Reply to Macreadie et al Comment on 'Geoengineering with seagrasses: is credit due where credit is given?'

M2017 pointed out that the units were missing from the global carbon burial estimates in table 1 of J&M2016 and that they could not reproduce our global fluxes from the values presented in the table. They thought that the rationale behind the calculation was unclear, that we had misrepresented the Kennedy et al (2010) global carbon burial rates, and that we had incorrectly calculated global carbon burial.

We did inadvertently omit the units for the global carbon sequestration values from the table, which should have been listed as kg C yr⁻¹ (as they were on the first page of the text). In addition, in the table we incorrectly reported the minimum sediment accumulation rate for our calculation as 0.003 g cm⁻² yr⁻¹, rather than 0.01 g cm⁻² yr⁻¹. We appreciate the opportunity to correct these errors. We have included a corrected version of that table in box 1, together with an explanation of the calculations and assumptions. Our calculated global carbon sequestration rate was correctly reported in J&M2016, as was the overestimate factor. The Kennedy et al (2010) values were correctly represented in the original table, but they were in kg C yr⁻¹, not g C yr⁻¹, as M2017 assumed.

Macreadie et al (2017) suggested that we should have used only the autochthonous portion of Kennedy's rate to represent earlier estimates of carbon burial by seagrasses. That would be reasonable, given that we had excluded part (though not all) of the allochthonous carbon from our own calculation by subtracting the average 'outside-patch' carbon concentration from the average 'inside-patch' concentration that Kennedy et al (2010) had reported (box 1). If we substitute the autochthonous carbon burial rates provided by M2017 (7–40 Tg C yr⁻¹), the factor of the overestimate (min/min, max/max) is 4—465 times. We could also have calculated a wider range of overestimate (5–7400 times) if we had calculated the ratio as max/min, min/max (see box 1).

The exact factor by which existing global rates are overestimated is not the important point. That factor cannot be calculated with any certainty from the data currently available from seagrass meadows. We included the calculation only to illustrate the effect of the methodological problems raised in J&M2016. The identified problems with the current methods systematically bias the calculation high, for the reasons explained in J&M2016 and in sections 2.2 and 2.3 of this paper.

M2017 stated that, in citing only Marbà et al (2015) as an example of the correct use of ²¹⁰Pb in seagrass sediments, J&M2017 had missed multiple papers (Mateo et al 1997, Serrano et al 2012, Macreadie et al 2015a, Macreadie et al 2015b, Serrano et al 2016a, Macreadie et al 2012, Greiner et al 2013, Serrano et al 2016b). We did not miss these papers. Of the eight papers, four were published after or at the same time as our paper was submitted (25 June 2015), so we could not have cited them. In addition, of the eight papers, only four actually include ²¹⁰Pb dating. In J&M2016 our intention was not to enumerate widespread flaws in individual papers; rather we explained the inadequacies of the methods and concentrated on specific directions on how to improve the international protocols. However, since Macreadie et al have raised this specific subset of papers, we will discuss below each of the four that used ²¹⁰Pb.

Before doing that, we need to explain the effect of bioturbation and other surface mixing on radionuclide profiles in sediments. For laminated sediments (e.g. oligotrophic lake sediments or anoxic sediments) an assumption is frequently made that no post-depositional processes affect the ²¹⁰Pb profile other than radio-decay. This assumption is not valid where surface mixing occurs (e.g. Appleby 2001). For more than five decades, geochemists have recognized the mixing problem in ocean sediments, and have developed models to deal with its effect on radionuclide profiles (e.g. Goldberg and Koide 1962, Boudreau 1997). The surface layer of oxic marine sediment is almost always mixed.

Neglecting the effects of mixing leads to sedimentation rate estimates that are too high (Silverberg et al 1986). Where surface mixing occurs, any specific layer within the sediments will contain material deposited over a range of years approximated by the depth of the mixed layer (cm) divided by the sedimentation rate (cm yr⁻¹) (e.g. see Guinasso and Schink 1975, Johannessen and Macdonald 2012). Therefore, in mixed sediments specific layers cannot be identified with a particular year of deposition.

The cores presented by Greiner et al (2013) were not interpretable using ²¹⁰Pb. In the core for which they reported a sedimentation rate, older material had clearly been deposited on top of younger material, as evidenced by the inverted ²¹⁰Pb profile. It is not clear how the sedimentation velocity of 0.66 cm yr⁻¹ was derived, nor how the authors then arrived at a detailed, 100 year history of variability in sedimentation rate. The Year-4 core showed a lower inventory of ²¹⁰Pb over its whole length than did the Year-0 core, which, contrary to the authors' suggestion, could not be explained as the result of bioturbation. They also assigned discrete ages to each depth in the bioturbated core.

In the Macreadie et al (2012) paper, the sampling resolution (3 cm thick slices at 4 cm or 10 cm intervals) was insufficient to resolve the surface mixed layer. Despite commenting that the cores were bioturbated, based on the observed sediment mottling, the authors assigned a discrete age to each depth in the core and then used those ages to calculate a sedimentation velocity (cm yr⁻¹).

The remaining two papers cited by M2017 were published after J&M2016 was submitted. These more recent papers avoided many of the problems associated with the earlier papers, but they still contain errors in interpretation, as discussed below.

