The collapse and recovery potential of carbon sequestration by baleen whales in the Southern Ocean

Limiting climate warming below 2°C requires both reducing anthropic greenhouse gas emissions and 33 sequestering more atmospheric carbon. Natural Climate Solutions (NCS) rely on the ability of 34 ecosystems to capture and store carbon. Despite the important role of marine megafauna on the 35 ocean carbon cycle, its potential as a NCS has not yet been explored. Here, we quantify the amount of 36 carbon potentially sequestered by five baleen whale species across the Southern Hemisphere between 37 1890 and 2100 through both the sinking of carcasses after natural death and the fertilisation of 38 phytoplankton by nutrients in faeces. At their pre-exploitation abundances, the five whales could 39 sequester 10.6 10 6 tonnes of carbon per year (tC.yr -1 ) but this natural carbon sink was reduced at 2 10 6 40 tC.yr -1 in 1965 due to commercial whaling. However, the restoration of whale populations could 41 sequester 8.7 10 6 tC.yr -1 at the end of the 21 st century suggesting an efficient but neglected NCS that 42 remains to be estimated globally including all marine vertebrates. 43 46 47 48 49 50 51


59
The concentration of atmospheric carbon dioxide (CO2) has dramatically increased since the beginning 60 6 concentrations of macronutrients (nitrates and phosphates) are high but primary productivity is low 157 21 . Primary productivity, through phytoplankton growth, is thus limited by the availability of trace 158 elements (Fe, Cu, Zn, Co, Cd), especially iron 21 . Iron-rich whale faeces thus stimulate phytoplankton 159 growth and, by extension, carbon sinking 33 . To quantify this sequestration pathway, we estimated the 160 amount of iron supplied in the euphotic zone by whales based on the egestion rate and the bioavailable 161 iron concentration in faeces. At their biotic capacity, the five species defecate about 9.3 10 3 tonnes 162 (range: 3.2 10 3 -18.4 10 3 tonnes) of iron. About 11 10 2 tonnes of this iron (12.2%) can be used by 163 phytoplankton (see Methods). The phytoplankton carbon flux at 200 metres depth is then 10.4 10 6 164 tC.yr -1 (range: 3.6 10 6 -20.5 10 6 tC.yr -1 ). This is about 40 times more than sequestration via carcasses 165 (Fig. 2b). The main contributing species are fin whales (47%), followed by blue and southern right 166 whales with 24% and 16% respectively. Finally, antarctic minke and humpback whales contribute at 167 only 8% and 5%, respectively (Fig. 2b). 168 Overall the five whales at their pre-exploitation abundances across the Southern Hemisphere can 169 sequester up to 10.6 10 6 tC.yr -1 (range: 3.8 10 6 -20.8 10 6 tC.yr -1 ). The indirect sequestration pathway, 170 via the stimulation of phytoplankton growth, represents about 98% of the total carbon flux towards 171 the deep sea. 172 173 Carbon sequestration dynamics from 1890 to 2100 174 We predicted carbon sequestration dynamics from 1890 to 2100 under various exploitation levels and 175 climate change scenarios. A stable phase from 1890 to 1912 was followed by a sharp drop in the 176 amount of carbon sequestered over the exploitation period (Fig. 3). Indeed, all species experienced 177 population declines, particularly the main contributors (fin and blue whales), which were reduced to 178 approximately 3% and 0.5% of their pre-exploitation stock, respectively ( Supplementary Fig. 2). As a 179 result, carbon sequestration from these Southern Hemisphere whales decreased to a minimum of 2 180 10 6 tC.yr -1 (range: 0.9 10 6 and 3.7 10 6 tC.yr -1 ) in 1965, i.e. 19% of the pre-exploitation level. In the model 181 without climate change, carbon sequestration would reach 8.9 10 6 tC.yr -1 (range: 3.4 10 6 and 17 10 6 182 tC.yr -1 ) in 2100 under the predicted increase of antarctic minke whale populations and the recovery of 183 all other species. However, in the model including the effects of climate change, no species, other than 184 the antarctic minke whale, would be able to recover to their pre-exploitation level before the end of 185 the 21 st century. The antarctic minke whale would increase rapidly and reach a population size greater 186 than that predicted in the model without climate change ( Supplementary Fig. 2). However, this 187 population increase alone would not be sufficient to recover pre-exploitation level for carbon 7 5.2 10 6 tC.yr -1 (range: 2.2 10 6 -9.6 10 6 tC.yr -1 ), i.e. only half of the pre-exploitation level, under the 190 major influence of antarctic minke whales. 