Economic and biophysical limits to seaweed farming for climate change mitigation

Monte Carlo analysis

Seaweed production costs and net costs of climate benefits were estimated on the basis of outputs of the biophysical and technoeconomic models described below. The associated uncertainties and sensitivities were quantified by repeatedly sampling from uniform distributions of plausible values for each cost and economic parameter (n = 5,000 for each nutrient scenario from the biophysical model, for a total of n = 10,000 simulations; see Supplementary Figs. 14 and 15)^{47,48,49,50,51,52}. Parameter importance across Monte Carlo simulations (Fig. 3 and Supplementary Fig. 9) was determined using decision trees in LightGBM, a gradient-boosting machine learning framework.

Biophysical model

G-MACMODS is a nutrient-constrained, biophysical macroalgal growth model with inputs of temperature, nitrogen, light, flow, wave conditions and amount of seeded biomass^30,53, that we used to estimate annual seaweed yield per area (either in tons of carbon or tons of dry weight biomass per km² per year)^33,34. In the model, seaweed takes up nitrogen from seawater, and that nitrogen is held in a stored pool before being converted to structural biomass via growth⁵⁴. Seaweed biomass is then lost via mortality, which includes breakage from variable ocean wave intensity. The conversion from stored nitrogen to biomass is based on the minimum internal nitrogen requirements of macroalgae, and the conversion from biomass to units of carbon is based on an average carbon content of macroalgal dry weight (~30%)⁵⁵. The model accounts for farming intensity (sub-grid-scale crowding) and employs a conditional harvest scheme, where harvest is optimized on the basis of growth rate and standing biomass³³.

The G-MACMODS model is parameterized for four types of macroalgae: temperate brown, temperate red, tropical brown and tropical red. These types employed biophysical parameters from genera that represent over 99.5% of present-day farmed macroalgae (Eucheuma, Gracilaria, Kappahycus, Sargassum, Porphyra, Saccharina, Laminaria, Macrocystis)³⁹. Environmental inputs were derived from satellite-based and climatological model output mapped to 1/12-degree global resolution, which resolves continental shelf regions. Nutrient distributions were derived from a 1/10-degree resolution biogeochemical simulation led by the National Center for Atmospheric Research (NCAR) and run in the Community Earth System Model (CESM) framework³⁵.

Two nutrient scenarios were simulated with G-MACMODS and evaluated using the technoeconomic model analyses described below: the ‘ambient nutrient’ scenario where seaweed growth was computed using surface nutrient concentrations without depletion or competition, and ‘limited nutrient’ simulations where seaweed growth was limited by an estimation of the nutrient supply to surface waters (computed as the flux of deep-water nitrate through a 100 m depth horizon). For each Monte Carlo simulation in the economic analysis, the technoeconomic model randomly selects either the 5th, 25th, 50th, 75th or 95th percentile G-MACMODS seaweed yield map from a normal distribution to use as the yield map for that simulation. Figures and numbers reported in the main text are based on the ambient-nutrient scenario; results based on the limited-nutrient scenario are shown in Supplementary Figures.

Technoeconomic model

An interactive web tool of the technoeconomic model is available at https://carbonplan.org/research/seaweed-farming.

We estimated the net cost of seaweed-related climate benefits by first estimating all costs and emissions related to seaweed farming, up to and including the point of harvest at the farm location, then estimating costs and emissions related to the transportation and processing of harvested seaweed, and finally estimating the market value of seaweed products and either carbon sequestered or GHG emissions avoided.

Production costs and emissions

Spatially explicit costs of seaweed production ($ tDW⁻¹) and production-related emissions (tCO₂ tDW⁻¹) were calculated on the basis of ranges of capital costs ($ km⁻²), operating costs (including labour, $ km⁻²), harvest costs ($ km⁻²) and transport emissions per distance travelled (tCO₂ km⁻¹) in the literature (Table 1, Supplementary Tables 1 and 2); annual seaweed biomass (tDW km⁻², for the preferred seaweed type in each grid cell), line spacing and number of harvests (species-dependent) from the biophysical model; as well as datasets of distances to the nearest port (km), ocean depth (m) and significant wave height (m).

