231 Pa and 230 Th in the ocean model of the Community Earth System Model ( CESM 1 . 3 )

The sediment 231Pa / 230Th activity ratio is emerging as an important proxy for deep ocean circulation in the past. In order to allow for a direct model–data comparison and to improve our understanding of the sediment 231Pa / 230Th activity ratio, we implement 231Pa and 230Th in the ocean component of the Community Earth System Model (CESM). In addition to the fully coupled implementation of the scavenging behavior of 231Pa and 230Th with the active marine ecosystem module (particle-coupled: hereafter p-coupled), another form of 231Pa and 230Th have also been implemented with prescribed particle flux fields of the present climate (particle-fixed: hereafter p-fixed). The comparison of the two forms of 231Pa and 230Th helps to isolate the influence of the particle fluxes from that of ocean circulation. Under present-day climate forcing, our model is able to simulate water column 231Pa and 230Th activity and the sediment 231Pa / 230Th activity ratio in good agreement with available observations. In addition, in response to freshwater forcing, the p-coupled and p-fixed sediment 231Pa / 230Th activity ratios behave similarly over large areas of low productivity on long timescales, but can differ substantially in some regions of high productivity and on short timescales, indicating the importance of biological productivity in addition to ocean transport. Therefore, our model provides a potentially powerful tool to help the interpretation of sediment 231Pa / 230Th reconstructions and to improve our understanding of past ocean circulation and climate changes.


Introduction
The sediment 231 Pa / 230 Th activity ratio has been one major proxy for ocean circulation in the past (e.g., Yu et al., 1996;McManus et al., 2004;Gherardi et al., 2009).Quantities of 231  Pa (32.5 ka half-life) and 230 Th (75.2 ka half-life) are produced at a constant rate approximately uniformly in the ocean by the α decay of 235 U and 234 U, respectively, with a production activity ratio of 0.093 (Henderson and Anderson, 2003).Water column 231 Pa and 230 Th are subject to particle scavenging and transport to sediments (Bacon and Anderson, 1982;Nozaki et al., 1987).Different scavenging efficiency results in different ocean residence time: 231 Pa has a residence time of approximately 111 years and 230 Th has a residence time of approximately 26 years (Yu et al., 1996).The longer residence time of 231 Pa than 230 Th makes 231 Pa more subject to ocean transport, and therefore in the modern ocean about 45 % of 231 Pa produced in the North Atlantic is transported to the Southern Ocean (Yu et al., 1996), resulting in a sediment 231 Pa / 230 Th activity ratio lower than 0.093 in the North Atlantic and a sediment 231 Pa / 230 Th activity ratio higher than 0.093 in the Southern Ocean.
The application of the above principle to interpret sediment 231 Pa / 230 Th as the strength of Atlantic meridional overturning circulation (AMOC), however, can be complicated by other factors, leading to uncertainties in using 231 Pa / 230 Th as a proxy for past circulation (Keigwin and Boyle, 2008;Lippold et al., 2009;Scholten et al., 2008).In addition to the ocean transport, sediment 231 Pa / 230 Th is also influenced by particle flux and composition (Chase et al., 2002;Geibert and Usbeck, 2004;Scholten et al., 2008; Published by Copernicus Publications on behalf of the European Geosciences Union. S. Gu and Z. Liu: 231 Pa and 230 Th in the ocean model of the Community Earth System Model (CESM1.3)Siddall et al., 2007;Walter et al., 1997).The region of a higher particle flux tends to have a higher 231 Pa / 230 Th (Kumar et al., 1993;Yong Lao et al., 1992), which is referred to as the "particle flux effect" (Siddall et al., 2005).Regional high particle flux in the water column will favor the removal of isotopes into the sediment, which leads to more isotopes transported into this region due to the down-gradient diffusive flux and subsequently more removal of isotopes into the sediment.Since 231 Pa has a longer residence time, this effect is more prominent on 231 Pa than on 230 Th and therefore sediment 231 Pa / 230 Th will be higher in high-productivity regions.Also, opal is able to scavenge 231 Pa much more effectively than 230 Th, leading to higher 231 Pa / 230 Th in high-opal-flux regions such as the Southern Ocean (Chase et al., 2002).Moreover, sediment 231 Pa / 230 Th is suggested to record circulation change only within 1,000 m above the sediment, instead of the whole water column, complicating the interpretation of sediment 231 Pa / 230 Th reconstructions (Thomas et al., 2006).For example, sediment 231 Pa / 230 Th approaching 0.093 during Heinrich Stadial event 1 (HS1) from the subtropical North Atlantic is interpreted as the collapse of AMOC (McManus et al., 2004).If sediment 231 Pa / 230 Th only the records deepest water mass, it is possible that during HS1, AMOC shoals as opposed to fully collapsing, and yet an increase of deep water imported from the Southern Ocean featuring high 231 Pa / 230 Th can increase the sediment 231 Pa / 230 Th approaching the production ratio (0.093) (Thomas et al., 2006).Therefore, it is important to incorporate 231 Pa and 230 Th into climate models for a direct model-data comparison and to promote a thorough understanding of sediment 231 Pa / 230 Th as well as past ocean circulation.
The presence of 231 Pa and 230 Th has been simulated in previous modeling studies (Dutay et al., 2009;Luo et al., 2010;Marchal et al., 2000;Rempfer et al., 2017;Siddall et al., 2005).Marchal et al. (2000) simulates 231 Pa and 230 Th in a zonally averaged circulation model, using the reversible scavenging model of Bacon and Anderson (1982).Going one step further, Siddall et al. (2005) extends Marchal et al. (2000) by including particle dissolution with prescribed particle export production in a 3-D circulation model.Rempfer et al. (2017) further couples 231 Pa and 230 Th with an active biogeochemical model and includes boundary scavenging and sediment resuspensions to improve model performance in simulating water column 231 Pa and 230 Th activity.Here we follow previous studies to implement 231 Pa and 230 Th into the Community Earth System Model (CESM).Our standard 231 Pa and 230 Th are coupled with an active marine ecosystem model ("p-coupled") and therefore is influenced by both ocean circulation change and particle flux change.To aid understanding of the influence of the particle flux, we have also implemented an auxiliary version of 231 Pa and 230 Th ("p-fixed") for which the particle fluxes are fixed at prescribed values.Therefore, p-fixed 231 Pa / 230 Th is only influenced by ocean circulation change.By comparing the p-fixed 231 Pa / 230 Th with the p-coupled 231 Pa / 230 Th, we will be able to separate the effect of circulation change from particle flux change.In addition, the p-fixed 231 Pa and 230 Th can be run without the marine ecosystem module, reducing computational cost by a factor of 3 in the ocean-alone model simulation, making it a computationally efficient tracer for sensitivity studies.
This paper describes the details of 231 Pa and 230 Th in CESM and serves as a reference for future studies using this tracer module.In Sect.2, we describe the model and the implementation of 231 Pa and 230 Th.In Sect.3, we describe the experimental design.We will finally compare simulated 231 Pa and 230 Th fields with observations, show model sensitivities to model parameters and also the sediment 231 Pa / 230 Th ratio response to freshwater forcing in Sect. 4.

