Marine-cloud brightening: an airborne concept

Marine Cloud Brightening (MCB) is a proposed Solar Radiation Modification (SRM) geoengineering technique to enhance Marine Boundary Layer (MBL) cloud albedo. Extant proposals consider 104 − 105 autonomous ships spraying seawater, generating and dispersing sea salt nanoparticles. Alternatively, this paper proposes industrially manufacturing NaCl nanoparticles using ethanol anti-solvent brine precipitation. With desiccation, size optimization and narrowed size distribution, aerosol mass flux reduces by ∼500× (17× for dry mass flux). This facilitates Unmanned Aerial Vehicle delivery (e.g. MQ-9 Reaper Unmanned Aerial Vehicle). Increased speed and wake turbulence improves areal coverage per vehicle versus ships—reducing fleet size. Utilizing extant airframe designs improves vehicle Technology Readiness Level (TRL)—potentially improving system operational cost (est. $40B · yr −1) and lead time. This approach further reduces energy requirements (5× less), technical risk and system complexity. Increased readiness amplifies proliferation risk—particularly for inexpensive regional heatwave and hurricane suppression—making governance more urgent.


Introduction
MCB proposes injecting sea salt (mostly NaCl) nanoparticles in the marine boundary layer (MBL, 0-1 km altitude).Airborne micro-and nanoparticles of salt exist naturally in very high concentrations, up to 1 mm −3 in some regions [1].MCB aims to increase local number concentrations approximately 2 × − 4 × [1].These aerosols act as CCN, raising the concentration of cloud droplets-thus brightening MBL clouds, due to the Twomey effect [2].
MCB-together with the technically related Cirrus Cloud Thinning (CCT)-is one of two major proposed solar geoengineering techniques.The alternative is Stratospheric Aerosol Injection (SAI) [3]-which generally requires higher altitude (20-25km-albeit with altitudes comparable to tropical CCT for seasonal polar SAI).MCB offers finer spatial and temporal control of climate than SAI.Fine control potentially increases governance challenges; increasing degrees of freedom adds potential points of contention.Tropospheric salt particles remain local and last 10 days-versus ∼2 years for hemispherically mobile SAI particles.
MCB requires development and implementation of a scalable deployment system, before use at scale is possible.This system should have the capacity to achieve negative radiative forcing sufficient to offset anthropogenic greenhouse warming; however regional MCB has also been proposed (e.g.Great Barrier Reef [4]).Various researchers [1,5,6] propose 10 4 -10 5 ships spraying seawater aerosols in the MBL.This paper instead proposes centralized industrial manufacturing of monodisperse salt nanoparticles (in a few production facilities)-with UAV dispersal suggested, facilitated by aerosol mass reductions.The overall concept is illustrated in figures 1 and 2.


• Optimal particle diameter of 30-40 nm (dry) . Top: proposed MCB system.Salt nanoparticles are generated industrially, using waste desalination brine.Bottom: accessible area (navy blue), assuming a 2,200 km UAV range, from present Extended-range Twin engine Operations Performance Standards (ETOPS) airfields-diversion airfields for commercial twin-engine airliners, with significant maintenance facilities extant.Note that larger areas could be accessed by using some less equipped airfields (for example in the Galapagos islands), or building new airfields (for example in the Kerguelen Islands or in South Georgia).
implying a minimum dry salt mass flux ∼10 2 kg s −1 ; seawater mass is ∼30 × higher.Any upwards deviation from optimal diameter disproportionately affects mass efficiency, due to the cubic mass diameter relationship.