191 For both models, fin whales were the major contributors until 1959 then antarctic minke whales 192 became the main contributors ( Fig. 3)  Finally, we present the cumulated deficit in carbon sequestration over the whole period (1890 to 2100) 202 due to whaling and under two scenarios of climate change. This deficit was estimated as the difference 203 between the total sequestration (direct and indirect ways) estimated through time under population 204 fluctuations (Fig. 3) and the total sequestration at the carrying capacity (pre-exploitation level). It thus 205 expresses the amount of carbon that has not been sequestered since 1890 and will not be till 2100 due 206 to the combined effects of whaling and climate change. This deficit would reach more than 1Gt.C in 207 2100 without climate change ( Fig. 4a) but 1.2 Gt.C under the more realistic "business as usual" scenario 208 with climate change (Fig. 4b) The five whale populations at their carrying capacity (pre-exploitation level), represent an annual 222 sequestration potential of 2.5 10 5 ± 0.5 10 5 tC.yr -1 via carcasses sinking toward the deep ocean. This 223 estimate is 30% higher than that of Pershing et al. 14 who estimate this sequestration at 1.6 10 5 tC.yr -224 1 with four additional species (grey whale, sei whale, Bryde's whale, bowhead whale) taken into 225 account. This difference is explained by their lower estimates of carrying capacities. Our estimates 226 were based on hindcasting population dynamics given historical whaling and fitted to current surveys, 227 so are likely more robust estimates of historical carrying capacities. 228 For the five whale species, the fertilization-induced sequestration reaches up to 10.4 10 6 tC.yr -1 with a 229 range of 3.6 10 6 -20.5 10 6 tC.yr -1 at carrying capacity or pre-exploitation level. Previous estimate of 230 the indirect sequestration by sperm whales in the Southern Ocean was 0.4 10 6 tC.yr -1 , which is 231 consistent with our results given the population size of sperm whales compared to those of the five 232 species of baleen whales 13 . Given the difference of population size between sperm whales and our 233 five baleen whale species, our results are of the same order of magnitude as this previous assessment. 234 Indeed, these baleen whales represent a biomass about 100 times larger than that of sperm whales 235 and they consume prey (krill) richer in iron (1.7.10-4 Kg iron/Kg dry weight 23 ) than sperm whales which 236 mainly consume cephalopods (0.75.10-5 Kg iron/Kg dry weight 13 ). 237 We show that the role of whales in carbon sequestration resides more in their capacity to boost other 238 biological carbon pumps (like marine snow) through fertilization than in exporting their own biomass 239 (carcasses) in the deep sea (98% against 2% of the total flux). Thanks to these two sequestration 240 pathways, the annual carbon flux induced by whales prior to their exploitation (10.6 10 6 tC.yr -1 on 241 average) was comparable to the carbon fluxes observed in other ecosystems, especially coastal ones, 242 such as mangroves (31.2 10 6 -34.4 10 6 tC.yr -1 ) or salt marshes (4.8 10 6 -87.2 10 6 tC.yr -1 ) (Supplementary 243 Table 1). 244 Due to their long-life cycles, the recovery of many baleen whale populations after over-exploitation 245 has been a very slow process. Therefore, the consequences of whaling extend well beyond the 246 exploitation period and currently limit sequestration of these five baleen whales to 3.4 10 6 tC.yr -1 (1.4 247 10 6 -6.1 10 6 tC.yr -1 ), i.e. 32% of the pre-exploitation level. In the model without climate change, the 248 carbon sink could be restored at 84% of its pre-exploitation level by 2100, then reaching 8.9 10 6 tC.yr -249 1 . However, the recovery of whale populations and of the carbon pump may be delayed and weakened 250 by climate change 17,18 . This can be explained by changes in the abundance and distribution of krill due 251 to changing primary productivity patterns in the Southern Ocean 17 . Furthermore, the distribution of 252 krill is expected to contract southward due to increasing temperature and reduced sea-ice