Physical ocean model
We implement 231 Pa and 230 Th in the ocean model (Parallel Ocean Program version 2, POP2) (Danabasoglu et al., 2012) of CESM (Hurrell et al., 2013).CESM is a state-ofthe-art coupled climate model and studies describing model components and analyzing results can be found in a special collection in Journal of Climate (http://journals.ametsoc.org/topic/ccsm4-cesm1).We run the ocean-alone model, which is coupled to data atmosphere, land, ice and river runoff under the normal year forcing of CORE-II data (Large and Yeager, 2008), using the low-resolution version of POP2 with a nominal 3 • horizontal resolution and 60 vertical layers.

Biogeochemical component (BGC)
CESM has incorporated a marine ecosystem module that simulates biological variables (Moore et al., 2013).The marine ecosystem module has been validated against presentday observations extensively (e.g., Doney et al., 2009;Long et al., 2013;Moore et al., 2002Moore et al., , 2004;;Moore and Braucher, 2008).The implementation of 231 Pa and 230 Th requires particle fields: CaCO 3 , opal and particulate organic carbon (POC).These particle fields can be obtained through the ecosystem driver from the ecosystem module (Jahn et al., 2015).The ecosystem module simulates the particle fluxes in reasonable agreement with the present-day observations.The pattern and magnitude of the annual mean particle fluxes (CaCO 3 , opal, POC) leaving the euphotic zone at 105 m are similar to those of the satellite observations (Figs. 7.2.5 and 9.2.2 in Sarmiento and Gruber, 2006) (Fig. 1a-c): particle fluxes are higher in the high-productivity regions such as high latitudes and equatorial Pacific; opal flux is high in the Southern Ocean.The remineralization scheme of particles is based on the ballast model of Armstrong et al. (2002).Detailed parameterizations for parti-cle remineralization are documented in Moore et al. (2004) with temperature-dependent remineralization length scales for POC and opal.We do not consider dust because it is suggested to be unimportant for 231 Pa and 230 Th fractionation (Chase et al., 2002;Siddall et al., 2005).
Another important process that contributes to 231 Pa and 230 Th activity is the reversible scavenging by sinking particles (Bacon and Anderson, 1982), which describes the adsorption of isotopes onto sinking particles and desorption after the dissolution of particles.This process transports 231 Pa and 230 Th downward and leads to a general increase of 231 Pa and 230 Th activity with depth.The reversible scavenging considers total isotope activity (A i t ) as two categories (Eq.1): dissolved isotopes (A i d ) and particulate isotopes (A i p ) (superscript i refers to 231 Pa and 230 Th) and A i p is the sum of the isotopes associated with different particle types (A i j,p ) (subscript j refers to different particle types: CaCO 3 , opal and POC): Dissolved and particulate isotopes are assumed to be in equilibrium, which is a reasonable assumption in the open ocean (Bacon and Anderson, 1982;Henderson et al., 1999;Moore and Hunter, 1985).The ratio between the particulate isotope activity and the dissolved isotope activity is set by a partition coefficient, K (Eq.2): where R j is the ratio of particle concentration (C j ) to the density of seawater (1024.5 kg m −3 ).Subscript j refers to different particle types (CaCO 3 , opal and POC).Values of partition coefficient K used in our control simulation follow Chase et al. (2002) and Siddall et al. (2005) (Table 2).Particulate isotopes (A i p ) will be transported by sinking particles, which is described by w s ∂A i p ∂z (Eq.3), where w s is the sinking velocity.We do not differentiate between slowly sinking small particles and rapidly sinking large particles as in Dutay et al. (2009) and consider all particles as slowly sinking small particles with sinking velocity of w s = 1000 m yr −1 (Arsouze et al., 2009;Dutay et al., 2009;Kriest, 2002), which is similar to Rempfer et al. (2017) and Siddall et al. (2005).Any particulate isotopes (A i p ) at the ocean bottom layer are removed from the ocean as sediment, which is the sink for the isotope budget.A detailed vertical differentiation scheme to calculate this term in the model is provided in the supplementary material.The reversible scavenging scheme applied here is the same as the neodymium implementation in POP2 (Gu et al., 2017).Therefore, the conservation equation for 231 Pa and 230 Th activity can be written as www.geosci-model-dev.net/10/4723/2017/Geosci.Model Dev., 10, 4723-4742, 2017 K CaCO 3 2.5 × 10 5 1.0 × 10 7 5 × 10 4 2 × 10 6 1.25 × 10 6 5 × 10 7 K opal 1.67 × 10 6 5 × 10 5 3.33 × 10 5 1 × 10 5 8.33 × 10 6 2.5 × 10 6 K POC 1.0 × 10 7 1.0 × 10 7 2 × 10 6 2 × 10 6 5 × 10 7 5 × 10 7 τ (yr) 118 33 501 143 27 9 Partition coefficients used in CTRL follows (Chase et al., 2002;Siddall et al., 2005).Both p-coupled and p-fixed versions are enabled in CTRL, which yields identical results (discussed in Sect.4.1).Only the p-fixed version is enabled in Exp_1 and Exp_2.The residence time (τ ) is for the p-fixed version in each experiment.
where the total isotope activity is controlled by decay from U (first term), radioactive decay (second term), reversible scavenging (third term) and physical transport by the ocean model (fourth term, including advection, convection and diffusion).A i p can be calculated by combining Eq. ( 1) and Eq. ( 2): which leads to When this is related back to Eq. (1), we get Particle fields used in the reversible scavenging can be either prescribed or simultaneously generated from the marine ecosystem module.Therefore, two forms of 231 Pa and 230 Th are implemented in POP2: "p-fixed" and "p-coupled".The pfixed 231 Pa and 230 Th use particle fluxes prescribed as annual mean particle fluxes generated from the marine ecosystem module under present-day climate forcing (Fig. 1).The pcoupled 231 Pa and 230 Th use particle fluxes computed simultaneously from the marine ecosystem module.The p-fixed and p-coupled 231 Pa and 230 Th can be turned on at the case build time and the p-coupled 231 Pa and 230 Th requires the ecosystem module to be turned on at the same time.
Compared with previous studies of modeling 231 Pa and 230 Th, our p-fixed version is the same as Siddall et al. (2005), except that different prescribed particle fluxes are used.The p-coupled version allows coupling to a biogeochemical module, which is similar to Rempfer et al. (2017), but we do not include boundary scavenging and sediment resuspensions as in Rempfer et al. (2017) because boundary scavenging and sediment resuspensions are suggested to be unimportant in influencing the relationship between 231 Pa p / 230 Th p and AMOC strength (Rempfer et al., 2017).