Particle generation inputs
The process outlined below is based on [8], and is experimentally validated with the parameters outlined below.This process could be made considerably more efficient by reusing the output 97.5% pure ethanol solution before reconcentration, though this possibility requires further lab testing.The lower limit of purity has not been experimentally established, and multiple reuse cycles cannot be ruled out.Hence, the process energy requirements outlined below are 'worst-case', and can therefore be considerably reduced-to perhaps below a third-with modest process changes.
Particle generation requires 2.7 • 10 9 capillaries, running 2M NaCl at 2.4 • 10 3 m 3 h −1 into 99.6% purity (anhydrous) ethanol at 9.6 • 10 4 m 3 h −1 , with potential ethanol purity reductions discussed above.A 2M NaCl solution is chosen to ensure relatively uniform particle size distributions.A higher concentration would reduce the required number of capillaries, but it would also broaden the particle size distribution [8].
Ethanol flux is 9.6 • 10 4 m 3 h −1 .Given a 40:1 volumetric ethanol-brine ratio [8], final volumetric ethanol concentration is ∼97.5%.The energy cost [10] of desiccating azeotropic ethanol (i.e. which cannot be purified by distillation-96% ethanol by vol.) is 1.25 MJ kg −1 (1.134 MJ kg −1 for the pervaporation system [10]).Pervaporation is similar to distillation, but uses a selectively permeable membrane to reduce contamination of the vapor phase with the unwanted volatile.At the required flow rate, total power is 18.1 GW-mostly low temperature industrial heat (>95%, 70 • C).This is available from cogeneration, solar thermal, geothermal or heat pump inputs.Notably, the low cost and ease of storing ethanol mean that large volumes can be stored, enabling the system to run off inexpensive off-peak power, or stranded intermittent renewables.
Generating the requisite 2M NaCl flux (2.4 • 10 3 m 3 h −1 ) requires less energy than desiccation.Reverse osmosis desalination generates brines 75 kg m −3 -i.e.60% of the required 2M concentration (117 kg m −3 ).Weak brine can be concentrated in evaporation ponds; a 7 km 2 pond can provide the required evaporation rate (in typical south western US humidity conditions: AZ) to vaporize 1.6 • 10 3 m 3 • h −1 [11].Brine flow rate is 40% of the outfall capacity of individual large plants; Ashkelon (Israel) generates 10 4 m 3 • h −1 [12].Separation of nanoparticles from ethanol can be achieved using cross-flow filtration [13].Centrifugal drying cannot be used, due to the small particle size and low density differential (resulting in very low terminal velocity).In cross-flow filtration, the feed flow is tangential to the filter surface.The liquid phase exits through the filter tube walls, normal to the mixture flow through the pipes.This enables continuous filtration in industrial processes-e.g. using ceramic filters with pore dimensions 5 nm [14,15].27MW average mechanical power is required, at pressure 10 6 Pa.Filtration requires 80, 000 m 2 surface area-assuming a specific flux of 120l • h −1 m −2 bar −1 [14], e.g.1200 tubes of length 100 m and radius 10 cm.Additional steps are needed to store nanoparticles as a powder within a pressurized inert gas or supercritical fluid, such as nitrogen or carbon dioxide.The latter benefits from continuous dimensional expansion under depressurisation, as opposed to droplet shrinking; this aids dispersion and reduces clumping.Vaporization of the concentrated ethanol solution, and the addition of nitrogen requires 200MW (e.g. with an airtight pond of surface area less than 0.5 km 2 ; and 1MW to generate the required N 2 ).This assumes 1:1 ethanol(l):NaCl(s) by volume, after filtration.The use of carbon dioxide (for example captured from the ocean) would slightly raise power requirements to 4MW [16].
Pumping losses are negligible: capillary pressure drop is only 140Pa [17] requiring 1MW.The power budget is summarized in figure 3 below.
To increase stability, colloidal stabilizers such as PVP (Polyvinylpyrrolidone) can be used in small amounts during the antisolvent precipitation process.However, the environmental impact and the effects on particle dispersion, hydrophilic properties, and cost remain for later work.
Finally, a packing stage is envisaged, loading the material into reusable dispensers for attachment to aircraft.This protects the cargo, reduces safety risks loading at pressure, allows materials stockpiling, and simplifies the skillset required for aircraft operators.The energetics and cost of packing are envisaged to be relatively trivial, and are not explored further.
Particles emitted per meter travelled is 10 16 m −1 -as per ships [7].This number concentration constraint is to reduce coagulation and agglomeration of the salt particles after emission, and to optimize the concentration of CCN.Aircraft range limits (several thousand km) suggest a platform with several tonnes payload, unless operated from ships.Fixed wing UAVs are ideal; rotorcraft UAVs have poorer endurance, range and payload; wide body tanker airplanes have excess payload, wasting energy and capital; crop dusters and small fire-fighting aircraft have littoral applications, briefly discussed later.
Fixed-wing UAVs are designed for endurance, with high lift to drag ratios.Their cruising speed (100 m/s) and operating altitude (0-10 km) are appropriate for efficient spraying.UAV speeds allow a higher particle emission rate per vehicle than ships, reducing fleet size.Effective discharge airspeed is enhanced by wingtip vortices and general wake turbulence; propeller wash contains significant residual kinetic energy (both translation and rotation).Potentially, distribution of particles suspended in heterogeneously sized droplets of a volatile carrier fluid could avoid excess drift, whilst broadening the distribution track [18].Deployment must be near clouds; small aerosols do not reliably settle [19].The effect of gravity on particles can be neglected, since their terminal velocity in the air is less than 1 μm • s −1 .
Wood et al [7] consider 10 4 -10 5 boats; particle injection rates are limited by local winds (average 7 m s −1 ), precluding a smaller fleet.UAVs travel 14 × faster than ships (100 m s −1 ) [20].With injection in the propeller wash, effective wind speed is further increased by 40% (propeller propulsive efficiency of 0.8 [21]); performance does not increase by the same factor-due to commuting time from airfields to spray zones, reducing spraying duration.Later design work can optimize for wake turbulence and wingtip vorticity at the expense of rangecreating long and short range UAV variants.However, a radial distribution pattern around dispersed airfields risks proximal saturation coverage-arguably negating any short range variant advantages, absent additional runways.An alternative system design can include mid-air resupply (e.g.Global Hawk UAV); this involves various complexities not further considered.Finally, one-way flights between airfields may prove optimal given a sufficiently dense airfield network.

