Downloaded by guest on October 2, 2021 rpclpeatlands tropical on storage. networks carbon drainage and and geomorphology of peatland patterns effects tropical rainfall long-term predict temporal to in used changes be can mathematical models Our numerical to accumulation. and hours from long-term timescales reduce find can on We years precipitation networks. net drainage controlled in and study are fluctuations level, to that sea fluxes model climate, and in this storage changes carbon use by carbon area of We peatland the interior. rate tropical to dome the how proportional still-growing that is peat- the so growth of that rivers, their domes find by peat accompanying of We bounded uptake edges canals. are the at or they first where rivers shape ultimate of their reach network peat- lands ultimate, tropical a a the of within regime, storage, land rainfall carbon hence given and eleva- a morphology, surface stable under peat the specifies, Brunei of that Laplacian deter- in tion) (the flow parameter peatland groundwater shape and disturbance. a pristine rainfall mine and how a show climate from we by Darussalam, data affected comprehensive subsidence are and Using peatlands growth car- tropical accompanying of release patterns of how and for How- storage existed hydrology. bon has groundwater feedbacks theory the quantitative and no of carbon accumulation ever, disruption peat of human link megatons of that because of and year Denmark, hundreds M., per Odense emit dioxide Denmark, Southern now of University peatlands 2017) Evolution, Tropical 6, Earth February for review Center for Nordic (received and 2017 Biology 5, of May Institute approved Canfield, E. Donald by Edited Darussalam Brunei Darussalam; Brunei 20560; Germany; BB3910, DC and Potsdam, Begawan Washington, 14476 History, Seri Potsdam, Natural Bandar of Besar, of University Menteri Museum Jalan National Resources, Institution, Primary Smithsonian Anthropology, and Industry of Ministry Department, ae.I hswy h aeo etacmlto sdetermined is accumulation peat of rate accumu- the peat way, and this decay, its In of lack exceeds lates. by peat aer- inhibited of is low, production the decomposition is oxygen, aerobic when high, table by dioxide; is water carbon table controlled the water releasing are When occurs, establishes peat 9). decomposition studies (4, obic tropical of dynamics of range table loss a water and from accumulation Evidence and carbon 8). that growth of 4, the fluxes (1, quantifies of also dioxide description peatlands a tropical Because carbon, of 7). subsidence (6, organic emissions of mostly are change 20% is cover or alone peat land emissions and fuel year peatlands use fossil per land global megatons Asian global of 2% of Southeast about hundreds to of from equivalent rates Emissions reemis- at causing carbon (2–5): now is that agriculture which of for sion drainage waterlogging, and peat- (2) by fire tropical by years trees of lands disturbance of Human has thousands decomposition. peat domes suppresses for of these in preserved photosynthesis peat and by as stored sequestered carbon been The (1). high meters T storage carbon peatland 02139; MA Cambridge, Technology, Universit of Institute Massachusetts Engineering, Environmental a www.pnas.org/cgi/doi/10.1073/pnas.1701090114 Kai Ming Fuu Cobb R. Alexander peatlands tropical of fluxes carbon the and determine geomorphology rainfall in patterns temporal How etrfrEvrnetlSnigadMdln,SnaoeMTAlac o eerhadTcnlg,180 Singapore; 138602 Technology, and Research for Alliance Singapore–MIT Modeling, and Sensing Environmental for Center h oia etad tr iaoso abni etdomes, peat more or in 10 and carbon across kilometers forms of land mounded gigatons gently store peatlands ropical rniDrsaa er fBre ete iityo nutyadPiayRsucs aa etr ea,Bna eiBgwnBB3910, Begawan Seri Bandar Besar, Menteri Jalan Resources, Primary and Industry of Ministry Centre, Borneo of Heart Darussalam Brunei eTuos,CR,Isiu ainlPltcnqed olue Universit Toulouse, de Polytechnique National Institut CNRS, Toulouse, de e ´ a,2 | u aia aiSu’ut Haji Salihah Nur , etadgeomorphology peatland a,1 | lsnM Hoyt M. Alison , lmt-abnccefeedbacks cycle climate-carbon | b etadhydrology peatland ar Gandois Laure , g h ilg rgam,Uiest rniDrsaa,Bna eiBgwnB11,Bue Darussalam; Brunei BE1410, Begawan Seri Bandar Darussalam, Brunei Universiti Programme, Biology n hre .Harvey F. Charles and , | c agrnEri Jangarun , h oesgnl uvtr htacut o h abnstorage carbon boundary. is the drainage for it the accounts Nonetheless, within that (1). curvature m/km gentle 1 dome’s surface about the peat the are the Gradients of to subtle: doming very corresponding The in is surface. mound, peat shape the groundwater domed of slightly shape or a domed is table, implies faster. table boundaries water is water water near the of the gradient flow in steeper the gradient where A gradient the boundaries The lim- dome follows near table. rate flow—and steeper efficiency water a lateral peat—the at the conducts the occurs in it of flow transmissivity which This hydraulic with rivers. the by by bounded ited dome, is peat each it of edge where the toward groundwater downslope flows and Water evapotranspiration, flow. rainfall, between table balance water low a by exposed is peat that 1). time (Fig. of fraction the by 1073/pnas.1701090114/-/DCSupplemental at online information supporting contains article This Singapore. 118221 Research, and Technology ence, 2 Dig- Dryad the in 1 deposited been have paper (dx.doi.org/10.5061/dryad.18q5n). this Repository in ital reported data The option. deposition: access Data open PNAS the through online available Freely Submission. Direct PNAS a is article This interest. of conflict no declare authors The A.R.C. research; paper. and performed the data; analyzed wrote N.S.H.S. C.F.H. C.F.H. and and and A.M.H., F.M.K., A.R.C., K.A.S., tools; analytic R.D., new J.E., contributed A.R.C. L.G., A.M.H., A.R.C., site; field uhrcnrbtos ...adCFH eindrsac;ARC n ..etbihdthe established J.E. and A.R.C. research; designed C.F.H. and A.R.C. contributions: Author a,b rsn drs:GsMtooyLbrtr,Ntoa erlg etr gnyfrSci- for Agency Center, Metrology National Laboratory, Metrology Gas address: [email protected]. Present Email: addressed. be should correspondence whom To rieudsubdpa . oth- peat from undisturbed sequestration, erwise of to instead lead emissions, could dioxide models carbon climate the that by suggest calculate projected results cli- seasonality The to greater in drainage. used changes peatland tropical by be and driven mate can emissions peat- dioxide the principle the carbon This defines long-term in time. that stored over surface carbon land sequestered peat naturally the of determine of of amount together curvature pattern peat particular the The of a years. permeability of the of subsidence and thousands rainfall and over growth peatlands the tropical affect timescales yearly on shorter rainfall in and swamps fluctuations peat how tropical reveals protected Asia last Southeast in the of one from dataset A Significance h ae al ie n al napaln codn othe to according peatland a in falls and rises table water The alSbte,F336Csae-ooa,France; Castanet-Tolosan, F-31326 Sabatier, Paul e ´ d Ren , c Laboratoire Dommain e ´ f PNAS nttt fErhadEvrnetlScience, Environmental and Earth of Institute clgeFntonlee Environnement, et Fonctionnelle Ecologie ´ | ulse nieJn 2 2017 12, June online Published e,f aaihAuSalim Abu Kamariah , . b www.pnas.org/lookup/suppl/doi:10. eateto ii and Civil of Department e eatetof Department d Forestry g | , E5187–E5196

