Intact rainforest and Land-use Change in Amazônia: Scott R. Saleska Harvard University: S.C. Wofsy, L. Hutyra, A.H. Rice, B.C. Daube, D.M. Matross, E.H. Pyle, J.W. Munger University of Sao Paulo: P.B. de Camargo, H. de Rocha; INPE: V.W. Kirchhoff; U.C. Irvine: M.L. Goulden, S.D. Miller Seeing both the forest and the trees

Land-use change I (obvious): Biomass burning, conversion to agriculture

Land-use change II (subtle): Treefall destroys research site Outhouse Outhouse remnants Fallen tree

How can both kinds of changes be important?

“Changed” forest: High intensity  relatively smaller area (e.g. burning, total clearing) (1.9 million ha/year in 1990s) (Source: Brazilian Space Agency, INPE) “Intact” forest: small effect  huge area (~600 million ha)

How can both kinds of changes be important? “Changed” forest: High intensity  relatively smaller area (e.g. burning, total clearing) (1.9 million ha/year in 1990s) (Source: Brazilian Space Agency, INPE) “Intact” forest: small effect  huge area (~600 million ha)  ~0.3% per year change in what that intact forest is doing is comparable to all the land use change

Land-use change II (subtle): Treefall destroys research site Outhouse Outhouse remnants Fallen tree

Santarém Rio Amazonas Rio Tapajós Km 67 site Km 83 site (logged Sept ’01) Intact rainforest research sites in a protected reserve: the Tapajós National Forest, near Santarém, Brazil

Profile system inlets (8 levels) 58 m: Level 1 Eddy flux 47 m: Level 2 Eddy flux Part 1: Eddy flux (“the forest”) 64m tall tower: Tapajos forest, Km 67

Flux CO 2 = w ’ CO 2 ’ w’ = vertical wind CO 2 = (CO 2 + CO 2 ’) Eddy covariance m s o C Seconds CO 2 rich Cool CO 2 depleted Warm  mol CO 2 mol -1 a) w b) T c) CO 2

(A)Eddy flux of CO 2 for eddy1 (58m) and eddy2 (47m); (B)friction velocity (u*) (turbulence); (C)mean CO 2 concentration 0- 60m ("canopy storage"); (D)net ecosystem exchange (NEE = Eddy flux+ ) Initial Results: Hourly time series from the Primary Forest eddy flux tower at km 67 On windy nights (10-11 April, u*>0.2 m/s (B)) CO 2 efflux (A) is strongly positive, CO 2 storage (C) is low, and NEE (D)  flux (A). On calm nights (12-14 April), flux (A) and u* (B) are virtually zero, and CO 2 storage is high, causing NEE to be significantly higher than eddy flux. calm nights windy nights 10 April 12 April14 April Date in 2001

uptake loss to atmosphere start of dry season logging at km 83 Sum hourly fluxes to get ecosystem carbon balance over time: Summary: (1) towers agree with each other; (2) Forest ecosystem is showing net carbon loss (first in Amazonia) Biometry Cumulative NEE (loss to atmosphere), kg C ha Cumulative Net Ecosystem Exchange (NEE) and precip Average annual C-balance over two years:

Some Questions 1.Why is the forest not in carbon balance? 2.How does this flux (~ 1 Mg C/ha/yr) compare to those from land-use change?

Part 2 (“the trees”): Biometric Study of Tapajós Forest, km 67 dendrometers, 1000 trees dbh tapes, 2800 trees quantify woody debris, coarse fine, and litter

net uptake | net C emission growth/loss rate (Mg C ha -1 yr -1 ) Live Biomass ( tC/ha) mortality growth recruit Recruitment, growth, mortality, and dead wood stock directly measured live wood net gain: 1.4  0.6 MgC ha -1 yr -1 Carbon fluxes from live and dead biomass,