UAV platform
A particle injection rate (excluding carrier gas or fluid) of 6 • 10 17 s −1 (or 280 kg • h −1 ) per UAV is considered, scaled up from [7] by 1-2 orders since: (i) Particles are stored as powder in pressurized inert gas or supercritical fluid (lowering coagulation risk) (ii) Effective wind velocity is more than 1 order of magnitude greater than average sea level wind speed.With assumptions as above, 500 on-station UAVs are required-1-2 orders of magnitude fewer than ships.This number includes payload constraints, in particular the weight of empty aerosol tanks (10% of the payload).However, total active fleet size is ∼3 × higher than this number, considering maintenance and commuting time.This paper assumes a 4:1 ratio of total airframes to on-station airframes, to account for other factors: varying demand, training, crashes, upgrades, unplanned outages, environmental monitoring, experimentation, and aborted missions.Notably, previous MCB literature tends to omit discussion of these additional vessels.
Comparable and less expensive alternatives are available from other nations-Chengdu Wing Loong III (China), for which no cost is publicly available; TAI Anka (Turkey), $ 25M.
Alternative platforms are possible.Unmanned seaplanes provide superior high seas coverage, even using short range aircraft-but require support ships for refueling, resupply, storm shelter, and maintenance.Additionally they have more stringent maintenance requirements [23] with associated costs implications.Conceivably, fixed wing UAVs could also be launched from existing or dedicated cargo ships, using catapults and recovery nets.Rotorcraft UAVs could also be launched from ships without requiring hardware modifications for launch and recovery; this option notably allows inexpensive testing and scaling, as such aircraft are available in the $10,000 price range and can also be rented.Alternatively, manned crop-dusters and firefighters (e.g.Air Tractor AT 802-F) are immediately and inexpensively available ($8200 h −1 ); these are likely useful for littoral use (e.g.reef protection), but are less suited to remote high seas operations.With appropriate support and configuration, aircraft based on these types could be used as high seas seaplanes, or could fly from marine runways-carriers or fixed platforms Blue water operations using single engine aircraft would add substantial support costs and crew risks, necessitating the provision of costly infrastructure: search-and-rescue, crew accommodation, medical facilities, deck cranes, and storm-proof aircraft storage.Even with such marine infrastructure, any high seas piloted work would likely remain dangerous, lonely, and tedious-giving a locus for industrial action or political protest.Further, the alternative approaches above introduce substantial system and regulatory complexities unnecessary for discussion in this preliminary analysis.
MQ-9 UAVs can readily use sustainable aviation fuel (Fisher-Tropsch synthesized or biofuel).An alternative is green hydrogen, requiring a major redesign-i.e. a hydrogen powerplant, with compressed or cryogenic hydrogen storage.A brief discussion of hydrogen power options is merited, not least because this simplifies airside energy use calculations.Several options, including fuel cells [24] or hydrogen gas turbines [25] are currently available, all with very high TRL.For the proposed application, fuel cells are most suitable: they have a very high efficiency (50-60%), and a very high time between overhaul (TBO) of between 5000-25000 h [24]-a particular advantage at remote airfields or for marine UAVs.Fuel cells have been demonstrated in flight in 2023 by universal Hydrogen, using a fuel cell of dimensions, mass, and power compatible with the MQ-9 [26].Fuel cells are sensitive to salt [27], though most of the flight would occur at relatively high altitudes, where salt concentrations are much lower, and the UAV paths can calculated to ensure that UAVs do not fly back in either their own wake, or the wakes of other aircraft.If this salt poisoning problem cannot be solved, UAVs can fly with existing sustainable aviation fuel or with hydrogen gas turbines-albeit with reduced energy efficiency.
UAV fleet requirement (2, 000) is six times the total production of MQ-9s to date (319); this may result in costs savings.For clarity, discussion of specific aircraft types does not mandate use of these actual designsmerely airframes with comparable capabilities.Accordingly, this allows practical production scaling-even if existing manufacturers fail to increase production, or do not offer civilian versions.Absolute scaling requirement is modest, compared to all reasonable historic metrics: global aircraft tonnage, global aircraft count, single-model aircraft production volume, single-model aircraft production rate [28].
A military MCB program is not proposed.In common with other joint civil/military airframes (e.g.A-330 and A-330 MRTT [29]), civilian versions are already made for the MQ-9 [30]; this eliminates national security or threat misidentification concerns.Significant civilian savings are expected from the $30M military CapEx unit price-omitting the following: • military-grade secure communications