EARTH, ATMOSPHERIC, AND ECOLOGY PNAS PLUS PLANETARY SCIENCES tion under constant rainfall (17, 18). Although these subsequent works simulate the dynamics of peat production and decompo- sition in increasing detail, a strength of Ingram’s model was that it provided quantitative intuition for how peat dome morphol- ogy depends on peat hydrologic properties and average rain- fall. Could a principle like Ingram’s exist that describes peatland dynamics as well as statics and remains applicable with realistic drainage networks and rainfall regimes? We established a field site in one of the last pristine peat swamp forests in Southeast Asia and then used measurements Fig. 1. feedback leading to peat accumulation. Peat accumu- from this site to develop a mathematical model for the geomor- lation occurs because of waterlogging of plant remains and is therefore phic evolution of tropical peatlands that is simpler, yet more gen- determined by the proportion of time that peat is protected from aero- eral than Ingram’s model for high-latitude peatlands. Our model bic decomposition by a high water table. Over time, peat builds up into makes it possible to predict effects of changes in rainfall regime gently mounded land forms, or domes, bounded by rivers. The slopes in a and drainage networks on carbon storage and fluxes in tropical peat dome, although very small, govern groundwater flow toward bound- peatlands. The model predicted, perhaps surprisingly, that sur- ing rivers at rates limited by the transmissivity of the peat. face peat would be older near dome margins. We tested these predictions by radiocarbon dating core samples and comparing the age of each sample to the simulated age at its location and Once the peatland surface is sufficiently domed, water is shed depth. Finally, we explored the future of tropical peatlands under so rapidly that no more organic matter can be waterlogged within climate projections by simulating the geomorphic evolution of an the confines of the drainage network, and peat accumulation idealized peat dome under projected changes in rainfall patterns stops (10). This maximally domed shape sets a limit on how much and drainage. carbon a peat dome can sequester and preserve under a given rainfall regime (11). If the peat dome is flatter than its stable Methods shape for the current climate, it will sequester carbon and grow; Field Measurements. We established a field site in a pristine peat forest if it is more domed than its stable shape, it will release carbon in Brunei Darussalam (Borneo) to study a peat dome where current pro- and subside as peat decomposes. [In the tropical peat literature, cesses affecting peat accumulation are essentially similar to those during “subsidence” is used for a decline in the peat surface elevation, its long-term development (Fig. 2). At the site, we installed 5 piezometers regardless of mechanism (5).] The volume of this stable shape along a 2.5-km trail, 12 piezometers along a 180-m transect, and 3 through- times the average carbon density of the peat defines a capacity fall gauges. We completed a total station survey of peat surface elevation for storage of carbon as peat within the drainage boundary. along the transect to characterize peat surface microtopography. To char- acterize large-scale peatland morphology, we also obtained LiDAR data for If we can predict the stable shapes of peat domes and how the entire study area. To study peat dome development, we collected nine they evolve over time in a given climate, we can determine how peat cores from which we obtained 35 radiocarbon dates. To test whether peatland carbon storage capacity and carbon fluxes are affected our undisturbed site behaved similarly to sites studied by other groups, we by changes in rainfall regime, drainage network, and sea level. installed four respiration chambers and a piezometer at a nearby logged However, when predicting the stable shapes of peat domes and but undrained site. their evolution toward these shapes, there are two complicat- ing factors: (i) The boundaries imposed by drainage networks Morphology vs. Microtopography. Superimposed on the gross morphology have complex shapes and (ii) rainfall is intermittent and vari- of a peat dome is a fine microtopography of meter-scale depressions, or able. The water table rises during rainstorms and falls during hollows, separated by higher areas, or hummocks (19, 20). The hummocks dry periods, even when the peat surface is stable. These fluc- consist of partly decomposed logs, branches, and leaves lodged among liv- ing buttresses, stilt roots, pneumatophores, and giant rhizomes. Whereas tuations in the water table seem to be important because it is the microtopography in high-latitude peat may have regular and ori- widely believed that seasonality of rainfall affects tropical peat ented patterns (21), surveys by Lampela et al. (20) in a tropical peat swamp accumulation (12, 13). But how should we take these fluctuations in Central Kalimantan showed no orientation or regularity. Similarly, our into account to predict the slow development and stable shapes microtopography survey and other observations revealed no regular pat- of peat domes? Understanding the global impact of changes in terns or channels in peat dome microtopography. rainfall amount and variability, drainage networks, and sea level In describing the evolution of peat dome morphology, we want to cap- on tropical peatland carbon storage and fluxes requires a theory ture the effects of the hummock-and-hollow microtopography without that can accommodate the complicated drainage networks and explicitly simulating its details. Measurements from the 12 piezometers intermittent rainfall of the real world. along our microtopography transect showed that the water table is rela- tively smooth, even though the peat surface is highly irregular on a spatial Ingram (10) made the first prediction of the limiting shape of scale of centimeters to meters (Fig. 3). We therefore represent the peat sur- a temperate peat dome imposed by the balance between rainfall face by a reference surface p, smooth like the water table, that underlies and groundwater flow. Assuming constant rainfall, he computed the actual peat surface p˜. We refer to this reference surface p as the land the steady-state shape of a peat dome with uniform permeabil- surface. The peat surface p˜ is a “texture” that sits on the smooth land sur- ity between parallel rivers. Clymo (14) later developed a simple face p. The bottoms of hollows provide the most readily identifiable local dynamic model for accumulation of peat at a single point in the reference elevation (20), so we define the land surface p as a smooth surface landscape. Clymo’s model assumed that the thickness of peat fit through the bottoms of hollows (local minima in the peat surface p˜). On above the water table would not change and focused on anaer- the basis of this definition, we determined the current land surface at our obic decomposition in deeper waterlogged peat. Hilbert et al. site by smoothing a raster map obtained from local minima in LiDAR last- return points. We also used the transect survey and piezometer data to find (15) later built on Clymo’s model to allow a varying thickness the land surface p along the microtopography survey transect (SI Methods). of peat above the water table via a simple water balance whereby drainage increases linearly with peat surface elevation. Hilbert’s Groundwater Flow. We model the dynamics of the water table H sub- model inspired a series of increasingly sophisticated models for ject to net precipitation qn (rainfall intensity R minus evapotranspiration, vegetation dynamics and peat accumulation at a point. The most ET), using Boussinesq’s equation for essentially horizontal groundwater recent of these point models computes water table depth from flow monthly rainfall, using a site-specific model (16). Meanwhile, ∂H Sy = qn + ∇ · (T∇H), [1] numerical models have been used to simulate peat accumula- ∂t

