assessing impacts of selective logging on water, energy

Biogeosciences 17 4999ndash5023 2020httpsdoiorg105194bg-17-4999-2020copy Author(s) 2020 This work is distributed underthe Creative Commons Attribution 40 License

Assessing impacts of selective logging on water energy and carbonbudgets and ecosystem dynamics in Amazon forests using theFunctionally Assembled Terrestrial Ecosystem SimulatorMaoyi Huang1 Yi Xu12 Marcos Longo34 Michael Keller345 Ryan G Knox6 Charles D Koven6 and RosieA Fisher78

1Atmospheric Sciences and Global Change Division Pacific Northwest National Laboratory Richland WA USA2School of Geography Nanjing Normal University Nanjing China3Embrapa Agricultural Informatics Campinas SP Brazil4Jet Propulsion Laboratory California Institute of Technology Pasadena CA USA5International Institute of Tropical Forestry USDA Forest Service Rio Piedras Puerto Rico USA6Earth amp Environmental Sciences Division Lawrence Berkeley National Laboratory Berkeley CA USA7Climate and Global Dynamics Laboratory National Center for Atmospheric Research Boulder CO USA8Centre Europeacuteen de Recherche et de Formation Avanceacutee en Calcul Scientifique Toulouse France

Correspondence Maoyi Huang (maoyihuangpnnlgov)

Received 9 April 2019 ndash Discussion started 23 April 2019Revised 29 June 2020 ndash Accepted 11 August 2020 ndash Published 19 October 2020

Abstract Tropical forest degradation from logging fireand fragmentation not only alters carbon stocks and car-bon fluxes but also impacts physical land surface prop-erties such as albedo and roughness length Such impactsare poorly quantified to date due to difficulties in access-ing and maintaining observational infrastructures as well asthe lack of proper modeling tools for capturing the inter-actions among biophysical properties ecosystem demogra-phy canopy structure and biogeochemical cycling in trop-ical forests As a first step to address these limitations weimplemented a selective logging module into the Function-ally Assembled Terrestrial Ecosystem Simulator (FATES) bymimicking the ecological biophysical and biogeochemicalprocesses following a logging event The model can spec-ify the timing and aerial extent of logging events splittingthe logged forest patch into disturbed and intact patches de-termine the survivorship of cohorts in the disturbed patchand modifying the biomass and necromass (total mass ofcoarse woody debris and litter) pools following logging Weparameterized the logging module to reproduce a selectivelogging experiment at the Tapajoacutes National Forest in Braziland benchmarked model outputs against available field mea-surements Our results suggest that the model permits the co-

existence of early and late successional functional types andrealistically characterizes the seasonality of water and carbonfluxes and stocks the forest structure and composition andthe ecosystem succession following disturbance Howeverthe current version of FATES overestimates water stress inthe dry season and therefore fails to capture seasonal varia-tion in latent and sensible heat fluxes Moreover we observeda bias towards low stem density and leaf area when comparedto observations suggesting that improvements are needed inboth carbon allocation and establishment of trees The ef-fects of logging were assessed by different logging scenariosto represent reduced impact and conventional logging prac-tices both with high and low logging intensities The modelsimulations suggest that in comparison to old-growth foreststhe logged forests rapidly recover water and energy fluxes in1 to 3 years In contrast the recovery times for carbon stocksforest structure and composition are more than 30 years de-pending on logging practices and intensity This study laysthe foundation to simulate land use change and forest degra-dation in FATES which will be an effective tool to directlyrepresent forest management practices and regeneration inthe context of Earth system models

Published by Copernicus Publications on behalf of the European Geosciences Union

5000 M Huang et al Assessing impacts of selective logging in tropical forests using FATES

1 Introduction

Land cover and land use in tropical forest regions are highlydynamic and nearly all tropical forests are subject to signif-icant human influence (Martiacutenez-Ramos et al 2016 Dirzoet al 2014) While old-growth tropical forests have beenreported to be carbon sinks that remove carbon dioxidefrom the atmosphere through photosynthesis these forestscould easily become carbon sources once disturbed (Luys-saert et al 2008) Using data from forest inventory andlong-term ecosystem carbon studies from 1990 to 2007 Panet al (2011) suggested a net tropical forest can be a netsource of carbon source of 13plusmn07 Pg C yrminus1 from land usechange consisting of a gross tropical deforestation loss of29plusmn05 Pg C yrminus1 that is partially offset by a carbon uptakeby tropical secondary forest regrowth of 16plusmn05 Pg C yrminus1These estimates however do not account for a tropical for-est that has been degraded through the combined effects ofselective logging (cutting and removal of merchantable tim-ber) fuelwood harvest understory fires and fragmentation(Nepstad et al 1999 Bradshaw et al 2009) To date the ef-fects of forest degradation remain poorly quantified Recentstudies suggested that degradation may contribute to carbonloss 40 as large as clear-cut deforestation (Berenguer et al2014) and the emission from selective logging alone couldbe equivalent tosim 10 to 50 of that from deforestation inthe tropical countries (Pearson et al 2014 Huang and As-ner 2010 Asner et al 2009) Selective logging of tropicalforests is an important contributor to many local and nationaleconomies and corresponds to approximately one-eighth ofglobal timber (Blaser et al 2011) The integrated impact oftimber production and other forest uses has been posited asthe cause of up tosim 30 of the difference between potentialand actual biomass stocks globally comparable in magnitudeto the effects of deforestation (Erb et al 2017) Selective log-ging includes cutting large trees and additional degradationthrough widespread damage to remaining trees subcanopyvegetation and soils (Asner et al 2004 Asner et al 2005)Selective logging accelerates gap-phase regeneration withinthe degraded forests (Huang et al 2008)

Over half of all tropical forests have been cleared orlogged and almost half of standing old-growth tropicalforests are designated by national forest services for tim-ber production (Sist et al 2015) Disturbances that resultfrom logging are known to cause forest degradation at thesame magnitude as deforestation each year in terms of bothgeographic extent and intensity with widespread collateraldamage to remaining trees vegetation and soils leading todisturbance to water energy and carbon cycling as well asecosystem integrity (Keller et al 2004b Asner et al 2004Huang and Asner 2010)

In most Earth system models (ESMs) that couple terres-trial and atmospheric processes to investigate global change(eg the Community Earth System Model or the Energy Ex-ascale Earth System Model) selective logging is typically

represented as simple fractions of affected area or an amountof carbon to be removed on a coarse grid (eg 05) Oneexception is the representation of wood harvest in the LM3Vland model that explicitly accounts for postdisturbance landage distribution as part of the Geophysical Fluid DynamicsLaboratory (GFDL) Earth system model (Shevliakova et al2009) In the ESMs grid cell fractional areas are typicallybased on timber production rates estimated from sawmillsales and export statistics (Hurtt et al 2011 Lawrence et al2012) This approach while practical does not effectivelydifferentiate selective logging that retains forest cover fromdeforestation

The realistic representation of wood harvest was absentin most ESMs because the models generally did not repre-sent the demographic structure of forests (tree size and stemnumber distributions Bonan 2008) But progress over thepast two decades in ecological theory and observations (Bus-tamante et al 2016 Strigul et al 2008 Hurtt et al 1998Moorcroft et al 2001) has made it feasible to include veg-etation demography more directly into Earth system modelsthrough individual to cohort-based vegetation in land models(Sato et al 2007 Watanabe et al 2011 Smith et al 2001Smith et al 2014 Weng et al 2015 Baidya Roy et al 2003Hurtt et al 1998 Fisher et al 2015) These vegetation de-mography modules are relatively new in land models so ef-forts are still underway to improve their parameterizations ofresource competition for light water nutrients recruitmentmortality and disturbance including both natural and anthro-pogenic components (Fisher et al 2017)

In this study we aim to (1) describe the development ofa selective logging module implemented into The Function-ally Assembled Terrestrial Ecosystem Simulator (FATES)for simulating anthropogenic disturbances of various inten-sities to forest ecosystems and their short-term and long-term effects on water energy carbon cycling and ecosystemdynamics (2) assess the capability of FATES in simulatingsite-level water energy and carbon budgets as well as for-est structure and composition (3) benchmark the simulatedvariables against available observations at the Tapajoacutes Na-tional Forest in the Amazon thus identifying potential direc-tions for model improvement and (4) assess the simulatedrecovery trajectory of tropical forest following disturbanceunder various logging scenarios In Sect 2 we provide abrief summary of FATES introduce the new selective log-ging module and describe numerical experiments performedat two sites with data from field surveys and flux towers InSect 3 FATES-simulated water energy and carbon fluxesand stocks in intact and disturbed forests are compared toavailable observations and the effects of logging practice andintensity on simulated forest recovery trajectory in terms ofcarbon budget size structure and composition in plant func-tional types are assessed Conclusions and future work arediscussed in Sect 4

Biogeosciences 17 4999ndash5023 2020 httpsdoiorg105194bg-17-4999-2020

M Huang et al Assessing impacts of selective logging in tropical forests using FATES 5001

2 Model description and study site

21 The Functionally Assembled Terrestrial EcosystemSimulator

The Functionally Assembled Terrestrial Ecosystem Simu-lator (FATES) has been developed as a numerical terres-trial ecosystem model based on the ecosystem demographyrepresentation in the community land model (CLM) for-merly known as CLM(ED) (Fisher et al 2015) FATES isan implementation of the cohort-based ecosystem demog-raphy (ED) concept (Hurtt et al 1998 Moorcroft et al2001) that can be called as a library from an ESM landsurface scheme currently including the CLM (Oleson etal 2013) or Energy Exascale Earth System Model (E3SM)land model (ELM) (httpsclimatemodelingscienceenergygovprojectsenergy-exascale-earth-system-model last ac-cess 28 June 2020) In FATES the landscape is discretizedinto spatially implicit patches each of which represents landareas with a similar age since last disturbance The discretiza-tion of ecosystems along a disturbancendashrecovery axis allowsthe deterministic simulation of successional dynamics withina typical forest ecosystem Within each patch individualsare grouped into cohorts by plant functional types (PFTs)and size classes (SCs) so that cohorts can compete for lightbased on their heights and canopy positions Following dis-turbance a patch fission process splits the original patchinto undisturbed and disturbed new patches A patch fusionmechanism is implemented to merge patches with similarstructures which helps prevent the number of patches fromgrowing too big In addition to the ED concept FATES alsoadopted a modified version of the perfect plasticity approxi-mation (PPA) (Strigul et al 2008) concept by splitting grow-ing cohorts between canopy and understory layers as a con-tinuous function of height designed for increasing the proba-bility of coexistence (Fisher et al 2010) An earlier versionof FATES CLM(ED) has been applied regionally to explorethe sensitivity of biome boundaries to plant trait representa-tion (Fisher et al 2015)

In this study we specified two plant functional types(PFTs) in FATES corresponding to early successional andlate successional plants representative of the primary axisof variability in tropical forests (Reich 2014) The earlysuccessional PFT is light demanding and grows rapidly un-der high-light conditions common prior to canopy closureThis PFT has low-density woody tissues shorter leaf androot lifetimes and a higher background mortality comparedto the late successional PFT that has dense woody tissueslonger leaf and root lifetimes and lower background mor-tality (Brokaw 1985 Whitmore 1998) and thus can surviveunder deep shade and grow slowly under closed canopy

The key parameters that differentiate the two PFTs inFATES are listed in Table 1 including specific leaf area atthe canopy top (SLA0) the maximum rate of carboxylationat 25 C (Vcmax25) specific wood density background mor-

Table 1 FATES parameters that define early and late successionalPFTs

Parameter names Units Early Latesuccessional successionalPFT PFT

Specific leaf area m2 g Cminus1 0015 0014Vcmax at 25 C micromol mminus2 sminus1 65 50Specific wood density g cmminus3 05 09Leaf longevity yr 09 26Background mortality rate yrminus1 0035 0014Leaf C N g C g Nminus1 20 40Root longevity yr 09 26

tality leaf and fine root longevity and leaf C N ratio Theparameter ranges were selected based on literature for tropi-cal forests Specifically it has been reported that SLA valuesranges from 0007 to 0039 m2 g Cminus1 (Wright et al 2004)and Vcmax25 ranges between 101 and 1057 micromol mminus2 sminus1

(Domingues et al 2005) The specific wood densities wereset to be 05 and 09 g cm3 and the background mortalityrates were set to 0035 and 0014 yrminus1 for early and late suc-cession PFTs respectively consistent with those used in theEcosystem Demography Model version 2 for Amazon forests(Longo et al 2019) For simplicity leaf longevity and rootlongevity were set to be the same for each PFT (ie 09 and26 years for early and late successional PFTs) following therange in Trumbore and Barbosa De Camargo (2009)