extent 35 . mid-latitudes areas (humpback whales, fin whales and southern right whales) although copepods also 255 make up a large proportion of the diet of southern right whales. Antarctic minke whales and blue 256 whales could benefit most, especially in the Pacific area, of the ice-extent reduction in the Southern 257 Ocean because of their ice-dependency, assuming they can shift their distribution southwards to 258 follow the krill 17 . However, since antarctic minke whales would increase under marked climate change 259 scenario (due to increasing biomass of krill at high latitudes where they are distributed), prey 260 availability for other species may be reduced. Our MICE model includes interspecific competition 261 between the five whale species, and thus accounts for associated effects of changing prey availability 262 on whale recovery given changing whale densities. As a consequence, the recovery of some species is 263 predicted to slow down, with estimated declines again during the 21 st century for humpback, fin and 264 southern right whales 17 . As a result, despite a predicted increase of antarctic minke whale populations, 265 the total carbon flux would not return to its pre-exploitation level due to the negative impact of climate 266 change on other species. A negative feedback loop between climate and whale populations could 267 therefore occur in the southern hemisphere. However, these results should be taken with caution as 268 they present several uncertainties. 269 270

Limits and uncertainties 271
This first estimate of the carbon flux generated by whales is restricted to five baleen species in the 272 southern hemisphere, whereas there are fifteen species of baleen whales globally. Indeed, the number 273 of species considered here was restricted to those included in the whale populations models 16,17 (i.e. 274 species commercially exploited in Antarctic waters, in most cases feeding predominantly on Antarctic 275 krill, for which enough survey data were available). Therefore, our model may significantly 276 underestimate the importance of carbon sequestration mediated by whales in the Southern Ocean 277 and at global scale by excluding other southern species (Bryde's whale, Pygmy right whale and Dwarf 278 antarctic minke whale), northern species (bowhead whale, gray whale, omuras whale, northern right 279 whale) and toothed whales. Moreover, the indirect carbon sequestration mechanism was only 280 considered during the summer period, when the whales are located in the Southern Ocean. During 281 winter, they migrate towards tropical regions to breed. However, other nutrients (nitrogen, 282 phosphorus) also excreted by whales 34 limit phytoplankton productivity in these areas 21 . Thus, they 283 could indirectly promote carbon sequestration during the breeding season and migration. The carbon 284 sequestration induced by whales is therefore likely to be much larger and extends towards the tropics. 285 tissues have the same carbon concentration 36 and some (fat tissue, muscle) may be consumed 288 primarily by scavengers 37 . In order to gain precision, it seems essential to determine the carbon level 289 in the different types of tissue (bone, muscle, blubber, viscera) for each species. On the other hand, 290 the proportion of biomass reaching the deep ocean before being consumed or remineralised is 291 uncertain and probably highly variable, depending on the presence of scavengers or currents for 292 example. Finally, these migratory species experience significant weight variations during the year 38 : 293 they may gain several tonnes during the summer and be considerably thinner at the end of the 294 breeding season. The amount of carbon sequestered therefore depends on the seasonality of natural 295 mortality, which is not taken into account in our study. 