E5188 | www.pnas.org/cgi/doi/10.1073/pnas.1701090114 Cobb et al. Downloaded by guest on October 2, 2021 Downloaded by guest on October 2, 2021 h aeo etls ydcmoiin(p decomposition constant by loss peat p dynam- of the rate modeled the we peat production 22), of peat loss (4, rate or of others the accumulation Like determines peat 22). of table ics (4, water loss the or above accumulation exposed peat of ness Balance. Carbon Local with uniform areas is expect in behavior would Discussion). height table and we water table (Results nonlaminar, instead water but were same gradients, flow table the water if at different addition, behavior in In flow peatland site. never local the We the different at hummocks. within if y through case hollows 6 flow the connecting our laminar is channels by which ephemeral gradient, limited observed head is the flow to overall as proportional peat the peatland only is local requires the flow the equation Boussinesq’s lateral through above periods. that water occurs wet during of flow hollows, flow some in surface, the though hollows to even and flow” refer hummocks “groundwater and than meters) larger of much apply (tens scales We pools. at isolated these equation in through between Boussinesq’s flow scale hummocks by the large limited of a is matrix on peatland porous the the flow through open-channel chan- flow allow Instead, into not peatland. connected the do not therefore are and pools (20) These nels pools. small forming below, peat thick. are domains they than flow wider for much equation are that modeling gradient peatlands groundwater head like standard particular a a is by equation driven perimeter per flow obe al. et Cobb transmissivity and elevation, table water in increment yield specific the where Fig. north) in explained (Damit, are (blue). (colors (E site simulations site 6). hydrologic Mendaram degraded the for at used a (triangles) tube Piezometers and flow (D) the south) of boundaries (Mendaram, the site and points) primary (green cores a peat in at radiocarbon-dated site showing Field data, ( LiDAR (B) island. airborne Malaysia. Borneo Peninsular on and Darussalam, Sumatra, Brunei Borneo, in peatlands of 2. Fig. A − thg ae als olw eoefloe rmstrto fthe of saturation from flooded become hollows tables, water high At H uvypit nmcooorpytasc Fg 3B). (Fig. transect microtopography in points Survey ) fpa xoe bv h ae al ie eopsto rate decomposition a times table water the above exposed peat of Distribution (A) Darussalam. Brunei in collection data field of Site α, C f p ra ag fsuisdmntae httethick- the that demonstrates studies of range broad A S hntewtrtbei tteln ufc and surface land the at is table water the when y steaon fwtrrqie o differential a for required water of amount the is ∂ ∂ p t = ∂ f p/∂ p − t (p C stedfeec ewe h rate the between difference the as otu a fsuyae from area study of map Contour ) − B H − E D )α H hc stethickness the is which )α, T ∇H stevolumetric the is Boussinesq’s . [2] eemndteefc fwtrtbehih ntransmissivity on height table water of flow effect groundwater the of determined divergence the controls water the divergence events, the flow. groundwater and rain of evapotranspiration Between of because yield. declines table specific table water in the increase gives of rate height vs. neg- intensity are rainfall flow the outward and and ligible, evapotranspiration of effects the 5 rain, (Fig. table water rains water the between the intervals of of decline dry rise uniform during uniform the and the rain describing heavy curves during table of summa- pair be a can by behavior rized table water the uniform, eleva- approximately table use water the from we distinct tion as follows, surface, land what the as to In falls, relative and (28). height” rises Hooijer table table “water water by the approximately elsewhere as remains observed piezometers surface all water land the in the second, uniform to and when rapidly; fea- relative First, very height falls distinctive peatlands. it table tropical two high, is in show table behavior water 5) surface. the table (Fig. piezome- water the five transect of in near 2.5-km tures height a flows table water along by on ters data dominated of is months Eighteen balance water Peatlands. Local Tropical of Capacity Storage Carbon Discussion and under Results microtopography in changes or near river conductivity compaction in hydraulic by agriculture. changes in coa- caused long-term changes peat dome surface consider these any of not the model did how process be to also to attempt long-term apparent We tend networks. not a not domes did However, suggesting is We area. peatlands, it larger lescence. older so a fill in dome, to larger peat expand now a could by dome dome between occupied area peat every is tropical complexes, rivers peatland of Asian Southeast predictions most quantitative In dynamics. reasonable make can that Approach. Modeling of Limitations perturbation general after answer peatlands drainage. to and tropical change model from climate the by fluxes used cores carbon peat then about from and dates questions peatland radiocarbon the against measured from model and surface extracted our Brunei peat in tested the site We to field LiDAR. surface our by peat at modern dome simulated peat the a matching of accumulation evolution Dis- peat and 2,700-y for (Results the spells parameters ulating dry the and fitted rain then heavy We to cussion). table water the of response the non- severe the handle Simulation). function to Dome transmissivity designed the features of special linearity with S1) (Fig. scheme (iv and the (i parameters: by four require coupled equations function The are space. and equations time in two These elevation table fluxes. water carbon and Eqs. phogenesis on based accumulation peat Simulations. Numerical water- of loss detectable decomposition. show anaerobic not do from (27), tropical peat site from logged our cores including decom- peat (2), of anaerobic Asia analyses include in pre- because sites not nor table did water site also the our We below extreme site. at position our observed these of neither simulations because were by simulations depths dicted and affect heights not table did water However, ulti- (16). effects tables production) water (primary these high trees very including by that uptake likely carbon also net is limit it mately and 26), (25, tables water low brackets angle where is production peat that way time a over such decomposition in by fluctuates balanced table subsiding water nor growing the neither wherever stable, 0 is surface peat The 4). (Fig. Transmissivity functions transmissivity and yield specific the determined We eea te tde aesonalvln f fsi CO soil of off leveling a shown have studies other Several H eopsto aeconstant rate decomposition a ) bv ensalvl eas h ae al height table water the Because level. sea mean above S y (ii , rnmsiiyfunction transmissivity a ) T H h·i ebitanmrclmdlo aelgigand waterlogging of model numerical a built We n h etsraeelevation surface peat the and safnto fwtrtbeheight table water of function a is ζ niaeatm average. time a indicate = PNAS H f p − = 1 u olwst ul h ipetmodel simplest the build to was goal Our | p hp and ulse nieJn 2 2017 12, June online Published orfrt h ae al height table water the to refer to − T h oe ssafiievolume finite a uses model The α. H T 2 (iii , IEpne ecito fPeat of Description Expanded (SI i osmlt etdm geomor- dome peat simulate to α, aeo etproduction peat of rate a ) E and oho hc vary which of both p, F · ∇ .Drn heavy During ). pcfi yield specific a ) (T 2 fu tvery at efflux f ∇ p h∂ α , S H y p/∂ , | We ). ζ ysim- by T E5189 from t and ζ i [3] f = is p ,

EARTH, ATMOSPHERIC, AND ECOLOGY PNAS PLUS PLANETARY SCIENCES AB C

Fig. 3. Microtopography and water table dynamics in a tropical peatland. (A) Cartoon of tropical peat cross-section showing variables p˜, the peat surface; p, the “land surface,” a smooth surface fit through local minima in p˜; H, water table elevation; and ζ, water table height relative to the land surface, ζ = H − p. The peat surface p˜ is irregular on a spatial scale of meters, with higher areas (hummocks) separating local depressions (hollows) that are not connected into channels. (B) Total station survey of peat elevation p˜ (black circles) along a transect and the land surface p (dashed black line). The minimum, median, and maximum water table elevations H from each of 12 piezometers along the transect are also shown (dashed blue lines). The absolute elevation of the survey points comes from matching local minima among survey points within 20-m × 20-m squares (white diamonds) with local minima in LiDAR last return data within the same squares (red diamonds). The land surface p is represented by the dashed horizontal black line. (C) Water table dynamics along a survey transect (B) in late 2012, relative to the land surface p. What appears to be a single blue line is superimposed data from the 12 piezometers shown in B. Also shown are the average minimum, median, and maximum water table elevations above the land surface during the same time period for all 12 piezometers.