Experiments
We run a control experiment (CTRL) and two experiments with different partition coefficients to show model sensitivity.We have both p-fixed and p-coupled 231 Pa and 230 Th in CTRL, but only p-fixed 231 Pa and 230 Th in sensitivity experiments.Equilibrium partition coefficients for 231 Pa and 230 Th vary among different particle types and the magnitude of the partition coefficients for different particle types remains uncertain (Chase et al., 2002;Chase and Robert, 2004;Luo and Ku, 1999).Since the control experiment in Siddall et al. (2005) is able to simulate major features of 231 Pa and 230 Th distributions, we use the partition coefficients from the control experiment in Siddall et al. (2005) in our CTRL (Table 2).Two sensitivity experiments are performed with decreased (EXP_1) and increased (EXP_2) partition coefficients by a factor of 5 (Table 2).
All the experiments are ocean-alone experiments with the normal year forcing by CORE-II data (Large and Yeager, 2008).The 231   -4).The p-fixed and p-coupled dissolved and particulate 231 Pa and 230 Th in CTRL are highly correlated with each other, with correlations greater than 0.995, and regression coefficients are all near 1.0 (R 2 > 0.995).The correlation coefficient between p-fixed and pcoupled sediment 231 Pa / 230 Th activity ratios in CTRL is 0.99 and the regression coefficient is 0.9 (R 2 = 0.98).This is expected because the particle fields used in the p-fixed version are prescribed as the climatology of the particle fields used in the p-coupled version.Therefore, under the same climate forcing, p-fixed and p-coupled version of 231 Pa and 230 Th should be very similar.For the discussion of results in CTRL below, we only discuss the p-fixed 231 Pa and 230 Th.
The residence time of both 231 Pa and 230 Th in CTRL are comparable with observations.The residence time is calculated as the ratio of global-average total isotope activity and the radioactive ingrowth of the isotope.Residence time in CTRL is 118 yr for 231 Pa and 33 yr for 230 Th (Table 2), which are of the same magnitude as 111 yr for 231 Pa and 26 yr for 230 Th in observation (Yu et al., 1996).
CTRL can simulate the general features of dissolved water column 231 Pa and 230 Th activities.Dissolved 231 Pa and 230 Th activities increase with depth in CTRL, as shown in two GEOTRACES transects (Deng et al., 2014;Hayes et al., 2015) in the Atlantic (Figs. 2 and 3).The dissolved 231 Pa and 230 Th activities in CTRL are also of the same order of magnitude as in observations in most of the ocean, except that simulated values are larger than observations in the abyssal, which is also the case in Siddall et al. (2005) and Rempfer et al. (2017) (their Figs. 2 and 3, experiment Re3d).Our model is unable to simulate the realistic dissolved 231 Pa and 230 Th activities in the abyssal, probably because boundary scavenging and sediment resuspensions are not included in our model.In Rempfer et al. (2017), without boundary scavenging and sediment resuspension, dissolved 231 Pa and 230 Th activities are quite large in the deep ocean.However, if boundary scavenging and sediment resuspension are included, the water column dissolved 231 Pa and 230 Th activity is in the right magnitude compared with observation.Therefore, we hypothesize that with boundary scavenging and sediment resuspensions added, dissolved 231 Pa and 230 Th activities in the abyssal should be greatly reduced.
A more quantitative model-data comparison is shown in Fig. 5.The linear regression coefficient between model results and observations (references of observations are listed in Table 3), an indication of model ability to simulate 231 Pa and 230 Th activity (Dutay et al., 2009), is near 1.0 for dissolved 231    of 231 Pa from the North Atlantic to the Southern Ocean by AMOC (Gherardi et al., 2009;Lippold et al., 2011Lippold et al., , 2012a;;Luo et al., 2010;Rempfer et al., 2017).The sediment 231 Pa / 230 Th in CTRL is overall consistent with observations (references of observations are listed in Table 3).The North Atlantic shows a low sediment 231 Pa / 230 Th activity ratio as in observations because 231 Pa is more subject to the southward transport by active ocean circulation than 230 Th because of its longer residence time.The Southern Ocean maximum in the sediment 231 Pa / 230 Th activity ratio is also simulated in CTRL.High opal fluxes in the Southern Ocean, which preferentially removes 231 Pa into sediment (K 231 Pa opal > K 230 Th opal ) (Chase et al., 2002), leading to an increased sediment 231 Pa / 230 Th activity ratio.In addition, upwelling in the Southern Ocean brings up deep water enriched with 231 Pa, which is transported from the North Atlantic, to shallower depth and further contributes to the scavenging.CTRL can also produce a higher sediment 231 Pa / 230 Th activity ratio in regions with high particle production (e.g., the eastern equatorial Pacific, the North Pacific and the Indian Ocean) due to the "particle flux effect".Specifically, in the North Atlantic, the distribution of sediment 231 Pa / 230 Th matches the distribution of particle, especially opal, production: sediment 231 Pa / 230 Th is higher where opal production is high, and vice versa (Fig. 4 and Fig. 1c).Quantitatively, the regression coefficient between sediment 231 Pa / 230 Th in CTRL and observation in the Atlantic is 0.86, which is larger than in other basins.This suggests that sediment 231 Pa / 230 Th is better simulated in the Atlantic than in other basins.One possible explanation is that sediment 231 Pa / 230 Th in the Atlantic is controlled by both ocean circulation and particle flux, while in other basins sediment 231 Pa / 230 Th is controlled almost only by particle flux.With active AMOC, the north-south gradient of sediment 231 Pa / 230 Th can be simulated.However, for example, in the Southern Ocean, sediment 231 Pa / 230 Th is dominantly controlled by opal flux, which varies on small scales and is difficult for simulation.Therefore, model performance in simulating sediment 231 Pa / 230 Th in the Southern Ocean is not as good as in the Atlantic.

