Significance
Understanding of local weather “tipping factors” is extraordinarily restricted. Their illustration inside Earth System fashions (ESMs) is totally unconstrained due to an absence of any precise occurrences of those occasions in latest instances. For example, most ESMs fail to simulate the fast desertification of the Sahara in the course of the Holocene. Right here, we overcome this by tuning unsure points of 1 ESM utilizing the local weather of 6,000 y in the past. This method reveals a hitherto-absent tipping level inside this ESM that produces glorious settlement between the modeled and noticed timing of abrupt change in North Africa. This demonstrates how paleoclimate data can be utilized to radically enhance the illustration of abrupt modifications within the ESMs which can be employed for future projections.
Summary
Whereas paleoclimate data present that the Earth System is characterised by a number of totally different tipping factors, their illustration inside Earth System fashions (ESMs) stays poorly constrained. It’s because historic observations don’t embody variations massive sufficient to impress such regime modifications, and paleoclimate circumstances are hardly ever used to assist develop and tune ESMs, which doubtlessly ignores a wealthy supply of data on abrupt local weather change. A essential instance is the early to mid-Holocene “greening” and subsequent fast desertification of the Sahara, which most ESMs fail to breed, casting doubt on the illustration of land–environment coupling and monsoon dynamics. Right here, we present that this greening and abrupt termination will be efficiently simulated with one ESM after optimizing unsure mannequin elements utilizing each present-day observations and crucially mid-Holocene (6,000 y earlier than current) reconstructions. The optimized mannequin shows abrupt threshold habits, which exhibits glorious settlement with lengthy paleoclimate data that weren’t used within the authentic optimization. These outcomes recommend that with a purpose to realistically seize climate-system thresholds, ESMs first should be conditioned with applicable paleoclimate data.
The steadiness of Earth’s local weather is essential for understanding previous and future environments (1
⇓
⇓
⇓–5). The later a part of the Holocene epoch was steady, maybe offering the distinctive environmental background for agriculture and civilization. In distinction, paleoclimate data reveal how the soundness of the Holocene was uncommon. The local weather system is able to fast transitions that tremendously outpace the speed of change of driving components. Below the affect of accelerating greenhouse-gas radiative forcing, the potential for abrupt local weather change is more likely to improve sooner or later (4
⇓–6). Nevertheless, understanding of the likelihood and potential prevalence of tipping components is extraordinarily restricted (4, 6). A significant motive for that is that Earth System fashions (ESMs) utilized in future projections usually are not evaluated towards any precise abrupt climate-change occasions (7). As a substitute, fashions are tuned towards the spatial and seasonal distribution of key variables, corresponding to radiation stability, temperature, and precipitation (8). Which means poorly constrained ESMs have for use to quantify the chance of abrupt thresholds (5).
A distinguished instance of an abrupt local weather change occurred in North Africa in the course of the mid-Holocene. A number of impartial traces of proof help a major greening of the Sahara in the course of the early to mid-Holocene. Fossil pollen data present a widespread northward growth of shrub and savannah biomes (9
⇓
⇓–12). Sedimentary data point out the reactivation of aquifers and deepening and growth of lakes (13). Ocean cores present a 60 to 80% discount in mineral mud flux reaching websites so far as the Caribbean (14, 15), and leaf waxes from sediment cores close to West Africa have a destructive hydrogen isotope tour according to enhanced rainfall (16).
The Holocene African humid interval (AHP) got here to an finish round 4,000 y (4 ka) to six ka earlier than current (BP) (17, 18). Early proof recommended that it terminated comparatively quickly in contrast with the gradual millennial-scale decline in insolation (19), however a marine sediment mud file from the area (14) supplied the primary compelling proof for centennial and even decadal scale transition out of the AHP. Reconstructions from a wider vary of proxies and over a bigger space (16, 17) show an abrupt transition, however in areas additional to the south or east, the hydrological response is rather more gradual (20, 21).
Charney et al. (22) first recommended a constructive suggestions mechanism between vegetation and the monsoon in North Africa. A fast termination of the AHP is according to this course of as a result of small modifications in exterior forcings will be amplified, accelerating the tempo of the system response. Theoretically, this may very well be linked to bistability within the system and the opportunity of abrupt transitions between states (23, 24). A significant barrier to testing this principle in additional element is the underestimation of mid-Holocene rainfall in North Africa by practically all normal circulation fashions (GCMs) (25, 26) and ESMs (27). In consequence, the dynamics of this technique can’t be evaluated with the detailed three-dimensional fashions used to challenge future change (7).