net uptake | net C emission growth/loss rate (Mg C ha -1 yr -1 ) Live Biomass ( tC/ha) Dead Wood (30 – 45 tC/ha) mortality growth recruit mortality decomp- osition Best estimate decomposition k (0.12  0.02 yr -1 ) based on field studies near Manaus (Chambers et al. 2001) Recruitment, growth, mortality, and dead wood stock directly measured live wood net gain: 1.4  0.6 MgC ha -1 yr -1 dead wood net loss: +3.3  1.1 Carbon fluxes from live and dead biomass,

net uptake | net C emission growth/loss rate (Mg C ha -1 yr -1 ) Live Biomass ( tC/ha) Dead Wood (30 – 45 tC/ha) whole- forest net loss: +1.9  1.0 mortality growth recruit mortality decomp- osition Best estimate decomposition k (0.12  0.02 yr -1 ) based on field studies near Manaus (Chambers et al. 2001) Recruitment, growth, mortality, and dead wood stock directly measured live wood net gain: 1.4  0.6 MgC ha -1 yr -1 dead wood net loss: +3.3  1.1 Carbon fluxes from live and dead biomass,

net uptake | net C emission growth/loss rate (Mg C ha -1 yr -1 ) Live Biomass ( tC/ha) Dead Wood (30 – 45 tC/ha) whole- forest net loss: +1.9  1.0 mortality growth recruit mortality decomp- osition Best estimate decomposition k (0.12  0.02 yr -1 ) based on field studies near Manaus (Chambers et al. 2001) Recruitment, growth, mortality, and dead wood stock directly measured Eddy Flux +1.3  0.9 Net Flux is sensitive to the decomposition rate of the dead wood, k. (Note: This forest would be losing carbon even if decomposition was as low as in a temperate forest in the southern U.S.) Biometry agrees with eddy flux live wood net gain: 1.4  0.6 MgC ha -1 yr -1 dead wood net loss: +3.3  1.1 Carbon fluxes from live and dead biomass,

Question 1: Why is this forest not in carbon balance? Three observations to consider: (1) The balance for live wood and dead wood is in opposite directions net uptake | net C emission Carbon fluxes from live and dead biomass, growth/loss rate (Mg C ha -1 yr -1 ) Live Biomass ( tC/ha) Dead Wood (30 – 45 tC/ha) whole- forest net loss: +1.9  1.0 mortality growth recruit mortality decomp- osition Best estimate decomposition k (0.12  0.02 yr -1 ) based on field studies near Manaus (Chambers et al. 2001) Recruitment, growth, mortality, and dead wood stock directly measured Eddy Flux, +1.3  0.9 live wood net gain: 1.4  0.6 MgC ha -1 yr -1 dead wood net loss: +3.3  1.1

Question 1: Why is this forest not in carbon balance? Three observations to consider: (1) The balance for live wood and dead wood is in opposite directions (2) Stock of above- ground dead wood is exceptionally large: in comparison to other sites; relative to what is needed for steady-state (  decade of mortality inputs to accumulate the excess dead wood stock) net uptake | net C emission Carbon fluxes from live and dead biomass, growth/loss rate (Mg C ha -1 yr -1 ) Live Biomass ( tC/ha) Dead Wood (30 – 45 tC/ha) whole- forest net loss: +1.9  1.0 mortality growth recruit mortality decomp- osition Best estimate decomposition k (0.12  0.02 yr -1 ) based on field studies near Manaus (Chambers et al. 2001) Recruitment, growth, mortality, and dead wood stock directly measured Eddy Flux, +1.3  0.9 live wood net gain: 1.4  0.6 MgC ha -1 yr -1 dead wood net loss: +3.3  1.1

Recruitment Growth Net = 1.5 Mg C ha -1 yr -1 Mortality Outgrowth Flux from Biomass & Changes in tree number density, by size class Observation 3: Demographic shift: (a) The increase in flux to biomass is in the smaller size classes (b) Corresponding increase in density of tree stems in the smaller size classes Question 1 (cont’d): Why is this forest not in carbon balance?  95% Conf. on each bin uptake loss to atmosphere

