The Weather Research and Forecasting model [18] inline coupled with chemistry (WRF/Chem) [19] was used in the setup of [12] , [20] , [11] . A further-developed Grell-Dévényi cumulus-ensemble scheme [21] and the WRF- Single-Moment 5-class scheme [22] determined cloud processes on the cumulus and resolvable scale, respectively. Shortwave and long-wave radiation processes were accounted for by the Goddard two-stream multi-band scheme [23] and Rapid Radiative Transfer Model [24] , respectively. Aerosol-radiation and cloud-radiation feedbacks followed [25] . Surface and atmospheric boundary layer (ABL) processes followed [26] . The exchange of heat and matter at the atmosphere-surface interface was calculated by the modified NOAH land-surface scheme [27] , [18] , which also predicted soil-moisture and temperature, frozen ground and snow conditions.

The chemical mechanism [28] used inline-calculated photolysis rates [29] . Dry deposition of gases was calculated following [30] with the modifications for Alaska by [31] . The Modal Aerosol Dynamics Model for Europe [32] and Secondary Organic Aerosol Model [33] considered aerosol chemistry, dynamics and physics, some aqueous phase reactions, and aerosol removal.

In all simulations, biogenic emissions were determined inline based on the meteorological and soil conditions following [34] [35] . Furthermore, WRF/Chem considered sea-salt uptake. This means all simulations consider these natural emissions.

In 2008, 153 cruise ships entered Glacier Bay during the 92-day peak season (1 June - 31 August) and 184 in total between May 15 and September 15. This timeframe is hereafter referred to as the tourist season or season. The reference simulation (REF) utilized this timeframe in 2008 to make inferences on activity-based emissions [12] .

The 2008 activity-based ship emissions were calculated based on each cruise ship’s characteristics (number and type of engines, engine power, maximum speed, fuel-type used, capacity) and the 2008 actual voyage data (ship position, cruise speed, operation mode) [12] . Hourly emission rates were determined for NO_x, SO₂, VOC, carbon monoxide (CO), ammonia (NH₃), and particulate matter (PM). VOC and PM-emission rates represented bulk-emission rates of several different species. PM was split into particulate matter of less than 2.5 mm (PM_2.5) and 10 mm (PM₁₀) in aerodynamic diameter at a 9:1 ratio following [3] . Furthermore, PM_2.5 was split into unspecified PM_2.5, sulfate, nitrate, organic matter, and black carbon following [3] [36] . VOCs were split into hexane, ethylene, propylene, acetylene, benzene, toluene, xylene, and others following [3] .

The emission inventory for increased quotas encompassed the activity-based emissions of 2008 [12] plus the emissions due to additional cruises. The simulation using this emission inventory and its results is hereafter referred to as QTA. As we realistically assumed that the “peak season” quota would be met should the NPS increase it from 153 to 184 (two ships per day), we also assumed in accord with [1] that demand would result in two ships per day during May and September. Thus, our QTA could be considered a “worst-case” scenario where 248 total ships enter the park, a 35% increase from the same period in 2008.

For simplicity, we assumed that increases in ship entries reflected the fraction of the current fleet. In the following, we call the cruise-ships of the two major companies, which currently are allowed in Glacier Bay, group 1 and group 2, while we combined the cruise ships from all the other companies into a group referred to as group 3. Again, for simplicity, we applied the 2008 fractional share of entrances of 37%, 50%, and 13% for the additional cruise ships in these groups. Consequently, group 1, group 2, and group 3 had 24, 32, and 8 additional entrances, respectively.

The cruise-ship type with the most entrances in 2008 in each group was chosen as representative for the additional cruise ships of that group. The representative ships of the three groups had total auxilliary engine powers of 17,700, 14,000, 10,100 kW and total main engine powers of 40,000, 39,000, and 40,000 kW with maximum cruise speeds of 12.3, 11.8, and 12.9 m・s⁻¹, respectively. In cruising mode, the auxiliary engine loads amount 50%, 60%, and 30%, respectively. All additional ships used intermediate fuel oil (IFO380) or marine gas oil (MGO) for their main engines, and heavy fuel oil (HFO) or marine diesel oil (MDO) for their auxilliary engines.