A comparatively abrupt discount in vegetation cowl and precipitation at round 5,500 y BP was simulated with the intermediate-complexity mannequin CLIMBER-2 (2). Subsequent research with different ESMs of intermediate complexity or low-resolution GCMs have proven both no abrupt change (28) or solely regional impacts within the Jap Sahara (29). The latter examine seems to be in contradiction with subsequent reconstructions, which as a substitute present gradual hydrological change within the Jap area (20, 21). One higher-resolution ESM exhibits a sensible greening and desertification (30), however the transition is just not as abrupt as some paleoclimate data recommend (14).
The explanations for the failure of ESMs to simulate the AHP stay unclear, and it’s attainable that new processes should be thought-about (31
⇓–33). Some research have argued that it’s due to lacking processes corresponding to mud, however subsequent work has proven that this was as a consequence of overestimated shortwave absorption by mud particles in older observational datasets which can be nonetheless employed in some ESMs (34, 35). Alternatively, current fashions might already embody sufficient representations of the important thing processes, however these usually are not presently optimized or tuned appropriately, which means that feedbacks are too robust or too weak, and emergent properties are biased. The latter would indicate that paleoclimate is uniquely helpful for narrowing appropriate parameter ranges in ESMs/GCMs (36, 37).
Right here, we analyze a collection of 4 transient simulations of the Holocene from 10 ka BP to current day utilizing variations of HadCM3-M2.1 (the Hadley Centre Coupled Mannequin Model 3 coupled to Model 2.1 of the Met Workplace Floor Alternate Scheme [MOSES]) (38
⇓–40). These configurations are primarily based on the prevailing launch mannequin model (STD), however embody paleoclimate-informed modifications to atmospheric convection (+CONV) and dynamic vegetation (+VMS). +CONV improve the sensitivity of convection to summer time insolation (37), and +VMS improves the simulation of vegetation cowl in semiarid areas, which is required as a result of many vegetation fashions have incorrect sensitivity (32). A fourth setup combines these (+CONV+VMS). Conceptually, this method is exclusive as a result of the parameters have been optimized towards each present-day and mid-Holocene local weather reconstructions (37). The +CONV configuration of HadCM3 produces practically twice the precipitation improve over North Africa relative to STD. Thus, we’re doubtlessly capable of examine the dynamics of the AHP termination intimately and to look at whether or not totally different parameterizations alter the basic properties of the mannequin on this area.
Outcomes
“Greening” and Abrupt Change
The 4 simulations all present a worldwide imply warming of ∼1.2 Okay over the course of the Holocene (
SI Appendix, Figs. S1 and S2). That is pushed by reductions in ice quantity and rising greenhouse-gas forcing, principally as a consequence of rising CO2, as proven in
SI Appendix, Fig. S2. Whole radiative forcing will increase by round 1.2 Wm– 2 over the course of the simulation. The gradual discount in Northern Hemisphere summer time insolation forces a discount within the energy of the Northern Hemisphere monsoons (Fig. 1 and
SI Appendix, Fig. S3).
Simulated and reconstructed Holocene precipitation (precip.) in North Africa. (A) Annual imply simulated precipitation with 4 variations of HadCM3-M2.1. (B) The 100-y working imply as compared with the four-core common reconstruction (16) proven in thick grey. (C–F) Comparability with particular person cores. The uncertainty vary for the reconstructed precipitation is proven by skinny grey traces.
Over North Africa, the precipitation response could be very totally different within the 4 fashions, as proven in Fig. 1 for the Western area from 20 to 30∘ N by 20∘ W to five∘ E. In STD and +CONV, there’s a very small enhancement of the West African monsoon within the early Holocene in comparison with the late Holocene. This solely penetrates to round 15∘ N. STD is according to HadCM3-M2.1 simulations of 6 ka BP submitted to the Paleoclimate Modelling Intercomparison Mission (PMIP) Part II and is much like different PMIP simulations (41). It exhibits a northward extension of precipitation over North Africa by only some levels, equal to roughly one mannequin grid cell. The response is far stronger in +VMS, however +CONV+VMS exhibits the best precipitation improve and vegetation growth (mentioned beneath). A singular characteristic in +CONV+VMS are the abrupt reductions in precipitation at round 7.7, 6.6, and 6.1 ka BP. +VMS is considerably drier than +CONV+VMS till round 6 ka BP, and there’s no proof of this abrupt habits within the VMS configuration. The VMS mannequin is dryer within the early Holocene and wetter within the later Holocene, demonstrating that the convection modifications in +CONV+VMS improve the mannequin’s sensitivity to the utilized forcings, significantly the summer time insolation decline, as proven additionally in
SI Appendix, Fig. S3, and by the identified land–environment coupling coefficients proven in
SI Appendix, Figs. S4 and S5.