Recruitment Growth Net = 1.5 Mg C ha -1 yr -1 Mortality Outgrowth Recruitment Net Mortality Outgrowth size class, cm DBH density change, stems ha -1 Flux from Biomass & Changes in tree number density, by size class Question 1 (cont’d): Why is this forest not in carbon balance?  95% Conf. on each bin uptake loss to atmosphere Observation 3: Demographic shift: (a) The increase in flux to biomass is in the smaller size classes (b) Corresponding increase in density of tree stems in the smaller size classes

Hypothesis: Tapajós forest site is recovering from recent episode(s) of disturbance which: (1) Caused sharply elevated mortality preceeding onset of this study. (2) Caused a large increase in dead wood pool (to the point where losses exceed inputs) (3) Opened canopy gaps causing significant new growth and recruitment into smaller size classes of live wood (making overall growth uptake exceptionally high) Question 1: Why is this forest not in carbon balance?

Hypothesis: Tapajós forest site is recovering from recent episode(s) of disturbance which: (1) Caused sharply elevated mortality preceeding onset of this study. (2) Caused a large increase in dead wood pool (to the point where losses exceed inputs) (3) Opened canopy gaps causing significant new growth and recruitment into smaller size classes of live wood (making overall growth uptake exceptionally high) Question 1: Why is this forest not in carbon balance? Perhaps contributed to by recent El Niňo droughts? Condit et al. (1995), Williamson et al. (2000) link El Nino to elevated tree mortality

Summary: Net Carbon loss, because respiration losses from excess dead wood dominate gains from tree growth transient disturbance-recovery dynamic – possibly typical of old-growth forest, but not seen in eddy flux studies so far

Source: Moorcroft, Hurtt & Pacala (2001) Modeled flux following disturbance in gap of a “balanced-biosphere” rainforest near Manaus Gap Age (time since disturbance in years) Mg (C) ha -1 yr Tapajos Km 67 site (loss) npp = net primary productivity (uptake) rh = heterotrophic respiration (loss) nep = net ecosystem productivity (change in carbon balance)

Summary: Net Carbon loss, because respiration losses from excess dead wood dominate gains from tree growth transient disturbance-recovery dynamic – possibly typical of old-growth forest, but not seen in eddy flux studies so far Implications for Land-use change: Intact forest is dynamic: small changes in dynamics caused by human-induced changes in global climate can have effects comparable to more obvious land-use change:

Summary: Net Carbon loss, because respiration losses from excess dead wood dominate gains from tree growth transient disturbance-recovery dynamic – possibly typical of old-growth forest, but not seen in eddy flux studies so far Implications for Land-use change: Intact forest is dynamic: small changes in dynamics caused by human-induced changes in global climate can have effects comparable to more obvious land-use change: –rising CO 2 could stimulate forest, causing more carbon uptake

Summary: Net Carbon loss, because respiration losses from excess dead wood dominate gains from tree growth transient disturbance-recovery dynamic – possibly typical of old-growth forest, but not seen in eddy flux studies so far Implications for Land-use change: Intact forest is dynamic: small changes in dynamics caused by human-induced changes in global climate can have effects comparable to more obvious land-use change: –rising CO 2 could stimulate forest, causing more carbon uptake –Global warming could increase El Niño droughts, and forest disturbance, causing more carbon loss

Source: Moorcroft, Hurtt & Pacala (2001) Modeled flux following disturbance in gap of a “balanced-biosphere” rainforest near Manaus Gap Age (time since disturbance in years) Mg (C) ha -1 yr Tapajos Km 67 site (loss) npp = net primary productivity (uptake) rh = heterotrophic respiration (loss) nep = net ecosystem productivity (change in carbon balance)