• stealth capabilities
• electronic counter measures

• crew vetting
OpEx costs are less flexible, but will still benefit from omission of costly spares and the maintenance requirements of specialist components.HR & management costs are also likely to be substantially lower, given the routine, lower-risk nature of the work and the reduced vetting requirements resulting from civilian use.General maintenance could be done at operational airfields; centralized major maintenance would primarily involve powerplant replacement or refurbishment.With near 50% uptime, powerplants would have to be changed annually (a drop-in process), assuming a conservative TBO of 8000 h [24].This requires the production or refurbishment of 10 3 powerplants annually.This refurbishment and major maintenance could be done in a limited number of specific locations to minimize costs, requiring only minimum routine maintenance capabilities in all airfields.Transfer flights are not specifically costed.Powerplant replacement can also be done on base, but limited supply visits to remote bases implies amortization of a costly stock of spares will compete on cost with transfer flights to maintenance depots.
Operational civilian flight cost per flight hour is around $800 for US civilian agencies [31], though this corresponds only to the marginal cost of an added flight hour.Military cost per flight hour (all inclusive) is around $3500.Taking approximately 60% of flight hours as commuting, transfer, monitoring, training, and abortive missions gives military on-station cost of $8750 per hour, per airframe.Civilian costings are likely to be no more than 75% of military-particularly given the large volume of orders, international market for airframes, simpler piloting work (4 flying airframes per on-shift pilot is assumed), and the above list of removed subsystems and capabilities.Hence, hourly on station cost is estimated to be around $6, 500.This gives annual airside costs of $29B to reverse a doubling of CO 2 [7].

Cirrus cloud thinning
UAVs are also capable of Cirrus Cloud Thinning (CCT) operations; the MQ-9 has a 15 km ceiling.CCT may require a change in particle size, by modifying the anti-solvent flow ratio [8].Optimal performance may mandate alternative materials-e.g.BiI 3 [32], or some acids or proteins.CCT is less studied than MCB; consensus on materials, particle flux, seeding conditions, etc is not currently established.Accordingly, affordable, safe and effective solvent/anti-solvent pairings may not exist for CCT, so further discussion is curtailed.

Stratospheric aerosol injection
Due to altitude limitations, the aircraft discussed are unsuitable for SAI, beyond limited high latitude deployments [33].However, the proposed approach may impact SAI development and deployment.
Firstly, antisolvent precipitation may conceivably be used for SAIas an alternative to environmental condensation from SO 2 or H 2 S. Cost, material efficiency, and particularly size control advantages are equally applicable to SAI.Solids considered for SAI include titania, alumina, and diamonds [34]; candidate materials might feasibly be made via antisolvent precipitation, achieving optimal particle size distribution.
Secondly, there is no specific altitude limit preventing appropriately designed successor UAVs from performing SAI [35].UAV programs for MCB will necessarily create technical and organizational capacity, partially applicable to SAI.As such, any UAV MCB program poses general SRM proliferation risks.

Implementation
The proposed approach reduces lead time and implementation risks.Centralized production facilities can manufacture nanoparticles-leveraging existing power, desalination, and ethanol facilities.Similarly, international shipping and existing UAV platforms facilitate delivery.Therefore, the proposed system could be assembled using high TRL equipment.Hypothetical program pacing overlooks entirely the financial, environmental, and governance development requirement-and therefore does not identify a critical path.