The mannequin outcomes are in contrast with the person and 4 core-mean rainfall reconstructions by Tierney et al. (16) in Fig. 1. The reconstruction depicts annual imply rainfall, however is in contrast right here with the simulated June–July–August–September (JJAS) imply, since many of the precipitation within the mannequin, round 70%, happens on this season. Nevertheless, the simulated annual imply charges are decrease than reconstructed. Latest isotope modeling has recommended that the reconstructions could also be too excessive (42). We judged that the temporal traits are sturdy, and these are helpful for evaluating the timing and rapidity of the rainfall modifications.
The common precipitation reconstruction from the 4 cores agrees remarkably nicely with the +CONV+VMS simulation, each when it comes to the amplitude and the timing of discount round 6.1 ka BP. Three of the person data exhibits glorious settlement with the +CONV+VMS mannequin within the timing of the abrupt collapse at round 6 ka BP. The remaining core (GC37 at 27 ∘ N) exhibits an out-of-phase response. It’s unclear whether or not this may very well be as a consequence of chronological uncertainty, which is round 75 y for this a part of file, with 14 C dates each 1 to five ka, or due to potential site-specific components.
The reconstructions additionally present an abrupt precipitation discount at or simply after 8 ka BP, which roughly coincides with later phases of the 8.2 ka BP North Atlantic melt-water occasion, although the drying is considerably longer in length (16). This centennial-scale drying occasion is supported by data from throughout the continent (12, 43
⇓
⇓–46). A really comparable occasion happens round 7.7 ka BP within the +CONV+VMS simulation independently of any melt-water forcing within the mannequin. This seems to be a simulated “flickering” of the land–environment system, probably a precursor to the ultimate collapse at 6.5 ka BP. The bistability diagnostic (mentioned beneath) has already elevated considerably by this level within the +CONV+VMS simulation, supporting this interpretation. The similarities between the reconstructed and simulated rainfall discount at or after 8 ka BP due to this fact current a hitherto-unexplored rationalization for the noticed occasion—as an indicator or symptom of system instability. Tighter chronological controls within the paleoclimate data and/or additional ensembles of simulations are wanted for a definitive analysis of this speculation.
We additionally evaluate the simulated vegetation protection with the mid-Holocene biome reconstruction (47) in Fig. 2. The simulated fractional protection of plant useful varieties (PFTs) has been translated to mega-biomes following ref. 48. The comparability is carried out for six.5 ka BP as a result of the vegetation cowl and precipitation collapse simply earlier than 6.0 ka BP. The +CONV+VMS simulated biomes evaluate most favorably with the mid-Holocene pollen-based reconstruction, though there’s nonetheless too little grass or temperate forest within the Central and Jap Sahara round 20 to 30∘ E, as much as round 24∘ N. Different reconstructions (49) additionally help this as the utmost northward extension of the monsoon precipitation in the course of the Holocene. The opposite mannequin configurations have an excessive amount of naked soil throughout the entire continent between 20 and 35∘ N. Each the fractional protection and leaf space index are correlated with the precipitation anomalies, suggesting a decent coupling, which we focus on additional beneath. The dearth of wetter circumstances within the jap area of the Sahara is according to earlier mannequin simulations (50, 51) and is because of descending air lots on this area in Northern Hemisphere summer time, according to a distant forcing by the Asian monsoon (52). The descent is just not strengthened within the early and mid-Holocene, regardless of the stronger monsoon to the east, which means that there are competing influences working on this area, together with, for instance, tropical plumes (30, 53).
Mid-Holocene simulated and reconstructed (47) vegetation biomes (A–D), simulated vegetation protection (E–H), and leaf space index (shading) and precipitation anomalies (contours; I–L). Temp., temperate; trop., tropical.