using our water table data. The water table declines during dry We observe that water table height is uniform (∇ζ = 0). If trans- intervals because of a combination of evapotranspiration and missivity T is also spatially uniform, the groundwater divergence the divergence of groundwater flow; however, the two are easily term simplifies to the transmissivity times the peat surface Lapla- distinguished at low water tables because evapotranspiration cian (∇ · [T ∇(p + ζ)] = T ∇2p). The time derivative ∂p/∂t of ceases at night (Fig. 5D). Therefore, we can obtain the diver- the land surface elevation is negligible because peat accumula- gence of groundwater flow from the declining water table dur- tion or loss is much slower than rise or fall of the water table, ing dry intervals after accounting for evapotranspiration (refs. so the term p can be dropped from the time derivative. We 28, 29; further details are in SI Methods). We find that transmis- observe that the fluctuations in water table height ∂ζ/∂t are uni- sivity increases exponentially at high water tables, when water form, as is net precipitation qn , so the groundwater divergence rises into hollows and flows through hummocks, but decreases term T ∇2p must also be spatially uniform. Thus, Boussinesq’s dramatically at low water tables when water flows through fine equation simplifies to an ordinary differential equation (ODE) pores in the peat matrix (Fig. 5C). Very high permeability near describing the uniform fluctuation of the water table relative to the peat surface is consistent with our observations of more void the peat surface space higher in the peat profile and also with recent data from other tropical peatlands (30). The water table curves (Fig. 5 E dζ S = q + T ∇2p, [5] and F) indicate that the near-surface permeability is so great that y dt n the total thickness of deeper peat is unimportant for groundwa- ter flow. Therefore, transmissivity is approximately independent where the peat surface Laplacian ∇2p is uniform. of peat depth and depends only on the water table height ζ, which is uniform in space (although highly variable in time). Morphology of peat surface explains uniform water table behav- ior. According to Boussinesq’s equation, uniform transmissivity AB is not, by itself, enough to explain the uniform fluctuation of the water table. Even in hydrologic systems where hydraulic proper- ties are uniform, the water table can behave differently at differ- ent locations because of topography. For example, in most hydro- logic systems a rainstorm drives a different water table response at a topographic divide than it does near where groundwater dis- charges to a river. To understand the uniform water table behavior in peatlands, we refer back to Boussinesq’s equation (Eq. 1). If both the spe- cific yield Sy and the transmissivity T depend only on the local water table height relative to the surface and not on position within the peatland, uniform water table movement occurs if the divergence of the peat surface gradient, or the peat surface Laplacian ∇2p, is uniform (Fig. S2 C–E). (The “Laplacian of Fig. 4. Peat accumulation and CO2 flux vs. water table height in tropical the peat surface” ∇2p, or just “Laplacian,” is the scalar result peatlands. (A) Peat accumulation represents the balance between peat pro- duction and decomposition. (B) Aerobic decomposition is one of the two of applying the Laplacian operator ∇2 to the land surface ele- main sources of peat surface CO2 flux; the other source is root respiration. A vation p.) To see why a uniform land surface Laplacian explains shows peat accumulation or loss vs. water table height from model calibra- uniform water table behavior, we rewrite Boussinesq’s equation tion (solid line) and from literature subsidence data (circles, ref. 4; triangles, (Eq. 1) in terms of the water table height relative to the land sur- ref. 22). The straight line was not fitted to these data, but rather arose nat- face (ζ = H − p), instead of the water table elevation H : urally from calibration to match the modern surface of the Mendaram peat dome (Fig. 7). In B, soil surface CO2 flux vs. water table height at our site in Brunei Darussalam (white circles) was very similar to fluxes in other tropical ∂(p + ζ) peatlands (squares, ref. 23; diamonds, ref. 19; triangles, ref. 24; pentagons, Sy = qn + ∇ · [T ∇(p + ζ)]. [4] ∂t ref. 9; and hexagons, ref. 25).

E5190 | www.pnas.org/cgi/doi/10.1073/pnas.1701090114 Cobb et al. Downloaded by guest on October 2, 2021 Downloaded by guest on October 2, 2021 tbepaln,teei oln-emcag ntewtrtable a water In the uniform. uni- in change also spatially long-term is is no accumulation is height peat there table of peatland, stable rate peat- water entire local the an the if make form, because could that stable, behavior land simplest the is (they height elevation long-term table water no to or stationary; is surface are continue there peat but the table in climate, water change mor- of average and vagaries a surface the mean peat with a we fluctuate the for morphology,” which morphology “stable in stable By phology a dome. compute peat to tropical way peatland simple tropical remarkably stable determines morphology. Laplacian water surface the 6). and Laplacian, 5 Uniform Figs. in uniform piezometer” plain of (“ 6). differently region behaves (Fig. table the behavior outside table water contrast, uniform In of surface region uniform the a in indicating Laplacian map, LiDAR- elevation our a surface in contour peat find the derived by we enclosed area Indeed, the along and 6). gradient contour each nor- normal (Fig. integrated integrated area the between the enclosed relationship linear between the regression and by a gradient Laplacian mal of surface the slope of the There- uniformity normal area. studying the enclosed the examine the can of by we integral divided fore, the contour to the equal along eleva- is gradient surface contour closed the Laplacian any average of the within theorem, derivative because divergence the second map by However, the tion. elevation uses surface Laplacian the the in noise microtopographic maps. elevation surface against of tested be uniformity can where This curvature surface uniform. elevation are peat fluctuations the table of derivatives water curvature second direc- uniform the predicts horizontal of dynamics perpendicular sum two in the (∇ tions elevation to surface equal the is of It surface: peat 6). (Fig. region plain bog flatter intensity (B precipitation transmissivity net and rain, yield heavy specific During depth. table water ( overlapping R with sequences of alignment by (E evapotranspiration. of cycles from selected (B piezometer site field for curves transmissivity obe al. et Cobb local maximum and median, (B minimum, the 3). are surface (Fig. different shown transect a Also survey with 6). region microtopography a (Fig. 180-m in accumulation a lies peat (red) from current river lines) the of horizontal from area (dashed farthest an piezometer elevation The to surface interval. corresponding peat 10-m plain”), a “bog over gauges (the rain Laplacian automated three from lines) (vertical 5. Fig. aso h etsraeLpainaehgl estv to sensitive highly are Laplacian surface peat the of Maps the of curvature the describes Laplacian surface peat The E ;wt ori,ntpeiiaincnit nyo vptasiainE (F ET evapotranspiration of only consists precipitation net rain, no with ); ueipsdwtrtbehih jge lelns rmfiepeoeessann . madrifl intensity rainfall and km 2.5 spanning piezometers five from lines) blue (jagged height table water Superimposed (A) calibration. and hydrology Site DEF A 2 p = ∂ h nfr etsraeLpainpoie a provides Laplacian surface peat uniform The 2 h∂ p /∂ p /∂ x 2 t i A. + 0 = D, ∂ 2 , Inset and p h∂ /∂ and H and F y hw elnn ae alsdrn h a usae)adsed ae alsa ih sae)die ydiurnal by driven (shaded) night at tables water steady and (unshaded) day the during tables water declining shows atrrcag uv (E curve recharge Master ) C 2 /∂ ,adbu rnlcn ie r sebe rmfil aain data field from assembled are lines translucent blue and ), .Tu,aayi fwtrtable water of analysis Thus, ). C t ,dtrie rmrcag n eeso uvs( curves recession and recharge from determined ), i 0 = .Uiomwtrtable water Uniform ). n eeso uv (F curve recession and ) .Dse lc ie in lines black Dashed ). egt oaywtraddb e rcptto uteventually must precipitation flow groundwater net by by removed added be water any so height, oee,i h lmt ean iia oteciaedrn its C during (t) climate tons m the to 3.88 metric similar of 1,535 remains depth about climate peat the stores mean if dome and however, a peat m) has the 4.92 currently example, (max site carbon For primary the peatland. gives our peat the at the of of capacity density carbon storage mean equa- given the Poisson’s satisfying any times surface tion in carbon the and under peat volume boundary The drainage the climate. any gives inside capacity principle storage uniform-Laplacian the car- raphy, peatland capacity. tropical storage determine bon network drainage drainage any and in Climate regime rainfall given network. a under morphology (∇ parameter land shape a finds decomposition one balances production peat (Eq. until value sta- Laplacian the find (Eq. can fluctuations one regime, Laplacian rainfall surface ble any hydrologic–biological the for and However, rainfall system. of pattern temporal the of transmissivity average The climate. peatland tropical (Eq. morphology any equation of Poisson’s topography solving stable by the compute can We hs h Laplacian the Thus, smnsteaeaentpeiiaindvddb h average the by divided precipitation net average transmissivity the minus is .I hsway, this In Methods). (SI peatland the in everywhere 3) sebe rmitraso ev anadn an respectively, rain, no and rain heavy of intervals from assembled ) E p and ∞ = 0 sn h prpit alca au o that for value Laplacian appropriate the using , ihatilLaplacian trial a with 5) F sin As A. hr nevlo ae al aafo single a from data table water of interval Short (D) ).  E S and ∇ ∇ y yseiyn h tbepaln topog- peatland stable the specifying By d ∇ 2 2 dt PNAS p , p ζ h e uv sfo h izmtri the in piezometer the from is curve red the A, F 2 q C, ∞ ∞  p hwwtrtbersos optdfrom computed response table water show n ∞ and E, = yrpael iuaigwtrtable water simulating repeatedly by ftesal etadsurface peatland stable the of = | R = ulse nieJn 2 2017 12, June online Published hq − 2 BC − p n Ti oiae yrifl intensity rainfall by dominated is ET F ∞ i o h tbepa surface peat stable the for 7) hT hq illp-cl pcfi il and yield specific Hillslope-scale ) hT + htdsrbssal peat- stable describes that ) n i i i hT sacmlctdfunction complicated a is . i∇ ∇ 2 2 p p n dutn the adjusting and ∞ . | · ha E5191 p −1 [7] [6] ∞ ;

EARTH, ATMOSPHERIC, AND ECOLOGY PNAS PLUS PLANETARY SCIENCES carbon at our site (31), which suggests a maximum downward A velocity of water of about 1 m/y or at most a 1.4-mm water table decline during a single 12-h night, 1/16th of the 22-mm water table decline from evapotranspiration during the day (Fig. 5). (Evapotranspirative flux is about 1/10th of the rate of decline of the water table from evapotranspiration because about 1/10th of the deep peat cross-section is available for water flow; see spe- cific yield curve in Fig. 5B.) A shape parameter related to our stable peatland Laplacian (Eq. 7) appeared in Ingram’s model for temperate peatland mor- phology (10) assuming constant precipitation, uniform hydraulic conductivity, and simple river geometry (Ingram’s parameter is net precipitation divided by hydraulic conductivity, instead of average transmissivity). Our result is more general, because it handles varying rainfall and arbitrary landscapes, but is also B mathematically simpler, because of our finding that transmis- sivity in tropical peatlands is approximately independent of peat depth.