Cost, electricity and fuel usage
Most power required for recycling the ethanol solution and generating the 2M NaCl solution is low-temperature industrial heat (defined as 165 • C).This 20 GW th (thermal power) can be co-generated using existing thermal power plants (combined heat & power; CHP), geothermal heating plants, solar concentration heating plants (mirror arrays), or heat pumps with geothermal low temperature sources.In particular, under-utilized geothermal-rich areas could be used.Ideal locations include the Pacific Rift Valley or Iceland-mirroring the approach taken by Climeworks for direct air capture [36].
With co-generation from existing thermal power, demand can be expressed as effective electrical poweri.e.electricity production loss due to co-generation.This is 3.3GW e -since typically for each unit of electrical power lost, approximately six units of thermal power (<90°C) are generated [37].The exact figure depends on the hot source T h and cold source temperatures T c , and is bounded by the Carnot limit - . Given average worldwide electricity cost of $ 190 MWh −1 [38] (business users), this amounts to a yearly program cost (for process energy, not UAVs) of USD 5.4 B. However, realistic programs locate for optimal input prices; such optimization significantly lowers costs, by choosing locations where power is produced inexpensively, and by using power during low-demand hours, since the rate of production does not have to be constant.The cost of this system can be estimated from [10], which mentions the operating cost of the pervaporation system as EUR 0.0112 • kg −1 ; with the required amount of ethanol to reprocess, a yearly cost of $8.9B is expected (with a $1.1 for 1 EUR conversion rate (August 2023).
Yearly electrical equivalent energy requirement 28.6 TWh e is ∼0.13% of world [39] electricity generation (2.2 • 10 4 TWh e ; 2022).Again, opportunities to reuse waste ethanol one or more times before desiccating would reduce this figure.
MCB ships require approximately 24.6 GW e for the same numerical emission rate (


), based on 41 kW for 10 15 s −1 [40]-neglecting navigation, control, and other subsystems loads.The energy required by ships thus amounts to 210 TWh yr −1 , ∼0.9% of the yearly world electricity production (2022)-approximately 5 × less efficient than UAVs (even when considering UAV propulsion requirements).The electricity cost is $41B • yr −1 , which does not include ship (structure, control and propulsion costs) amortization and maintenance.Decentralized electricity production (i.e. the ship system) is usually more expensive than centralized electricity production (i.e. the UAV system).Particularly, maintenance of a decentralized electricity production system is also more complex and costly-especially so when operating unmanned and in challenging marine conditions.
While conventional jet fuel is likely to be used in the near term, analysis of hydrogen fuel requirements helps scale system energy requirements.Total annual UAV H 2 demand-conservatively assuming constant full throttle, and a typical fuel cell efficiency of 55%-is 3.2 • 10 8 kg; 0.4% of current worldwide hydrogen production of 7.5 • 10 10 kg • yr −1 [41].The total average power required to generate this flux as green hydrogen is 1.6 GW, assuming an electrolysis efficiency of 0.75 [42].Hydrogen production rates will also likely increase in the near future.Electrolysis, compression, and storage efficiencies will also continue to increase-further reducing costs and energy demand, due to experience curve effects and economies of scale.