The timing of the top of the humid interval is estimated from the modeled naked soil fraction following ref. 30 and is proven in Fig. 3. The +CONV+VMS mannequin exhibits an earlier hydrological cycle decline within the north and a later transition additional south, particularly within the western equatorial area. This sample is broadly supported by the reconstructed timing, which has been compiled from a collection of paleo-hydrological reconstructions (17, 30). Further reconstructions in West Africa at round 10 to twenty∘ N are required to resolve the sharp gradient simulated within the +CONV+VMS mannequin model.
Simulated and noticed time of termination of the Holocene AHP. The mannequin outcomes are primarily based on the simulated naked soil fraction following ref. 30.
Diagnostics of Bistability
Techniques that include intrinsic abrupt thresholds ought to, in principle, present proof of essential slowing as they method a selected tipping level (e.g., ref. 3). One signature of this is a rise in variability because the system reaches a threshold (23). The variance of the vegetation fraction (utilizing the ±500-y filtered sign) averaged over the north Western Sahara (20 to 30∘ N, 20∘ W to five∘ E) will increase between 9.5 and 6.5 ka BP within the +CONV+VMS simulation. The variance improve in +CONV+VMS is according to the abrupt habits at round 7.5 and 6.5 ka BP and seems to suggest bistability within the system. Conversely, the remaining three mannequin configurations present fixed and even reducing variance (Fig. 4), suggesting a monostable system with no tipping factors. This demonstrates that refined parameter modifications in ESMs/GCMs can strongly affect the emergent properties related to abrupt local weather phenomena.
Simulated north African vegetation protection. Imply (A) and working ±500-y variance (B) for northwest Africa (20 to 30∘ N, 20∘ W to five∘ E and the spatial change within the variance throughout the early Holocene in CONV+VMS (C).
The spatial change in variance throughout the early Holocene is proven in Fig. 4C. The rise is strongest within the Western Sahara, the place the land–environment feedbacks are robust as a consequence of pronounced albedo contrasts between vegetation and the high-mineral-content soils and the place precipitation is weakest immediately. The abrupt modifications listed below are according to different paleoclimate proof that means that additional eastward and to the south, the reconstructed hydrological modifications are usually rather more gradual (20, 21).
We carried out extra sensitivity simulations with out interactive vegetation to guage the affect of vegetation cowl on the precipitation response over North Africa (Supplies and Strategies). These diagnostic simulations present a robust dependence of the precipitation over North Africa (20 to 30∘ N, 20∘ W to fifteen∘ E) on the fraction of vegetation (see additionally
SI Appendix, Fig. S6). The speed of precipitation improve per fractional vegetation cowl is 276 mm/y for the 8 ka simulation and 217 mm/y at 6 ka after the vegetation has contracted. These values are according to a robust land–environment coupling on this area (24). Along with the obvious threshold mentioned above, this factors to a dominant management by vegetation on the abrupt dynamics within the system. The cheap settlement with the hydrogen-isotope precipitation reconstruction proven in Fig. 1 means that this will even be the case for the actual system.
Dialogue
Valdes (2011) (7) argued that local weather fashions are overly steady and, therefore, most likely unable to realistically seize abrupt occasions, corresponding to these seen within the paleoclimate file. This can be as a result of GCMs are implicitly biased towards stability by their analysis solely towards modern observational targets (8). Whereas there was some progress in simulating abrupt habits of the Atlantic ocean circulation (54, 55), it stays unclear how mannequin boundary circumstances, parameter settings, or the span of resolved (bio-)bodily processes will mix to offer a sensible illustration of abrupt local weather change extra extensively. That is due to this fact a essential analysis query, given the funding in GCM/ESM improvement and the potential for high-impact or catastrophic system modifications sooner or later (4, 6).
Right here, we present {that a} real looking and spontaneous collapse of the greening of the Sahara happens solely within the mannequin model, which has been optimized with each present-day observations and mid-Holocene reconstructions. The profitable replication of this occasion by the mannequin demonstrates that coupled GCMs, just like the one used right here, most likely resolve the required spectrum of biophysical processes. These embody initiation and environmental mixing of convection, biophysical coupling to land floor, and coupling with the large-scale circulation. Nevertheless, in the usual mannequin model, these processes have doubtlessly been misspecified, most likely as a result of present-day observations present restricted constraints for a way these coupled processes function throughout abrupt occasions (37).
