Commenton Tropical forests are a ... · forests experiencing degradation (i.e., sites spread...
5
TECHNICAL COMMENT ◥ CARBON CYCLE Comment on “Tropical forests are a net carbon source based on aboveground measurements of gain and loss” Matthew C. Hansen*, Peter Potapov, Alexandra Tyukavina Baccini et al. (Reports, 13 October 2017, p. 230) report MODIS-derived pantropical forest carbon change, with spatial patterns of carbon loss that do not correspond to higher-resolution Landsat-derived tree cover loss.The assumption that map results are unbiased and free of commission and omission errors is not supported. The application of passive moderate-resolution optical data to monitor forest carbon change overstates our current capabilities. B accini et al.(1) report net tropical forest aboveground carbon stock change from Moderate Resolution Imaging Spectro- radiometer (MODIS) data and purport to capture all forest carbon dynamics result- ing from both natural and anthropogenic pro- cesses. We believe their method and results overstate current monitoring capabilities and may confuse the global community of practi- tioners working to establish robust and defen- sible forest carbon monitoring systems. Our methods for mapping tree cover (2, 3) are very similar to those of Baccini et al. for mapping aboveground carbon; we use the same passive optical data (they use a simpler set of annual composite images that do not cap- ture phenological variation) and the same data- mining tool (regression trees in a random forest implementation). The approaches differ princi- pally in the dependent variable. Our research characterizing global tree cover uses MODIS and Landsat passive optical sensing systems. Land cover, including tree cover, is directly sensed in passive optical Earth observations because the multitemporal, multispectral re- flectance variations directly relate to biophysical surface properties and vegetation types. Forest carbon is a more challenging variable to char- acterize, requiring knowledge about the volume and carbon density of vegetation—parameters that are not directly related to surface reflec- tance. Although the empirical model built by Baccini et al. suggests a relationship between passive optical imagery and forest carbon, it is likely the very same signal we use in mapping tree cover. Tree cover is largely discriminated using a spectral signature that is a combination of high reflectance in the near-infrared, asso- ciated with dense vegetation cover, and low reflectance in other bands due to light extinc- tion from canopy shadowing and absorption from healthy vegetation cover. There is no physical basis for passive optical reflectance data being sensitive to changes in forest carbon parameters such as diameter at breast height or stem wood density. Land change maps derived from satellite data consist of accurately mapped change along with inevitable omission and commission errors, whether the mapped variables are categorical or continuous. For data derived from MODIS, accurate mapping of land change is limited by its coarse spatial resolution (~500 m), as sub- stantial areas of subpixel land change are not reliably quantified. For example, Hansen et al. (4) showed that area estimation of stand- replacement tree cover loss from Landsat sam- ples at 30-m spatial resolution was twice that of a MODIS-derived tree cover loss map product. If MODIS data are forced to reproduce all of the area detected by Landsat, change maps quickly turn into noise, or more correctly, unconstrained omission and commission errors. The use of independent probability-based reference data to assess error rates and bias is standard prac- tice and has been recommended by the IPCC (5–8). Baccini et al. attempt to track changes in forest carbon stocks directly with propagated model uncertainties presented in place of ac- curacy assessments. Admittedly, an accuracy assessment of forest carbon change is difficult and would need to consist of a probability- based sample of time 1 and time 2 carbon stock measurements at the scale of observation. Re- gardless, in a modeling approach like that of Baccini et al., map biases and associated omis- sion and commission errors are undetermined. This is critical, as the mapping challenge is greater for Baccini et al., given a dependent variable (aboveground carbon change) that has a much more tenuous relation to the MODIS- derived independent spectral reflectance varia- bles. Baccini et al. state that their map is free from bias because they use a “single predictive model” applied to carbon stock gain and loss estimation. However, the distribution of map commission and omission errors is a function not only of model choice, but also of spatio- temporal variations of forest change, sensitivity RESEARCH Hansen et al, Science 363, eaar3629 (2019) 11 January 2019 1 of 4 Department of Geographical Sciences, University of Maryland, College Park, MD 20740, USA. *Corresponding author. Email: [email protected] Table 1. Disaggregation of Baccini et al. aboveground loss by percent tree cover, tree cover loss, forest intactness, and colocated tree cover loss and gain. Baccini et al.(1) aboveground loss Tropical America Tropical Africa Tropical Asia By percent tree cover category of Hansen et al.(2) ≥60% tree cover 86.1% 40.9% 74.1% ............................................................................................................................................................................................................. 30–59% tree cover 7.9% 24.4% 15.0% ............................................................................................................................................................................................................. 