Sensitivity to partition coefficient K
In this section, we show model sensitivity to partition coefficients by increasing and decreasing the partition coefficient, K, by a factor of 5, but keeping the relative ratio for different particles the same (Table 2).Our model shows similar model sensitivity as in Siddall et al. (2005) as discussed below.
As stated in Siddall et al. (2005), the isotope decay term in Eq. ( 3) is 3 orders of magnitude less than the production term.If we neglect the transport term and the decay term in Eq. ( 3) and assume particulate phase activity at the surface as 0, when equilibrium is reached, the activity of particulate phase will be as in Eq. ( 7).By combining Eq. ( 7) with Eq. ( 2) and R i = F w s •ρ , we can obtain Eq. ( 8).Under the assumption that there is isotope decay and ocean transport, Eq. ( 7) suggests that the particulate isotope activity depends on the production rate and settling velocity and will increase linearly with depth.Equation (8) suggests that the dissolved isotope activity depends on the production rate, partition coefficient K and particle flux and will also increase linearly with depth.Any departure from this linear relationship with depth is due to ocean transport, which is suggested by observations (Bacon and Anderson, 1982;Roy-Barman et al., 1996).Results of Eqs. ( 7) and ( 8) can aid understanding of the differences in Exp_1 and Exp_2.
Increasing K will decrease water column dissolved 231 Pa and 230 Th activities but will not change particulate 231 Pa and 230 Th too much (Fig. 6).Magnitude of dissolved 231 Pa and 230 Th in Exp_1 (smaller K) is at least 1 order larger than that in Exp_2 (larger K), while magnitude of particulate 231 Pa and 230 Th in Exp_1 and Exp_2 is of the same order.As suggested by Eq. ( 8), if there is no isotope decay and no ocean transport, larger K will lead to smaller dissolved isotope activity but unchanged particulate activity.Intuitively, larger K will lead to more 231 Pa and 230 Th attached to particles and further buried into sediment, which increases the sink for the 231 Pa and 230 Th budget.With the sources for 231 Pa and 230 Th staying the same, dissolved 231 Pa and 230 Th will be reduced.Increasing K will also reduce the vertical gradient of dissolved 231 Pa and 230 Th, because reversible scavenging acts as the vertical transport and increases this vertical transport and can also decrease the vertical gradient.However, changes in the particulate 231 Pa and 230 Th are relatively small (Fig. 6).Equation ( 7) suggests that particulate phase activity is independent of K. Therefore, changing K will have limited influence on particulate phase activity.A i p (z) = Increasing K will also reduce the spatial gradient in the sediment 231 Pa / 230 Th activity ratio and vice versa (Fig. 7).Larger K will decrease the 231 Pa and 230 Th residence time and most isotopes produced in the water column are removed into sediment locally (Table 2).Therefore, the sediment 231 Pa / 230 Th ratio becomes more homogeneous and approaching the production ration of 0.093 (Fig. 7b).The deviation (the root mean squared error) of sediment 231 Pa / 230 Th is 0.0726 in CTRL, 0.0770 in Exp_1 and 0.0739 in Exp_2.The linear regression coefficients between sediment 231 Pa / 230 Th in the model and the observations are listed in Table S1 in the Supplement.Although the performance of global sediment 231 Pa / 230 Th in Exp_1 is better than CTRL, the performance of Atlantic 231 Pa / 230 Th in Exp_1 is worse.We consider better simulation of sediment 231 Pa / 230 Th in the Atlantic to be more important since the most important application of sediment 231 Pa / 230 Th is using sediment 231 Pa / 230 Th in the North Atlantic to reconstruct past AMOC.In addition, water column isotope activity is too large in Exp_1 compared with observations.Therefore, the partition coefficient in CTRL is of the right order of magnitude.AMOC strength will change and will influence sediment 231 Pa / 230 Th in different ways.Our model with p-fixed and p-coupled 231 Pa and 230 Th can help to detangle these two effects.In this section, we examine the sediment 231 Pa / 230 Th (p-fixed and p-coupled) response in the North Atlantic to idealized freshwater perturbation.