296 The main uncertainty in our estimated indirect sequestration is the amount of bioavailable iron 297 provided by whales at sea surface. First, iron concentration values in whale faeces were obtained with 298 few replicates, and those of antarctic minke whales were estimated from available data for other 299 species. Surprisingly, the iron concentration in southern right whale faeces 27 is higher than that of 300 other species, whereas it would be expected to be lower. Indeed, the proportion of krill, an iron 301 accumulator, in their diet is lower compared to other species 23 . This may be indicative of high 302 intraspecies, temporal or spatial variability in faecal iron concentration. On the other hand, individual 303 variability within species has also been ignored. Iron retention in the body varies with age and 304 reproductive status 39 . To address this issue, the population could be divided into different categories 305 (juveniles, adults, pregnant or lactating females) and the iron concentration in the faeces could be 306 estimated for each category. 307 Secondly, to estimate the response of phytoplankton to the iron supply, the fraction of iron contained 308 in faeces, which is finally incorporated by phytoplankton, needs to be estimated. This is influenced by 309 both bioavailability of the iron supply and the fate of this iron. On the one hand, iron bioavailability in 310 the ocean is influenced by many processes (dissolved or particulate form, degree of oxidation, 311 complexation with organic ligands, etc.) 40 . No studies have estimated the bioavailable fraction of iron 312 released by whales. We considered the dissolved iron or the iron dissolving in the first 12 hours to be 313 bioavailable but our values are probably underestimated. Indeed, iron particles are small (a few 314 micrometres) and their density is close to water density so they sink slowly and could remain for 315 several days in the euphotic zone where dissolution may continue 41 . In addition, the dissolution 316 experiment of Ratnarajah et al. 41 was conducted in the dark. Light can increase dissolution (photo-317 dissolution) 42,43 . On the other hand, both heterotrophic bacteria and autotrophic phytoplankton 318 depend on the available iron pool for growth. So, auto-and heterotrophs could compete for iron when 319 dissolved organic carbon limitation of bacterial growth is alleviated 44 . However, heterotrophic bacteria 320 can also recycle iron, increasing iron solubility and availability by recycling particulate iron into 321 dissolved iron, producing organic ligands binding to the iron and increasing iron suspension and 322 bioavailability 45 . 323 These various processes can either increase or decrease the availability of iron for phytoplankton 324 making difficult to assess the bioavailability and fate of iron supplied by whales in the euphotic zone. 325 Therefore, a better estimation of the amount of iron released by the different whale species and its 326 availability for phytoplankton seems essential to precisely quantify the indirect sequestration pathway. 327 Finally, to quantify the additional carbon sequestered owing to iron inputs by whales, experimental 328 measurements of carbon exported at 200 meters in response to iron addition were used 33 . These data 329 may be a source of overestimation since we assume that all carbon exported at 200 metres will 330 sediment and be sequestered. However, organic matter may be remineralised before reaching the 331 deep ocean 46 or be broken down into smaller particles that sink more slowly 47 . 332 333

Importance of whale diversity for carbon sequestration 334
We studied five species of baleen whales, diverse in size, longevity and life cycles, each contributing 335 differently to the two carbon sequestration pathways. Sequestration via carcasses sinking is mainly 336 supported by the most massive species such as blue and fin whales, which can reach approximately 337 117 and 65 tonnes per individual, respectively. In contrast, the indirect sequestration pathway 338 depends mainly on the smaller but more abundant antarctic minke whale. This is explained by the fact 339 that prey consumption (determining the amount of faeces) is a hypo-allometric function of body mass 340 48 , whereas in the case of carcasses, the carbon sequestration is a linear function of body mass. 341 Considering that biomass and abundance are the two components that determine the relative 342 contribution of a species to carbon sequestration, the direct pathway is more dependent on individual 343 body mass while the indirect pathway is more dependent on the number of individuals. Southern right 344 whales are also weaker contributors to the direct sequestration pathway because of their tendency to 345 float after death. Conversely, their strong iron concentration in faeces make them very efficient in 346 sequestering carbon via the fertilization pathway. 