10–29% tree cover 3.4% 30.4% 6.8% ............................................................................................................................................................................................................. 0–9% tree cover 2.6% 4.2% 4.1% ............................................................................................................................................................................................................................................................................................................................................ Within Hansen et al.(2) mapped tree cover loss ≥60% tree cover 27.1% 13.4% 47.9% ............................................................................................................................................................................................................. 30–59% tree cover 3.0% 8.9% 6.6% ............................................................................................................................................................................................................. 10–29% tree cover 0.9% 6.1% 2.0% ............................................................................................................................................................................................................. 0–9% tree cover 0.2% 0.0% 0.4% ............................................................................................................................................................................................................................................................................................................................................ Outside of Hansen et al.(2) mapped tree cover loss 68.7% 71.5% 43.2% ............................................................................................................................................................................................................................................................................................................................................ Within an intersection of intact (9) and hinterland (10) forest maps 25.3% 7.9% 2.5% ............................................................................................................................................................................................................................................................................................................................................ Within hinterland forests (10) 28.9% 10.5% 3.7% ............................................................................................................................................................................................................................................................................................................................................ For colocated tree cover loss and tree cover gain from Hansen et al.(2) 0.65% 0.00% 5.45% ............................................................................................................................................................................................................................................................................................................................................ on June 16, 2020 http://science.sciencemag.org/ Downloaded from
Transcript of Commenton Tropical forests are a ... · forests experiencing degradation (i.e., sites spread...
TECHNICAL COMMENT◥
CARBON CYCLE
Comment on “Tropical forests are anet carbon source based on abovegroundmeasurements of gain and loss”Matthew C. Hansen*, Peter Potapov, Alexandra Tyukavina
Baccini et al. (Reports, 13 October 2017, p. 230) report MODIS-derived pantropicalforest carbon change, with spatial patterns of carbon loss that do not correspond tohigher-resolution Landsat-derived tree cover loss. The assumption that map results areunbiased and free of commission and omission errors is not supported. The application ofpassive moderate-resolution optical data to monitor forest carbon change overstates ourcurrent capabilities.
Baccini et al. (1) report net tropical forestaboveground carbon stock change fromModerate Resolution Imaging Spectro-radiometer (MODIS) data and purport tocapture all forest carbon dynamics result-
ing from both natural and anthropogenic pro-cesses. We believe their method and resultsoverstate current monitoring capabilities andmay confuse the global community of practi-tioners working to establish robust and defen-sible forest carbon monitoring systems.Our methods for mapping tree cover (2, 3)
are very similar to those of Baccini et al. formapping aboveground carbon; we use the samepassive optical data (they use a simpler setof annual composite images that do not cap-ture phenological variation) and the same data-mining tool (regression trees in a random forestimplementation). The approaches differ princi-
pally in the dependent variable. Our researchcharacterizing global tree cover uses MODISand Landsat passive optical sensing systems.Land cover, including tree cover, is directlysensed in passive optical Earth observationsbecause the multitemporal, multispectral re-flectance variations directly relate to biophysicalsurface properties and vegetation types. Forestcarbon is a more challenging variable to char-acterize, requiring knowledge about the volumeand carbon density of vegetation—parametersthat are not directly related to surface reflec-tance. Although the empirical model built byBaccini et al. suggests a relationship betweenpassive optical imagery and forest carbon, it islikely the very same signal we use in mappingtree cover. Tree cover is largely discriminatedusing a spectral signature that is a combinationof high reflectance in the near-infrared, asso-ciated with dense vegetation cover, and lowreflectance in other bands due to light extinc-tion from canopy shadowing and absorptionfrom healthy vegetation cover. There is nophysical basis for passive optical reflectance
