Scott Saleska (U. of Arizona) Lucy Hutyra, Elizabeth Hammond-Pyle, Dan Curran, Bill Munger, Greg Santoni, Steve Wofsy (Harvard University) Kadson Oliveira (LBA-Santarem) Simone Vieira, Plinio Camargo (CENA, U. of S ã o Paulo) From source to sink: Tracing the effects of natural disturbance on tropical forest carbon balance

How do we estimate large-scale carbon balance?

How do we estimate large-scale carbon balance (given that disturbance is intrinsic to old-growth forest dynamics)?

I.Introduction: Two sub-questions II.Prior findings from an eddy flux study: disturbance-induced carbon loss in old- growth forest? III.A preliminary test of predictions from prior findings IV.Conclusion: Understanding disturbance

Introduction: Modeled carbon flux following disturbance in gap of “balanced-biosphere” Amazon rainforest Mg (C) ha -1 yr -1 Forest Gap Age (time since disturbance in years) NPP (net primary productivity) R het (heterotrophic respiration) Positive NEP (=sink atm CO2) Negative NEP (=source to atm) NEP (net ecosystem productivity) = NPP - R het Source: Moorcroft, Hurtt, & Pacala (2001)

Mg (C) ha -1 yr -1 Forest Gap Age (time since disturbance in years) NPP (net primary productivity) R het (heterotrophic respiration) NEP (net ecosystem productivity) = NPP - R het Landscape in carbon balance, but… Individual gaps may never be in carbon balance Key point: Introduction: Modeled carbon flux following disturbance in gap of “balanced-biosphere” Amazon rainforest

What is large-scale Carbon balance? Two sub-questions: Mg (C) ha -1 yr -1 Forest Gap Age (time since disturbance in years) NPP (net primary productivity) R het (heterotrophic respiration) NEP (net ecosystem productivity) = NPP - R het (1)What is the distribution of gap ages across the landscape? (2)What are the detailed dynamics of forest response to disturbance?

What is large-scale Carbon balance? Two sub-questions: Mg (C) ha -1 yr -1 Forest Gap Age (time since disturbance in years) NPP (net primary productivity) R het (heterotrophic respiration) NEP (net ecosystem productivity) = NPP - R het (1)What is the distribution of gap ages across the landscape? (2)What are the detailed dynamics of forest response to disturbance? (see Jeff Chambers talk coming up…)

Km 67 site Km 83 site Santarém Rio Amazonas Rio Tapajós LBA project near Santarem: Controls on Carbon Balance in the Tapajós National Forest, Brazil

Profile system inlets (8 levels) 58 m: Eddy flux Methods (1) Eddy Flux: 64m tall tower at Km67 in Tapajos forest (2) Biometry: ~3000 trees and dead wood measured in 20 ha in footprint of eddy tower

Prior findings 1: eddy flux measurement show net loss of C in Tapajos National forest of Amazônia, uptake loss to atmosphere Accumulated Mg(C) ha -1 Source: Saleska et al. (2003) Science Average Annual C- Balance

Two observations suggest disturbance- recovery process: (1) The balance for live wood and dead wood is in opposite directions net C loss | net C uptake Prior findings 2: Carbon fluxes to biomass and dead wood  suggests C-loss is transient consequence of disturbance Live Biomass ( MgC/ha) Dead Wood (30–45 MgC/ha) whole- forest net: -1.9  1.0 (loss) growth/loss rate (Mg C ha -1 yr -1 ) mortality decomp- osition live wood change: +1.4  0.6 MgC ha -1 yr -1 dead wood change: -3.3  1.1 (loss)

uptake loss to atmosphere size class, cm DBH flux, MgC ha -1 yr -1 Recruitment Growth Net Accumulation (1.4 Mg C ha -1 yr -1 ) Mortality Outgrowth Observation 2: Demographic shift: The increase in flux to biomass is in the smaller size classes Corresponding increase in density of tree stems in the smaller size classes Prior findings 2: Carbon fluxes to biomass by size-class  suggests C-loss is transient consequence of disturbance

Mg (C) ha -1 yr -1 Forest Gap Age (time since disturbance in years) NPP (net primary productivity) R het (heterotrophic respiration) NEP (net ecosystem productivity) = NPP - R het (1) What is the distribution of gap ages across the landscape? (2) What are the detailed dynamics of forest response to disturbance? Predictions for continued observations following disturbance