Given that both SLA0 and Vcmax25 span wide ranges andhave been identified as the most sensitive parameters inFATES in a previous study (Massoud et al 2019) we per-formed one-at-a-time sensitivity tests by perturbing themwithin the reported ranges Based on these tests it is evidentthat these parameters not only affect water energy and car-bon budget simulations but also the coexistence of the twoPFTs In the version of FATES used in this study (Interestedreaders are referred to the ldquoCode and data availabilityrdquo sec-tion for details) coexistence of PFTs is not assured for allparameter combinations even if they are both within reason-able ranges on account of competitive exclusion feedbackprocesses that prevent coexistence in the presence of largediscrepancies in plant growth and reproduction rates (Fisheret al 2010 Bohn et al 2011) In order to demonstrateFATESrsquo capability in simulating water energy carbon bud-gets and forest structure and composition in a holistic waywe chose to report results based on a set of parameter valuesthat produces reasonable stable fractions of two PFTs as re-ported in Table 1 Nevertheless we have included a summaryof all sensitivity tests performed in the Supplement for com-pleteness The sensitivity tests demonstrated that by tuningSLA0 and Vcmax25 for the different PFTs FATES is not onlycapable of capturing coexistence of PFTs but also capable ofreproducing observed water energy and carbon cycle fluxesin the tropics

httpsdoiorg105194bg-17-4999-2020 Biogeosciences 17 4999ndash5023 2020


Figure 1 (a) Landscape components of selective logging (b) location of the Tapajoacutes National Forest in the Amazon and (c) a typical loggingblock showing tree-fall location skid trail road and log deck coverages Panels (b) and (c) are from Asner et al (2008)

22 The selective logging module

The new selective logging module in FATES mimics the eco-logical biophysical and biogeochemical processes follow-ing a logging event The module (1) specifies the timing andareal extent of a logging event (2) calculates the fractions oftrees that are damaged by direct felling collateral damageand infrastructure damage and adds these size-specific plantmortality types to FATES (3) splits the logged patch intodisturbed and intact new patches (4) applies the calculatedsurvivorship to cohorts in the disturbed patch and (5) trans-ports harvested logs off-site by reducing site carbon poolsand adds remaining necromass to coarse woody debris andlitter pools

The logging module structure and parameterization arebased on detailed field and remote sensing studies (Putz etal 2008 Asner et al 2004 Pereira et al 2002 Asner etal 2005 Feldpausch et al 2005) Logging infrastructureincluding roads skids trails and log decks are conceptuallyrepresented (Fig 1) The construction of log decks used tostore logs prior to road transport leads to large canopy open-ings but their contribution to landscape-level gap dynamicsis small In contrast the canopy gaps caused by tree felling

are small but their coverage is spatially extensive at the land-scape scale Variations in logging practices significantly af-fect the level of disturbance to tropical forests following log-ging (Pereira et al 2002 Macpherson et al 2012 Dykstra2002 Putz et al 2008) Logging operations in the tropics areoften carried out with little planning and typically use heavymachinery to access the forests accompanied by constructionof excessive roads and skid trails leading to unnecessary treefall and compaction of the soil We refer to these typical op-erations as conventional logging (CL) In contrast reducedimpact logging (RIL) is a practice with extensive preharvestplanning where trees are inventoried and mapped out for themost efficient and cost-effective harvest and seed trees aredeliberately left on-site to facilitate faster recovery Throughplanning the construction of skid trails and roads soil com-paction and disturbance can be minimized Vines connectingtrees are cut and tree-fall directions are controlled to reducedamages to surrounding trees Reduced impact logging re-sults in consistently less disturbance to forests than conven-tional logging (Pereira et al 2002 Putz et al 2008)

The FATES logging module was designed to representa range of logging practices in field operations at a land-scape level Both CL and RIL can be represented in FATES



by specifying mortality-rate-associated direct felling collat-eral damages and mechanical damages as follows once log-ging events are activated we define three types of mortalityassociated with logging practices ndash direct-felling mortality(lmortdirect) collateral mortality (lmortcollateral) and mechan-ical mortality (lmortmechanical) The direct-felling mortalityrepresents the fraction of trees selected for harvesting that aregreater than or equal to a diameter threshold (this threshold isdefined by the diameter at breast height (DBH)= 13 m de-noted as DBHmin) collateral mortality denotes the fractionof adjacent trees that are killed by felling of the harvestedtrees and the mechanical mortality represents the fractionof trees killed by construction of log decks skid trails androads for accessing the harvested trees as well as storing andtransporting logs off-site (Fig 1a) In a logging operationthe loggers typically avoid large trees when they build logdecks skids and trails by knocking down relatively smalltrees as it is not economical to knock down large trees There-fore we implemented another DBH threshold DBHmaxinfra so that only a fraction of treesleDBHmaxinfra (called mechan-ical damage fraction) are removed for building infrastructure(Feldpausch et al 2005)

To capture the disturbance mechanisms and degree ofdamage associated with logging practices at the landscapelevel we apply the mortality types following a workflow de-signed to correspond to field operations In FATES as illus-trated in Fig 2 individual trees of all plant functional types(PFTs) in one patch are grouped into cohorts of similar-sizedtrees whose size and population sizes evolve in time throughprocesses of recruitment growth and mortality For the pur-pose of reporting and visualizing the model state these co-horts are binned into a set of 13 fixed size classes in terms ofthe diameter at the breast height (DBH) (ie 0ndash5 5ndash10 10ndash15 15ndash20 20ndash30 30ndash40 40ndash50 50ndash60 60ndash70 70ndash80 80ndash90 90ndash100 and ge 100 cm) Cohorts are further organizedinto canopy and understory layers which are subject to dif-ferent light conditions (Fig 2a) When logging activities oc-cur the canopy trees and a portion of big understory treeslose their crown coverage through direct felling for harvest-ing logs or as a result of collateral and mechanical damages(Fig 2b) The fractions of the canopy trees affected by thethree mortality mechanisms are then summed up to spec-ify the areal percentages of an old (undisturbed) and a new(disturbed) patch caused by logging in the patch fission pro-cess as discussed in Sect 21 (Fig 2c) After patch fissionthe canopy layer over the disturbed patch is removed whilethat over the undisturbed patch stays untouched (Fig 2d) Inthe undisturbed patch the survivorship of understory treesis calculated using an understory death fraction consistentwith the default value corresponding to that used for natu-ral disturbance (ie 05598) To differentiate logging fromnatural disturbance a slightly elevated logging-specific un-derstory death fraction is applied in the disturbed patch in-stead at the time of the logging event Based on data fromfield surveys over logged forest plots in the southern Ama-

Figure 2 The mortality types (direct felling mechanical and col-lateral) and patch-generating process in the FATES logging moduleThe white fraction in panels (c) (d) and (f) indicates mortality asso-ciated with other disturbances in FATES (a) Canopy and understorylayers in each cohort in FATES (b) mortality applied at the time ofa logging event (c) the patch fission process following a given log-ging event (d) canopy removal in the disturbed patch following thelogging event (e) calculation of the understory survivorship basedon the understory death fraction in each patch (d) the final states ofthe intact and disturbed patches

zon (Feldpausch et al 2005) the understory death fractioncorresponding to logging is now set to be 065 as the defaultbut can be modified via the FATES parameter file (Fig 2e)Therefore the logging operations will change the forest fromthe undisturbed state shown in Fig 2a to a disturbed state inFig 2f in the logging module It is worth mentioning thatthe newly generated patches are tracked according to the agesince disturbance and will be merged with other patches ofsimilar canopy structure following the patch fusion processesin FATES in later time steps of a simulation pending the in-clusion of separate land use fractions for managed and un-managed forest

Logging operations affect forest structure and composi-tion as well as also carbon cycling (Palace et al 2008) bymodifying the live biomass pools and flow of necromass



Figure 3 The flow of necromass following logging

(Fig 3) Following a logging event the logged trunk prod-ucts from the harvested trees are transported off-site (as anadded carbon pool for resource management in the model)while their branches enter the coarse woody debris (CWD)pool and their leaves and fine roots enter the litter pool Sim-ilarly trunks and branches of the dead trees caused by collat-eral and mechanical damages also become CWD while theirleaves and fine roots become litter Specifically the densi-ties of dead trees as a result of direct felling collateral andmechanical damages in a cohort are calculated as follows

Ddirect = lmortdirecttimesnA

Dcollateral = lmortcollateraltimesnA

Dmechanical = lmortmechanicaltimesnA

(1)

where A stands for the area of the patch being logged and n

is the number of individuals in the cohort where the mortalitytypes apply (ie as specified by the size thresholds DBHminand DBHmaxinfra ) For each cohort we denote Dindirect =

Dcollateral+Dmechanical and Dtotal =Ddirect+DindirectLeaf litter (Litterleaf kg C) and root litter (Litterroot kg C)

at the cohort level are then calculated as

Litterleaf =DtotaltimesBleaftimesA (2)Litterroot =Dtotaltimes (Broot+Bstore)timesA (3)

where Bleaf and Broot are live biomass in leaves and fineroots respectively and Bstore is stored biomass in the labilecarbon reserve in all individual trees in the cohort of interest

Following the existing CWD structure in FATES (Fisher etal 2015) CWD in the logging module is first separated intotwo categories aboveground CWD and belowground CWDWithin each category four size classes are tracked based ontheir source following Thonicke et al (2010) trunks largebranches small branches and twigs Aboveground CWDfrom trunks (CWDtrunk_agb kg C) and branches and twigs(CWDbranch_agb kg C) are calculated as follows

CWDtrunk_agb =DindirecttimesBstem_agbtimes ftrunktimesA (4)CWDbranch_agb =DtotaltimesBstem_agbtimes fbranchtimesA (5)

where Bstem_agb is the amount of aboveground stem biomassin the cohort and ftrunk and fbranch represent the fraction of

trunks and branches (eg large branches small branchesand twigs) Similarly the belowground CWD from trunks(CWDtrunk_bg kg C) and branches and twigs (CWDbranch_bgkg C) are calculated as follows

CWDtrunk_bg =DtotaltimesBroot_bgtimes ftrunktimesA (6)CWDbranch_bg =DtotaltimesBroot_bgtimes fbranchtimesA (7)

where Bcroot (kg C) is the amount of coarse root biomass inthe cohort Site-level total litter and CWD inputs can thenbe obtained by integrating the corresponding pools over allthe cohorts in the site To ensure mass conservation the totalloss of live biomass due to logging 1B (ie carbon in leaffine roots storage and structural pools) needs to be balancedwith increases in litter and CWD pools and the carbon storedin harvested logs shipped off-site as follows

1B =1Litter+1CWD+ trunk_product (8)

where 1litter and 1CWD are the increments in litter andCWD pools and trunk_product represents harvested logsshipped off-site

Following the logging event the forest structure and com-position in terms of cohort distributions as well as thelive biomass and necromass pools are updated Followingthis logging event update to forest structure the native pro-cesses simulating physiology growth and competition forresources in and between cohorts resume Since the canopylayer is removed in the disturbed patch the existing under-story trees are promoted to the canopy layer but in generalthe canopy is incompletely filled in by these newly promotedtrees and thus the canopy does not fully close Thereforemore light can penetrate and reach the understory layer inthe disturbed patch leading to increases in light-demandingspecies in the early stage of regeneration followed by a suc-cession process in which shade-tolerant species dominategradually

23 Study site and data

In this study we used data from two evergreen tropi-cal forest sites located in the Tapajoacutes National ForestBrazil (Fig 1b) These sites were established during theLarge-Scale Biosphere-Atmosphere Experiment in Amazo-nia (LBA) and are selected because of data availability in-cluding those from forest plot surveys and two flux towers es-tablished during the LBA period (Keller et al 2004a) Thesesites were named after distances along the BR-163 highwayfrom Santareacutem km67 (251prime S 5458primeW) and km83 (33prime S5456primeW) They are situated on a flat plateau and were es-tablished as a control-treatment pair for a selective loggingexperiment Tree felling operations were initiated at km83 inSeptember 2001 for a period of about 2 months Both sitesare similar with a mean annual precipitation of sim 2000 mmand mean annual temperature of 25 C on nutrient-poor clayOxisols with low organic content (Silver et al 2000)



Table 2 Distributions of stem density (N haminus1) basal area (m2 haminus1) and aboveground biomass (Kg C mminus2) before and after logging atkm83 separated by diameter of breast height (normal text) and aggregated across all sizes (bold text)

Time Before logging After logging

Variables Early Late Total Early Late Total

Stem density (N haminus1) 264 195 459 260 191 443Stem density (10ndash30 cm N haminus1) 230 169 399 229 167 396Stem density (30ndash50 cm N haminus1) 18 12 30 17 12 29Stem density (ge 50 cm N haminus1) 16 14 30 14 12 18

Basal area (m2 haminus1) 116 92 210 103 83 185Basal area (10ndash30 cm m2 haminus1) 22 17 42 22 17 38Basal area (30ndash50 cm m2 haminus1) 24 16 42 24 16 39Basal area (gt= 50 cm m2 haminus1) 70 59 126 58 51 108

AGB (Kg C mminus2) 76 89 165 68 79 147AGB (10ndash30 cm kg C mminus2) 18 20 38 18 20 38AGB (30ndash50 cm kg C mminus2) 11 11 23 11 11 22AGB (gt= 50 cm kg C mminus2) 46 58 104 38 49 87