4.
In HOSING, after applying freshwater forcing to the North Atlantic, AMOC strength quickly decreases to a minimum of 2 Sv (AMOC_off) (Fig. 9a).During the AMOC_off state, compared with CTRL with active AMOC (AMOC_on), pfixed sediment 231 Pa / 230 Th shows an overall increase in the North Atlantic and a decrease in the South Atlantic (Fig. 10b) because of the reduced southward transport of 231 Pa from the North Atlantic by AMOC, consistent with paleo-proxy evidence there (e.g., Gherardi et al., 2005Gherardi et al., , 2009;;McManus et al., 2004).The overall increase of the sediment 231 Pa / 230 Th ratio in the North Atlantic in response to the AMOC collapse can be seen more clearly in the time evolution of the sediment 231 Pa / 230 Th ratio averaged from 20 to 60 • N in the North Atlantic (Fig. 9b, green).Quantitatively, the 231 Pa / 230 Th increases from 0.074 in AMOC_on to 0.098 in AMOC_off in the p-fixed version, approaching the production ration of 0.093.This increase of 231 Pa / 230 Th is also observed in the subtropical North Atlantic from the two sites near the Bermuda Rise (Fig. 9e and f), which is of comparable magnitude with the change from LGM to HS1 in reconstructions there (McManus et al., 2004).In addition, the pattern of the p-fixed (Fig. 10a) sediment 231 Pa / 230 Th ratio in the Atlantic in AMOC_off state is similar to the opal distribution (Fig. 1b) because, without active circulation, the sediment 231 Pa / 230 Th ratio is more controlled by particle flux effect, which is similar to the Pacific in CTRL.It is further noted that our p-fixed sediment 231 Pa / 230 Th ratio in HOS-ING behaves similarly to that in Siddall et al. (2007).
The overall increase in the p-fixed sediment 231 Pa / 230 Th ratio in the North Atlantic is not homogenous and the magnitude of the change between AMOC_on and AMOC_off varies with location depending on the distribution of particle flux, especially the opal flux (Figs. 9 and 10).The maximum increase in the p-fixed sediment 231 Pa / 230 Th ratio occurs near 40 • N in the western Atlantic (Fig. 10a), where the opal production in our model is maximum in the North Atlantic (Fig. 1b).The sediment 231 Pa / 230 Th ratio in this region during AMOC_on is larger than the production ratio of 0.093 because opal maximum provides extra 231 Pa to this region ("particle flux effect"), which overwhelms the active ocean circulation transporting 231 Pa southward outside this region (Fig. 9d, green).During AMOC_off, without active ocean circulation, the particle flux effect becomes even stronger because less 231 Pa is transported out of the North Atlantic and the p-fixed sediment 231 Pa / 230 Th ratio becomes even larger.It should be noted that the opal maximum in this region is not in the observation (Fig. 7.2.5 in Sarmiento and Gruber, 2006).However, our sediment 231 Pa / 230 Th response in HOSING is self-consistent with the particle flux in our model since the location of the maximum 231 Pa / 230 Th increase matches the location of opal flux in our model.
In most regions of the Atlantic, p-coupled sediment 231 Pa / 230 Th shows a similar response to p-fixed 231 Pa / 230 Th in HOSING.The evolutions of p-fixed and pcoupled sediment 231 Pa / 230 Th activity ratios in HOSING are highly correlated (Fig. 11a).The change in the sediment 231 Pa / 230 Th ratio from AMOC_on to AMOC_off are similar in both the p-fixed and p-coupled version (Fig. 11b).The correlation between p-fixed and p-coupled sediment 231 Pa / 230 Th ratio change from AMOC_on to AMOC_off is 0.72 (1455 points) and the linear regression coefficient is 0.71 (R 2 = 0.52).A high correlation between p-fixed and pcoupled response mainly happens over low-productivity regions (Figs. 1,10,and 11), where the circulation effect on   sediment 231 Pa / 230 Th is more important than the particle flux change in HOSING.
In spite of these similarities discussed above, the responses of p-fixed and p-coupled sediment 231 Pa / 230 Th to the freshwater forcing can differ significantly in high-productivity regions because of the productivity change.With persistent freshwater forcing over the North Atlantic, most regions in the North Atlantic show reduced production of CaCO 3 , opal and POC (Fig. 8).Productivity in the North Atlantic is suggested to be halved during AMOC collapse because of increased stratification, which reduces nutrient supply from the deep ocean (Schmittner, 2005).In our model, the productivity in the midlatitude North Atlantic is indeed greatly reduced after the freshwater forcing is applied.For example, opal production from 30 to 50 • N in the Atlantic at the end of HOSING is reduced by 50-90 % of its original value in CTRL.However, opal production increases in the highlatitude North Atlantic (north of 50 • N).The pattern of opal production changes with high-opal-production regions shifting northward in HOSING (Fig. 8d, e and f).These particle flux changes will influence sediment 231 Pa / 230 Th as discussed below.