UAVs advantages:
(i) Centralized (in several production facilities) particle generation enables economies of scale and simpler plant maintenance (vs marine electricity generation), reducing CapEx and OpEx.
(ii) Reduced total electrical equivalent power (5 GW e versus 24.6 GW e for marine).Power generation is simplified (centralized) and more economical (capital & maintenance costs).Ships require onboard power generation; this is potentially difficult to achieve continuously using renewables, unless costly and bulky storage is used.By contrast, centralized powerplants are less expensive to build and maintain, and do not need to store energy for use during nights or adverse weather.For the UAV system, aerosol and fuel production facilities could be installed near stranded geothermal or solar energy resources.In the case of the ship system, intermittency of the energy source would require either in-ship energy storage with sufficient capacity, or fleet scaling.
(iii) The proposed system is more weather-robust, particularly to hurricanes or typhoons.UAVs can evade extreme weather events, or shelter in hardened hangars-ships need more time to move.
(iv) Dry particle ejection and higher UAV speed & altitude allows UAVs to operate directly in clouds.This solves the problem of negative plume buoyancy due to evaporative cooling in the ship system.
(v) Injection altitude and location can be dynamically optimized by adjusting the UAV location in 3Dleveraging local wind, atmospheric and cloud conditions.
(vi) UAVs can be operated from carriers (e.g.TCG Anadolu, commissioned April 2023), marine platforms, tethered barges, icebergs, and seaplane support vessels-albeit with diverse regulations, logistics requirements, and additional costs.
(vii) Greatly simplified UAV maintenance.Ships would require design robustness and frequent maintenance, due to the harsh marine environment: corrosion, biofouling, guano, high winds & waves, cetacean strikes, vertebrate colonization, etc.Furthermore, power generation systems, pumps, motors, sensors and actuators require periodic maintenance.This maintenance is more costly in ships than UAVs, due to fleet size (5 − 50 × ) and accessibility issues.Ships would either have to travel thousands of miles to ports, or be maintained by long-distance repair ships.By comparison with established UAV programs, the operational robustness of complex autonomous remote ships is somewhat speculative.
(viii) Accessibility of marine platforms makes them vulnerable to piracy, theft, scavenging, sabotage, and military attacks.
UAVs disadvantages: (i) Lower systemic robustness, due to centralization and concentrated points of failure; However the system has redundancy, and strategies exist to increase robustness, in particular bulk storage of ethanol, salt nanoparticles and hydrogen ensuring continuous operation even with supply chain disruptions.Note also that while the marine system appears more robust, some large scale events could defeat its high redundancy; a major hurricane or cyber attack could be particularly disruptive.
(ii) Lower uptime: UAVs must avoid harsh weather, while unmanned ships are comparatively robust in moderately severe weather.
(iii) The MQ-9 has approximately 1.8 crashes per 10 5 flight hours [43].With 4.38 • 10 6 flight hours annually, 80 crashes per year would be expected-representing substantial fleet attrition, comparable to airframe retirements (45 retirements per year).However most MQ-9 flights are military; these are inherently more risky than routine civilian operations and less able to sustain precautionary outages in challenging conditions.In civilian use, the crash rate should be greatly reduced-through better preparation, use of more robust systems, and avoidance of severe weather areas.Airbags and parachutes could be used to recover ditched airframes, though they would increase aircraft weight and complexity, meriting a costbenefit analysis versus payload and fuel carriage.
(iv) Requirement for specific geographically dispersed airfields gives significant political leverage to an unlikely collection of states.This imposes political risk and enables rent-seeking behavior [44].(vi) Requirement to ship nanoparticles and fuel from production facilities to airfields will add small energy and resource costs-outweighed by manufacturing centralization savings.However, logistics to very remote bases may be far more costly than global averages, and also necessitate substantial stockpiling.

(vii) Centralized supply chain increases military vulnerability
The key differences between ships and UAVs are summarized in table 1 below.

Conclusions
Ethanol anti-solvent precipitation of brine leads 500× aerosol mass reduction for MCB, enabling UAV distribution.This is due to desiccation (30×), and a combination size optimization and homogeneity (17×).Furthermore, the proposed system is less expensive than the classical ship-based system.The proposed approach relies on extant aircraft platforms and a terrestrial supply chain that is either extant or high TRL.This approach conserves energy (5×), compared to marine operations-with potential further savings available due to antisolvent reuse.This systemic approach is potentially adaptable to CCT (cirrus) operations, and may later influence SAI geoengineering.
This high TRL and cross-applicable technology approach poses proliferation risks, and invites a governance response.The ready applicability of MCB to regional scale cooling using limited fleets further increases nearterm proliferation risks-particularly since UAVs are difficult to detect and inspect-in contrast with ships.

Figure 2 .
Figure 2. Comparison of particle size distributions for sprayers (blue) and anti-solvent precipitation (orange; normal).The spray system is a lognormal distribution with mean 70 nm and geometric standard deviation 1.57 [6], while the anti-solvent precipitation offers a normal distribution with mean 40 nm and standard deviation 3 nm [8].Note the different vertical scales.

Figure 3 .
Figure 3. Top: process used to generate the salt nanoparticles.Bottom: overall system power usage (5 GW).

( v )
Accessing the entire oceanic surface requires the creation of remote airstrips or the use of local airports in some locations (e.g.Galapagos Islands, Antarctica, Kerguelen Island).Alternatively, mid-air resupply (per Global Hawk UAV) is possible.Currently accessible locations are shown in figure 1(b).