Data in this table obtained based on inventory during the LBA period (Menton et al 2011 de Sousa et al 2011)

Prior to logging both sites were old-growth forests withlimited previous human disturbances caused by huntinggathering Brazil nuts and similar activities A compre-hensive set of meteorological variables as well as landndashatmosphere exchanges of water energy and carbon fluxeshave been measured by an eddy covariance tower at an hourlytime step over the period of 2002 to 2011 including precip-itation air temperature surface pressure relative humidityincoming shortwave and longwave radiation latent and sen-sible heat fluxes and net ecosystem exchange (NEE) (Hayeket al 2018) Another flux tower was established at km83the logged site with hourly meteorological and eddy covari-ance measurements in the period of 2000ndash2003 (Miller et al2004 Goulden et al 2004 Saleska et al 2003) The towersare listed as BR-Sa1 and BR-Sa3 in the AmeriFlux network(httpsamerifluxlblgov last access 28 June 2020)

These tower- and biometric-based observations were sum-marized to quantify logging-induced perturbations on old-growth Amazonian forests in Miller et al (2011) and are usedin this study to benchmark the model-simulated carbon bud-get Over the period of 1999 to 2001 all trees ge 35 cm inDBH in 20 ha of forest in four 1 km long transects within thekm67 footprint were inventoried as well as trees ge 10 cm inDBH on subplots with an area of sim 4 ha At km83 inventorysurveys on trees ge 55 cm in DBH were conducted in 1984and 2000 and another survey on trees gt10 cm in DBH wasconducted in 2000 (Miller et al 2004) Estimates of above-ground biomass (AGB) were then derived using allometricequations for Amazon forests (Rice et al 2004 Chambers etal 2004 Keller et al 2001) Necromass (ge 2 cm diameter)production was also measured approximately every 6 monthsin a 45-year period from November 2001 through Febru-

ary 2006 in a logged and undisturbed forest at km83 (Palaceet al 2008) Field measurements of ground disturbance interms of number of felled trees as well as areas disturbedby collateral and mechanical damages were also conductedat a similar site in Paraacute state along multitemporal sequencesof postharvest regrowth of 05ndash35 years (Asner et al 2004Pereira et al 2002)

Table 2 provides a summary of stem density and basal areadistribution across size classes at km83 based on the biomasssurvey data (Menton et al 2011 de Sousa et al 2011) Tofacilitate comparisons with simulations from FATES we di-vided the inventory into early and late succession PFTs usinga threshold of 07 g cmminus3 for specific wood density consis-tent with the definition of these PFTs in Table 1 As shownin Table 2 prior to the logging event in year 2000 this forestwas composed of 399 30 and 30 trees per hectare in sizeclasses of 10ndash30 30ndash50 and ge 50 cm respectively follow-ing logging the numbers were reduced to 396 29 and 18trees per hectare losing sim 13 of trees ge 10 cm in sizeThe changes in stem density (SD) were caused by differentmechanisms for different size classes The reduction in stemdensity of 2 haminus1 in the ge 50 cm size class was caused bytimber harvest directly while the reductions of 3 and 1 haminus1

in the 10ndash30 and 30ndash50 cm size classes were caused by col-lateral and mechanical damages Corresponding to the lossof trees in logging operations the basal area (BA) decreasedfrom 39 40 and 129 m2 haminus1 to 38 39 and 108 m2 haminus1and aboveground biomass (AGB) decreased from 38 23and 104 kg C mminus2 to 38 22 and 87 kg C mminus2 in the 10ndash30 30ndash50 and ge 50 cm size class respectively



Table 3 Cohort-level fractional damage fractions in different logging scenarios

Scenarios Conventional logging Reduced impact logging

High Low High (km83times 2) Low (km83)

Experiments CLhigh CLlow RILhigh RILlow

Direct-felling fraction (DBHge DBHmin1) 018 009 024 012

Collateral damage fraction (DBHge DBHmin) 0036 0018 0024 0012Mechanical damage fraction (DBHlt DBHmaxinfra

2) 0113 0073 0033 0024Understory death fraction3 065 065 065 065

1 DBHmin = 50 cm 2 DBHmaxinfra = 30 cm 3 Applied to the new patch generated by direct felling and collateral damage

Table 4 Comparison of energy fluxes (meanplusmn standard deviation)between eddy covariance tower measurements and FATES simula-tions

Variables LH (W mminus2) SH (W mminus2) Rn (W mminus2)

Observed (km83) 1016plusmn 80 256plusmn 52 1293plusmn 185Simulated (Intact) 876plusmn 132 394plusmn 212 1128plusmn 123

Simulated (RILlow) 873plusmn 133 396plusmn 212 1129plusmn 124Simulated (RILhigh) 870plusmn 133 398plusmn 213 1129plusmn 124Simulated (CLlow) 871plusmn 133 397plusmn 213 1128plusmn 124Simulated (CLhigh) 868plusmn 133 397plusmn 212 1129plusmn 124

24 Numerical experiments

In this study the gap-filled meteorological forcing data forthe Tapajoacutes National Forest processed by Longo et al (2019)are used to drive the CLM(FATES) model Characteristicsof the sites including soil texture vegetation cover fractionand canopy height were obtained from the LBA-Data ModelIntercomparison Project (de Gonccedilalves et al 2013) Specif-ically soil at km67 contains 90 clay and 2 sand whilesoil at km83 contains 80 clay and 18 sand Both sites arecovered by tropical evergreen forest at sim 98 within theirfootprints with the remaining 2 assumed to be covered bybare soil As discussed in Longo et al (2018) who deployedthe Ecosystem Demography Model version 2 at this site soiltexture and hence soil hydraulic parameters are highly vari-able even with the footprint of the same eddy covariancetower and they could have significant impacts on not onlywater and energy simulations but also simulated forest com-position and carbon stocks and fluxes Further generic pedo-transfer functions designed to capture temperate soils typi-cally perform poorly in clay-rich Amazonian soils (Fisher etal 2008 Tomasella and Hodnett 1998) Because we focuson introducing the FATES logging we leave for forthcomingstudies the exploration of the sensitivity of the simulations tosoil texture and other critical environmental factors

CLM(FATES) was initialized using soil texture at km83(ie 80 clay and 18 sand) from bare ground and spunup for 800 years until the carbon pools and forest struc-

ture (ie size distribution) and composition of PFTs reachedequilibrium by recycling the meteorological forcing at km67(2001ndash2011) as the sites are close enough The final statesfrom spin-up were saved as the initial condition for follow-upsimulations An intact experiment was conducted by runningthe model over a period of 2001 to 2100 without logging byrecycling the 2001ndash2011 forcing using the parameter set inTable 1 The atmospheric CO2 concentration was assumed tobe a constant of 367 ppm over the entire simulation periodconsistent with the CO2 levels during the logging treatment(Dlugokencky et al 2018)

We specified an experimental logging event in FATES on1 September 2001 (Table 3) It was reported by Figueiraet al (2008) that following the reduced-impact-loggingevent in September 2001 9 of the trees greater than orequal to DBHmin = 50 cm were harvested with an associ-ated collateral damage fraction of 0009 for trees geDBHminDBHmaxinfra is set to be 30 cm so that only a fraction of treesle 30 cm are removed for building infrastructure (Feldpauschet al 2005) This experiment is denoted as the RILlow ex-periment in Table 2 and is the one that matches the actuallogging practice at km83

We recognize that the harvest intensity in September 2001at km83 was extremely low Therefore in order to study theimpacts of different logging practices and harvest intensi-ties three additional logging experiments were conducted aslisted in Table 3 conventional logging with high intensity(CLhigh) conventional logging with low intensity (CLlow)and reduced impact logging with high intensity (RILhigh)The high-intensity logging doubled the direct-felling fractionin RILlow and CLlow as shown in the RILhigh and CLhigh ex-periments Compared to the RIL experiments the CL exper-iments feature elevated collateral and mechanical damagesas one would observe in such operations All logging experi-ments were initialized from the spun-up state using site char-acteristics at km83 previously discussed and were conductedover the period of 2001ndash2100 by recycling meteorologicalforcing from 2001 to 2011



Figure 4 Simulated energy budget terms and leaf area indices in intact and logged forests compared to observations from km67 (left) andkm83 (right) (Miller et al 2011) The dashed vertical line indicates the timing of the logging event The shaded areas in panels (a)ndash(f) areuncertainty estimates based on ulowast-filter cutoff analyses in Miller et al (2011)



3 Results and discussions

31 Simulated energy and water fluxes

Simulated monthly mean energy and water fluxes at the twosites are shown and compared to available observations inFig 4 The performances of the simulations closest to siteconditions were compared to observations and summarizedin Table 4 (ie intact for km67 and RILlow for km83) Theobserved fluxes as well as their uncertainty ranges notedas Obs67 and Obs83 from the towers were obtained fromSaleska et al (2013) consistent with those in Miller et al(2011) As shown in Table 4 the simulated mean (plusmn standarddeviation) latent heat (LH) sensible heat (SH) and net radi-ation (Rn) fluxes at km83 in RILlow over the period of 2001ndash2003 are 902plusmn 101 396plusmn 212 and 1129plusmn 124 W mminus2compared to tower-based observations of 1016plusmn80 256plusmn52 and 1293plusmn 185 W mminus2 Therefore the simulated andobserved Bowen ratios are 035 and 020 at km83 respec-tively This result suggests that at an annual time step theobserved partitioning between LH and SH is reasonablewhile the net radiation simulated by the model can be im-proved At seasonal scales even though net radiation is cap-tured by CLM(FATES) the model does not adequately par-tition sensible and latent heat fluxes This is particularly truefor sensible heat fluxes as the model simulates large sea-sonal variabilities in SH when compared to observations atthe site (ie standard deviations of monthly mean simulatedSH aresim 212 W mminus2 while observations aresim 52 W mminus2)As illustrated in Fig 4c and d the model significantly over-estimates SH in the dry season (JunendashDecember) while itslightly underestimates SH in the wet season It is worthmentioning that incomplete closure of the energy budget iscommon at eddy covariance towers (Wilson et al 2002 Fo-ken 2008) and has been reported to be sim 87 at the twosites (Saleska et al 2003)

Figure 4j shows the comparison between simulated andobserved (Goulden et al 2010) volumetric soil moisturecontent (m3 mminus3) at the top 10 cm This comparison revealsanother model structural deficiency that is even though themodel simulates higher soil moisture contents compared toobservations (a feature generally attributable to the soil mois-ture retention curve) the transpiration beta factor the down-regulating factor of transpiration from plants fluctuates sig-nificantly over a wide range and can be as low as 03 in thedry season In reality flux towers in the Amazon generallydo not show severe moisture limitations in the dry season(Fisher et al 2007) The lack of limitation is typically at-tributed to the plantrsquos ability to extract soil moisture fromdeep soil layers a phenomenon that is difficult to simulateusing a classical beta function (Baker et al 2008) and po-tentially is reconcilable using hydrodynamic representationof plant water uptake (Powell et al 2013 Christoffersenet al 2016) as in the final stages of incorporation into theFATES model Consequently the model simulates consis-

Figure 5 Simulated (a) aboveground biomass and (b) coarsewoody debris in intact and logged forests in a 1-year period be-fore or after the logging event in the four logging scenarios listedin Table 3 The observations (Obsintact and Obslogged) were derivedfrom inventory (Menton et al 2011 de Sousa et al 2011)

tently low ET during dry seasons (Fig 4e and f) while ob-servations indicate that canopies are highly productive owingto adequate water supply to support transpiration and pho-tosynthesis which could further stimulate coordinated leafgrowth with senescence during the dry season (Wu et al2016 2017)

32 Carbon budget as well as forest structure andcomposition in the intact forest

Figures 5ndash7 show simulated carbon pools and fluxes whichare tabulated in Table 5 as well As shown in Fig 5 prior tologging the simulated aboveground biomass and necromass(CWD+ litter) are 174 and 50 Mg C haminus1 compared to 165and 584 Mg C haminus1 based on permanent plot measurementsThe simulated carbon pools are generally lower than obser-vations reported in Miller et al (2011) but are within reason-able ranges as errors associated with these estimates couldbe as high as 50 due to issues related to sampling and al-lometric equations as discussed in Keller et al (2001) Thelower biomass estimates are consistent with the finding ofexcessive soil moisture stress during the dry season and lowleaf area index (LAI) in the model

Combining forest inventory and eddy covariance mea-surements Miller et al (2011) also provides estimatesfor net ecosystem exchange (NEE) gross primary pro-duction (GPP) net primary production (NPP) ecosystemrespiration (ER) heterotrophic respiration (HR) and au-



Figure 6 Simulated carbon fluxes in intact and logged forests compared to observed fluxes from km67 (left) and km83 (right) The dashedblack vertical line indicates the timing of the logging event while the dashed red horizontal line indicates estimated fluxes derived basedon eddy covariance measurements and inventory (Miller et al 2011) The shaded areas in panels (a)ndash(f) are uncertainty estimates based onulowast-filter cutoff analyses in Miller et al (2011) Panels (g)ndash(i) show comparisons between annual fluxes as only annual estimates of thesefluxes are available from Miller et al (2011)