Capital costs were calculated as:

where c_cap is the total annualized capital costs per km², c_capbase is the annualized capital cost per km² (for example, cost of buoys, anchors, boats, structural rope) before applying depth and wave impacts, k_d and k_w are the impacts of depth and waviness on capital cost, respectively, each expressed as a multiplier between 0 and 1 modelled using our Monte Carlo method and applied only to grid cells with depth >500 m and/or significant wave height >3 m, respectively, and c_sl is the total annual cost of seeded line calculated as:

where c_slbase is the cost per metre of seeded line, and p_sline is the total length of line per km², based on the optimal seaweed type grown in each grid cell.

Operating and maintenance costs were calculated as:

where c_op is the total annualized operating and maintenance costs per km², c_ins is the annual insurance cost per km², c_lic is the annual cost of a seaweed aquaculture license per km², c_lab is the annual cost of labour excluding harvest labour, and c_opbase is all other operating and maintenance costs.

Harvest costs were calculated as:

where c_harv is the total annual costs associated with harvesting seaweed per km², c_harvbase is the cost per harvest per km² (including harvest labour but excluding harvest transport), and n_harv is the total number of harvests per year.

Costs associated with transporting equipment to the farming location were calculated as:

where c_eqtrans is total annualized cost of transporting equipment, c_transbase is the cost to transport 1 ton of material 1 km on a barge, m_eq is the annualized equipment mass in tons and d_port is the ocean distance to the nearest port in km.

The total production cost of growing and harvesting seaweed was therefore calculated as:

where c_prod is total annual cost of seaweed production (growth + harvesting), c_cap is as calculated in equation (1), c_op is as calculated in equation (3), c_harv is as calculated in equation (4), c_eqtrans is as calculated in equation (5) and s_dw is the DW of seaweed harvested annually per km².

Emissions associated with transporting equipment to the farming location were calculated as:

where e_eqtrans is the total annualized CO₂ emissions in tons from transporting equipment, e_transbase is the CO₂ emissions from transporting 1 ton of material 1 km on a barge, m_eq is the annualized equipment mass in tons and d_port is the ocean distance to the nearest port in km.

Emissions from maintenance trips to/from the seaweed farm were calculated as:

where e_mnt is total annual CO₂ emissions from farm maintenance, d_port is the ocean distance to the nearest port in km, n_mnt is the number of maintenance trips per km² per year, a_mnt is the area tended to per trip, d_mnt is the distance travelled around each km² for maintenance and e_mntbase is the CO₂ emissions from travelling 1 km on a typical fishing maintenance vessel (for example, a 14 m Marinnor vessel with 2 × 310 hp engines) at an average speed of 9 knots (16.67 km h⁻¹), resulting in maintenance vessel fuel consumption of 0.88 l km⁻¹ (refs. 28,56).

Total emissions from growing and harvesting seaweed were therefore calculated as:

where e_prod is total annual emissions from seaweed production (growth + harvesting), e_eqtrans is as calculated in equation (7), e_mnt is as calculated in equation (8) and s_dw is the DW of seaweed harvested annually per km².

Market value and climate benefits of seaweed

Further transportation and processing costs, economic value and net emissions of either sinking seaweed in the deep ocean for carbon sequestration or converting seaweed into usable products (biofuel, animal feed, pulses, vegetables, fruits, oil crops and cereals) were calculated on the basis of ranges of transport costs ($ tDW⁻¹ km⁻¹), transport emissions (tCO₂-eq t⁻¹ km⁻¹), conversion cost ($ tDW⁻¹), conversion emissions (tCO₂-eq tDW⁻¹), market value of product ($ tDW⁻¹) and the emissions avoided by product (tCO₂-eq tDW⁻¹) in the literature (Table 1). Market value was treated as globally homogeneous and does not vary by region. Emissions avoided by products were determined by comparing estimated emissions related to seaweed production to emissions from non-seaweed products that could potentially be replaced by seaweed (including non-CO₂ greenhouse gas emissions from land use)²⁴. Other parameters used are distance to nearest port (km), water depth (m), spatially explicit sequestration fraction (%)⁵⁷ and distance to optimal sinking location (km; cost-optimized for maximum emissions benefit considering transport emissions combined with spatially explicit sequestration fraction; see ‘Distance to sinking point calculation’ below). Each Monte Carlo simulation calculated the cost of both CDR via sinking seaweed and GHG emissions mitigation via seaweed products.