North of 50 • N in the Atlantic, opal productivity increases during AMOC_off (Fig. 8f) and will result an increase in sediment 231 Pa / 230 Th.The increase caused by greater opal productivity enhances the sediment 231 Pa / 230 Th increase caused by reduced AMOC.Therefore, the increase in p-coupled sediment 231 Pa / 230 Th from AMOC_on to AMOC_off is larger than p-fixed sediment 231 Pa / 230 Th change (Fig. 9c).
In the midlatitude North Atlantic, opal productivity decreases during AMOC_off (Fig. 8f) and will lead to a decrease in sediment 231 Pa / 230 Th, which is opposite to the effect of reduced AMOC.The p-coupled sediment 231 Pa / 230 Th shows an initial decrease in first 200 years (Fig. 9d, e and f, red dash lines) caused by the reduced opal productivity.But this decreasing trend is reversed eventually, suggesting that the influence of particle flux change is overwhelmed by the effect of reduced AMOC.It the long run, most regions in the subtropical and midlatitude Atlantic show increased sediment 231 Pa / 230 Th in HOS-ING (Fig. 10d), indicating the dominant effect of reduced AMOC.However, sediment 231 Pa / 230 Th at 40 • N west Atlantic, where opal productivity is at its maximum during AMOC_on, show a decrease from AMOC_on to AMOC_off (Figs. 9d and 10d).During AMOC_on, the opal productivity maximum at 40 • N in the west Atlantic lead to regional maximum sediment 231 Pa / 230 Th because of the particle flux effect (Fig. 4).During AMOC_off, this opal productivity maximum is eliminated (Fig. 8e) and there is no more extra 231 Pa supplied by surroundings to this region, which leads to a decrease in sediment 231 Pa / 230 Th.This decrease in sediment 231 Pa / 230 Th caused by productivity change is greater than the increase caused by the reduced AMOC.Therefore, sediment 231 Pa / 230 Th experiences a decrease from AMOC_on to AMOC_off at this location (Fig. 9d and Fig. 10d).Our results suggest that although the circulation effect is more dominant than the particle flux change in controlling sediment 231 Pa / 230 Th on long timescales over most of North Atlantic (Fig. 11), particle flux change can be important on short timescales and in high-productivity regions.With pfixed and p-coupled 231 Pa and 230 Th, our model can help to detangle the circulation effect and particle flux effect.
It has been suggested that the particulate 231 Pa / 230 Th response to the change in AMOC depends on the location and depth.Above 2 km and high-latitude North Atlantic, particulate 231 Pa / 230 Th decreases with the increased AMOC (Rempfer et al., 2017).Our results are consistent with this finding (Fig. 12a and b).Both p-fixed and p-coupled particulate 231 Pa / 230 Th show similar patterns of change from AMOC_on to AMOC_off: a decrease in particulate 231 Pa / 230 Th at shallow depth and north of 60 • N and an increase in particulate 231 Pa / 230 Th below 2 km and south of 60 • N during AMOC_off.Therefore, sediment depth should also be taken into consideration when interpreting sediment 231 Pa / 230 Th.Since the pattern in the p-coupled ratio is similar to the pattern in the p-fixed ratio, the opposite particulate 231 Pa / 230 Th changes in the shallow and deep North Atlantic is associated with AMOC change.During AMOC_on, upper limb of AMOC (about upper 1 km) transport water northward, which provides extra 231 Pa to North Atlantic, and particulate 231 Pa / 230 Th is larger than the production ratio of 0.093.In contrast, the lower limb of AMOC (2-3 km) features southward transport, which transports 231 Pa to the Southern Ocean, and particulate 231 Pa / 230 Th is smaller than the production ratio of 0.093 (Fig. 12 solid).Particulate 231 Pa / 230 Th decreases with depth (Fig. 12c solid).During AMOC_off, ocean transport of 231 Pa is greatly reduced.Therefore, shallow (deep) depth experiences a decrease (increase) in particulate 231 Pa / 230 Th and the vertical gradient in the particulate 231 Pa / 230 Th is also greatly reduced (Fig. 12c dash).Our results support that the depth dependence of particulate 231 Pa / 230 Th is mainly caused by lateral transport of 231 Pa by circulation (Gherardi et al., 2009;Lippold et al., 2011Lippold et al., , 2012a;;Luo et al., 2010;Rempfer et al., 2017).
Overall, our model is able to simulate the correct magnitude of the sediment 231 Pa / 230 Th ratio response to the freshwater forcing.Our experiments suggest that the change in circulation is the dominant factor that influences sediment 231 Pa / 230 Th on long timescales over most of the globe in the idealized hosing experiment, although the detailed difference between p-fixed and p-coupled sediment 231 Pa / 230 Th ratio response to freshwater forcing in different locations can be complicated.