The typical paths, which these three cruise ships traveled within Southeast Alaska on their way to, in and from Glacier Bay in 2008 served as routes of the additional cruises. This means additional emissions not only occurred in Glacier Bay, but also in Southeast Alaska along the routes outbound to, and inbound from Glacier Bay. Since ship emissions not only depend on the ship’s characteristics, but also on cruise speed, operation mode, and engine load, activity-based emissions were calculated assuming a typical cruising behavior for the additional ships (Figure 1). Like for the 2008 activity-based emissions, the emissions of particulate matter and precursor gases for the additional ships were calculated in dependence of the ships’ individual characteristics and voyage data following [37] .

The days with only one or with no cruise-ship entrance in 2008 were assigned to the three groups under consideration of the aforementioned share. To implement the quota of two entrances in Glacier Bay per day on all days, six additional ships were needed as they have to cruise to, in and from Glacier Bay (Figure 1). The designed cruise-schedules ensured entrance of Glacier Bay on the assigned days. To realize the above share and quotas, each of the three groups was assigned two additional cruise ships. Depending on their assigned cruise paths, the additional ships cruised three to five days within the area covered by the model domain. Consequently, the realization of the in total 64 additional entrances into Glacier Bay during the season resulted occassionally in up to six additional ships cruising Southeast Alaska per day.

The WRF/Chem domain was the same as used by [20] . It encompassed the atmosphere over the eastern Gulf of Alaska, Southeast Alaska and western Canada centered at 58.5 N, 135.5 W with a 7 km increment of 120 ´ 120 grid-points and 28 vertically stretched layers to 100 hPa. To exclude lateral boundary effects from the analysis we discarded the five outermost grid-points leaving 110 ´ 110 grid-points for analysis (Figure 2). Doing so is common in limited area modeling [20] .

Three simulations were performed for 15 May 2008 to 15 September 2008 with identical model setup except for the cruise-ship emissions. One simulation did not consider emissions from cruise-ships. It only considered the natural emissions. This simulation is called CLN as it represents a clean atmosphere. It serves to provide context to understand the overall impacts of cruise-ship emissions on air quality and visibility. Note that the other two simulations considered biogenic emissions as well.

The reference simulation (REF) assumed the activity-based emission of 2008 [12] and is identical to the reference simulation in [12] [20] . For an evaluation of this simulation, see [11] [12] .

The third simulation (QTA) assumed an increase in cruise-ship volume to two ships per day every day from 15 May - 15 September in Glacier Bay. The goal for this simulation was to compare visibility and air-quality conditions between the existing (REF) conditions and if the park decided to increase the ship-entry quota.

All simulations were initialized every five days with respect to meteorology using downscaled 1˚ × 1˚, 6 h- resolution National Centers for Environmental Prediction global final analyses data [38] for sea-surface temperatures (SST), meteorological, snow and soil quantities. The meteorological data of these analyses also served as lateral boundary conditions for all three simulations.

All three simulations used identical idealized profiles of clean background concentrations for initialization. On all following days, the chemical fields at the end of the previous day served as initial conditions for the following

day. Identical idealized profiles of clean air background concentrations served as lateral boundary conditions in all three simulations. This design ensured that any differences in atmospheric trace-gas and particulate-matter concentrations, visibility as well as in meteorological fields between CLN and REF, CLN and QTA, as well as REF and QTA were only due to cruise-ship emissions.

The reference simulation was evaluated by means of twice-daily radiosonde data at Yakutat and hourly observations of 2 m temperature, 2 m dewpoint temperature, 10 m wind speed, and wind direction from 31 land and 11 buoy sites [11] . WRF/Chem performed well in predicting the vertical temperature profiles with overall bias (simulated minus observed), root-mean-square error (RMSE) and correlation (R) of 0.1˚C, 1.8˚C, and 0.84, respectively. A cold bias of −1.4˚C to −0.3˚C occurred within the first 600 m above ground. Over the entire vertical profile and all profiles of the season, dewpoint-temperature bias, RMSE and R were 0.8˚C, 4.6˚C and 0.60, respectively. Below 1500 m, dewpoint-temperature biases amounted 0.2˚C to 1.2˚C. Overall wind-speed bias, RMSE and R amounted to ~0 m・s⁻¹, 4.13 m・s⁻¹ and 0.69. Correlation R increased with increasing height. In the lower ABL, WRF/Chem often underestimated wind speed by up to 1.2 m・s⁻¹ due to subgrid-scale channeling effects. Over the entire vertical profile and all profiles, wind-direction bias, RMSE and correlation were −12˚, 9˚ and 0.66, respectively [11] .