Mg (C) ha -1 yr -1 Forest Gap Age (time since disturbance in years) NPP (net primary productivity) R het (heterotrophic respiration) NEP (net ecosystem productivity) = NPP - R het (1) What is the distribution of gap ages across the landscape? (2) What are the detailed dynamics of forest response to disturbance? Tapajos Km 67 site? (loss) Predictions for continued observations following disturbance

Mg (C) ha -1 yr -1 Forest Gap Age (time since disturbance in years) NPP (net primary productivity) R het (heterotrophic respiration) NEP (net ecosystem productivity) = NPP - R het Tapajos Km 67 site? (loss) Predictions for continued observations following disturbance

Mg (C) ha yr -1 (1) Whole-Ecosystem flux Predictions (a) NEP should move from source to sink R het (respiration) NEP (net ecosystem productivity) = NPP - R het NPP Positive NEP (=sink) Negative NEP (=source) Predictions for continued observations following disturbance

Mg (C) ha yr -1 (1) Whole-Ecosystem flux Predictions (a) NEP should move from source to sink (b) Due to decline in respiration (as deadwood substrate decays away) R het (respiration) NEP (net ecosystem productivity) = NPP - R het NPP Positive NEP (=sink) Negative NEP (=source) Predictions for continued observations following disturbance

net C loss | net C uptake Predictions for continued observations following disturbance Live Biomass Dead Wood growth/loss rate Mort. growth recruit Mort. De- comp. prediction (2) Forest Demography Predictions (a) Shift from uptake towards balance in live biomass (i.e. decrease in growth/ recruitment, increase in mortality) (a)

net C loss | net C uptake Predictions for continued observations following disturbance Live Biomass Dead Wood growth/loss rate Mort. growth recruit Mort. De- comp. prediction (2) Forest Demography Predictions (a) Shift from uptake towards balance in live biomass (i.e. decrease in growth/ recruitment, increase in mortality) (b) Shift from loss towards balance in dead wood (a)(b)

net C loss | net C uptake Predictions for continued observations following disturbance Live Biomass Dead Wood growth/loss rate Mort. growth recruit Mort. De- comp. prediction (2) Forest Demography Predictions (a) Shift from uptake towards balance in live biomass (i.e. decrease in growth/ recruitment, increase in mortality) (b) Shift from loss towards balance in dead wood (c) Shift in tree growth from smaller to middle size classes Tree Size-class (a)(b) (c)

III. Test of predictions with new observations

Test of predictions Cumulative km67 Carbon Flux, Mg (C) ha -1 uptake loss to atmosphere Jan Jul source

Test of predictions Cumulative km67 Carbon Flux, Mg (C) ha -1 uptake loss to atmosphere Jan Jul sourcesink

Day of Year Cumulative Mg (C) ha -1 uptake loss to atmosphere Test of predictions Cumulative km67 Carbon Flux,

Test of predictions: Trends in Annual sums of carbon flux Component fluxes (MgC/ha/yr) Resp (declining at 0.9 MgC/ha/yr 2 ) GPP (increasing at 0.6 MgC/ha/yr 2 ) Net Exchange (loss) (= Resp – GPP) (MgC/ha/yr)

Precip controls short-term, but not long-term, integrated Resp flux Respiration (MgC/ha/yr) Monthly resp vs precip Annual resp vs precip Monthly precip (mm/month)Annual precip (mm/year) Resp = (±.003)·Precip (R 2 = 0.5)

Test of predictions: Forest Demography (a): Live biomass pool shifts toward balance Recruitment Growth Mortality Net Uptake MgC / ha / yr Net Uptake Net Uptake (near zero) Carbon fluxes to Live Biomass at km 67 net C loss | net C uptake

Test of predictions: Forest Demography (c): “Grow-in” shifts tree growth from smaller to larger size-classes Size-class (cm DBH) annual growth increment (MgC) by size class MgC/ha/year size class, cm DBH MgC growth/ MgC live

Summary Preliminary pattern confirms disturbance recovery hypothesis: (1) for eddy flux predictions: –Net ecosystem exchange shifts from source towards sink –Due in part to decline in ecosystem respiration (2) for demographic predictions: -Live biomass pool shifts towards balance -Grow-in shifts growth fluxes towards larger size classes

IV. Conclusions (1)Multiple measures (demography, component fluxes, net C-balance) can be integrated to detect and track whole-forest response to disturbance (2)Detailed quantification of: - Flux magnitudes & Response times - demographic state allows tests of models of important forest dynamics (3)We are making progress on understanding key factors for predicting large-scale forest carbon balance

Rolim et al. (2005) Aboveground Biomass timeseries in 0.5-ha plots of Linhares Atlantic rainforest Large-impact ENSO events