Summary
231 Pa and 230 Th have been implemented in the ocean model of the CESM in both the p-coupled and p-fixed forms.Our control experiment under present-day climate forcing is able to simulate most 231 Pa and 230 Th water column activity and the sediment 231 Pa / 230 Th activity ratio consistently with observations by using the parameters that are suggested by Chase et al. (2002) and used in Siddall et al. (2005).Our sensitivity experiments with varying parameters suggest that these parameters are of the right order of magnitude.
Furthermore, our model is able to simulate the overall sediment 231 Pa / 230 Th ratio change in the North Atlantic with a magnitude comparable to the reconstruction in response to the collapse of AMOC, although the detailed response can be complicated in different regions.Finally, the p-fixed form is able to capture many major features of that of the p-coupled form over large ocean areas on long timescales, although the two forms can also differ significantly in some regions, especially the region with high opal productivity.
Much remains to be improved in our 231 Pa and 230 Th module in the future.For example, the model can be further improved by including nepheloid layers to better simulate water column 231 Pa and 230 Th activity as in Rempfer et al. (2017).In addition, a partition coefficient for different particles can be further tuned, which can improve our understanding of the affinity of 231 Pa and 230 Th to different particles, complementing the limited observational studies available (e.g., Chase et al., 2002;Scholten et al., 2005;Walter et al., 1997).At present, as the first attempt to implement 231 Pa and 230 Th in the CESM with both p-fixed and p-coupled versions, our model can serve as a useful tool to improve our understanding of the processes of 231 Pa and 230 Th as well as interpretations of sediment 231 Pa / 230 Th reconstructions for past ocean circulation and climate changes.