For seaweed CDR, after the seaweed is harvested, it can either be sunk in the same location that it was grown, or be transported to a more economically favourable sinking location where more of the seaweed carbon would remain sequestered for 100 yr (see ‘Distance to optimal sinking point’ below). Immediately post-harvest, the seaweed still contains a large amount of water, requiring a conversion from dry mass to wet mass for subsequent calculations³³:

where s_ww is the annual wet weight of seaweed harvested per km² and s_dw is the annual DW of seaweed harvested per km².

The cost to transport harvested seaweed to the optimal sinking location was calculated as:

where c_swtsink is the total annual cost to transport harvested seaweed to the optimal sinking location, c_transbase is the cost to transport 1 ton of material 1 km on a barge, d_sink is the distance in km to the economically optimized sinking location and s_ww is the annually harvested seaweed wet weight in t km⁻² as in equation (10).

The costs associated with transporting replacement equipment (for example, lines, buoys,

anchors) to the farming location and hauling back used equipment at the end of its assumed lifetime (1 yr for seeded line, 5–20 yr for capital equipment by equipment type) in the sinking CDR pathway were calculated as:

where c_eqtsink is the total annualized cost to transport both used and replacement equipment, c_transbase is the cost to transport 1 ton of material 1 km on a barge, m_eq is the annualized equipment mass in tons, d_sink is the distance in km to the economically optimized sinking location and d_port is the ocean distance to the nearest port in km. We assumed that the harvesting barge travels from the farming location directly to the optimal sinking location with harvested seaweed and replaced (used) equipment in tow (including used seeded line and annualized mass of used capital equipment), sinks the harvested seaweed, returns to the farm location and then returns to the nearest port (see Supplementary Fig. 16). These calculations assumed the shortest sea-route distance (see Distance to optimal sinking point).

The total value of seaweed that is sunk for CDR was therefore calculated as:

where v_sink is the total value (cost, if negative) of seaweed farmed for CDR in $ tDW⁻¹, v_cprice is a theoretical carbon price, c_swtsink is as calculated in equation (11), c_eqtsink is as calculated in equation (12) and s_dw is the annually harvested seaweed DW in t km⁻². We did not assume any carbon price in our Monte Carlo simulations (v_cprice is equal to zero), making v_sink negative and thus representing a net cost.

To calculate net carbon impacts, our model included uncertainty in the efficiency of using the growth and subsequent deep-sea deposition of seaweed as a CDR method. The uncertainty is expected to include the effects of reduced phytoplankton growth from nutrient competition, the relationship between air–sea gas exchange and overturning circulation (hereafter collectively referred to as the ‘atmospheric removal fraction’) and the fraction of deposited seaweed carbon that remains sequestered for at least 100 yr. The total amount of atmospheric CO₂ removed by sinking seaweed was calculated as:

where e_seqsink is net atmospheric CO₂ sequestered annually in t km⁻², k_atm is the atmospheric removal fraction and k_fseq is the spatially explicit fraction of sunk seaweed carbon that remains sequestered for at least 100 yr (see ref. 57).

The emissions from transporting harvested seaweed to the optimal sinking location were calculated as:

where e_swtsink is the total annual CO₂ emissions from transporting harvested seaweed to the optimal sinking location in tCO₂ km⁻², e_transbase is the CO₂ emissions (tons) from transporting 1 ton of material 1 km on a barge (tCO₂ per t-km), d_sink is the distance in km to the economically optimized sinking location and s_ww is the annually harvested seaweed wet weight in t km⁻² as in equation (10). Since the unit for e_transbase is tCO₂ per t-km, the emissions from transporting seaweed to the optimal sinking location are equal to (e_{mathrm{transbase}} times d_{mathrm{sink}} times s_{mathrm{ww}}), and the emissions from transporting seaweed from the optimal sinking location back to the farm are equal to 0 (since the seaweed has already been deposited, the seaweed mass to transport is now 0). Note that this does not yet include transport emissions from transport of equipment post-seaweed-deposition (see equation 16 below and Supplementary Fig. 16).