Dynamics of Tropical Peatland Topography and Carbon Fluxes. Peat accumulation parameters regulate dome dynamics. Our analysis shows how the rate of peat production fp and decom- position rate constant α affect both the stable morphology and C the dynamics of tropical peat domes. These parameters of the peat accumulation function (Eq. 2) have an indirect but strong effect on the stable peat surface Laplacian and hence peat- land carbon storage capacity via the mean transmissivity hT i (Eq. 7) because the mean water table depth must be equal to the ratio of the peat production rate to the decomposition rate constant (fp /α; Eq. 3). A higher decomposition rate constant implies a higher mean water table in stable peat domes, meaning Fig. 6. Estimation of peat surface Laplacian. (A) Regions of different mor- phology and water table behavior within the flow tube used for field site simulations and locations of piezometers (triangles). Farthest from the river, the land surface is relatively flat (bog plain), next there is a region in which A the Laplacian of the land surface elevation is uniform (“stable”), and finally there is a narrow region near the river where hydrologic processes and peat accumulation are affected by the rise and fall of the bounding river (“river flooding influence”). (B) Profile of LiDAR land surface elevation from A, showing piezometer locations (vertical dashed lines). (C) Normal gradient driving efflux, integrated along contours, vs. enclosed area. The slope in the stable region gives the average land surface Laplacian of the land surface there and was used for calibration of hydrologic parameters.

2,300-y development, we predict that in about 2,500 y it will reach a stable shape with a mean peat depth of 4.54 m (max 7.10 m) and store 1,800 t C · ha−1 (Fig. S3; simulations of dynamics are described in the next section). BC The uniformity of the stable peat surface Laplacian is an approximation that requires that (i) peat accumulation rate ∂p/∂t is a nondecreasing function of water table height, (ii) flow of water is proportional to water table gradient (Boussinesq’s equation), and (iii) transmissivity is independent of location because flow through deep peat is negligible compared with near-surface flow. In reality, groundwater flow through deeper peat will result in a deviation of the stable peat dome surface from the uniform-Laplacian shape in very large peat domes. Specifically, groundwater flow through deep, low-permeability peat will tend to flatten the dome center, because of slow infiltra- tion of water into the deep peat, and steepen the dome margin, Fig. 7. Morphogenesis of Mendaram peat dome. (A) Shape of peat dome because of exfiltration of water back into the high-permeability over time, including modeled peat surface (contours), modern peat surface near-surface peat near the boundary. Deep groundwater flow from LiDAR (dashed black line), and calibrated radiocarbon dates from peat should be manifested as a downward (dome center) or upward core samples (colored circles). The deepest peat layers before 2,250 cal y BP represent a uniformly deposited peat on a gently sloping clay (dome margin) trend in the water table during nights without plain (27). (B) Simulated age of peat vs. calibrated radiocarbon ages from rain when the water table is low; no such trend is apparent in samples in the Mendaram peat dome. (C) Age of shallow peat samples (25– our piezometer data (Fig. 5D), suggesting that deep groundwa- 65 cm depth) vs. distance from river at a primary site (solid circles) and a ter flow is small. A small deep groundwater flow term is further nearby deforested site (open circles). Note the old peat near the surface supported by radiocarbon dating of porewater dissolved organic close to the river as predicted by the model.

E5192 | www.pnas.org/cgi/doi/10.1073/pnas.1701090114 Cobb et al. Downloaded by guest on October 2, 2021 Downloaded by guest on October 2, 2021 ow olce 2adtoa aicro ae rmbsland basal from dates radiocarbon additional 22 collected we so the near peat so growing, there. stopped older has is and margin surface shape the stable carbon, not con- its sequester has center and reached that dome peat the region accumulate Whereas central to Laplacian. the tinues stable as its dome reached peat bog yet tropical the define a bog we of unforested instead, the plain (21); unlike peatlands (1), high-latitude distinct of be plains trop- not of may vegetation plains bog The ical plain. bog (smaller- at central flat growing Laplacian) relatively continues a magnitude forming dome rate, peat uniform approximately the an and of 8 interior and 7 the (Figs. there Meanwhile, lowest because is boundaries, surface its dome at stable first the shape stable margins. its and reaches bog dome at plain peat surface bog the central near peat a old predicts tropical other principle in uniform-Laplacian and The site effect our the at 4B). similar that (Fig. is peatlands suggesting fluxes sites, on table other water from of those to 4A; similar very (Fig. exactly the α matched almost nonetheless data surface—but peat subsidence modern literature— the the from to data subsidence only these vege- function to accumulation calibrated drained, peat not linear in was Our accumula- subsidence 22). (4, peat on swamps data peat calibrated tated literature our to measured compared function surface then tion peat We modern LiDAR. the by and the surface between peat difference least-squares simulated accumu- with the minimizing peat also and obtained 7) and (Fig. We sites section). other (next parameters lation data from field field Brunei data our other published at our dome with peat a agree of site topography the to fitted data. eters faster. literature are match dynamics parameters dome Fit peat but change, not does constant rate sition obe al. et Cobb and Laplacian, production peat surface both stable If storage. smaller carbon less a transmissivity, higher a here. presented (smaller- findings Gray boxes, flatter plain. black bog a central results; of the established area middle, boxes, the the in to growing, area proportional Laplacian) is rate surface dome magnitude a peat balanced at a carbon is When sequesters dome. production it any the the peat in inside Laplacian, and everywhere uniform dome uniformly, decomposition a by peat fluctuates has a height surface table of dome water capacity the storage When carbon boundary. drainage and shape stable the uniform a transmissivity by hydraulic average described to completely shape a Laplacian toward surface evolves dome peat ical 8. Fig. 1 = le etna oemrishsntbe rdce before, predicted been not has margins dome near peat Older . 80 oe ftoia etdm eeomn.Tesurface The development. dome peat tropical of Model d −1 .Orsi CO soil Our ). ∇ 2 p f ∞ p α α , ie ytertoo vrg e precipitation net average of ratio the by given nraetgte,cro trg capacity storage carbon together, increase ysmltn h vlto ftedome the of evolution the simulating by 2 hT hme esrmnswr also were measurements chamber h ufc Laplacian surface The i. etacmlto param- accumulation Peat efidta tropical a that find We f p f p n h decompo- the and 1 = .46 ∇ 2 p p mm ∞ i.S2). Fig. fatrop- a of defines · y hq −1 n i , o eusrto ogl rprinlt h rao h central the of car- area the of to rate proportional the roughly makes sequestration development bon area. dome plain of bog offset pattern to centripetal proportional the is in rate with sequestration deposited Carbon consistent our by first 27), 7B). simulated (Fig. ages was ref. model and to material dates in radiocarbon peat 11 when measured deep between (figure than of stratum later peat dates study, that younger radiocarbon y earlier by bias 500 an peat would about In older pools peat. of tip-up tip- younger replacement form in with to that fill peat (iii) estimated older then and we remove that surface, forests the pools swamp below up growth peat root peat in have tree into falls (ii) may carbon tree growth, network young dome simulated drainage of inject y may the match 2,300 (i) to the during because did cores shifted We exactly from 7B). ages dates (Fig. at peat well radiocarbon ages matched simulated expect and depths and 7 not and dates (Fig. locations Radiocarbon forest same peat 27). swamp the locations mangrove (1, peat same the Methods) the the of SI from at establishment samples ages interior the basal before simulated the excluding to depths, toward peat and depths 7 deeper (Fig. same in domes dates the near same older at the was of than peat radio- margins near-surface These that prediction. dome confirmed this test dates to carbon samples peat near-surface etacmlto rls assutk rrlaeo abn respec- carbon, of release or uptake t −24 causes CO initial loss The tively. or accumulation in as peat simulations dotted- for (long time signal vs. ENSO tive) increased (F or line). line) or (solid dashed sea- line), conditions increased short-dashed in line), (D, decomposition. change dotted-dashed drainage no aerobic line), (B, from dotted-dotted-dashed rise ( C, loss level sonality sea peat line), drainage rapid long-dashed Sustained drives (A, (D) cm peat. 50 of of loss (E causes depth mm/y a 1,095 to rivers to tidal mm/y as elevation 902 surface (C from peat rise. in dome shift peat upward the an bounding to level Sea leads (B) m steps. morphology. 