Source: Moorcroft, Hurtt & Pacala (2001) Modeled flux following disturbance in gap of a “balanced-biosphere” rainforest near Manaus Gap Age (time since disturbance in years) Mg (C) ha -1 yr Tapajos Km 67 site (loss) Amount of change in nep, that, spread across the intact Amazon, would be ~twice that from deforestation: ±1 Mg ha -1 yr -1 x 600 million ha = ±0.6 Pg yr -1 ` npp = net primary productivity (uptake) rh = heterotrophic respiration (loss) nep = net ecosystem productivity (change in carbon balance)

Source: Moorcroft, Hurtt & Pacala (2001) Modeled flux following disturbance in gap of a “balanced-biosphere” rainforest near Manaus Gap Age (time since disturbance in years) Mg (C) ha -1 yr Tapajos Km 67 site (loss) Amount of change in nep, that, spread across the intact Amazon, would be ~twice that from deforestation: ±1 Mg ha -1 yr -1 x 600 million ha = ±0.6 Pg yr -1 versus 0.3 Pg yr -1 lost from Amazonian deforestation (~1.4 Pg yr -1 global deforest.) ` npp = net primary productivity (uptake) rh = heterotrophic respiration (loss) nep = net ecosystem productivity (change in carbon balance)

Summary: Net Carbon loss, because respiration losses from excess dead wood dominate gains from tree growth transient disturbance-recovery dynamic – possibly typical of old-growth forest, but not seen in eddy flux studies so far Implications for Land-use change: Intact forest is dynamic: small changes in dynamics caused by human-induced changes in global climate can have effects comparable to more obvious land-use change: –rising CO 2 could stimulate forest, causing more carbon uptake –Global warming could increase El Niño droughts, and forest disturbance, causing more carbon loss Integrated analysis of different kinds of change needed for full understanding of consequences of land-use change

Biometry plots upwind of flux tower, with locations & sizes of all trees >35cm, subset>10cm DBH Transects (from 35 cm DBH) (inset) Coarse Woody Debris Plots Legend for CWD Plots plot size of wood # area > 30cm cm cm m 2 25 m 2 1 m 2 tower (20 ha, 2800 trees, stratified) N

Questions Land-use Change: change relative to what? (i.e. what did the land do before its use was changed?) What is “change”? -forest conversion, often by fire, to agricultural or grazing lands -Logging -Global climate change

Q: Is there “lost flux”? We expect total nighttime NEE (which depends only on the physiology of forest respiration), to be essentially independent of atmospheric turbulence. Definition: Net Ecosystem Exchange: NEE = Eddy Flux + Flux out the top“Storage flux ” “Turbulence” (e.g. U* = friction velocity) Idealized Relations: NEE components, however, are expected to depend on turbulence, but in opposite directions NEE Eddy Flux “shut-off” of eddy transport at low turbulence… Storage Flux … leads to compensating build-up of CO 2 in canopy

Q: Is there “lost flux”? We expect total nighttime NEE (which depends only on the physiology of forest respiration), to be essentially independent of atmospheric turbulence. Observed Relations: Fall-off in NEE for U*<0.2 Answer: Yes, it looks like it. As U*  0, eddy flux decreases and storage flux increases as expected, but their sum (NEE) declines for U* < 0.2 m/sec:

Q: Is there “lost flux”? We expect total nighttime NEE (which depends only on the physiology of forest respiration), to be essentially independent of atmospheric turbulence. Observed Relations: Fall-off in NEE for U*<0.2 Answer: Yes, it looks like it. As U*  0, eddy flux decreases and storage flux increases as expected, but their sum (NEE) declines for U* < 0.2 m/sec: We take this as evidence of lost flux. Solution: filter out data below some U* threshold, then fill by interpolation “lost flux”

uptake loss to atmosphere uncorrected eddy flux km67 km83 Cumulative Net Ecosystem Exchange (NEE) and precip Average annual C-balance start of dry season logging at km 83 Summary: (1) with U* correction, towers agree with each other; (Part 1)