The emissions associated with transporting replacement equipment (for example, lines, buoys, anchors) to the farming location and hauling back used equipment at the end of its assumed lifetime (1 yr for seeded line, 5–20 yr for capital equipment by equipment type)^28,41 in the sinking CDR pathway were calculated as:

where e_eqtsink is the total annualized CO₂ emissions in tons from transporting both used and replacement equipment, e_transbase is the CO₂ emissions from transporting 1 ton of material 1 km on a barge, m_eq is the annualized equipment mass in tons, d_sink is the distance in km to the economically optimized sinking location and d_port is the ocean distance to the nearest port in km. We assumed that the harvesting barge travels from the farming location directly to the optimal sinking location with harvested seaweed and replaced (used) equipment in tow (including used seeded line and annualized mass of used capital equipment), sinks the harvested seaweed, returns to the farm location and then returns to the nearest port. These calculations assumed the shortest sea-route distance (see Distance to optimal sinking point).

Net CO₂ emissions removed from the atmosphere by sinking seaweed were thus calculated as:

where e_remsink is the net atmospheric CO₂ removed per ton of seaweed DW, e_seqsink is as calculated in equation (14), e_swtsink is as calculated in equation (15), e_eqtsink is as calculated in equation (16) and s_dw is the annually harvested seaweed DW in t km⁻².

Net cost of climate benefits

Sinking

To calculate the total net cost and emissions from the production, harvesting and transport of seaweed for CDR, we combined the cost and emissions from the sinking-pathway cost and value modules. The total net cost of seaweed CDR per DW ton of seaweed was calculated as:

where c_sinknet is the total net cost of seaweed for CDR per DW ton harvested, c_prod is the net production cost per DW ton as calculated in equation (6) and v_sink is the net value (or cost, if negative) per ton seaweed DW as calculated in equation (13).

The total net CO₂ emissions removed per DW ton of seaweed were calculated as:

where e_sinknet is the total net atmospheric CO₂ removed per DW ton of seaweed harvested annually (tCO₂ tDW⁻¹ yr⁻¹), e_remsink is the net atmospheric CO₂ removed via seaweed sinking annually as calculated in equation (17) and e_prod is the net CO₂ emitted from production and harvesting of seaweed annually as calculated in equation (9). For each Monte Carlo simulation, locations where e_sinknet is negative (that is, net emissions rather than net removal) were not included in subsequent calculations since they would not be contributing to CDR in that location under the given scenario. Note that these net emissions cases only occur in areas far from port in specific high-emissions scenarios. Even in such cases, most areas still contribute to CO₂ removal (negative emissions), hence costs from locations with net removal were included.

Total net cost was then divided by total net emissions to get a final value for cost per ton of atmospheric CO₂ removed:

where c_pertonsink is the total net cost per ton of atmospheric CO₂ removed via seaweed sinking ($ per tCO₂ removed), c_sinknet is total net cost per ton seaweed DW harvested as calculated in equation (18) ($ tDW⁻¹) and e_sinknet is the total net atmospheric CO₂ removed per ton seaweed DW harvested as calculated in equation (19) (tCO₂ tDW⁻¹).

GHG emissions mitigation

Instead of sinking seaweed for CDR, seaweed can be used to make products (including but not limited to food, animal feed and biofuels). Replacing convention products with seaweed-based products can result in ‘avoided emissions’ if the emissions from growing, harvesting, transporting and converting seaweed into products are less than the total greenhouse gas emissions (including non-CO₂ GHGs) embodied in conventional products that seaweed-based products replace.

When seaweed is used to make products, we assumed it is transported back to the nearest port immediately after being harvested. The annualized cost to transport the harvested seaweed and replacement equipment (for example, lines, buoys, anchors) was calculated as:

where c_transprod is the annualized cost per ton seaweed DW to transport seaweed and equipment back to port from the farm location, c_transbase is the cost to transport 1 ton of material 1 km on a barge, m_eq is the annualized equipment mass in tons, d_port is the ocean distance to the nearest port in km, s_ww is the annual wet weight of seaweed harvested per km² as calculated in equation (10) and s_dw is the annual DW of seaweed harvested per km².