Serrano et al (2016a) calculated the sedimentation rate below the surface mixed layer and explicitly accounted for supported ²¹⁰Pb in the core. These are appropriate steps. However, a comparison with the ²¹⁰Pb inventory shows that some of their calculated sedimentation rates cannot be correct. In one core, for example ('P. sinuosa 2 m'), ²¹⁰Pb was indistinguishable from background by 15 cm depth; since ²¹⁰Pb decays to background after ∼110 years (5 half-lives), the maximum sedimentation velocity possible (even in the absence of mixing, which pushes younger ²¹⁰Pb deeper into the core) is 0.14 cm yr⁻¹, but the authors reported a sedimentation velocity of 0.3 cm yr⁻¹.

Serrano et al (2016b) appear to have calculated sedimentation velocity correctly, but then, after stating that there was a 15 cm mixed layer, they assigned dates (13 years apart at the tops of the cores) to depths down the core and used those dates to determine the depth for the 100 year inventory. With the reported sedimentation velocity of 0.27 cm yr⁻¹, mixing in the 15 cm surface mixed layer would have smeared together about 56 years of accumulated sediment.

M2017 also questioned our use of sediment accumulation rates from general coastal areas. They suggested that we should have used rates calculated specifically from seagrass ecosystems, citing, in addition to the papers discussed above, Miyajima et al (1998), Serrano et al (2012), Serrano et al (2014) and Macreadie et al (2015a).

Miyajima et al (1998) did not estimate a sediment accumulation rate. Serrano et al (2012, 2014) calculated the rate of Posidonia plant material burial down the core, by measuring the ¹⁴C age of pieces of plant material throughout the core. This seems like a reasonable approach for this particular type of seagrass that forms enormous root mattes, although it would be impractical for other seagrasses. From the rate of burial of plant material, they calculated a sediment accumulation rate of 0.09 g cm⁻² yr⁻¹, which fits within the global range that we used.

The Macreadie et al (2015a) paper was published after J&M2016 was submitted. The authors reported sedimentation velocities of 0.01–0.19 cm yr⁻¹ but did not provide information that would permit the calculation of sediment accumulation rate in g cm⁻² yr⁻¹.

The rates that we used in our calculation were drawn from coastal sites all over the world, which made them more representative than a set of rates measured only in coastal Australia. We only included rates given in or convertible to units of g cm⁻² yr⁻¹, which permitted a direct calculation of carbon burial (g C cm⁻² yr⁻¹). Including a wider range of sediment accumulation rates, as we did, actually yielded a higher global carbon burial rate and a lower overestimate factor than would have been derived using only seagrass sites.

Macreadie et al stated that they agreed with J&M 2016 'that carbon stock estimates in combination with ²¹⁰Pb age dating is one of the best approaches to accurately calculate carbon accumulation rates in seagrass meadows.' They appear to have misunderstood the method that we recommend. (see section 3). It is not correct to divide carbon inventory, or 'stock,' over a particular depth range by the age calculated for that depth range, as presented in several recent studies (e.g. Serrano et al 2016a, 2016b). This method has two problems. (1) The inventory includes the enriched layer of labile organic carbon at the surface of the sediments, which is recycled to CO₂, not buried, on the timescale of 100 years. (According to Miyajima et al the top 10 cm of sediment in a seagrass meadow on the Great Barrier Reef was turned over / resuspended on the timescale of 17–170 d.) (2) In sediments that are physically or biologically mixed, this method also underestimates the age of the sediments in the interval considered, because younger (higher-activity) ²¹⁰Pb is mixed downward in the core. Each of these errors overestimates the rate of carbon burial. Multiplied together, the overestimates are compounded.

M2017 disagreed with our statement that it was the long-term sequestration rate, rather than instantaneous inventory, that ultimately mattered. They stated that the real potential of seagrass protection was to avoid 'emissions from disturbed sediments after canopy loss,' and that 'the burial capacity of seagrass meadows is small in terms of potential crediting.'

In fact, burial is more important than inventory, because the inventory in the surface sediments is volatile. It can be wiped out at any time by moorings, trawling, or other destructive processes. According to Macreadie et al (2015b), at one site half of the lost inventory was recovered within 5–10 years of replanting. Such volatility is not equivalent to burial on the timescale of 100 years, as required for carbon credits.

We agree that both the surface sediment inventory and the seagrass meadows themselves are valuable and should be preserved where possible, but to qualify for carbon credits, it is long-term burial, not deposition, that matters. Net community production, as raised by M2017, is irrelevant to long-term sequestration, since most of the new organic carbon is remineralized on the timescale of weeks to years.

M2017 stated that the Kennedy et al (2010) global estimates (48–112 Tg total C yr⁻¹ globally, or 12–40 Tg C yr⁻¹ autochthonous only) were reasonable, because they were 'within the range of estimates based on seagrass carbon burial data published in peer-reviewed literature.' Claiming that the global estimates summarized by Kennedy are reasonable because they agree with the earlier estimates on which they were based seems like a circular argument. We would suggest, rather, that since the earlier estimates were based on incorrect assumptions which systematically overestimated the carbon burial rates, the Kennedy estimates are also too high. For example, it is not reasonable to estimate global carbon burial rates in seagrass meadows based only on measurements made in Posidonia beds, since this genus produces unusually large root mattes that are not found in other types of seagrasses, nor is it correct to base carbon burial rates on carbon concentrations measured in the top 5–10 cm of the sediment.