0.5 stable of new time a rise accu- reaches subsequent peat dome causes peat at mm/y the 2,430 until morphology to mulation mm/y 2,237 dome from and increase peat rivers, rainfall Annual parallel the (A) two indicates give between line dome initially lines peat dashed an colored the The for of perturbations. morphology time different stable vs. the after elevation dome surface peat peat stable Simulated (A–D) peatlands. 9. Fig. CD AB ptal vrgdpa et s iefrsmltoswt oerain more with simulations for time vs. depth peat averaged Spatially ) F E · ha yai fet fciaecag ncro trg ntropical in storage carbon on change climate of effects Dynamic −1 · y −1 vrg CO Average ) . 2 msinfrtedang cnroi f h hr at chart the off is scenario drainage the for emission PNAS 2 msin(eaie rsqetain(posi- sequestration or (negative) emission .W locmae radiocarbon compared also We C). nraei esnlflcuto nrainfall in fluctuation seasonal in Increase ) eas eti otyognccarbon, organic mostly is peat Because E. | ulse nieJn 2 2017 12, June online Published | E5193 The

EARTH, ATMOSPHERIC, AND ECOLOGY PNAS PLUS PLANETARY SCIENCES bog plain (Fig. 8). Under a given climate, the rate of sequestra- area p¯∞ is proportional to the stable peat surface Laplacian 2 tion decreases as the dome approaches its stable shape and the p¯∞∝ ∇ p∞. central region of peat accumulation—the bog plain—shrinks in Dynamic simulations converge to new stable morphologies after area. For example, our simulations imply that the current rate changes in conditions. Our simulations of peat dome dynam- −1 −1 of CO2 sequestration at our site (0.80 t · ha · y , 100-y aver- ics demonstrate the convergence of initially stable domes to age) is less than one-quarter of its initial rate about 2,300 y ago new, stable, uniform-Laplacian morphologies after perturba- −1 −1 (3.81 t · ha · y ), and CO2 sequestration is more than five tions (Fig. 9). The simulations show the effect of increased times faster at the dome interior (1.89 t · ha−1 · y−1, 6.37 km total rainfall (Fig. 9 A and E), which is a recognized climate from the river) than at its edge (0.36 t · ha−1 · y−1, 1 km from feedback for tropical peatlands (12), and also show that artifi- the river; Fig. S3). The mechanism of tropical peat dome devel- cial drainage for agriculture (Fig. 9D) can dominate all natural opment that we describe therefore creates landscape-scale pat- feedbacks if not curtailed (4, 16). In addition, our simulations terns in local carbon fluxes and radiocarbon date profiles. Local demonstrate a third feedback: The increase in rainfall variabil- measurements of carbon fluxes or radiocarbon dates cannot be ity from warming climates (33) can cause peat loss if not com- upscaled to regional fluxes without considering dome morphol- pensated by an increase in total rainfall (Fig. 9 C and F). For ogy because the flatter interior of each peat dome sequesters car- these simulations, we generated new rainfall time series as sim- bon whereas the margins do not (Fig. 8). Old peat near the peat- ilar to current rainfall as possible but with larger annual and land surface (2), although in some cases caused by local climate El Nino–Southern˜ Oscillation (ENSO) fluctuations (Fig. S4 A change or disturbance, also can be expected at the margin of any and B and SI Setting Annual and ENSO Amplitudes of Rain- peat dome. fall). Either greater seasonality or a stronger ENSO decreased peatland carbon storage capacity, but an increase in seasonal- Future Effects of Changes in Drainage Networks and Climate. Our ity had a larger maximum effect, partly because the magnitude analysis provides a simple way of predicting long-term change of the ENSO fluctuation is smaller. In contrast, sea level rise in peat dome morphology and carbon storage in response to could drive peat accumulation in the long term by elevating the changes in drainage network, climate, or sea level because the tidal rivers draining most peat domes (Fig. 9 B and E). In gen- stable peat surface Laplacian completely specifies the stable eral, losses can be much more rapid than accumulation (Fig. peat topography with given drainage boundary conditions. If the 9E), because subsidence of drained peatlands can be far faster drainage network changes, we can solve Poisson’s equation in than typical accumulation rates (4). For example, the estimated the new drainage boundary to compute the gain or loss of peat, area-averaged current CO2 sequestration rate at our site is and the net carbon emissions, as the peat surface approaches its −1 −1 0.80 t · ha · y , whereas Hooijer et al. (5) estimated CO2 emis- new stable topography. If the climate changes, we can compute a sions of at least 73 t · ha−1 · y−1 from tropical peatlands under new stable Laplacian value for the new climatic conditions and plantation agriculture. determine how much a currently stable peatland will grow or Intermittency of rainfall reduces tropical peatland carbon stor- subside. age. We find that fluctuations in net precipitation on timescales Subdivision of a peatland by drainage canals reduces carbon from hours to years can reduce long-term peat accumulation. We storage. The average surface elevation of a stable peat dome is further explored the effects of variability in rainfall seen in our proportional to the area of the dome because of the uniform- Laplacian principle. If we scale the area of a peat dome√ by some factor k by multiplying both x and y coordinates by k, the sur- face elevation p must increase by the same factor k to keep the AB same Laplacian. Therefore, the carbon storage capacity of a peat dome scales with its area. For example, a peat dome that is cut into halves of approximately the same shape as the original dome will have one-half the carbon storage capacity (half the mean sta- ble peat depth) of the original dome. This provides a straightfor- ward way to estimate the long-term impacts of artificial drainage networks that are now affecting over 50% of the peatlands of Southeast Asia (32) and from which a robust quantification of carbon emissions is urgently needed (6). The dynamic response of a peat dome to changes in rainfall and sea level also depends on its area because of the centripetal pattern of dome development (Fig. 8). Because of their higher stable mean depth, larger-area domes reach their stable shape more slowly than smaller-area domes. Relative effects of on carbon storage capacity are Fig. 10. Effects of climate change on carbon storage capacity of tropi- cal peatlands. (A) Simulated carbon storage capacity (contours) vs. time- independent of drainage network. Although peatland drainage averaged rainfall and interval between storms in a simple rainfall model networks play a central role in determining absolute carbon stor- (Poisson process for storm incidents, exponentially distributed rain depth age and dynamics, we can calculate the proportional effect of cli- per storm). The balance between rainfall and groundwater flow sets a limit mate change on long-term carbon storage of a tropical peatland on the curvature of the peat surface and therefore limits the amount of car- independent of the drainage network. Poisson’s equation (Eq. 7) bon that can be stored as peat in a peatland. This carbon storage capacity must be solved in each drainage boundary to obtain the topog- is proportional to the Laplacian of the stable peat surface elevation (Results raphy of the stable peat surface. However, we can then predict and Discussion), so the relative effect of changes in rainfall patterns on car- the effects of changes in climate independent of the drainage bon storage capacity can be calculated independent of the drainage net- network because of the linearity of the Laplacian operator. By work. Higher rainfall increases carbon storage capacity, whereas increased time between storms reduces it. ( ) Carbon storage capacity (contours) as the definition of linearity for a mathematical operator, a peat B in A, but driven by rainfall at our site (diamond) or with a weakened or surface Laplacian that is larger by some factor k corresponds strengthened annual or El Nino–Southern˜ Oscillation fluctuation in rainfall. to a peat surface that is vertically stretched by the same factor The vertical shift to lower carbon storage with increased annual variation in 2 2 (k∇ p = ∇ kp) and therefore has a mean peat depth that rainfall (up arrow) corresponds to the simulated effect of increased season- is larger by the same factor. Thus, carbon storage capacity per ality in Fig. 9.