Figure 7 Trajectories of carbon pools in intact (left) and logged (right) forests The dashed black vertical line indicates the timing of thelogging event The dashed red horizontal line indicates observed prelogging (left) and postlogging (right) inventories respectively (Mentonet al 2011 de Sousa et al 2011)

totrophic respiration (AR) As shown in Table 5 the modelsimulates reasonable values in GPP (304 Mg C haminus2 yrminus1)and ER (297 Mg C haminus2 yrminus1) when compared to val-ues estimated from the observations (326 Mg C haminus2 yrminus1

for GPP and 319 Mg C haminus2 yrminus1 for ER) in the in-tact forest However the model appears to overesti-mate NPP (135 Mg C haminus2 yrminus1 as compared to theobservation-based estimate of 95 Mg C haminus2 yrminus1) andHR (128 Mg C haminus2 yrminus1 as compared to the estimatedvalue of 89 Mg C haminus2 yrminus1) while it underestimates AR(168 Mg C haminus2 yrminus1 as compared to the observation-basedestimate of 231 Mg C haminus2 yrminus1) Nevertheless it is worthmentioning that we selected the specific parameter set to il-lustrate the capability of the model in capturing species com-

position and size structure while the performance in captur-ing carbon balance is slightly compromised given the limitednumber of sensitivity tests performed

Consistent with the carbon budget terms Table 5 lists thesimulated and observed values of stem density (haminus1) in dif-ferent size classes in terms of DBH The model simulates471 trees per hectare with DBHs greater than or equal to10 cm in the intact forest compared to 459 trees per hectarefrom the observed inventory In terms of distribution acrossthe DBH classes of 10ndash30 30ndash50 and ge50 cm 339 73and 59 N haminus1 of trees were simulated while 399 30 and30 N haminus1 were observed in the intact forest In general thisversion of FATES is able to reproduce the size structure andtree density in the tropics reasonably well In addition to



Table 5 Comparison of carbon budget terms between observation-based estimateslowast and simulations at km83

Variable Obs Simulated

Prelogging 3 years postlogging Intact Disturb level 0 yr 1 yr 3 yr 15 yr 30 yr 50 yr 70 yr

AGB 165 147 174 RILlow 156 157 159 163 167 169 173(Mg C haminus1) RILhigh 137 138 142 152 158 163 168

CLlow 154 155 157 163 167 168 164CLhigh 134 135 139 150 156 163 162

Necromass 584 744 50 RILlow 73 67 58 50 50 53 51(Mg C haminus1) RILhigh 97 84 67 48 49 52 51

CLlow 76 69 59 50 50 54 54CLhigh 101 87 68 48 49 51 54

NEE minus06plusmn 08 minus10plusmn 07 minus069 RILlow minus050 165 183 minus024 027 minus023 minus016(Mg C haminus1 yrminus1) RILhigh minus043 391 384 minus033 013 minus035 minus027

CLlow minus047 202 204 minus027 027 004 03CLhigh minus039 453 417 minus037 014 minus055 023

GPP 326plusmn 13 320plusmn 13 304 RILlow 300 295 305 300 304 301 299(Mg C haminus1 yrminus1) RILhigh 295 285 300 300 303 301 300

CLlow 297 292 303 300 304 298 300CLhigh 295 278 297 300 305 304 300

NPP 95 98 135 RILlow 135 135 140 133 136 134 132(Mg C haminus1 yrminus1) RILhigh 135 133 138 132 136 134 132

CLlow 135 135 139 132 136 132 131CLhigh 136 132 138 132 136 135 131

ER 319plusmn 17 310plusmn 16 297 RILlow 295 312 323 298 307 298 298(Mg C haminus1 yrminus1) RILhigh 292 324 339 297 304 297 297

CLlow 294 312 323 297 307 298 302CLhigh 291 324 338 297 306 299 301

HR 89 104 128 RILlow 130 152 158 13 139 132 130(Mg C haminus1 yrminus1) RILhigh 131 172 177 129 137 131 129

CLlow 130 155 160 130 139 132 134CLhigh 132 177 179 129 1377 129 134

AR 231 201 168 RILlow 165 160 166 168 168 167 167(Mg C haminus1 yrminus1) RILhigh 162 152 162 168 168 167 168

CLlow 163 157 164 168 168 166 167CLhigh 159 146 159 168 168 170 167

lowast Source of observation-based estimates Miller et al (2011) Uncertainty in carbon fluxes (GPP ER NEE) are based on ulowast-filter cutoff analyses described in the same paper

size distribution by parametrizing early and late successionalPFTs (Table 1) FATES is capable of simulating the coex-istence of the two PFTs and therefore the PFT-specific tra-jectories of stem density basal area canopy and understorymortality rates We will discuss these in Sect 34

33 Effects of logging on water energy and carbonbudgets

The response of energy and water budgets to different levelsof logging disturbances is illustrated in Table 4 and Fig 4Following the logging event the LAI is reduced proportion-ally to the logging intensities (minus9 minus17 minus14 andminus24 for RLlow RLhigh CLlow and CLhigh respectivelyin September 2001 see Fig 4h) The leaf area index recov-ers within 3 years to its prelogging level or even to slightly

higher levels as a result of the improved light environmentfollowing logging leading to changes in forest structure andcomposition (to be discussed in Sect 34) In response tothe changes in stem density and LAI discernible differencesare found in all energy budget terms For example less leafarea leads to reductions in LH (minus04 minus07 minus06 minus10 ) and increases in SH (06 10 08 and 20 )proportional to the damage levels (ie RLlow RLhigh CLlowand CLhigh) in the first 3 years following the logging eventwhen compared to the control simulation Energy budget re-sponses scale with the level of damage so that the biggestdifferences are detected in the CLhigh scenario followed byRILhigh CLlow and RILlow The difference in simulated wa-ter and energy fluxes between the RILlow (ie the scenariothat is the closest to the experimental logging event) and in-



tact cases is the smallest as the level of damage is the lowestamong all scenarios

As with LAI the water and energy fluxes recover rapidlyin 3ndash4 years following logging Miller et al (2011) com-pared observed sensible and latent heat fluxes between thecontrol (km67) and logged sites (km83) They found that inthe first 3 years following logging the between-site differ-ence (ie loggedndashcontrol) in LH reduced from 197plusmn 24 to157plusmn 10 W m2 and that in SH increased from 36plusmn 11 to54plusmn 04 W m2 When normalized by observed fluxes dur-ing the same periods at km83 these changes correspond to aminus4 reduction in LH and a 7 increase in SH comparedto the minus05 and 4 differences in LH and SH betweenRLlow and the control simulations In general both observa-tions and our modeling results suggest that the impacts ofreduced impact logging on energy fluxes are modest and thatthe energy and water fluxes can quickly recover to their prel-ogging conditions at the site

Figures 6 and 7 show the impact of logging on carbonfluxes and pools at a monthly time step and the correspond-ing annual fluxes and changes in carbon pools are summa-rized in Table 5 The logging disturbance leads to reductionsin GPP NPP AR and AGB as well as increases in ER NEEHR and CWD The impacts of logging on the carbon budgetsare also proportional to logging damage levels Specificallylogging reduces the simulated AGB from 174 Mg C haminus1 (in-tact) to 156 Mg C haminus1 (RILlow) 137 Mg C haminus1 (RILhigh)154 Mg C haminus1 (CLlow) and 134 Mg C haminus1 (CLhigh) whileit increases the simulated necromass pool (CWD+ litter)from 500 Mg C haminus1 in the intact case to 73 Mg C haminus1

(RILlow) 97 Mg C haminus1 (RILhigh) 76 Mg C haminus1 (CLlow)and 101 Mg C haminus1 (CLhigh) For the case closest to the ex-perimental logging event (RILlow) the changes in AGB andnecromass from the intact case are minus18 Mg C haminus1 (10 )and 230 Mg C haminus1 (46 ) in comparison to observedchanges ofminus22 Mg C haminus1 in AGB (12 ) and 16 Mg C haminus1

(27 ) in necromass from Miller et al (2011) respectivelyThe magnitudes and directions of these changes are reason-able when compared to observations (ie decreases in GPPER and AR following logging) On the other hand the simu-lations indicate that the forest could be turned from a carbonsink (minus069 Mg C haminus1 yrminus1) to a larger carbon source in 1ndash5 years following logging consistent with observations fromthe tower suggesting that the forest was a carbon sink or amodest carbon source (minus06plusmn 08 Mg C haminus1 yrminus1) prior tologging

The recovery trajectories following logging are also shownin Figs 6 7 and Table 5 It takes more than 70 years for AGBto return to its prelogging levels but the recovery of carbonfluxes such as GPP NPP and AR is much faster (ie within5 years following logging) The initial recovery rates of AGBfollowing logging are faster for high-intensity logging be-cause of increased light reaching the forest floor as indi-cated by the steeper slopes corresponding to the CLhigh andRILhigh scenarios compared to those of CLlow and RILlow

(Fig 9h) This finding is consistent with previous observa-tional and modeling studies (Mazzei et al 2010 Huang andAsner 2010) in that the damage level determines the num-ber of years required to recover the original AGB and theAGB accumulation rates in recently logged forests are higherthan those in intact forests For example by synthesizing datafrom 79 permanent plots at 10 sites across the Amazon basinRuttishauser et al (2016) and Piponiot et al (2018) show thatit requires 12 43 and 75 years for the forest to recover withinitial losses of 10 25 or 50 in AGB Correspondingto the changes in AGB logging introduces a large amountof necromass to the forest floor with the highest increases inthe CLhigh and RILhigh scenarios As shown in Fig 7d andTable 5 necromass and CWD pools return to the prelogginglevel in sim 15 years Meanwhile HR in RILlow stays elevatedin the 5 years following logging but converges to that fromthe intact simulation in sim 10 years which is consistent withobservations (Miller et al 2011 Table 5)

34 Effects of logging on forest structure andcomposition

The capability of the CLM(FATES) model to simulate vege-tation demographics forest structure and composition whilesimulating the water energy and carbon budgets simultane-ously (Fisher et al 2017) allows for the interrogation of themodeled impacts of alternative logging practices on the for-est size structure Table 6 shows the forest structure in termsof stem density distribution across size classes from the sim-ulations compared to observations from the site while Figs 8and 9 further break it down into early and late successionPFTs and size classes in terms of stem density and basal ar-eas As discussed in Sect 22 and summarized in Table 3the logging practices reduced impact logging and conven-tional logging differ in terms of preharvest planning and ac-tual field operation to minimize collateral and mechanicaldamages while the logging intensities (ie high and low)indicate the target direct-felling fractions The correspondingoutcomes of changes in forest structure in comparison to theintact forest as simulated by FATES are summarized in Ta-bles 6 and 7 The conventional-logging scenarios (ie CLhighand CLlow) feature more losses in small trees less than 30 cmin DBH when compared to the smaller reduction in stemdensity in size classes less than 30 cm in DBH in the reduced-impact-logging scenarios (ie RILhigh and RILlow) Scenar-ios with different logging intensities (ie high and low) re-sult in different direct-felling intensity That is the numbersof surviving large trees (DBH ge 30 cm) in RILlow and CLloware 117 haminus1 and 115 haminus1 but those in RILhigh and CLhighare 106 and 103 haminus1

In response to the improved light environment after theremoval of large trees early successional trees quickly es-tablish and populate the tree-fall gaps following logging in2ndash3 years as shown in Fig 8a Stem density in the lt10 cmsize classes is proportional to the damage levels (ie ranked



Figure 8 Changes in total stem densities and the fractions of the early successional PFT in different size classes following a single loggingevent on 1 September 2001 at km83 The dashed black vertical line indicates the timing of the logging event while the solid red line andthe dashed cyan horizontal line indicate observed prelogging and postlogging inventories respectively (Menton et al 2011 de Sousa et al2011)

as CLhighgtRILhighgtCLlowgtRILlow) followed by a transi-tion to late successional trees in later years when the canopyis closed again (Fig 8b) Such a successional process is alsoevident in Fig 9a and b in terms of basal areas The numberof early successional trees in the lt10 cm size classes thenslowly declines afterwards but is sustained throughout thesimulation as a result of natural disturbances Such a shiftin the plant community towards light-demanding species fol-lowing disturbances is consistent with observations reportedin the literature (Baraloto et al 2012 Both et al 2018) Fol-lowing regeneration in logging gaps a fraction of trees winthe competition within the 0ndash10 cm size classes and are pro-moted to the 10ndash30 cm size classes in about 10 years fol-lowing the disturbances (Figs 8d and 9d) Then a fractionof those trees subsequently enter the 30ndash50 cm size classesin 20ndash40 years following the disturbance (Figs 8f and 9f)and so on through larger size classes afterwards (Figs 8hand 9h) We note that despite the goal of achieving a deter-ministic and smooth averaging across discrete stochastic dis-

turbance events using the ecosystem demography approach(Moorcroft et al 2001) in FATES the successional processdescribed above as well as the total numbers of stems in eachsize bin shows evidence of episodic and discrete waves ofpopulation change These arise due to the required discretiza-tion of the continuous time-since-disturbance heterogeneityinto patches combined with the current maximum cap onthe number of patches in FATES (10 per site)