347 This complementarity between species, resulting from the diversity of their traits, helps to maintain 348 the different sequestration pathways. More generally, this functional complementarity is key to 349 support ecosystem multifunctionality 49,50 and increases the associated ecosystem services 51 . 350 Moreover, it is generally considered that the stability of a system increases with the diversity of its To consider whale population restoration as a Natural Climate Solution (NCS), several criteria must be 369 met: effectiveness, the presence of co-benefits and the limitation of associated disadvantages, and the 370 governability, i.e. the ability to implement this solution, while managing the conflicts and benefits 371 generated by its implementation 55 . 372 To maintain the global warming at a maximum of +2°C, total emissions must be less than 810 10 9 tCO2-373 eq between 2016 and 2100 6 and achieve zero net emissions by 2075 56 . This implies not only reducing 374 GHG emissions but also offsetting unavoidable emissions. For example, in the field of road transport, 375 our annual emissions are expected to decrease from 5.75 10 9 tCO2-eq.yr -1 in 2015 to 2.6 10 9 tCO2-eq.yr -376 1 in 2050 thanks to energy efficiency gains and cleaner fuels 57 . By 2050, southern baleen whales should 377 be able to offset 0.7% of these persistent road transport emissions by sequestering an average of 17 378 10 6 tCO2-eq.yr -1 . In addition, by 2100, GHG emissions from all transports should not exceed 1.8 10 9 379 tCO2-eq.yr -1 57 . According to the model predictions without climate change, these remaining emissions 380 could be compensated up to 1.8% by southern baleen whales, which would sequester about 32.6 10 6 381 tCO2.yr -1 in 2100. On the other hand, if climate change continues to follow a RCP 8.5 scenario 20 , 382 southern whales would be able to sequester only 19 10 6 tCO2.yr -1 , offsetting 1.1% of global transport 383 emissions in 2100. Moreover, if they are managed to be restored to their pre-exploitation levels by the 384 end of the 21 st century, they could compensate up to 2.1% of these emissions. Thus, although many 385 whale populations have been severely depleted by whaling, restoring their populations could In addition, the presence of whales is associated with many other benefits, promoting the good health 388 of ecosystems and some services to human societies. Indeed, whale carcasses are an essential source 389 of food for abyssal ecosystems 10,58 . Thanks to their enriching action on their planktonic environment, 390 they can be described as ecosystem engineers, favouring lower trophic levels 59 . Whales enhance krill 391 growth through increased phytoplankton production 39 , which also participates very efficiently in 392 carbon sequestration through the production of fast-sinking particulate faeces 24   The weight of whales, and by extension the amount of carbon they contain, depends on their age. The 461 age structure of the population in a given year, i.e. the number of individuals in an age class, was 462 constructed using demographic parameters for each species derived from the MICE model 463 Table 2 ; 3 = 34@ . @ ⋯ @43 = 43 . @ ; @ ; @>3 = . @ ⋯ K = K4@ . @

479
This age structure calculation is applied every year from 1890 to 2100. We assume that age structure 480 does not vary over time. Indeed, the mass of individuals is almost constant in adulthood. It is therefore 481 assumed that no adult year class is more impacted by whaling than another. Among juveniles, very 482 few catches have been reported for the whale species in this study 68 . 483 (2) (3) (4) In order to calculate the number of individuals dying naturally each year, the natural mortality rate (1-484 S or 1-Sjuv) was applied to the numbers in each age class (Supplementary Table 2). This gives the 485 number of individuals dying per age class in a given year. 486 To obtain the biomass of an age class, the number of individuals in this class was multiplied by the 487 corresponding individual body mass. The mass of individuals at each age follows the Von Bertalanffy 488 equation, whose parameters, depend on both species and sex 14 (Supplementary Table 3). Let a the 489 age, minf the maximum size of individuals, k the growth rate and a0 the theoretical age at which the 490 mass is zero, the mass m of an individual aged a is: 491 The total population biomass (Btot) was calculated from the biomass of each age class (Ba) as follows:

