magnitude and size distribution.