WRF/Chem captured the temporal evolutions of spatial mean 2 m air and 2 m dewpoint temperatures well, and those of 10 m wind speed and direction acceptably [11] . On average over all surface sites and the tourist season, biases of 2 m air and 2 m dewpoint temperatures, 10 m wind speed and 2 m relative humidity were −0.6˚C, 0.2˚C, 1.75 m・s⁻¹ and 2% (absolute), respectively. Performance varied among months.

The analysis of the impacts of increased quotas on air quality and visibility focused generally on Glacier Bay National Park and Preserve and specifically on Glacier Bay proper (the main fjord area that cruise ships target for itineraries).The total increase in ship emissions into the region and Glacier Bay was determined (Figure 2, Table 1). Daily mean concentrations over the tourist season were calculated for CLN, REF and QTA.

We compared the visibility and air-quality relevant concentrations obtained by REF to those obtained by CLN to assess the impacts of current cruise-ship emissions. Comparison of these quantities gained by REF and QTA served to gain insight into the impacts of increased quotas.

Relative humidity contributes to the reduction of extinction by swelling of soluble sulfates and nitrates [39] , while wind speed affects transport of pollutants. Therefore, we examined the relations of meteorological quantities, particular matter concentrations and visibility to understand the governing processes for degraded visibility in Glacier Bay.

The haze index (HI) was determined from the WRF/Chem data at ship-height as described by [12] . HI values close to 0 deciview (dv) represent pristine conditions under Rayleih scattering conditions, i.e. HI increases as visibility degrades. Comparisons of the HI values of REF and CLN or QTA and CLN permit separation of meteorology and emission-related visibility degradation. Differences were tested for significance at the 95% confidence level by a two-tailed Student t-test. The word significant only occurs in this context hereafter.

In 2008, winds were on average relatively calm in the first half of the season (cf. Figure 3(a)). The season

started out cool followed by a warm episode. After a short cold snap, temperatures increased over summer with absolute day-to-day variations of daily means up to about ±5˚C. September daily means averaged over Glacier Bay ranged between 9.7˚C and 11.9˚C. Except for four short events early in the season, bay-wide daily mean relative humidity always exceeded 75% (cf. Figure 3(b)). As the season progressed, relative humidity tended to increase as more cyclones reached Glacier Bay.

Radiation-cloud-aerosol interactions may affect meteorological quantities [25] [40] . Thus, simulated meteorological quantities showed differences between the simulations on an hourly basis, locally at various heights in Southeast Alaska. Typically, absolute values of differences between CLN and QTA or CLN and REF were largest in the downwind of areas with high emissions of radiative active pollutants and/or water-uptaking aerosols in QTA and REF, respectively.

In Glacier Bay, typically, absolute differences |CLN-REF| and |CLN-QTA| in near-surface meteorological quantities exceeded those of |REF-QTA|. Hourly absolute differences varied in time and space in Glacier Bay. The absolute differences in meteorological quantities remained relatively low, because the concentrations of pollutants relevant in the interactions were relatively low as compared to the concentrations in the studies on aerosols over the continental US [41] , on Alaska wildfire-smoke impacts [42] , or on severe haze in eastern China [43] . Furthermore, our study only considered cruise-ship and biogenic emissions.

Daily bay-averaged means of 2 m air temperature, 2 m relative humidity, 10 m wind speed, and sea-level pressure differed among CLN, REF, and QTA on the order of about ±0.1˚C, ±1%, less than ±1 cm and ±1 hPa, respectively for most of the season. This result was partly due to increases and decreases in time and space canceling each other out. However, over the Glacier Bay water, temporally maximum differences ranged between −1.1˚C and 0.8˚C, −22% and 5%, −66 cm・s⁻¹ and 87 cm・s⁻¹, and −1.8 hPa and 0.8 hPa among simulations. Since the impacts of increased entry quotas on radiation-cloud-aerosol interactions were beyond the scope of this study, they are not discussed any further.