The total value of seaweed that is used for seaweed-based products was calculated as:

where v_product is the total value (cost, if negative) of seaweed used for products ($ tDW⁻¹), v_mkt is how much each ton of seaweed would sell for, given the current market price of conventional products that seaweed-based products replace ($ tDW⁻¹), c_transprod is as calculated in equation (21) and c_conv is the cost to convert each ton of seaweed to a usable product ($ tDW⁻¹).

The annualized CO₂ emissions from transporting harvested seaweed and equipment back to port were calculated as:

where e_transprod is the annualized CO₂ emissions per ton seaweed DW to transport seaweed and equipment back to port from the farm location, e_transbase is the CO₂ emissions from transporting 1 ton of material 1 km on a barge, m_eq is the annualized equipment mass in tons, d_port is the ocean distance to the nearest port in km, s_ww is the annual wet weight of seaweed harvested per km² as calculated in equation (10) and s_dw is the annual DW of seaweed harvested per km².

Total emissions avoided by each ton of harvested seaweed DW were calculated as:

where e_avprod is total CO₂-eq emissions avoided per ton of seaweed DW per year (including non-CO₂ GHGs using a GWP time period of 100 yr), e_subprod is the annual CO₂-eq emissions avoided per ton seaweed DW by replacing a conventional product with a seaweed-based product, e_transprod is as calculated in equation (23) and e_conv is the annual CO₂ emissions per ton seaweed DW from converting seaweed into usable products. e_subprod was calculated by converting seaweed DW to caloric content⁵⁸ for food/feed and comparing emissions intensity per kcal to agricultural products²⁴, or by converting seaweed DW into equivalent biofuel content with a yield of 0.25 tons biofuel per ton DW⁵⁹ and dividing the CO₂ emissions per ton fossil fuel by the seaweed biofuel yield.

To calculate the total net cost and emissions from the production, harvesting, transport and conversion of seaweed for products, we combined the cost and emissions from the product-pathway cost and value modules. The total net cost of seaweed for products per ton DW was calculated as:

where c_prodnet is the total net cost per ton DW of seaweed harvested for use in products, c_prod is the net production cost per ton DW as calculated in equation (6) and v_product is the net value (or cost, if negative) per ton DW as calculated in equation (22).

The total net CO₂-eq emissions avoided per ton DW of seaweed used in products were calculated as:

where e_prodnet is the total net CO₂-eq emissions avoided per ton DW of seaweed harvested annually (tCO₂ tDW⁻¹ yr⁻¹), e_avprod is the net CO₂-eq emissions avoided by seaweed products annually as calculated in equation (24) and e_prod is the net CO₂ emitted from production and harvesting of seaweed annually as calculated in equation (9). For each Monte Carlo simulation, locations where e_prodnet is negative (that is, net emissions rather than net emissions avoided) were not included in subsequent calculations since they would not be avoiding any emissions in that scenario.

Total net cost was then divided by total net emissions avoided to get a final value for cost per ton of CO₂-eq emissions avoided:

where c_pertonprod is the total net cost per ton of CO₂-eq emissions avoided by seaweed products ($ per tCO₂-eq avoided), c_prodnet is total net cost per ton seaweed DW harvested for products as calculated in equation (25) ($ tDW⁻¹) and e_prodnet is total net CO₂-eq emissions avoided per ton seaweed DW harvested for products as calculated in equation (26) (tCO₂ tDW⁻¹).

Parameter ranges for Monte Carlo simulations

For technoeconomic parameters with two or more literature values (see Supplementary Table 1), we assumed that the maximum literature value reflected the 95th percentile and the minimum literature value represented the 5th percentile of potential costs or emissions. For parameters with only one literature value, we added ±50% to the literature value to represent greater uncertainty within the modelled parameter range. Values at each end of parameter ranges were then rounded before Monte Carlo simulations as follows: capital costs, operating costs and harvest costs to the nearest $10,000 km⁻², labour costs and insurance costs to the nearest $1,000 km⁻², line costs to the nearest $0.05 m⁻¹, transport costs to the nearest $0.05 t⁻¹ km⁻¹, transport emissions to the nearest 0.000005 tCO₂ t⁻¹ km⁻¹, maintenance transport emissions to the nearest 0.0005 tCO₂ km⁻¹, product-avoided emissions to the nearest 0.1 tCO₂-eq tDW⁻¹, conversion cost down to the nearest $10 tDW⁻¹ on the low end of the range and up to the nearest $10 tDW⁻¹ on the high end of the range, and conversion emissions to the nearest 0.01 tCO₂ tDW⁻¹.