This instance ought to inspire additional work with totally different paleoclimate occasions masking a wider vary of methods and feedbacks. It might be helpful to use these developments in different ESMs to check the universality of parameterizations and to see whether or not this might scale back intermodel uncertainty in future projections. Totally different statistical approaches to mannequin tuning can be helpful. Future work might think about an specific concentrate on transient local weather modifications and on model-emulation strategies. A key improvement can be to robustly combine throughout a hierarchy of fashions, from idealized atmospheric dynamical schemes (56) to lower-resolution GCMs (40) and extra computationally expensive ESMs, with, for instance, up to date cloud microphysics (57).
The absence of considerable abrupt local weather occasions in the course of the instrumental period and doubtless a number of previous centuries may very well be taken as proof that such occasions have been unlikely within the late Holocene preindustrial local weather state. Future local weather change will nearly actually improve the likelihood of encountering threshold and tipping factors. Our outcomes present the worth of paleoclimate data and that present-day local weather is a needed, however inadequate, constraint on the habits of ESMs, significantly for tipping occasions, the place there are not any parallels within the historic file. We have now demonstrated that “paleoclimate tuning” of ESMs can radically enhance their potential to simulate previous abrupt transitions, doubtlessly giving us extra confidence of their efficiency for future projections.
Supplies and Strategies
Coupled GCM
On this examine, we use the coupled GCM HadCM3-M2.1aD (40), which relies intently on HadCM3 (38, 58) coupled to model 2.1 of MOSES, a precursor of the present UK land floor mannequin Joint UK Land Setting Simulator (39). The atmospheric mannequin (HadAM3) has a horizontal decision of three.75× 2.5∘ (longitude–latitude) with 19 unequally spaced vertical ranges. It makes use of a Eulerian hydrostatic dynamical core with schemes for convection, turbulence, gravity waves, large-scale clouds and precipitation, and cloud microphysics (40). The ocean mannequin has a horizontal decision of 1.25× 1.25∘ with 20 vertical ranges. It solves the primitive equations utilizing a rigid-lid formulation. It contains parameterizations of sea-ice leads, isopycnal eddy mixing, and the ocean mixed-layer. The environment and ocean are coupled every day with no flux corrections (38).
MOSES 2.1 represents the land floor as a tiled patchwork of 9 land-cover varieties: 5 PFTs: broadleaf and needle-leaf bushes, C3 and C4 grasses, and shrubs; and 4 nonplant cowl varieties: lakes, city, naked floor, and ice. This mannequin contains the dynamic vegetation scheme TRIFFID (39) in order that plant protection, construction, and productiveness are coupled to the bodily local weather. PFT protection is up to date each 10 mannequin days utilizing a Lotka–Volterra competitors formulation with a hierarchy of bushes–shrubs–grasses (39). On this model, the vegetation respiration charge is decreased at greater temperatures following developments in MOSES 2.2. That is necessary in hotter climates, particularly over the Amazon (40). HadCM3-M2.1 doesn’t embody dynamic mud, however that is unlikely to be essential for the north African monsoon within the mid-Holocene. Ref. 34 confirmed that almost all local weather fashions overestimate the absorption of shortwave radiation by mud as a result of these fashions depend on outdated mud optical parameters (34), resulting in unrealistically massive impacts on the radiation finances and, therefore, precipitation.
HadCM3-M2.1aD is used right here in 4 configurations. Along with the usual model (STD) evaluated earlier than (40), we introduce modifications to convection (CONV), as described in ref. 37, and vegetation moisture stress (VMS), each described in additional element in
SI Appendix
. In CONV, convective entrainment/mixing detrainment is decreased at decrease atmospheric ranges and elevated additional aloft. This enhances the mass flux anomalies and thereby will increase the sensitivity of monsoonal precipitation to insolation. VMS makes use of an up to date parameterization of vegetation moisture stress, which has been optimized to breed the local weather–vegetation protection relationship within the tropics as compared with satellite-observed distribution of vegetation and to permit the dynamic simulation of a “inexperienced” Sahara for mid-Holocene (6 ka BP) circumstances. The fourth model combines each the convection and moisture stress modifications. These mannequin variations are labeled STD, +CONV, +VMS, and +CONV+VMS, respectively, and are summarized in Table 1.