As discussed in Sect 24 the early successional trees havea high mortality (Fig 10a c e and g) compared to the mor-tality (Fig 10b d f and h) of late successional trees as ex-pected given their higher background mortality rate Theirmortality also fluctuates at an equilibrium level because ofthe periodic gap dynamics due to natural disturbances whilethe mortality of late successional trees remains stable Themortality rates of canopy trees (Fig 11a c e and g) remainlow and stable over the years for all size classes indicat-ing that canopy trees are not light-limited or water-stressedIn comparison the mortality rate of small understory trees



Figure 9 Changes in basal area of the two PFTs in different size classes following a single logging event on 1 September 2001 at km83The dashed black vertical line indicates the timing of the logging event while the solid red line and the dashed cyan horizontal line indicateobserved prelogging and postlogging inventories respectively (Menton et al 2011 de Sousa et al 2011) Note that for the size class 0ndash10 cmobservations are not available from the inventory

(Fig 11b) shows a declining trend following logging consis-tent with the decline in mortality of the small early succes-sional tree (Fig 10a) As the understory trees are promotedto larger size classes (Fig 11d f) their mortality rates stayhigh It is evident that it is hard for the understory trees tobe promoted to the largest size class (Fig 11h) therefore themortality cannot be calculated due to the lack of population

4 Conclusion and discussions

In this study we developed a selective logging module inFATES and parameterized the model to simulate differentlogging practices (conventional and reduced impact) withvarious intensities This newly developed selective loggingmodule is capable of mimicking the ecological biophysicaland biogeochemical processes at a landscape level follow-ing a logging event in a lumped way by (1) specifying the

timing and areal extent of a logging event (2) calculatingthe fractions of trees that are damaged by direct felling col-lateral damage and infrastructure damage and adding thesesize-specific plant mortality types to FATES (3) splitting thelogged patch into disturbed and intact new patches (4) ap-plying the calculated survivorship to cohorts in the disturbedpatch and (5) transporting harvested logs off-site and addingthe remaining necromass from damaged trees into coarsewoody debris and litter pools

We then applied FATES coupled to CLM to the TapajoacutesNational Forest by conducting numerical experiments drivenby observed meteorological forcing and we benchmarkedthe simulations against long-term ecological and eddy co-variance measurements We demonstrated that the model iscapable of simulating site-level water energy and carbonbudgets as well as forest structure and composition holis-



Figure 10 Changes in mortality (5-year running average) of the (a) early and (b) late successional trees in different size classes following asingle logging event on 1 September 2001 The dashed black vertical line indicates the timing of the logging event

tically with responses consistent with those documented inthe existing literature as follows

1 The model captures perturbations on energy and waterbudget terms in response to different levels of loggingdisturbances Our modeling results suggest that loggingleads to reductions in canopy interception canopy evap-oration and transpiration and elevated soil temperatureand soil heat fluxes in magnitudes proportional to thedamage levels

2 The logging disturbance leads to reductions in GPPNPP AR and AGB as well as increases in ER NEEHR and CWD The initial impacts of logging on thecarbon budget are also proportional to damage levels asresults of different logging practices

3 Following the logging event simulated carbon fluxessuch as GPP NPP and AR recover within 5 years but ittakes decades for AGB to return to its prelogging lev-

els Consistent with existing observation-based litera-ture initial recovery of AGB is faster when the loggingintensity is higher in response to the improved light en-vironment in the forest but the time to full AGB recov-ery in higher intensity logging is longer

4 Consistent with observations at Tapajoacutes the prescribedlogging event introduces a large amount of necromassto the forest floor proportional to the damage level ofthe logging event which returns to prelogging level insim 15 years Simulated HR in the low-damage reduced-impact-logging scenario stays elevated in the 5 yearsfollowing logging and declines to be the same as theintact forest in sim 10 years

5 The impacts of alternative logging practices on foreststructure and composition were assessed by parameter-izing cohort-specific mortality corresponding to directfelling collateral damage mechanical damage in thelogging module to represent different logging practices



Figure 11 Changes in mortality (5-year running average) of the (a) canopy and (b) understory trees in different size classes following asingle logging event on 1 September 2001 The dashed black vertical line indicates the timing of the logging event

(ie conventional logging and reduced impact logging)and intensity (ie high and low) In all scenarios theimproved light environment after the removal of largetrees facilitates the establishment and growth of earlysuccessional trees in the 0ndash10 cm DBH size class pro-portional to the damage levels in the first 2ndash3 yearsThereafter there is a transition to late successional treesin later years when the canopy is closed The number ofearly successional trees then slowly declines but is sus-tained throughout the simulation as a result of naturaldisturbances

Given that the representation of gas exchange processes isrelated to but also somewhat independent of the representa-tion of ecosystem demography FATES shows great potentialin its capability to capture ecosystem successional processesin terms of gap-phase regeneration competition among light-demanding and shade-tolerant species following disturbanceand responses of energy water and carbon budget compo-nents to disturbances The model projections suggest that

while most degraded forests rapidly recover energy fluxesthe recovery times for carbon stocks forest size structureand forest composition are much longer The recovery tra-jectories are highly dependent on logging intensity and prac-tices the difference between which can be directly simulatedby the model Consistent with field studies we find throughnumerical experiments that reduced impact logging leads tomore rapid recovery of the water energy and carbon cyclesallowing forest structure and composition to recover to theirprelogging levels in a shorter time frame

5 Future work

Currently the selective logging module can only simulatesingle logging events We also assumed that for a site such askm83 once logging is activated trees will be harvested fromall patches For regional-scale applications it will be crucialto represent forest degradation as a result of logging fire



Table 6 The simulated stem density (N haminus1) distribution at km83

Years followinglogging

Disturbancelevel

Size classes (DBH cm)

lt10 cm 10ndash30 cm 30ndash50 cm ge 50 cm

Prelogging Intact 21 799 339 73 59

0 years RILlowRILhighCLlowCLhigh

19 10117 62818 03115 996

316306299280

68656662

49414941

1 year RILlowRILhighCLlowCLhigh

22 51822 45023 67323 505

316306303279

67666663

54465446


23 69925 96025 04828 323

364368346337

68666864

50435143


21 10520 61822 88622 975

389389323348

63676166

56535755


22 97921 33223 14023 273

291288317351

82876677

62596653


22 11923 36924 80626 205

258335213320

84616072

62667658


20 59422 14319 70519 784

356326326337

58635556

64616362

and fragmentation and their combinations that could repeatover a period Therefore structural changes in FATES havebeen made by adding prognostic variables to track distur-bance histories associated with fire logging and transitionsamong land use types The model also needs to include thedead tree pool (snags and standing dead wood) as harvest op-erations (especially thinning) can lead to live tree death frommachine damage and windthrow This will be more impor-tant for using FATES in temperate coniferous systems andthe varied biogeochemical legacy of standing versus downedwood is important (Edburg et al 2011 2012) To better un-derstand how nutrient limitation or enhancement (eg viadeposition or fertilization) can affect the ecosystem dynam-ics a nutrient-enabled version of FATES is also under test-ing and will shed more lights on how biogeochemical cy-cling could impact vegetation dynamics once available Nev-

ertheless this study lays the foundation to simulate land usechange and forest degradation in FATES leading the way todirect representation of forest management practices and re-generation in Earth system models

We also acknowledge that as a model development studywe applied the model to a site using a single set of parametervalues and therefore we ignored the uncertainty associatedwith model parameters Nevertheless the sensitivity studyin the Supplement shows that the model parameters can becalibrated with a good benchmarking dataset with variousaspects of ecosystem observations For example Koven etal (2020) demonstrated a joint team effort of modelers andfield observers toward building field-based benchmarks fromBarro Colorado Island Panama and a parameter sensitivitytest platform for physiological and ecosystem dynamics us-



Table 7 The simulated basal area (m2 haminus1) distribution at km83


Disturbancelevel



Prelogging Intact 32 81 85 440


31302927

80777671

83808178

383318379317


33333130

77757468

77767674

388328388327


33343232

84858079

84828380

384324383325


31343435

94958991

76817478

401353402354


33343231

70727787

90987778

420379425381


32323433

66765371

91706898

429418454384


32333837

84797670

73785870

449427428416

ing FATES We expect to see more of such efforts to betterconstrain the model in future studies

Code and data availability CLM(FATES) has two sep-arate repositories for CLM and FATES available athttpsgithubcomESCOMPctsm (last access 25 June 2020httpsdoiorg105281zenodo3739617 CTSM DevelopmentTeam 2020) and httpsgithubcomNGEETfates (last access25 June 2020 httpsdoiorg105281zenodo3825474 FATESDevelopment Team 2020)

Site information and data at km67 and km83 can be foundat httpsitesfluxdataorgBR-Sa1 (last access 10 Februray 2020httpsdoiorg1018140FLX1440032 Saleska 2002ndash2011) and

httpsitesfluxdataorgBR-Sa13 (last access 10 February 2020httpsdoiorg1018140FLX1440033 Goulden 2000ndash2004)

A README guide to run the model and formatted datasetsused to drive model in this study will be made available fromthe open-source repository httpsgithubcomhuangmyFATES_Logging_Manuscriptgit (Huang 2020)

Supplement The supplement related to this article is available on-line at httpsdoiorg105194bg-17-4999-2020-supplement

Author contributions MH MK and ML conceived the study con-ceptualized the design of the logging module and designed the nu-merical experiments and analysis YX MH and RGK coded the



module YX RGK CDK RAF and MH integrated the module intoFATES MH performed the numerical experiments and wrote themanuscript with inputs from all coauthors

Competing interests The authors declare that they have no conflictof interest

Acknowledgements This research was supported by the Next-Generation Ecosystem ExperimentsndashTropics project through theTerrestrial Ecosystem Science (TES) program within the US De-partment of Energyrsquos Office of Biological and Environmental Re-search (BER) The Pacific Northwest National Laboratory is op-erated by the DOE and by the Battelle Memorial Institute undercontract DE-AC05-76RL01830 LBNL is managed and operatedby the Regents of the University of California under prime con-tract no DE-AC02-05CH11231 The research carried out at the JetPropulsion Laboratory California Institute of Technology was un-der a contract with the National Aeronautics and Space Administra-tion (80NM0018D004) Marcos Longo was supported by the Sa˜oPaulo State Research Foundation (FAPESP grant 201507227-6)and by the NASA Postdoctoral Program administered by the Uni-versities Space Research Association under contract with NASA(80NM0018D004) Rosie A Fisher acknowledges the National Sci-ence Foundationrsquos support of the National Center for AtmosphericResearch

Financial support This research has been supported by the USDepartment of Energy Office of Science (grant no 66705) theSatildeo Paulo State Research Foundation (grant no 201507227-6)the National Aeronautics and Space Administration (grant no80NM0018D004) and the National Science Foundation (grant no1458021)

Review statement This paper was edited by Christopher Still andreviewed by two anonymous referees

References

Asner G P Keller M Pereira J R Zweede J C and Silva JN M Canopy damage and recovery after selective logging inamazonia field and satellite studies Ecol Appl 14 280ndash298httpsdoiorg10189001-6019 2004

Asner G P Knapp D E Broadbent E N OliveiraP J C Keller M and Silva J N Selective Log-ging in the Brazilian Amazon Science 310 480ndash482httpsdoiorg101126science1118051 2005

Asner G P Keller M Pereira R and Zweed J C LBA-ECOLC-13 GIS Coverages of Logged Areas Tapajos Forest ParaBrazil 1996 1998 ORNL DAAC Oak Ridge Tennessee USAhttpsdoiorg103334ORNLDAAC893 2008

Asner G P Rudel T K Aide T M Defries R and Emer-son R A Contemporary Assessment of Change in Humid Trop-ical Forests Una Evaluacioacuten Contemporaacutenea del Cambio en

Bosques Tropicales Huacutemedos Conserv Biol 23 1386ndash1395httpsdoiorg101111j1523-1739200901333x 2009

Baidya Roy S Hurtt G C Weaver C P and Pacala SW Impact of historical land cover change on the July cli-mate of the United States J Geophys Res-Atmos 108 4793httpsdoiorg1010292003JD003565 2003

Baker I T Prihodko L Denning A S Goulden M Miller Sand Da Rocha H R Seasonal drought stress in the AmazonReconciling models and observations J Geophys Res-Biogeo113 httpsdoiorg1010292007JG000644 2008

Baraloto C Heacuterault B Paine C E T Massot H Blanc LBonal D Molino J-F Nicolini E A and Sabatier D Con-trasting taxonomic and functional responses of a tropical treecommunity to selective logging J Appl Ecol 49 861ndash870httpsdoiorg101111j1365-2664201202164x 2012