We extended the minimum range values of capital costs to $10,000 km⁻² and transport emissions to 0 to reflect potential future innovations, such as autonomous floating farm setups that would lower capital costs and net-zero emissions boats that would result in 0 transport emissions. To calculate the minimum value of $10,000 km⁻² for a potential autonomous floating farm, we assumed that the bulk of capital costs for such a system would be from structural lines and flotation devices, and we therefore used the annualized structural line (system rope) and buoy costs from ref. 41 rounded down to the nearest $5,000 km⁻². The full ranges used for our Monte Carlo simulations and associated literature values are shown in Supplementary Table 1.

Distance to optimal sinking point

Distance to the optimal sinking point was calculated using a weighted distance transform (path-finding algorithm, modified from code in ref. 60) that finds the shortest ocean distance from each seaweed growth pixel to the location at which the net CO₂ removed is maximized (including impacts of both increased sequestration fraction and transport emissions for different potential sinking locations) and the net cost is minimized. This is not necessarily the location in which the seaweed was grown, since the fraction of sunk carbon that remains sequestered for 100 yr is spatially heterogeneous (see ref. 57). For each ocean grid cell, we determined the cost-optimal sinking point by iteratively calculating equations (11–20) and assigning d_sink as the distance calculated by weighted distance transform to each potential sequestration fraction (0.01–1.00) in increments of 0.01. Except for transport emissions, the economic parameter values used for these calculations were the averages of unrounded literature value ranges; we assumed that the maximum literature value reflected the 95th percentile and the minimum literature value represented the 5th percentile of potential costs or emissions, or for parameters with only one literature value, we added ±50% to the literature value to represent greater uncertainty within the modelled parameter range. For transport and maintenance transport emissions, we extended the minimum values of the literature ranges to zero to reflect potential net-zero emissions transport options and used the mean values of the resulting ranges. The d_sink that resulted in minimum net cost per ton CO₂ for each ocean grid cell was saved as the final d_sink map, and the associated sequestration fraction value that the seaweed is transported to via d_sink was assigned to the original cell where the seaweed was farmed and harvested (Supplementary Fig. 19). If the cost-optimal location to sink using this method was the same cell where the seaweed was harvested, then d_sink was 0 km and the sequestration fraction was not modified from its original value (Supplementary Fig. 18).

Comparison of gigaton-scale sequestration area to previous estimates

Previous related work estimating the ocean area suitable for macroalgae cultivation¹³ and/or the area that might be required to reach gigaton-scale carbon removal via macroalgae cultivation^13,19,36 has yielded a wide range of results, primarily due to differences in modelling methods. For example, Gao et al. (2022)³⁶ estimate that 1.15 million km² would be required to sequester 1 GtCO₂ annually when considering carbon lost from seaweed biomass/sequestered as particulate organic carbon (POC) and refractory dissolved organic carbon (rDOC), and assume that the harvested seaweed is sold as food such that the carbon in the harvested seaweed is not sequestered. The area (0.31 million km²) required to sequester 1 GtCO₂ in our study assumes that all harvested seaweed is sunk to the deep ocean to sequester carbon.

Additionally, Wu et al.¹⁹ estimates that roughly 12 GtCO₂ could be sequestered annually via macroalgae cultivation in approximately 20% of the world ocean area (that is, 1.67% ocean area per GtCO₂), which is a much larger area per GtCO₂ than our estimate of 0.085% ocean area. This notable difference arises for several reasons (including differences in yields, which in Wu et al. are around 500 tDW yr⁻¹ in the highest-yield areas, whereas yields in our cheapest sequestration areas from G-MACMODS average 3,400 tDW km⁻² yr⁻¹) that arise from differences in model methodology. First, Wu et al. model temperate brown seaweeds, while our study considers different seaweed types, many of which have higher growth rates, and uses the most productive seaweed type for each ocean grid cell. The G-MACMODS seaweed growth model we use also has a highly optimized harvest schedule, includes luxury nutrient uptake (a key feature of macroalgal nutrient physiology) and does not directly model competition with phytoplankton during seaweed growth. Finally, tropical red seaweeds (the seaweed type in our cheapest sequestration areas) grow year-round, while others, such as the temperate brown seaweeds modelled by Wu et al., only grow seasonally. These differences all contribute to higher productivity in our model, leading to a smaller area required for gigaton-scale CO₂ sequestration compared with Wu et al.