In 2008, 16 different cruise ships visited Glacier Bay. These ships varied in terms of passenger volume, main and auxillary engine types, fuel types used, load factors, and maximum cruise speeds. Thus, their emissions differed. Furthermore, their emissions differed in terms of cruise speed, manuvering, total time in the bay and berthing durations. Thus, daily emissions were not just a function of the volume of ships, but also which ship(s) entered Glacier Bay and their operations once in the bay.

The cruise ships of the two major companies (group 1, group 2) and the cruise ships of all the other companies (group 3) visiting Glacier Bay encompassed ships of different sizes. For simplicity, the additional ships in each of the three groups (cf. Section 2.2) were assumed to be of the same type as the cruise ship with the most entrances in that group in 2008.

Compared to an additional cruise ship of group 1, the typical emissions in Glacier Bay per visit of an additional cruise ship of group 2 were 51%, 38%, 418%, 35%, 61%, and 37% higher for PM, VOC, SO₂, NO_x, CO, and NH₃, respectively. The additional cruise ship of group 3 emitted 24%, 27%, ~1%, 24%, 36% and 24% less PM, VOC, SO₂, NO_x, CO, and NH₃ than the ship of group 1. The cruise ship of group 3 emitted 16%, 20%, 19%, 18%, 22% and 17% less PM, VOC, SO₂, NO, CO, and NH₃ than the cruise ship of group 2 (Figure 1).

These differences resulted from different ratios of fuel types used, and different engine sizes, load factors, cruise speeds, maximum cruise speed, cruise pattern, as well as bethering durations (for a detailed discussion see [12] ). Compared to the additional cruise ship of group 1 the emissions per person on the additional ship of group 2 (3) are 38% higher (18% lower). The emissions per person on the additional cruise ship of group 3 are 40% lower than for the additional ship of group 2. Note that the ship of group 2 (3) has the lowest (highest) capacity of passengers plus crew.

Table 1 gives the season-accumulated emissions into Glacier Bay under the REF and QTA scenarios. With the above-described quotas, daily mean emissions in Glacier Bay increased by 41%, 41%, 38%, 23%, 38%, and 36% for PM₁₀, PM_2.5, SO₂, VOC, NO, and NH₃, respectively.

During the “peak season”, six or more cruise ships are a common occurance in some areas of Southeast Alaska [44] . In our study, increased quotas for Glacier Bay entrances meant in QTA increased inbound and outbound traffic in the area covered by the model domain (Figure 1). In the region covered by the model domain, the increased quota meant additional emissions from up to six cruise ships lasting for three to five days per additional cruise-ship depending on their routes.

In QTA, daily emissions increased the highest when six additional ships cruised in Southeast Alaska. Season- accumulated cruise-ship emissions increased not only in Glacier Bay (Figure 4), but also along the coast of the Gulf of Alaska, in Icy Strait and the Inside Passage (Figure 2).

Only on seven days of the tourist season, QTA and REF had the same number of cruise ships in the model domain, i.e. identical emissions across the domain in QTA and REF (cf. e.g. Figure 4). However, QTA and REF concentrations differed on these seven days due to different emissions on the preceding days (cf. e.g. Figure 5(b)).

In Glacier Bay, daily-accumulated emissions varied across the season (Figure 4) due to different cruising- speeds and berthing times for glacier viewing [12] as well as different combinations of visiting ship types and

their engine powers [45] . For example, in mid June, several of the largest cruise ships visited the park on consecutive days resulting in elevated emissions as compared to other periods when the same number of cruise ships visited.

Comparing all daily-accumulated QTA emissions occurring on days with zero emissions in REF to each other indicates the variability of confounding factors in emissions. Days without cruise-ship visits show as days with zero emissions in Figure 4. In QTA, on some days, daily increases in PM₁₀, and PM_2.5 emissions into Glacier Bay amounted up to 1.6%, and 2.6% of the seasonal total accumulated emissions in REF.