Configuration of the 4 GCM setups examined on this examine
Transient Local weather Forcings
4 10,000-y transient Holocene simulations have been carried out with HadCM3-M2.1. The 4 configurations have been compelled with time-dependent evolution of land-ice and sea-level derived from ICE-6G (59, 60), hint gasoline concentrations of CO2, CH4, and N2 O from ice-cores adjusted to the Antarctic Ice Core Chronology 2012 (61) and with the post-Industrialization rises changed with the piControl mixing ratios really helpful for the Coupled Mannequin Intercomparison Mission 6 and modifications within the orbital parameters (62). The ice-sheet space and coastlines are up to date each 500 y. Different forcings are up to date each mannequin time step. The time collection of those forcings are proven in
SI Appendix, Figs. S1 and S2. All different mannequin boundary circumstances stay fixed by time at their preindustrial settings. The photo voltaic fixed was up to date to 1,361 Wm– 2. Volcanic eruptions and anthropogenic land use usually are not included. This setup is according to the PMIP4 deglaciation transient experiments described by ref. 63. The preliminary circumstances are similar in every case and are taken from the ten ka BP state of a PMIP4 deglaciation simulations with HadCM3-M2.1, by which freshwater is routed from Eurasian ice-sheets to the Arctic Ocean. The atmospheric state and vegetation modify quickly to the parameter modifications, in order that variations between the 4 simulations are seen from the primary few simulated months of Fig. 4. The affect from the definition of the seasons over the previous 10,000 y has been evaluated (64). This calendar adjustment utilized to the JJAS time collection for North Africa is smaller than 2% when a 100-y working imply is utilized.
A transient cooling occasion lasting round 200 y is seen in all mannequin simulations at 8 ka BP. In these transient simulations, the land–sea masks is up to date each 500 y. At 8 ka BP, the Hudson Bay turns into transiently related to the Arctic Ocean. This permits a comparatively massive quantity of freshwater to combine into the Atlantic, inflicting a weakening of the Atlantic Meridional Overturning circulation (AMOC) for round 100 y. The influence on precipitation over Africa is short-lived, with an ∼60-y discount in precipitation by round 1 mm×d– 1. The oscillations that precede the eventual collapse at 6 ka BP and which can be related to the biogeophysical albedo suggestions are longer in length and principally bigger in amplitude. For instance, occasions at 7.7 ka BP, 6.7 to six.55 ka BP, and 6.5 ka BP are bigger and final between 150 and 300 y. For the reason that AMOC recovers inside 150 y after 8 ka BP, we imagine that the freshwater enter is just not answerable for the hydrological cycle variations evaluated right here.
Evaluating the Position of Dynamic Vegetation
We reran 150-y segments each 1,000 y of the transient simulation (+CONV+VMS) to diagnose the function of vegetation change within the precipitation response. We deactivated dynamic vegetation in three units of simulations. The primary was initialized with the simulated vegetation protection and the second and third with the noticed present-day distribution. Within the third, the soil albedo over the Sahara was set equal to the clear-sky floor albedo produced by the vegetation area within the first set. The distinction between the primary and second configuration quantifies the vegetation suggestions to first order. The distinction between the second and third configurations permits a separation of the vegetation suggestions into elements as a consequence of vegetation-induced modifications in albedo and in land-surface moisture recycling.
Acknowledgments
P.O.H. was supported by a College of Birmingham Fellowship. This paper is Tipping Factors within the Earth System (TiPES) contribution 129: P.J.V. has acquired funding from the European Union’s Horizon 2020 analysis and innovation programme beneath Grant Settlement No 820970. Mannequin simulations have been carried out on the College of Birmingham’s Birmingham Setting for Educational Analysis (BEAR) facility (https://intranet.birmingham.ac.uk/it/teams/infrastructure/research/bear/index.aspx). We thank Gethin Williams for putting in the code and the BEAR workforce for help. We thank Anne Dallmeyer for sharing the Holocene North Africa database and Martin Claussen, Martin Widmann, Tom Dunkley-Jones, and Kirsty Edgar for informative discussions.
Footnotes
- Acquired Might 13, 2021.
- Accepted August 30, 2021.
-
Writer contributions: P.O.H. and P.J.V. designed analysis; P.O.H. carried out analysis; P.O.H. analyzed information; and P.O.H. and P.J.V. wrote the paper.
-
The authors declare no competing curiosity.
-
This text is a PNAS Direct Submission.
-
This text comprises supporting data on-line at https://www.pnas.org/lookup/suppl/doi:10.1073/pnas.2108783118/-/DCSupplemental.
- Copyright © 2021 the Writer(s). Printed by PNAS.