Conversely, the ocean areas we model for seaweed-based CO₂ sequestration or GHG emissions avoided are much larger than the 48 million km² that Froehlich et al.¹³ estimate to be suitable for macroalgae farming globally. Although our maps show productivity and costs everywhere, the purpose of our modelling was to evaluate where different types of seaweed grow best and how production costs and product values vary over space, to highlight the lowest-cost areas (which are often the highest-producing areas) under various technoeconomic assumptions.

Comparison of seaweed production costs to previous estimates

Although there are not many estimates of seaweed production costs in the scientific literature, our estimates for the lowest-cost 1% area of the ocean ($190–$2,790 tDW⁻¹) are broadly consistent with previously published results: seaweed production costs reported in the literature range from $120 to $1,710 tDW⁻¹ (refs. 40,41,61,62), but are highly dependent on assumed seaweed yields. For example, Camus et al.⁴¹ calculate a cost of $870 tDW⁻¹ assuming a minimum yield of 12.4 kgDW m⁻¹ of cultivation line (equivalent to 8.3 kgDW m⁻² using 1.5 m spacing between lines). Using the economic values from Camus et al. but with our estimates of average yield for the cheapest 1% production cost areas (2.6 kgDW m⁻²) gives a much higher average cost of $2,730 tDW⁻¹. Contrarily, van den Burg et al.⁴⁰ calculate a cost of $1,710 tDW⁻¹ using a yield of 20 tDW ha⁻¹ (that is, 2.0 kg m⁻²). Instead assuming the average yield to be that from our lowest-cost areas (that is, 2.6 kgDW m⁻² or 26 tDW ha⁻¹) would decrease the cost estimated by van den Burg et al. (2016) to $1,290 tDW⁻¹. Most recently, Capron et al.⁶² calculate an optimistic scenario cost of $120 tDW⁻¹ on the basis of an estimated yield of 120 tDW ha⁻¹ (12 kg m⁻²; over 4.5 times higher than the average yield in our lowest-cost areas). Again, instead assuming the average yield to be that in our lowest-cost areas would raise Capron et al.’s production cost to $540 tDW⁻¹ (between the $190–$880 tDW⁻¹ minimum to median production costs in the cheapest 1% areas from our model; Fig. 1a,b).

Data sources

Seaweed biomass harvested

We used spatially explicit data for seaweed harvested globally under both ambient and limited-nutrient scenarios from the G-MACMODS seaweed growth model³³.

Fraction of deposited carbon sequestered for 100 yr

We used data from ref. 57 interpolated to our 1/12-degree grid resolution.

Distance to the nearest port

We used the Distance from Port V1 dataset from Global Fishing Watch (https://globalfishingwatch.org/data-download/datasets/public-distance-from-port-v1) interpolated to our 1/12-degree grid resolution.

Significant wave height

We used data for annually averaged significant wave height from the European Center for Medium-range Weather Forecasts (ECMWF) interpolated to our 1/12-degree grid resolution.

Ocean depth

We used data from the General Bathymetric Chart of the Oceans (GEBCO).

Shipping lanes

We used data of Automatic Identification System (AIS) signal count per ocean grid cell, interpolated to our 1/12-degree grid resolution. We defined a major shipping lane grid cell as any cell with >2.25 × 10⁸ AIS signals, a threshold that encompasses most major trans-Pacific and trans-Atlantic shipping lanes as well as major shipping lanes in the Indian Ocean, the North Sea, and coastal routes worldwide.

Marine protected areas (MPAs)

We used data from the World Database on Protected Areas (WDPA) and defined an MPA as any ocean MPA >20 km².

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Source: Ecology - nature.com