In 2008, less cruise-ship visits occurred during the May and September “shoulder season” (averaging one per day or less) than during the June to August “peak season” (up to two per day). Consequently, with respect to time, accumulated emissions increased stronger at the beginning and towards the end of the tourist season than during the “peak season” (Figure 4).

In CLN, particulate matter was mainly PM_2.5 from sea-salt. In the fjords, and Icy Strait, season-mean PM_2.5 (PM₁₀) concentrations at ship height ranged between 0.7 (0.2) and 0.9 (0.3) mg・m⁻³ (cf. Figure 6(a)). Over Lynn Chanel and Chatham Strait, PM_2.5 concentrations reached up to 2.5 mg・m⁻³. Along the coast, mean PM_2.5 concentrations varied between 0.6 and 0.75 mg・m⁻³. Farther out in the Gulf of Alaska and inland mean PM_2.5 concentrations decreased with increasing distance from the coast. In CLN, daily mean PM₁₀ concentrations varied within their magnitude in response to the weather conditions (Figure 5(a), Figure 3).

In REF and QTA, Glacier Bay daily mean PM₁₀ concentrations remained below 6 mg・m⁻³ except for five occasions (Figure 5(a)). On these days, the means ranged from about 10 mg・m⁻³ to more than 18 mg・m⁻³. These days also had highest PM_2.5 concentrations (not shown). PM_2.5 concentrations often showed more peaks than PM₁₀ concentrations for both REF and QTA. Occassionally, hourly PM_2.5 concentrations exceeded the daily means by threefold.

In QTA, daily mean PM₁₀ and PM_2.5 concentrations typically increased on days with additional cruise-ship

visits with largest increases on days when two additional cruise ships entered Glacier Bay, i.e. when REF had no visits on that day (cf. Figure 4(a), Figure 5(a)). At the end of August, daily mean PM₁₀ concentrations decreased for a short time in Glacier Bay despite the additional visits in QTA. More small particles competed for precursor gases in QTA than REF, and, hence, fewer particles reached sizes beyond 2.5 mm in diameter. Consequently, despite enhanced precursor gas and PM emissions in QTA, PM₁₀ concentrations decreased as PM_2.5 concentrations increased. On some days early in the season, PM_2.5 decreased despite increased emissions while PM₁₀ increased due to increased emissions plus more PM_2.5 growing to PM₁₀ due to the additional precursor emissions. These findings suggest that additional emissions can alter the fine-to-coarse particle ratios in both directions.

There were also some notable, spatially specific differences in PM₁₀ concentrations between REF and QTA. For example, at the mouth of Glacier Bay, in the mid-bay and in front of the glacier as well as along the shores of the Gulf of Alaska, season-mean PM concentrations increased up to about 0.1 mg・m⁻³ throughout the ABL in response to the increased entry quotas (cf. Figure 6(b), Figure 6(c)).

In CLN, no emission sources for SO₂ existed. Thus, SO₂ concentrations in Glacier Bay represented pristine air background values (cf. Figure 5(b), Figure 6(d)). Dry and wet deposition, chemical reactions and gas-to-particle conversion are sinks for SO₂. These processes depend, among other things, on the meteorological conditions [14] . For example, chemical reaction rates increase nonlinear with increasing temperature; Solubility of gases in droplets increases with decreasing temperature and droplet size. Besides the micrometeorological conditions, dry deposition depends on surface type [46] .

The removal processes and physio-chemical fates modify the spatial-temporal distribution of SO₂. Consequently, in CLN on season-average, near-surface mean SO₂ concentrations were highest over the St. Elias and Coast Mountains (0.8 ppb to 1.6 ppb) due to the low temperatures there (Figure 6(d)). In the Gulf of Alaska away from the coast, SO₂ concentrations were less than 0.4 ppb. Elsewhere near-surface SO₂ concentrations ranged between 0.7 ppb and 0.8 ppb. On average, dry deposition of SO₂ was strongest over the Tongass National Forest (up to 50 g・km⁻²・h⁻¹).