494
To assess the amount of carbon sequestered by the sinking of whale carcasses, the biomass was 495 converted into carbon mass. Several estimates of carbon content in whale tissues are available: 10.5% 496 36 and 15% 9 . A carbon rate of 12.5% ± 2.5% was thus used, assuming that this rate is identical for all 497 individuals without distinction of species, sex or age. To calculate the amount of carbon, the biomass 498 was multiplied by this carbon rate. 499 The fate of the carcasses depends on several factors. Attacks by predators such as killer whales (Orcinus 500 orca) are rare and very rarely lethal 69,70 . Most dead individuals should therefore sink and sequester 501 carbon in the deep sea. However, carcasses do not sink in their entirety to the ocean floor because 502 they are partly consumed by scavengers like sharks 71 or killer whales 72 or degraded by 503 microorganisms. It was estimated that between 50% and 90% of a carcass reaches the ocean floor 73 . 504 The conservative estimate of 50% has been used for all species except southern right whales, which 505 tend to float after death due to a higher proportion of blubber 74 . For southern right whales we 506 considered that only 10% of carcass biomass is sequestered in the deep sea 14 . 507 508

Indirect carbon sequestration via fertilization 509
Whales play also a role in the carbon cycle through their ability to fertilize the ocean with nutrients 510 contained in their faeces (Fig. 1b) Table 4). For humpback and fin whales, only a mean value was given. We applied the Taylor's law 78 to 540 estimate the associated variance (V) of a biological measure from its mean (µ) using the power 541  Table 4). Finally, the mass of 547 iron (miron in mg) defecated by each whale species is obtained by converting the wet mass of faeces 548 (Dtot) to dry mass, considering that the faeces are 75% water, and multiplying this quantity by the iron 549 concentration ([Fe] in mg/Kg, dry mass): 550 The need for whales to return to the surface to breathe and the limited duration of their dives suggest 552 that all faeces are released into the euphotic zone 80 . However, only one part of the iron defecated can 553 benefit to phytoplankton productivity. Indeed, iron can take different chemical forms and many factors 554 influence its bioavailability (dissolved or particulate form, degree of oxidation, complexation with 555 organic ligands). Iron in dissolved form (< 0.2 µm) is generally considered to be the most bioavailable 556 fraction 81 . More than 87% of the iron in whale faeces is in particulate form (> 0.2 µm) 41 . However, 557 particulate iron can dissolve over time and become more bioavailable. The proportion of bioavailable 558 iron in whale faeces was estimated to be 12.2% after 12 hours based on the results of a dissolution 559 experiment of particulate iron from whale faeces 41 . This rate was used to obtain the amount of 560 bioavailable iron released by whales in the euphotic zone. 561 The iron provided to the sea surface by the whales stimulates phytoplankton growth. However, not all 562 of the fixed carbon is sequestered in the deep sea because a part is rapidly remineralised. Several   Dynamics of carbon sequestration generated by the ve baleen whale species. Via the two pathways (carcasses and fertilization) between 1890 and 2100 without climate change (a) and with climate change (b). Above the curves the species with the highest contribution to the total sequestration is represented.
Vertical bars represent the switch between two major contributing species. Shaded areas represent the high and low estimations for carbon sequestration. Cumulative carbon sequestration de cit from 1890 to 2100 in the deep Southern Ocean by baleen whales. The total amount of non-sequestered carbon is compared to the pre-exploitation levels of whale populations, without climate change (a) and with climate change (b). Cumulative carbon sequestration de cit for each species without climate change (C) and with climate change (D). Shaded areas represent the high and low estimations for carbon de cit.