Berenguer E Ferreira J Gardner T A Aragatildeo L E O C DeCamargo P B Cerri C E Durigan M Oliveira R C DVieira I C G and Barlow J A large-scale field assessment ofcarbon stocks in human-modified tropical forests Glob ChangeBiol 20 3713ndash3726 httpsdoiorg101111gcb12627 2014

Blaser J Sarre A Poore D and Johnson S Status of TropicalForest Management 2011 International Tropical Timber Organi-zation Yokohama Japan 2011

Bohn K Dyke J G Pavlick R Reineking B Reu B and Klei-don A The relative importance of seed competition resourcecompetition and perturbations on community structure Bio-geosciences 8 1107ndash1120 httpsdoiorg105194bg-8-1107-2011 2011

Bonan G B Forests and Climate Change Forcings Feedbacksand the Climate Benefits of Forests Science 320 1444ndash1449httpsdoiorg101126science1155121 2008

Both S Riutta T Paine C E T Elias D M O CruzR S Jain A Johnson D Kritzler U H Kuntz MMajalap-Lee N Mielke N Montoya Pillco M X OstleN J Arn Teh Y Malhi Y and Burslem D F R P Log-ging and soil nutrients independently explain plant trait ex-pression in tropical forests New Phytol 221 1853ndash1865httpsdoiorg101111nph15444 2018

Bradshaw C J A Sodhi N S and Brook B W Tropical tur-moil a biodiversity tragedy in progress Front Ecol Environ 779ndash87 httpsdoiorg101890070193 2009

Brokaw N Gap-Phase Regeneration in a Tropical Forest Ecology66 682ndash687 httpsdoiorg1023071940529 1985

Bustamante M M C Roitman I Aide T M Alencar A An-derson L Aragatildeo L Asner G P Barlow J Berenguer EChambers J Costa M H Fanin T Ferreira L G FerreiraJ N Keller M Magnusson W E Morales L Morton DOmetto J P H B Palace M Peres C Silveacuterio D Trum-bore S and Vieira I C G Towards an integrated monitoringframework to assess the effects of tropical forest degradation andrecovery on carbon stocks and biodiversity Glob Change Biol22 92ndash109 httpsdoiorg101111gcb13087 2016

Chambers J Q Tribuzy E S Toledo L C Crispim B FHiguchi N Santos J d Arauacutejo A C Kruijt B Nobre AD and Trumbore S E RESPIRATION FROM A TROPICALFOREST ECOSYSTEM PARTITIONING OF SOURCES AND725 LOW CARBON USE EFFICIENCY Ecol Appl 14 72ndash88 httpsdoiorg10189001-6012 2004



Christoffersen B O Gloor M Fauset S Fyllas N M Gal-braith D R Baker T R Kruijt B Rowland L Fisher RA Binks O J Sevanto S Xu C Jansen S Choat B Men-cuccini M McDowell N G and Meir P Linking hydraulictraits to tropical forest function in a size-structured and trait-driven model (TFS v1-Hydro) Geosci Model Dev 9 4227ndash4255 httpsdoiorg105194gmd-9-4227-2016 2016

CTSM Development Team ESCOMPCTSM Update documen-tation for release-clm50 branch and fix issues with no-anthrosurface dataset creation (Version release-clm5034) Zenodohttpsdoiorg105281zenodo3779821 2020

de Gonccedilalves L G G Borak J S Costa M H Saleska S RBaker I Restrepo-Coupe N Muza M N Poulter B Ver-beeck H Fisher J B Arain M A Arkin P Cestaro B PChristoffersen B Galbraith D Guan X van den Hurk B JJ M Ichii K Imbuzeiro H M A Jain A K Levine N LuC Miguez-Macho G Roberti D R Sahoo A SakaguchiK Schaefer K Shi M Shuttleworth W J Tian H YangZ-L and Zeng X Overview of the Large-Scale BiospherendashAtmosphere Experiment in Amazonia Data Model Intercompari-son Project (LBA-DMIP) Agr Forest Meteorol 182ndash183 111ndash127 httpsdoiorg101016jagrformet201304030 2013

de Sousa C A D Elliot J R Read E L Figueira AM S Miller S D and Goulden M L LBA-ECO CD-04 Logging Damage km 83 Tower Site Tapajos NationalForest Brazil ORNL DAAC Oak Ridge Tennessee USAhttpsdoiorg103334ORNLDAAC1038 2011

Dirzo R Young H S Galetti M Ceballos G Isaac N J Band Collen B Defaunation in the Anthropocene Science 345401ndash406 httpsdoiorg101126science1251817 2014

Dlugokencky E J Hall B D Montzka S A Dutton G MuumlhleJ and Elkins J W Atmospheric composition [in State of theClimate in 2017] B Am Meteorol Soc 99 S46ndashS49 2018

Domingues T F Berry J A Martinelli L A Ometto J P HB and Ehleringer J R Parameterization of Canopy Structureand Leaf-Level Gas Exchange for an Eastern Amazonian Trop-ical Rain Forest (Tapajoacutes National Forest Paraacute Brazil) EarthInteract 9 1ndash23 httpsdoiorg101175ei1491 2005

Dykstra D P Reduced impact logging concepts and issues Ap-plying Reduced Impact Logging to Advance Sustainable ForestManagement Food and Agriculture Organization of the UnitedNations Regional Office for Asia and the Pacific Bangkok Thai-land 23ndash39 2002

Edburg S L Hicke J A Lawrence D M and Thorn-ton P E Simulating coupled carbon and nitrogen dynam-ics following mountain pine beetle outbreaks in the west-ern United States J Geophys Res-Biogeo 116 G04033httpsdoiorg1010292011JG001786 2011

Edburg S L Hicke J A Brooks P D Pendall E G EwersB E Norton U Gochis D Gutmann E D and MeddensA J H Cascading impacts of bark beetle-caused tree mortalityon coupled biogeophysical and biogeochemical processes FrontEcol Environ 10 416ndash424 2012

Erb K-H Kastner T Plutzar C Bais A L S Carval-hais N Fetzel T Gingrich S Haberl H Lauk CNiedertscheider M Pongratz J Thurner M and Luys-saert S Unexpectedly large impact of forest managementand grazing on global vegetation biomass Nature 553 73ndash76httpsdoiorg101038nature25138 2017

FATES Development Team The FunctionallyAssembled Terrestrial Ecosystem Simulator(FATES) (Version sci1355_api1100) Zenodohttpsdoiorg105281zenodo3825474 2020

Feldpausch T R Jirka S Passos C A M Jasper F and RihaS J When big trees fall Damage and carbon export by reducedimpact logging in southern Amazonia Forest Ecol Manag 219199ndash215 httpsdoiorg101016jforeco200509003 2005

Figueira A M E S Miller S D de Sousa C A D Menton MC Maia A R da Rocha H R and Goulden M L Effects ofselective logging on tropical forest tree growth J Geophys Res-Biogeo 113 G00B05 httpsdoiorg1010292007JG0005772008

Fisher R McDowell N Purves D Moorcroft P Sitch S CoxP Huntingford C Meir P and Ian Woodward F Assessinguncertainties in a second-generation dynamic vegetation modelcaused by ecological scale limitations New Phytol 187 666ndash681 httpsdoiorg101111j1469-8137201003340x 2010

Fisher R A Williams M Lola da Costa A Malhi Y da CostaR F Almeida S and Meir P The response of an EasternAmazonian rain forest to drought stress Results and modelinganalyses from a through-fall exclusion experiment Glob ChangeBiol 13 2361ndash2378 2007

Fisher R A Williams M Ruivo M L Sombroek W G and Lolada Costa A Evaluating climatic and edaphic controls of droughtstress at two Amazonian rain forest sites Agr Forest Meteorol148 850ndash861 2008

Fisher R A Muszala S Verteinstein M Lawrence P Xu CMcDowell N G Knox R G Koven C Holm J RogersB M Spessa A Lawrence D and Bonan G Taking off thetraining wheels the properties of a dynamic vegetation modelwithout climate envelopes CLM45(ED) Geosci Model Dev8 3593ndash3619 httpsdoiorg105194gmd-8-3593-2015 2015

Fisher R A Koven C D Anderegg W R L Christoffersen BO Dietze M C Farrior C E Holm J A Hurtt G C KnoxR G Lawrence P J Lichstein J W Longo M Matheny AM Medvigy D Muller-Landau H C Powell T L Serbin SP Sato H Shuman J K Smith B Trugman A T ViskariT Verbeeck H Weng E Xu C Xu X Zhang T and Moor-croft P R Vegetation demographics in Earth System Models Areview of progress and priorities Glob Change Biol 24 35ndash54httpsdoiorg101111gcb13910 2017

Foken T THE ENERGY BALANCE CLOSURE PROB-LEM AN OVERVIEW Ecol Appl 18 1351ndash1367httpsdoiorg10189006-09221 2008

Goulden M FLUXNET2015 BR-Sa3 Santarem-Km83-LoggedForest Dataset httpsdoiorg1018140FLX1440033 2000ndash2004

Goulden M L Miller S D da Rocha H R Menton MC de Freitas H C e Silva Figueira A M and de SousaC A D DIEL AND SEASONAL PATTERNS OF TROP-ICAL FOREST CO2 EXCHANGE Ecol Appl 14 42ndash54httpsdoiorg10189002-6008 2004

Goulden M L Miller S D and da Rocha H R LBA-ECOCD-04 Soil Moisture Data km 83 Tower Site Tapajos NationalForest Brazil ORNL DAAC Oak Ridge Tennessee USAhttpsdoiorg103334ORNLDAAC979 2010

Hayek M N Wehr R Longo M Hutyra L R WiedemannK Munger J W Bonal D Saleska S R Fitzjarrald D R



and Wofsy S C A novel correction for biases in forest eddycovariance carbon balance Agr Forest Meteorol 250ndash251 90ndash101 httpsdoiorg101016jagrformet201712186 2018

Huang M Script Repository for the FATES Logging ManuscriptGitHub available at httpsgithubcomhuangmyFATES_Logging_Manuscriptgit (last access 28 June 2020) 2020

Huang M and Asner G P Long-term carbon lossand recovery following selective logging in Ama-zon forests Global Biogeochem Cy 24 GB3028httpsdoiorg1010292009GB003727 2010

Huang M Asner G P Keller M and Berry J A Anecosystem model for tropical forest disturbance and se-lective logging J Geophys Res-Biogeo 113 G01002httpsdoiorg1010292007JG000438 2008

Hurtt G C Moorcroft P R And S W P and Levin S ATerrestrial models and global change challenges for the futureGlob Change Biol 4 581ndash590 httpsdoiorg101046j1365-24861998t01-1-00203x 1998

Hurtt G C Chini L P Frolking S Betts R A Feddema J Fis-cher G Fisk J P Hibbard K Houghton R A Janetos AJones C D Kindermann G Kinoshita T Klein GoldewijkK Riahi K Shevliakova E Smith S Stehfest E ThomsonA Thornton P van Vuuren D P and Wang Y P Harmo-nization of land-use scenarios for the period 1500ndash2100 600years of global gridded annual land-use transitions wood har-vest and resulting secondary lands Climatic Change 109 117ndash161 httpsdoiorg101007s10584-011-0153-2 2011

Keller M Palace M and Hurtt G Biomass estimation in theTapajos National Forest Brazil Examination of sampling andallometric uncertainties Forest Ecol Manag 154 371ndash382httpsdoiorg101016S0378-1127(01)00509-6 2001

Keller M Alencar A Asner G P Braswell B Bustamante MDavidson E Feldpausch T Fernandes E Goulden M KabatP Kruijt B Luizatildeo F Miller S Markewitz D Nobre A DNobre C A Priante Filho N da Rocha H Silva Dias P vonRandow C and Vourlitis G L ECOLOGICAL RESEARCHIN THE LARGE-SCALE BIOSPHEREndashATMOSPHERE EX-PERIMENT IN AMAZONIA EARLY RESULTS Ecol Appl14 3ndash16 httpsdoiorg10189003-6003 2004a

Keller M Palace M Asner G P Pereira R and Silva J NM Coarse woody debris in undisturbed and logged forests inthe eastern Brazilian Amazon Glob Change Biol 10 784ndash795httpsdoiorg101111j1529-8817200300770x 2004b

Koven C D Knox R G Fisher R A Chambers J Q Christof-fersen B O Davies S J Detto M Dietze M C Fay-bishenko B Holm J Huang M Kovenock M KueppersL M Lemieux G Massoud E McDowell N G Muller-Landau H C Needham J F Norby R J Powell T RogersA Serbin S P Shuman J K Swann A L S VaradharajanC Walker A P Wright S J and Xu C Benchmarking andparameter sensitivity of physiological and vegetation dynamicsusing the Functionally Assembled Terrestrial Ecosystem Simula-tor (FATES) at Barro Colorado Island Panama Biogeosciences17 3017ndash3044 httpsdoiorg105194bg-17-3017-2020 2020