E5194 | www.pnas.org/cgi/doi/10.1073/pnas.1701090114 Cobb et al. Downloaded by guest on October 2, 2021 Downloaded by guest on October 2, 2021 1 lsrP,Hne C,See I ev S oi J(04 ae,ptwy and pathways Rates, (2004) PJ Morin AS, Reeve DI, Siegel model. BCS, geophysical A Hansen - PH, Glaser raised 11. in shape and Size (1982) HAP Ingram 10. h tbemrhlg n trg aaiyo abnwti the within carbon boundary. of series drainage capacity on peatland storage time pattern sur- and rainfall rainfall peatland morphology the stable any of stable the effects from This the summarizes computed uniform. completely be and is can surface Laplacian land Lapla- face the the which of in shape cian limiting that a shows approach approach car- predic- peatlands Our dome tropical fluxes. these peat carbon and quantify carbon, capacity also storage organic morphogenesis bon Because dome mostly domes. peat of are peat tions tropical domes of peat morphology tropical drainage and the regimes on rainfall in networks changes of predict effects here long-term presented the models numerical and mathematical The Conclusions more kt sustain (19.5 would storage intermit- carbon kt site the long-term 1.80 our as more example, at times intensity For 10 observed average than declines. same actually table the rainfall water at tent the drizzle compensated after steady not rates a is flow tables water low peak by The during flow. flow the groundwater outward in in high fluctuations changes because exponential rainfall, cause table mean water by only not variation by rainfall, demon- controlled in simulations is accumulation peat- The peat on 10). long-term that inter- fluctuations (Fig. strate capacity ENSO of storage and effect carbon annual the land and computing time by arrival 9) storm (Fig. simulations dynamic obe al. et Cobb 400%, an by to mo ( led 1 1,000% over d than more 100%, 1 by by y over wk 1 averaging over 1 and over but surface 20%, percent, stable by simulated few overestimate the a increased by min, hour, Laplacian 20 each every constant of Treating as aver- instead intervals. evapotranspiration after longer and and Laplacian intensity in hourly rainfall surface on fluctuations precipitation stable neglecting net the of aging computing in effects by fluctuations the of rainfall explored effects the We ignore rainfall. entirely that overestimated same capacity. severely simulations be storage the by can carbon give capacity the same storage can carbon the in series However, and important, time Laplacian not rainfall surface are distinct stable exact rainfall many The in that peatlands. fluctuations sense tropical the of of storage details long-term carbon the predict and correctly sub- evolution to of rainfall effects in the fluctuations consider diurnal must models that 10A). (Fig. show rainfall transpiration mean same the our with at h, as 24 average, every on or h 14 site, every on arrive depending storms one-third convective by whether decrease can storage: capacity carbon storage Carbon long-term affects significantly storms convective .Hrn ,Juiie ,IoeT aaah 20)Cnrl ntecro aac of balance carbon the on Controls (2009) H Takahashi T, of Inoue J, development Jauhiainen and T, Hirano hydrology, Asia. geomorphology, 9. Southeast the of of peatlands Models The (1994) fire: RB of Winston line in the 8. change In cover (2016) land A and Hooijer use S, land Page peatlands. from fluxes tropical 7. carbon drained Regional in (2016) al. loss et carbon L, Calle and Subsidence 6. (2012) tropical from al. fluxes et gas A, in Greenhouse Hooijer fires (2010) forest 5. H and Joosten peat R, from Dommain released J, carbon Couwenberg of amount 4. The (2002) al. et SE, Page 3. sequestration carbon and Development (2011) H Joosten J, Sarawak Couwenberg R, of Dommain swamps peat the 2. of development and structure The (1964) JAR Anderson 1. u iuain ihsote analitniyadevapo- and intensity rainfall smoothed with simulations Our rpclpeatlands. tropical bodies. peat domed Sci Biol B Lond Soc R Trans 1980–2009. Asia, Biogeosciences Asia. Southeast in peatlands 1997. during Indonesia and changes sea-level post-glacial to Links variability. Asia: climate South-east Holocene in domes peat tropical of Brunei. and rvr o etaddvlpeti h usnByLwad,nrhr Ontario, northern Lowlands, Bay Hudson the in Canada. development peatland for drivers Nature · 297:300–303. ha Ecol J −1 rpGeogr Trop J 92:1036–1053. ; 9:1053–1071. nio e Lett Res Environ i.S4 Fig. Ecosystems elScA Bull Am Soc Geol Nature 18:7–16. 371:20150176. D lbCag Biol Change Glob utSiRev Sci Quat 12:873–887. 420:61–65. and 074011:1–12. 106:1594–1604. .Teitritnyo tropical of intermittency The E). 30:999–1010. i.S4 Fig. 16:1715–1732. D and E). · ha −1 Philos vs. 2 alo M oda K a-oi C(05 oeigrltosisbetween relationships Modeling (2015) CC in May-Tobin Transitions LK, America: north Goodman eastern KM, in Carlson bogs 22. Raised (1986) JA patterns Janssens vegetation PH, and Glaser microtopography surface 21. Ground (2016) al. et Car- M, Lampela (2005) H 20. Vasanders PJ, Martikainen 1: JEP, model Heikkinen development H, peatland Takahashi DigiBog J, The Jauhiainen (2012) 19. LR Belyea PJ, Morris AJ, Baird 18. climate of seasonality the and occurrence peat World (1994) AM Ziegler AL, Lottes 12. 7 orsP,BidA,Ble R(02 h iio etaddvlpetmdl2: model development peatland DigiBog The (2012) LR Belyea AJ, millennia: Baird over PJ, peatlands Morris tropical of 17. accumulation Carbon (2015) al. et dynamical S, as peatlands Kurnianto of analysis 16. and Modelling (2000) T Moore N, Roulet growth. D, bog Hilbert peat 15. to limits differ? The deltas (1984) Mahakam and RS Rajang Clymo the do 14. Why peat: no or Peat (2010) RA Gastaldo 13. ainudrGat 145 n 146 t ...,adb grant a by of and Institute C.F.H.), Massachusetts at (to Initiative Foun- 1114161 Solutions Science Technology. and Environmental National 1114155 the US from Grants the Model- by National under and program, Sensing the dation for research Environmental by Alliance interdisciplinary for supported Singapore–MIT ing Center com- Technology’s the was detailed and through research their Research Singapore This for Muhammad Foundation also binti manuscript. reviewers We Research Mureh, anonymous the Sukmaria assistance. for two Rosaidi on field Rahayu and Chua for ments Pandai, Glaser Ferrant Amy Ng, Azlan Paul Sylvain Depart- staff; and thank Awok, Long of Forestry Ariffin, bin Jun Zainal release Darussalam Watu Wafiuddin Jamilah and Bernard Brunei