In Glacier Bay, synoptic and surface conditions governed SO₂ concentrations in CLN (cf. Figure 3, Figure 5(b), Figure 6(d)). In REF and QTA, SO₂ concentrations kept the broad patterns of synoptic-scale related decreases and increases, but at enhanced concentrations due to ship emissions (Figures 1-3, Figure 4(b), Figure 5(b), Figure 6(e), Figure 6(f)). In Glacier Bay, the cruising to and berthing in front of the glaciers were obvious in the SO₂ distributions at ship level and through the ABL in REF and QTA. At ship level, increased quota increased SO₂ concentrations up to about 20 ppt on season-average, particularly along the cruise path and in front of the glaciers. Except for a few occasions in early June, late August and around September 10, SO₂ concentrations in QTA exceeded those of REF.

These exceptions of decreased SO₂ concentrations despite increased SO₂ emissions (cf. e.g. Figure 4(b), Figure 5(b)) resulted from shifts in the competition of SO₂ and NO_x for NH₃ in the formation of aerosols. Other emission-impact studies also reported such shifts (e.g. [47] [48] ). For example, ammonimum nitrate forms under low sulfate, but ammonia and nitric acid-rich conditions. Ammonimum and nitrate exist in the aerosol phase when enough ammonia and nitric acid exist for saturation in the gas phase. Sulfuric acid also has an extremely low vapor pressure. Ammonium-sulfate formation needs two ammonia molecules for one aerosol to form. Thus, abondance of SO₂ shifts aerosol formation towards more amonimum-sulfate, and less amonimum-nitrate aerosols.

In contrast to SO₂, NO_x has two natural sources within the model domain even in absence of anthropogenic emissions (CLN). Soil bacteria emit NO_x depending on the soil-temperature and soil-moisture conditions as part of their metabolic activities (e.g. [34] [46] ). In the atmosphere, peroxiacetyl nitrate (PAN) is a reservoir for NO_x that releases (takes up) NO_x as temperatures go up (down) [14] . This means that in equilibrium between NO_x and PAN, NO_x (PAN) concentrations are higher (lower) during warm than cool weather [14] .

Season mean NO_x concentrations followed the topographic elevation in CLN with lowest values over the glaciers of Glacier Bay and highest over the Gulf of Alaska and the low elevations inland (e.g. Figures 6(g)-(i)). The opposite was true for PAN (not shown). Furthermore, NO_x concentrations increased/decreased with the advection of warm/cold air (cf. e.g. Figure 3(a), Figure 5(c)).

In contrast to CLN, this general chemical behavior was less obvious in REF and QTA. The cruise-ship emissions disturbed the PAN-NO_x equilibrium and it took time for the equilibrium to re-establish. In QTA, throughout the season, NO_x concentrations were higher on days with additional entries in Glacier Bay than in REF (cf. Figure 4(c), Figure 5(c)). Typically, QTA PAN-concentrations exceeded those of REF and CLN (Figure 5(c)). Increased quota increased season-mean PAN concentrations up to 20 ppt over Glacier Bay (not shown).

In CLN, biogenic emissions were the only source of VOCs. In CLN, VOC concentrations were high over the Tongass National Forest and low over glaciers and the Gulf of Alaska (not shown). In CLN, REF, and QTA, VOC concentrations increased as summer progressed. While VOC minima matched up well among the three simulations, in REF and QTA, the ship emissions governed the maxima with the latter having higher concen- trations on days with additional entries than the former (not shown).

The cruise-ship emissions of NO_x and VOCs also affected the ozone (O₃) concentrations and season-mean O₃ concentrations were lowest in CLN (Figure 5(d), Table 2). In the unpolluted atmosphere of CLN, less short- wave radiation was absorbed by aerosols during clear sky conditions. Consequently, actinic fluxes exceeded those of REF or QTA meaning more photolysis occurred in CLN than REF or QTA. Similar to SO₂, the season- mean NO_x distribution reflected the cruise paths and berthing in front of the glaciers (Figure 6(h), Figure 6(i)) at ship level and throughout the ABL. In QTA, season-mean near-surface NO_x concentrations increased up to about 50 ppt in these locations, in Icy Strait and along the coast of the Gulf of Alaska.