Figure 2 .
Figure 2. Dissolved 231 Pa, dissolved230 Th and particulate231 Pa / 230 Th in CTRL along GEOTRACES transect GA02S(Deng et al., 2014) (the track is indicated in Fig.S4in the Supplement) for both p-fixed (top row) and p-coupled (bottom row)231 Pa and 230 Th (colored contour).Observations of dissolved231 Pa and230 Th activity are superimposed as colored circles using the same color scale.

Figure 3 .
Figure3.Dissolved231 Pa, dissolved230 Th and particulate231 Pa / 230 Th in CTRL along GEOTRACES transect GA03(Hayes et al., 2015) (the track is indicated in Fig.S4) for both p-fixed (top row) and p-coupled (bottom row)231 Pa and 230 Th (colored contour).Observations of dissolved231 Pa and230 Th activity are superimposed as colored circles using the same color scale.

Figure 4 .
Figure 4.The sediment 231 Pa / 230 Th activity ratio in CTRL for both p-fixed (a) and p-coupled version (b).Observations are attached as filled circles using the same color map.The 231 Pa / 230 Th activity ratio is plotted relative to the production ratio of 0.093 on a log 10 scale.
Pa and 230 Th activities are initiated from 0 in CTRL and are integrated for 2000 model years until equilibrium is reached.EXP_1 and EXP_2 are initiated from model year 1400 in CTRL and are integrated for another 800 model years to reach equilibrium.Since sediment 231 Pa / 230 Th in North Atlantic has been used to reflect the strength of AMOC, to test how sediment 231 Pa / 230 Th in our model responds to the change in AMOC and the change in particle fluxes, we carried out a freshwater perturbation experiment (HOSING) with both p-fixed and p-coupled 231 Pa and 230 Th.Starting from model year 2000 of CTRL, a freshwater flux of 1 Sv is imposed over the North Atlantic region of 50 to 70 • N and the experiment is integrated for 1400 model years until both p-fixed and pcoupled sediment 231 Pa / 230 Th ratios have reached quasiequilibrium.The partition coefficients used in HOSING are the same as in CTRL. 4 Results 4.1 Control experiment P-fixed and p-coupled version of 231 Pa and 230 Th in CTRL show identical results (Figs. 2 Pa and 230 Th (1.02 for [ 231 Pa] d and 1.14 for [ 230 Th] d ), suggesting that CTRL can simulate the dissolved 231 Pa and 230 Th in good agreement with observa-

Figure 5 .
Figure5.Scatter plot of global dissolved and particulate231 Pa and230 Th between observation and CTRL (p-fixed) (unit: dpm m −3 ): (a) dissolved231 Pa, (b) particulate231 Pa, (c) dissolved230 Th, (d) particulate230 Th.Observations in different depth ranges are indicated by different colors: green for 0-100 m, red for 100-1000 m, blue for 1000-3000 m and yellow for deeper than 3000 m.The purple line is the least-squares linear regression line for all depth ranges, the slope of which is indicated at the top right of each plot.The green line is the least-squares linear regression line for depth from 0 to 100 m.The red line is the least-squares linear regression line for depth from 100 to 1000 m.The blue line is the least-squares linear regression line for depth from 1000 to 3000 m.The yellow line is the least-squares linear regression line for depth deeper than 3000 m.

Figure 7 .
Figure 7.The sediment 231 Pa / 230 Th activity ratio in Exp_1 (a) and Exp_2 (b).Observations are attached as filled circles using the same color map.The 231 Pa / 230 Th activity ratio is plotted relative to the production ratio of 0.093 on a log 10 scale.

Figure 9 .
Figure 9.Time evolutions in HOSING.(a) Freshwater forcing (black) and AMOC strength (navy), which is defined as the maximum of the overturning stream function below 500m in the North Atlantic.(b) The North Atlantic average sediment 231 Pa / 230 Th activity ratio from 20 • N to 60 • N: p-fixed (green) and p-coupled (red).The production ratio of 0.093 is indicated by a solid black line (similar to in c, d, e and f).(c) The sediment 231 Pa / 230 Th activity ratio at (55 • N, 30 • W).(d) The sediment 231 Pa / 230 Th activity ratio at (40 • N, 40 • W).(e) The sediment 231 Pa / 230 Th activity ratio at (35 • N, 58 • W).(f) The sediment 231 Pa / 230 Th activity ratio at (34 • N, 60 • W).Panels (e) and (f) show locations near the Bermuda Rise.Locations of each site are shown as dots in Fig. 8b.

Figure 10 .Figure 11 Figure 12 .
Figure 10.The sediment 231 Pa / 230 Th activity ratio during AMOC off state and the difference between AMOC off and CTRL.(a) P-fixed log 10 ([ 231 Pa / 230 Th]/0.093) in AMOC_off.(b) Difference in the p-fixed sediment 231 Pa / 230 Th activity ratio between AMOC_off and AMOC_on.Panels (c) and (d) are similar to (a) and (b) for the p-coupled sediment 231 Pa / 230 Th activity ratio.Black dots in (b) shows the locations of sites in Fig. 9 from north to south.

Table 2 .
Partition coefficients for different particle types and residence time for 231 Pa and 230 Th in different experiments.