Lawrence P J Feddema J J Bonan G B Meehl G A OrsquoNeillB C Oleson K W Levis S Lawrence D M KluzekE Lindsay K and Thornton P E Simulating the Biogeo-chemical and Biogeophysical Impacts of Transient Land CoverChange and Wood Harvest in the Community Climate System

Model (CCSM4) from 1850 to 2100 J Climate 25 3071ndash3095httpsdoiorg101175jcli-d-11-002561 2012

Longo M Knox R G Levine N M Alves L F Bonal D Ca-margo P B Fitzjarrald D R Hayek M N Restrepo-CoupeN Saleska S R da Silva R Stark S C Tapaj igraveos R PWiedemann K T Zhang K Wofsy S C and Moorcroft PR Ecosystem heterogeneity and diversity mitigate Amazon for-est resilience to frequent extreme droughts New Phytol 219914ndash931 httpsdoiorg101111nph 2018

Longo M Knox R G Levine N M Swann A L S MedvigyD M Dietze M C Kim Y Zhang K Bonal D BurbanB Camargo P B Hayek M N Saleska S R da Silva RBras R L Wofsy S C and Moorcroft P R The biophysicsecology and biogeochemistry of functionally diverse verticallyand horizontally heterogeneous ecosystems the Ecosystem De-mography model version 22 ndash Part 2 Model evaluation fortropical South America Geosci Model Dev 12 4347ndash4374httpsdoiorg105194gmd-12-4347-2019 2019

Luyssaert S Schulze E D Borner A Knohl A Hes-senmoller D Law B E Ciais P and Grace J Old-growth forests as global carbon sinks Nature 455 213ndash215httpsdoiorg101038nature07276 2008

Macpherson A J Carter D R Schulze M D Vidal E andLentini M W The sustainability of timber production fromEastern Amazonian forests Land Use Policy 29 339ndash350httpsdoiorg101016jlandusepol201107004 2012

Martiacutenez-Ramos M Ortiz-Rodriacuteguez I A Pintildeero D DirzoR and Sarukhaacuten J Anthropogenic disturbances jeop-ardize biodiversity conservation within tropical rainfor-est reserves P Natl Acad Sci USA 113 5323ndash5328httpsdoiorg101073pnas1602893113 2016

Massoud E C Xu C Fisher R A Knox R G Walker A PSerbin S P Christoffersen B O Holm J A Kueppers L MRicciuto D M Wei L Johnson D J Chambers J Q KovenC D McDowell N G and Vrugt J A Identification ofkey parameters controlling demographically structured vegeta-tion dynamics in a land surface model CLM45(FATES) GeosciModel Dev 12 4133ndash4164 httpsdoiorg105194gmd-12-4133-2019 2019

Mazzei L Sist P Ruschel A Putz F E MarcoP Pena W and Ferreira J E R Above-groundbiomass dynamics after reduced-impact logging in theEastern Amazon Forest Ecol Manag 259 367ndash373httpsdoiorg101016jforeco200910031 2010

Menton M C Figueira A M S de Sousa C A D MillerS D da Rocha H R and Goulden M L LBA-ECOCD-04 Biomass Survey km 83 Tower Site Tapajos NationalForest Brazil ORNL DAAC Oak Ridge Tennessee USAhttpsdoiorg103334ORNLDAAC990 2011

Miller S D Goulden M L Menton M C da Rocha H Rde Freitas H C Figueira A M e S and Dias de SousaC A BIOMETRIC AND MICROMETEOROLOGICAL MEA-SUREMENTS OF TROPICAL FOREST CARBON BAL-ANCE Ecol Appl 14 114ndash126 httpsdoiorg10189002-6005 2004

Miller S D Goulden M L Hutyra L R Keller M Saleska SR Wofsy S C Figueira A M S da Rocha H R and de Ca-margo P B Reduced impact logging minimally alters tropicalrainforest carbon and energy exchange P Natl Acad Sci USA



108 19431ndash19435 httpsdoiorg101073pnas11050681082011

Moorcroft P R Hurtt G C and Pacala S W AMETHOD FOR SCALING VEGETATION DYNAMICSTHE ECOSYSTEM DEMOGRAPHY MODEL (ED)Ecol Monogr 71 557ndash586 httpsdoiorg1018900012-9615(2001)071[0557AMFSVD]20CO2 2001

Nepstad D C Verssimo A Alencar A Nobre C Lima ELefebvre P Schlesinger P Potter C Moutinho P MendozaE Cochrane M and Brooks V Large-scale impoverishmentof Amazonian forests by logging and fire Nature 398 505ndash5081999

Oleson K W Lawrence D M Bonan G B Drewniak BHuang M Koven C D Levis S Li F Riley W J SubinZ M Swenson S C Thornton P E Bozbiyik A FisherR Kluzek E Lamarque J-F Lawrence P J Leung L RLipscomb W Muszala S Ricciuto D M Sacks W Sun YTang J and Yang Z-L Technical Description of version 45 ofthe Community Land Model (CLM) National Center for Atmo-spheric Research Boulder CONcar Technical Note NCARTN-503+STR 2013

Palace M Keller M and Silva H NECROMASSPRODUCTION STUDIES IN UNDISTURBED ANDLOGGED AMAZON FORESTS Ecol Appl 18 873ndash884httpsdoiorg10189006-20221 2008

Pan Y Birdsey R A Fang J Houghton R Kauppi P E KurzW A Phillips O L Shvidenko A Lewis S L CanadellJ G Ciais P Jackson R B Pacala S W McGuire A DPiao S Rautiainen A Sitch S and Hayes D A Large andPersistent Carbon Sink in the Worldrsquos Forests Science 333 988ndash993 2011

Pearson T Brown S and Casarim F Carbon emissions fromtropical forest degradation caused by logging Environ ResLett 9 034017 httpsdoiorg1010881748-9326930340172014

Pereira Jr R Zweede J Asner G P and Keller MForest canopy damage and recovery in reduced-impact andconventional selective logging in eastern Para Brazil For-est Ecol Manag 168 77ndash89 httpsdoiorg101016S0378-1127(01)00732-0 2002

Piponiot C Derroire G Descroix L Mazzei L RutishauserE Sist P and Heacuterault B Assessing timber vol-ume recovery after disturbance in tropical forests ndash anew modelling framework Ecol Model 384 353ndash369httpsdoiorg101016jecolmodel201805023 2018

Powell T L Galbraith D R Christoffersen B O Harper AImbuzeiro H M Rowland L Almeida S Brando P M daCosta A C L Costa M H and Levine N M Confrontingmodel predictions of carbon fluxes with measurements of Ama-zon forests subjected to experimental drought New Phytol 200350ndash365 2013

Putz F E Sist P Fredericksen T and DykstraD Reduced-impact logging Challenges and op-portunities Forest Ecol Manag 256 1427ndash1433httpsdoiorg101016jforeco200803036 2008

Reich P B The world-wide lsquofastndashslowrsquo plant economicsspectrum a traits manifesto J Ecol 102 275ndash301httpsdoiorg1011111365-274512211 2014

Rice A H Pyle E H Saleska S R Hutyra L Palace MKeller M de Camargo P B Portilho K Marques D Fand Wofsy S C CARBON BALANCE AND VEGETATIONDYNAMICS IN AN OLD-GROWTH AMAZONIAN FORESTEcol Appl 14 55ndash71 httpsdoiorg10189002-6006 2004

Saleska S FLUXNET2015 BR-Sa1 Santarem-Km67-PrimaryForest Dataset httpsdoiorg1018140FLX1440032 2002ndash2011

Saleska S R Miller S D Matross D M Goulden M LWofsy S C da Rocha H R de Camargo P B Crill PDaube B C de Freitas H C Hutyra L Keller M Kirch-hoff V Menton M Munger J W Pyle E H Rice A Hand Silva H Carbon in Amazon Forests Unexpected SeasonalFluxes and Disturbance-Induced Losses Science 302 1554ndash1557 httpsdoiorg101126science1091165 2003

Saleska S R da Rocha H R Huete A R NobreAD Artaxo P and Shimabukuro Y E LBA-ECOCD-32 Flux Tower Network Data Compilation BrazilianAmazon 1999ndash2006 Oak Ridge National Laboratory Dis-tributed Active Archive Center Oak Ridge Tennessee USAhttpsdoiorg103334ORNLDAAC1174 2013

Sato H Itoh A and Kohyama T SEIBndashDGVM A newDynamic Global Vegetation Model using a spatially ex-plicit individual-based approach Ecol Model 200 279ndash307httpsdoiorg101016jecolmodel200609006 2007

Shevliakova E Pacala S W Malyshev S Hurtt G CMilly P C D Caspersen J P Sentman L T FiskJ P Wirth C and Crevoisier C Carbon cycling un-der 300 years of land use change Importance of the sec-ondary vegetation sink Global Biogeochem Cy 23 GB2022httpsdoiorg1010292007GB003176 2009

Silver W L Neff J McGroddy M Veldkamp E KellerM and Cosme R Effects of Soil Texture on Be-lowground Carbon and Nutrient Storage in a LowlandAmazonian Forest Ecosystem Ecosystems 3 193ndash209httpsdoiorg101007s100210000019 2000

Sist P Rutishauser E Pentildea-Claros M Shenkin A HeacuteraultB Blanc L Baraloto C Baya F Benedet F da Silva KE Descroix L Ferreira J N Gourlet-Fleury S Guedes MC Bin Harun I Jalonen R Kanashiro M Krisnawati HKshatriya M Lincoln P Mazzei L Medjibeacute V Nasi RdrsquoOliveira M V N de Oliveira L C Picard N PietschS Pinard M Priyadi H Putz F E Rodney K Rossi VRoopsind A Ruschel A R Shari N H Z Rodrigues deSouza C Susanty F H Sotta E D Toledo M Vidal EWest T A P Wortel V and Yamada T The Tropical man-aged Forests Observatory a research network addressing the fu-ture of tropical logged forests Appl Veg Sci 18 171ndash174httpsdoiorg101111avsc12125 2015

Smith B Prentice I C and Sykes M T Representation ofvegetation dynamics in the modelling of terrestrial ecosys-tems comparing two contrasting approaches within Euro-pean climate space Global Ecol Biogeogr 10 621ndash637httpsdoiorg101046j1466-822X2001t01-1-00256x 2001

Smith B Waringrlind D Arneth A Hickler T Leadley P Silt-berg J and Zaehle S Implications of incorporating N cy-cling and N limitations on primary production in an individual-based dynamic vegetation model Biogeosciences 11 2027ndash2054 httpsdoiorg105194bg-11-2027-2014 2014



Strigul N Pristinski D Purves D Dushoff J and PacalaS SCALING FROM TREES TO FORESTS TRACTABLEMACROSCOPIC EQUATIONS FOR FOREST DYNAMICSEcol Monogr 78 523ndash545 httpsdoiorg10189008-008212008

Thonicke K Spessa A Prentice I C Harrison S P DongL and Carmona-Moreno C The influence of vegetation firespread and fire behaviour on biomass burning and trace gas emis-sions results from a process-based model Biogeosciences 71991ndash2011 httpsdoiorg105194bg-7-1991-2010 2010

Tomasella J and Hodnett M G Estimating soil water retentioncharacteristics from limited data in Brazilian Amazonia SoilSci 163 190ndash202 1998

Trumbore S and Barbosa De Camargo P Soil carbon dynam-ics Amazonia and global change American Geophysical UnionWashington DC 451ndash462 2009

Watanabe S Hajima T Sudo K Nagashima T Takemura TOkajima H Nozawa T Kawase H Abe M Yokohata TIse T Sato H Kato E Takata K Emori S and KawamiyaM MIROC-ESM 2010 model description and basic results ofCMIP5-20c3m experiments Geosci Model Dev 4 845ndash872httpsdoiorg105194gmd-4-845-2011 2011

Weng E S Malyshev S Lichstein J W Farrior C E Dy-bzinski R Zhang T Shevliakova E and Pacala S W Scal-ing from individual trees to forests in an Earth system mod-eling framework using a mathematically tractable model ofheight-structured competition Biogeosciences 12 2655ndash2694httpsdoiorg105194bg-12-2655-2015 2015

Whitmore T C An Introduction to Tropical Rain Forests OxfordUniversity Press Inc New York USA 1998

Wilson K Goldstein A Falge E Aubinet M BaldocchiD Berbigier P Bernhofer C Ceulemans R Dolman HField C Grelle A Ibrom A Law B E Kowalski AMeyers T Moncrieff J Monson R Oechel W Ten-hunen J Valentini R and Verma S Energy balance clo-sure at FLUXNET sites Agr Forest Meteorol 113 223ndash243httpsdoiorg101016S0168-1923(02)00109-0 2002

Wright I J Reich P B Westoby M and Ackerly D DThe worldwide leaf economics spectrum Nature 428 821ndash827httpsdoiorg101038nature02403 2004

Wu J Albert L P Lopes A P Restrepo-Coupe N Hayek MWiedemann K T Guan K Stark S C Christoffersen B Pro-haska N and Tavares J V Leaf development and demographyexplain photosynthetic seasonality in Amazon evergreen forestsScience 351 972ndash976 2016