Wet and dry deposition and atmospheric removal of aerosols are of concern for land and marine ecosystems. Acidic deposition may decrease the pH-values of waters. Since WRF/Chem does not have an ocean model, it cannot provide information on how the deposition fluxes would change the pH-values of the waters. However, WRF/Chem provided information on the amount of pollutants deposited into waters, onto glacier, and on land. In nature, in the case of deposition of soot on glaciers, increased quotas may affect the radiation budget. However, the simulations did not consider this impact.

Changes in deposition fluxes contributed to the nonlinear responses of concentrations to the emission increases. On average, deposition of PM was largest over forests, and lowest over water in all simulations. It was lowest in CLN as here PM only had natural sources (e.g. sea-spray, gas-to-particle conversion). In REF, average deposition of PM over forests reached locally up to 4 kg・km⁻²・h⁻¹, and remained less than 0.4 kg・km⁻²・h⁻¹ into waters. In QTA, the season-mean PM deposition increased up to about 0.5 kg・km⁻²・h⁻¹ along the coasts and on the shores of Glacier Bay in QTA as compared to REF due to increased cruise-ship traffic.

Likewise, season-mean SO₂ deposition increased up to 0.5 g・km⁻²・h⁻¹ in Glacier Bay in particular in the fjords of glacier viewing. However, due to advection of pollutants, notable increases also occurred in the areas of Glacier Bay where the NPS prohibits ship traffic. Along the southern shores of Icy Strait and the entrance of Glacier Bay both SO₂ and NO₂ deposition increased up to 10 g・km⁻²・h⁻¹, and about 5 g・km⁻²・h⁻¹, respectively, in QTA compared to REF (not shown).

Relative humidity, PM₁₀ and NO₂ concentrations affect visibility [49] . The haze index HI = 10・ln(0.1・b_ext) quantifies the perceived visibility as a function of light extinction, b_ext. Light extinction depends on the Rayleigh

scattering of NO₂, relative humidity, the concentrations of sulfate, nitrate, organic carbon (OC), elemental carbon (EC), fine soil particles of aerodynamic diameters less than 2.5 μm, and coarse particulate matter (CM) of aerodynamic diameters between 2.5 μm and 10 μm. Thus, any shifts in fine-to-coarse ratio of these aerosols affect visibility.

We calculated light extinction from the WRF/Chem results following [20] and used it to determine visibility at mean cruise-ship height as this is the elevation that may impact visitor experience. The US Environmental Protection Agency (EPA) usually calculates visibility at 1.8 km height [49] . However, due to the humidity of the maritime climate, the top of the ABL was often cloudy thereby blocking the view to this height.

In CLN, REF and QTA, high relative humidity was the main cause of low visibility (Figure 3(b), Figure 7(a)). At high relative humidity, soluble nitrates and sulfates swell thereby reducing extinction [49] . In all cases, visibility showed a degrading trend as summer progressed due to increasing relative humidity associated with more cyclones reaching Glacier Bay. Linear trends followed HI_CLN(t) = 0.006・t − 244.1, HI_REF(t) = 0.0056・t − 217.58, and HI_QTA(t) = 0.0059・t − 231.98 where t is time. Haze indices were highly correlated with relative humidity (R = 0.68) and PM_2.5 (R = 0.66) in CLN. In CLN, all PM₁₀ was less than 2.5 μm in diameter, i.e. PM_2.5. In REF and QTA, besides the emitted PM₁₀, PM₁₀ existed from particle growth by accumulation of precursor gases on natural and emitted PM_2.5.

On average over Glacier Bay, CLN haze indices for the 20% best, median (40% - 60%) and 20% worst visibility days were 1.0, 1.9, and 2.9 dv, respectively. The mid points of these visibility ranges are the 10^th, 50^th and 90^th percentile, as which we addressed these ranges hereafter. In REF, the 10^th, 50^th and 90^th percentile of HI were 1.8, 3.0, and 5.2 dv, respectively [12] . Despite increased emissions, the 10^th, 50^th and 90^th percentile of HI of QTA and REF were the same. In all cases, worst visibility days coincided with high relative humidity (cf. Figure 3(b), Figure 7(a)).