Sukri, Hajah work and the Haji field project; of support; of this Industry Ahmad logistical facilitation of of Ali for support Ministry Joffre ment their Darussalam and Brunei for the Jalil Resources and Primary Center and Borneo of Heart ACKNOWLEDGMENTS. levels. sea river long- rising saltwater in and from other composition, intrusion shifts community of as tree such in effects changes understudied, also networks, the remain here that including outlined processes for dis- approach term framework realistic The a a sizes. given dome provides storage, peat carbon of and level, peat tribution sea effi- tropical rainfall, to changing on of principle drainage effects stable sys- uniform-Laplacian the earth for the their account Improved complex use ciently reach conditions. could peatland in domes models change a tem smaller a after within because faster areas shapes peat important, and because dome also 8 dome, (Fig. peat is center whole of the the in distribution over- for dome fastest growing is flux a accumulation of average center the the in important mea- estimate flux example, are carbon For of (34). domes carbon models surements system peat ground-truth earth tropical for of considerations within locations mor- measurements dome the con- of flux areas pattern that This central the implies 8). while of (Fig. phogenesis first perimeter carbon accumulate the elevation to dome, steady tinue peat a growing reaches a dome On trop- peatlands. in fluxes ical carbon modeling and measuring when morphology 4A). those (Fig. calibration though for even used accu- not data, peat were literature (v) data and match 5C); is parameters (Fig. balance surface mulation water the local near that flow so by tables, dominated water the high increases within at transmissivity table gradients exponentially (iv) water differing 5A); (iii) with (Fig. 5A); areas region (Fig. in water uniform-Laplacian same interior (ii) the dome 6C); is the behavior (Fig. in uniform is different edge Laplacian is surface dome and the the where approximately uniform near is is behavior region table Laplacian a surface in peat uniform The (i) observations: u nlssudrcrsteiprac fcnieiggeo- considering of importance the underscores analysis Our of range a by supported is principle uniform-Laplacian The n vegetation. and ae al et n etsi abnls nSuhatAinplantations. Asian Southeast in loss carbon soil Lett peat Res and depth table water stratigraphy. gross and landforms forest. swamp peat tropical a floor. in forest swamp peat tropical a 1797. from fluxes bon basis. hydrological and model, conceptual Rationale, chdooia iuain n2D. in simulations Ecohydrological approach. modeling A systems. 303:605–654. Geol Coal J Int 10:074006. Ecol J 88:230–242. 83:162–172. aaoeg aaolmtlPalaeoecol Palaeoclimatol Palaeogeogr lbCag Biol Change Glob etakMhu usfo rniDarussalam Brunei of Yussof Mahmud thank We PNAS Catena a Bot J Can Ecohydrology | ulse nieJn 2 2017 12, June online Published 139:127–136. 21:431–444. 64:395–415. 5:256–268. hlsTasRScLn ilSci Biol B Lond Soc R Trans Philos Ecohydrology lbCag Biol Change Glob 106:23–37. .The S3). Fig. 5:242–255. | 11:1788– Environ E5195

EARTH, ATMOSPHERIC, AND ECOLOGY PNAS PLUS PLANETARY SCIENCES 23. Melling L, Hatano R, Goh KJ (2005) Soil CO2 flux from three ecosystems in tropical 33. Huang P, Xie SP, Hu K, Huang G, Huang R (2013) Patterns of the seasonal response of peatland of Sarawak, Malaysia. Tellus 57B:1–11. tropical rainfall to global warming. Nat Geosci 6:357–361. 24. Ali M, Taylor D, Inubushi K (2006) Effects of environmental variations on CO2 efflux 34. Quer´ e´ CL, et al. (2009) Trends in the sources and sinks of . Nat Geosci from a tropical peatland in eastern Sumatra. 26:612–618. 2:831–836. 25. Jauhiainen J, Hooijer A, Page SE (2012) Carbon dioxide emissions from an Acacia 35. Moore ID, Grayson RB (1991) Terrain-based catchment partitioning and runoff pre- plantation on peatland in Sumatra, Indonesia. Biogeosciences 9:617–630. diction using vector elevation data. Water Resour Res 27:1177–1191. 26. Hirano T, et al. (2012) Effects of disturbances on the carbon balance of tropical peat 36. Stuiver M, Reimer PJ (1993) Extended 14C data base and revised CALIB 3.0 14C age swamp forests. Glob Change Biol 18:3410–3422. calibration program. Radiocarbon 35:215–230. 27. Dommain R, et al. (2015) Forest dynamics and tip-up pools drive pulses of high carbon 37. Reimer PJ, et al. (2013) IntCal13 and MARINE13 radiocarbon age calibration curves accumulation rates in a tropical peat dome in Borneo (Southeast Asia). J Geophys Res 0–50000 years cal BP. Radiocarbon 55:1869–1887. Biogeosci 120:617–640. 38. Ferziger JH, Peric´ M (1999) Computational Methods for Fluid Dynamics (Springer, 28. Hooijer A (2005) Hydrology of tropical forests: Recent research results from Berlin), 2nd Ed. Sarawak peatswamps. Forests, Water and People in the Humid Tropics, eds Bonell M, 39. Patankar SV (1980) Numerical Heat Transfer and Fluid Flow (Taylor & Francis, Boca Bruijnzeel LA (Cambridge Univ Press, Cambridge, UK), pp 447–461. Raton, FL). 29. Dolan TJ, Hermann AJ, Bayley SE, Zoltek J, Jr (1984) Evapotranspiration of a Florida, 40. Moerman JW, et al. (2013) Diurnal to interannual rainfall δ18O variations in northern U.S.A., freshwater wetland. J Hydrol 74:355–371. Borneo driven by regional hydrology. Earth Planet Sci Lett 369–370:108–119. 30. Baird AJ, et al. (2017) High permeability explains the vulnerability of the carbon store 41. Partin JW, Cobb KM, Adkins JF, Clark B, Fernandez DP (2007) Millennial-scale trends in in drained tropical peatlands. Geophys Res Lett 44:1333–1339. west Pacific warm pool hydrology since the Last Glacial Maximum. Nature 449:25–30. 31. Gandois L, et al. (2014) Origin, composition, and transformation of dissolved organic 42. Moore JC, Grinsted A, Zwinger T, Jevrejeva S (2013) Semiempirical and process-based matter in tropical peatlands. Geochim Cosmochim Acta 137:35–47. global sea level projections. Rev Geophys 51:484–522. 32. Miettinen J, Shi C, Liew SC (2016) Land cover distribution in the peatlands of Peninsu- 43. Roundtable on Sustainable Palm Oil (2013) Principles and criteria for the produc- lar Malaysia, Sumatra and Borneo in 2015 with changes since 1990. Glob Ecol Conserv tion of sustainable palm oil. Available at www.rspo.org/key-documents/certification/ 6:67–78. rspo-principles-and-criteria#.

E5196 | www.pnas.org/cgi/doi/10.1073/pnas.1701090114 Cobb et al. Downloaded by guest on October 2, 2021