Wu J Guan K Hayek M Restrepo-Coupe N Wiedemann KT Xu X Wehr R Christoffersen B O Miao G da SilvaR de Araujo A C Oliviera R C Camargo P B MonsonR K Huete A R and Saleska S R Partitioning controls onAmazon forest photosynthesis between environmental and bioticfactors at hourly to interannual timescales Glob Change Biol23 1240ndash1257 httpsdoiorg101111gcb13509 2017


Abstract

Introduction

Model description and study site

The Functionally Assembled Terrestrial Ecosystem Simulator

The selective logging module

Study site and data

Numerical experiments

Results and discussions

Simulated energy and water fluxes

Carbon budget as well as forest structure and composition in the intact forest

Effects of logging on water energy and carbon budgets

Effects of logging on forest structure and composition

Conclusion and discussions

Future work

Code and data availability

Supplement

Author contributions

Competing interests

Acknowledgements

Financial support

Review statement

References


1 Introduction

Land cover and land use in tropical forest regions are highlydynamic and nearly all tropical forests are subject to signif-icant human influence (Martiacutenez-Ramos et al 2016 Dirzoet al 2014) While old-growth tropical forests have beenreported to be carbon sinks that remove carbon dioxidefrom the atmosphere through photosynthesis these forestscould easily become carbon sources once disturbed (Luys-saert et al 2008) Using data from forest inventory andlong-term ecosystem carbon studies from 1990 to 2007 Panet al (2011) suggested a net tropical forest can be a netsource of carbon source of 13plusmn07 Pg C yrminus1 from land usechange consisting of a gross tropical deforestation loss of29plusmn05 Pg C yrminus1 that is partially offset by a carbon uptakeby tropical secondary forest regrowth of 16plusmn05 Pg C yrminus1These estimates however do not account for a tropical for-est that has been degraded through the combined effects ofselective logging (cutting and removal of merchantable tim-ber) fuelwood harvest understory fires and fragmentation(Nepstad et al 1999 Bradshaw et al 2009) To date the ef-fects of forest degradation remain poorly quantified Recentstudies suggested that degradation may contribute to carbonloss 40 as large as clear-cut deforestation (Berenguer et al2014) and the emission from selective logging alone couldbe equivalent tosim 10 to 50 of that from deforestation inthe tropical countries (Pearson et al 2014 Huang and As-ner 2010 Asner et al 2009) Selective logging of tropicalforests is an important contributor to many local and nationaleconomies and corresponds to approximately one-eighth ofglobal timber (Blaser et al 2011) The integrated impact oftimber production and other forest uses has been posited asthe cause of up tosim 30 of the difference between potentialand actual biomass stocks globally comparable in magnitudeto the effects of deforestation (Erb et al 2017) Selective log-ging includes cutting large trees and additional degradationthrough widespread damage to remaining trees subcanopyvegetation and soils (Asner et al 2004 Asner et al 2005)Selective logging accelerates gap-phase regeneration withinthe degraded forests (Huang et al 2008)

Over half of all tropical forests have been cleared orlogged and almost half of standing old-growth tropicalforests are designated by national forest services for tim-ber production (Sist et al 2015) Disturbances that resultfrom logging are known to cause forest degradation at thesame magnitude as deforestation each year in terms of bothgeographic extent and intensity with widespread collateraldamage to remaining trees vegetation and soils leading todisturbance to water energy and carbon cycling as well asecosystem integrity (Keller et al 2004b Asner et al 2004Huang and Asner 2010)

In most Earth system models (ESMs) that couple terres-trial and atmospheric processes to investigate global change(eg the Community Earth System Model or the Energy Ex-ascale Earth System Model) selective logging is typically

represented as simple fractions of affected area or an amountof carbon to be removed on a coarse grid (eg 05) Oneexception is the representation of wood harvest in the LM3Vland model that explicitly accounts for postdisturbance landage distribution as part of the Geophysical Fluid DynamicsLaboratory (GFDL) Earth system model (Shevliakova et al2009) In the ESMs grid cell fractional areas are typicallybased on timber production rates estimated from sawmillsales and export statistics (Hurtt et al 2011 Lawrence et al2012) This approach while practical does not effectivelydifferentiate selective logging that retains forest cover fromdeforestation

The realistic representation of wood harvest was absentin most ESMs because the models generally did not repre-sent the demographic structure of forests (tree size and stemnumber distributions Bonan 2008) But progress over thepast two decades in ecological theory and observations (Bus-tamante et al 2016 Strigul et al 2008 Hurtt et al 1998Moorcroft et al 2001) has made it feasible to include veg-etation demography more directly into Earth system modelsthrough individual to cohort-based vegetation in land models(Sato et al 2007 Watanabe et al 2011 Smith et al 2001Smith et al 2014 Weng et al 2015 Baidya Roy et al 2003Hurtt et al 1998 Fisher et al 2015) These vegetation de-mography modules are relatively new in land models so ef-forts are still underway to improve their parameterizations ofresource competition for light water nutrients recruitmentmortality and disturbance including both natural and anthro-pogenic components (Fisher et al 2017)

In this study we aim to (1) describe the development ofa selective logging module implemented into The Function-ally Assembled Terrestrial Ecosystem Simulator (FATES)for simulating anthropogenic disturbances of various inten-sities to forest ecosystems and their short-term and long-term effects on water energy carbon cycling and ecosystemdynamics (2) assess the capability of FATES in simulatingsite-level water energy and carbon budgets as well as for-est structure and composition (3) benchmark the simulatedvariables against available observations at the Tapajoacutes Na-tional Forest in the Amazon thus identifying potential direc-tions for model improvement and (4) assess the simulatedrecovery trajectory of tropical forest following disturbanceunder various logging scenarios In Sect 2 we provide abrief summary of FATES introduce the new selective log-ging module and describe numerical experiments performedat two sites with data from field surveys and flux towers InSect 3 FATES-simulated water energy and carbon fluxesand stocks in intact and disturbed forests are compared toavailable observations and the effects of logging practice andintensity on simulated forest recovery trajectory in terms ofcarbon budget size structure and composition in plant func-tional types are assessed Conclusions and future work arediscussed in Sect 4




































(1)


















































































CLlow 154 155 157 163 167 168 164CLhigh 134 135 139 150 156 163 162


CLlow 76 69 59 50 50 54 54CLhigh 101 87 68 48 49 51 54




CLlow 297 292 303 300 304 298 300CLhigh 295 278 297 300 305 304 300


CLlow 135 135 139 132 136 132 131CLhigh 136 132 138 132 136 135 131


CLlow 294 312 323 297 307 298 302CLhigh 291 324 338 297 306 299 301


CLlow 130 155 160 130 139 132 134CLhigh 132 177 179 129 1377 129 134


CLlow 163 157 164 168 168 166 167CLhigh 159 146 159 168 168 170 167
















































5 Future work






Disturbancelevel





19 10117 62818 03115 996

316306299280

68656662

49414941


22 51822 45023 67323 505

316306303279

67666663

54465446


23 69925 96025 04828 323

364368346337

68666864

50435143


21 10520 61822 88622 975

389389323348

63676166

56535755


22 97921 33223 14023 273

291288317351

82876677

62596653


22 11923 36924 80626 205

258335213320

84616072

62667658


20 59422 14319 70519 784

356326326337

58635556

64616362








Disturbancelevel





31302927

80777671

83808178

383318379317


33333130

77757468

77767674

388328388327


33343232

84858079

84828380

384324383325


31343435

94958991

76817478

401353402354


33343231

70727787

90987778

420379425381


32323433

66765371

91706898

429418454384


32333837

84797670

73785870

449427428416















References










































































































Abstract

Introduction




Study site and data








Future work


Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References



































(1)


















































































CLlow 154 155 157 163 167 168 164CLhigh 134 135 139 150 156 163 162


CLlow 76 69 59 50 50 54 54CLhigh 101 87 68 48 49 51 54




CLlow 297 292 303 300 304 298 300CLhigh 295 278 297 300 305 304 300


CLlow 135 135 139 132 136 132 131CLhigh 136 132 138 132 136 135 131


CLlow 294 312 323 297 307 298 302CLhigh 291 324 338 297 306 299 301


CLlow 130 155 160 130 139 132 134CLhigh 132 177 179 129 1377 129 134


CLlow 163 157 164 168 168 166 167CLhigh 159 146 159 168 168 170 167
















































5 Future work






Disturbancelevel





19 10117 62818 03115 996

316306299280

68656662

49414941


22 51822 45023 67323 505

316306303279

67666663

54465446


23 69925 96025 04828 323

364368346337

68666864

50435143


21 10520 61822 88622 975

389389323348

63676166

56535755


22 97921 33223 14023 273

291288317351

82876677

62596653


22 11923 36924 80626 205

258335213320

84616072

62667658


20 59422 14319 70519 784

356326326337

58635556

64616362








Disturbancelevel





31302927

80777671

83808178

383318379317


33333130

77757468

77767674

388328388327


33343232

84858079

84828380

384324383325


31343435

94958991

76817478

401353402354


33343231

70727787

90987778

420379425381


32323433

66765371

91706898

429418454384


32333837

84797670

73785870

449427428416















References










































































































Abstract

Introduction




Study site and data








Future work


Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References
































































CLlow 154 155 157 163 167 168 164CLhigh 134 135 139 150 156 163 162


CLlow 76 69 59 50 50 54 54CLhigh 101 87 68 48 49 51 54




CLlow 297 292 303 300 304 298 300CLhigh 295 278 297 300 305 304 300


CLlow 135 135 139 132 136 132 131CLhigh 136 132 138 132 136 135 131


CLlow 294 312 323 297 307 298 302CLhigh 291 324 338 297 306 299 301


CLlow 130 155 160 130 139 132 134CLhigh 132 177 179 129 1377 129 134


CLlow 163 157 164 168 168 166 167CLhigh 159 146 159 168 168 170 167
















































5 Future work






Disturbancelevel





19 10117 62818 03115 996

316306299280

68656662

49414941


22 51822 45023 67323 505

316306303279

67666663

54465446


23 69925 96025 04828 323

364368346337

68666864

50435143


21 10520 61822 88622 975

389389323348

63676166

56535755


22 97921 33223 14023 273

291288317351

82876677

62596653


22 11923 36924 80626 205

258335213320

84616072

62667658


20 59422 14319 70519 784

356326326337

58635556

64616362








Disturbancelevel





31302927

80777671

83808178

383318379317


33333130

77757468

77767674

388328388327


33343232

84858079

84828380

384324383325


31343435

94958991

76817478

401353402354


33343231

70727787

90987778

420379425381


32323433

66765371

91706898

429418454384


32333837

84797670

73785870

449427428416















References










































































































Abstract

Introduction




Study site and data








Future work


Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References










































5 Future work






Disturbancelevel





19 10117 62818 03115 996

316306299280

68656662

49414941


22 51822 45023 67323 505

316306303279

67666663

54465446


23 69925 96025 04828 323

364368346337

68666864

50435143


21 10520 61822 88622 975

389389323348

63676166

56535755


22 97921 33223 14023 273

291288317351

82876677

62596653


22 11923 36924 80626 205

258335213320

84616072

62667658


20 59422 14319 70519 784

356326326337

58635556

64616362








Disturbancelevel





31302927

80777671

83808178

383318379317


33333130

77757468

77767674

388328388327


33343232

84858079

84828380

384324383325


31343435

94958991

76817478

401353402354


33343231

70727787

90987778

420379425381


32323433

66765371

91706898

429418454384


32333837

84797670

73785870

449427428416















References










































































































Abstract

Introduction




Study site and data








Future work


Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References




Disturbancelevel





19 10117 62818 03115 996

316306299280

68656662

49414941


22 51822 45023 67323 505

316306303279

67666663

54465446


23 69925 96025 04828 323

364368346337

68666864

50435143


21 10520 61822 88622 975

389389323348

63676166

56535755


22 97921 33223 14023 273

291288317351

82876677

62596653


22 11923 36924 80626 205

258335213320

84616072

62667658


20 59422 14319 70519 784

356326326337

58635556

64616362








Disturbancelevel





31302927

80777671

83808178

383318379317


33333130

77757468

77767674

388328388327


33343232

84858079

84828380

384324383325


31343435

94958991

76817478

401353402354


33343231

70727787

90987778

420379425381


32323433

66765371

91706898

429418454384


32333837

84797670

73785870

449427428416















References










































































































Abstract

Introduction




Study site and data








Future work


Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References




Disturbancelevel





31302927

80777671

83808178

383318379317


33333130

77757468

77767674

388328388327


33343232

84858079

84828380

384324383325


31343435

94958991

76817478

401353402354


33343231

70727787

90987778

420379425381


32323433

66765371

91706898

429418454384


32333837

84797670

73785870

449427428416















References










































































































Abstract

Introduction




Study site and data








Future work


Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References







References










































































































Abstract

Introduction




Study site and data








Future work


Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References
























































































Abstract

Introduction




Study site and data








Future work


Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References

assessing impacts of selective logging on water, energy

Documents