At ship height, on average over Glacier Bay across the tourist season, mean HI were 1.9, 3.2, and 3.3 dv in CLN, REF and QTA, respectively. This visibility is better than the 2018 target set by the US EPA for pristine areas. Whereas, season-mean haze indices varied marginally in space for CLN over Glacier Bay (Figure 8(a)), both REF and QTA showed a gradient of decreasing visibility from the glacier viewing fjords to the entrance of Glacier Bay. Close to the entrance, advection of pollutants from ships cruising thru Icy Strait contributed to reduced visibility (Figure 8(a)). In QTA, atmospheric transport increased HI in the Muir Inlet, an area in the upper east arm of the fjord void of cruise-ship traffic (cf. Figure 8(b), Figure 8(c)). Due to changes in concentrations (Figure 6), and deposition, HI decreased over the Brady Glacier in QTA as compared to REF (cf. Figure 8(b), Figure 8(c)).

It is clear that transport by wind can be an essential removal path for pollutants emitted by cruise ships in Glacier Bay. However, wind can also transport pollutants into Glacier Bay [11] . In REF, haze indices were highly correlated with wind speeds (R = 0.83), PM₁₀ (R = 0.83), and PM_2.5 concentrations (R = 0.72). Correlations of these quantities for QTA were within about ±2% of those of REF. Without any cruise-ship emissions (CLN), the correlation between haze index and wind speed was far less (R = 0.38) because the existing particles and gases were around background concentrations. Thus, any transport by wind in/out of Glacier Bay replaced air with air of similar chemical composition.

In summary, while humidity and natural PM_2.5 determined visibility in Glacier Bay absent anthropogenic emissions (CLN), wind speed and the PM₁₀ concentrations were the most influential in determining visibility in Glacier Bay in REF and QTA (Figure 7). For thermodynamical reasons, at the same high relative humidity, large particles are the first to swell. As compared to CLN, the increased number and size of PM in REF and QTA replaced PM_2.5 by PM₁₀ in reducing visibility. In contrast to CLN, in REF and QTA, transport of air in/out of Glacier Bay now exchanged air of different composition. Therefore, the correlation between HI and wind increased in REF and QTA as compared to CLN.

Comparison of the visibility on the seven days with identical emissions in REF and QTA showed bay-wide average increase of 0.2 dv in QTA (Figure 4, Figure 7). This means that on these days, visibility was on average 1.1% less in QTA than in REF. The enhanced emissions of the previous days (e.g. Figure 4) and hence modified concentrations (Figure 5) caused this degradation of visibility (Figure 7).

In Glacier Bay at ship height, mean concentrations of PM₁₀ and PM_2.5 were 2%, and 3% higher in QTA, while those of NO₂ were 5% lower in QTA than in REF on the seven days of domain-wide identical emissions. Concentrations of O₃, PAN or VOC differed less than 1% on average over these seven days (NO₂ decreased while PAN, VOC, and O₃ remained nearly constant). This means that the additional primary aerosols from emissions on previous days together with enhanced secondary aerosol formation degraded the visibility on those seven days with domain-wide identical emissions in REF and QTA.

We compared HI-values of days without cruise-ship visits in Glacier Bay in REF with those of CLN. Pollutants remaining from emissions in Glacier Bay on previous days and/or advection increased the HI by about 46% on these days. On average, concentrations on days without emissions in Glacier Bay in REF were by a factor of 12.2, 2.5, 1.4 2.2, 1.2, and 1.4 higher for PM₁₀, PM_2.5, SO₂, NO₂, PAN, and VOC than in CLN, while O₃ concentrations remained about the same. These findings suggest that cruise-ship emissions from previous days, along with formation of secondary PM from precursors (NO, SO₂, NH₃) caused the on average 1.5 dv higher haze index in REF on cruise-free days as compared to CLN (Figure 4, Figure 7).

Together these findings indicate that while ship emissions contributed to degraded visibility in REF and QTA in Glacier Bay, meteorology primarily governed visibility. Increases in ship-visit quotas as assumed here would have relatively small impacts on visibility and air quality as compared to the situation with the current quota.