data being sensitive to changes in forest carbonparameters such as diameter at breast heightor stem wood density.Land change maps derived from satellite data
consist of accurately mapped change along withinevitable omission and commission errors,whether the mapped variables are categoricalor continuous. For data derived from MODIS,accurate mapping of land change is limited byits coarse spatial resolution (~500 m), as sub-stantial areas of subpixel land change are notreliably quantified. For example, Hansen et al.(4) showed that area estimation of stand-replacement tree cover loss from Landsat sam-ples at 30-m spatial resolution was twice thatof a MODIS-derived tree cover loss map product.If MODIS data are forced to reproduce all of thearea detected by Landsat, change maps quicklyturn into noise, or more correctly, unconstrainedomission and commission errors. The use ofindependent probability-based reference datato assess error rates and bias is standard prac-tice and has been recommended by the IPCC(5–8).Baccini et al. attempt to track changes in
forest carbon stocks directly with propagatedmodel uncertainties presented in place of ac-curacy assessments. Admittedly, an accuracyassessment of forest carbon change is difficultand would need to consist of a probability-based sample of time 1 and time 2 carbon stockmeasurements at the scale of observation. Re-gardless, in a modeling approach like that ofBaccini et al., map biases and associated omis-sion and commission errors are undetermined.This is critical, as the mapping challenge isgreater for Baccini et al., given a dependentvariable (aboveground carbon change) that hasa much more tenuous relation to the MODIS-derived independent spectral reflectance varia-bles. Baccini et al. state that their map is freefrom bias because they use a “single predictivemodel” applied to carbon stock gain and lossestimation. However, the distribution of mapcommission and omission errors is a functionnot only of model choice, but also of spatio-temporal variations of forest change, sensitivity
RESEARCH
Hansen et al, Science 363, eaar3629 (2019) 11 January 2019 1 of 4
Department of Geographical Sciences, University ofMaryland, College Park, MD 20740, USA.*Corresponding author. Email: [email protected]
Table 1. Disaggregation of Baccini et al. aboveground loss by percent tree cover, tree cover loss, forest intactness, and colocated tree cover loss and gain.
Baccini et al. (1) aboveground loss Tropical America Tropical Africa Tropical Asia
By percent tree cover category of Hansen et al. (2) ≥60% tree cover 86.1% 40.9% 74.1%.. .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. .
30–59% tree cover 7.9% 24.4% 15.0%.. .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. .
10–29% tree cover 3.4% 30.4% 6.8%.. .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. .
0–9% tree cover 2.6% 4.2% 4.1%.. .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .
Within Hansen et al. (2) mapped tree cover loss ≥60% tree cover 27.1% 13.4% 47.9%.. .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. .
30–59% tree cover 3.0% 8.9% 6.6%.. .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. .
10–29% tree cover 0.9% 6.1% 2.0%.. .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. .
0–9% tree cover 0.2% 0.0% 0.4%.. .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .
Outside of Hansen et al. (2) mapped tree cover loss 68.7% 71.5% 43.2%.. .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .
Within an intersection of intact (9) and hinterland (10) forest maps 25.3% 7.9% 2.5%.. .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .
Within hinterland forests (10) 28.9% 10.5% 3.7%.. .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .
For colocated tree cover loss and tree cover gain from Hansen et al. (2) 0.65% 0.00% 5.45%.. .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .. ... ... .. ... .. ... ... .. ... ... .
on June 16, 2020
http://science.sciencemag.org/
Dow
nloaded from
to different land cover dynamics, data quality,and optical reflectance relationships with above-ground carbon.We compared our tree cover loss data with
Baccini et al.’s carbon loss estimates for theperiod 2003–2014 to assess map correspon-dence (Table 1 and Figs. 1 to 3). We found that69%, 72%, and 43% of estimated carbon lossesfor Latin America, Africa, and Southeast Asia,respectively, are not colocated with Landsat-scale forest disturbances, defined as any MODISgrid cell with ≥10% Landsat-derived tree coverloss. The carbon losses reported by Baccini et al.must therefore be due to factors other than
observable land cover and land use forest dis-turbances. Unfortunately, there is no evidencein their paper or in the literature to explainsuch large-scale biomass change dynamics forregions without tree cover loss. Other anom-alous results include more than 25% of carbonloss in South America being located within in-tact forest landscapes with no observable humanor natural disturbance (9, 10) and more than30% of tropical Africa’s aboveground carbonloss found in parklands of 10 to 30% tree cover(2). On the basis of our experience with MODIS(3, 4, 11), much of the carbon loss presented byBaccini et al. is likely commission error.
Moreover, Baccini et al. state that carbonchange per MODIS cell is a net estimate, re-sulting in conservative aggregate gain and lossvalues. To assess this supposition, we examinedMODIS grid cells with >10% loss and >10% gainfrom Landsat (2). Areas of both loss and gainaccounted for 6%, 7%, and 21% of Landsat-defined mapped tree cover loss for tropicalAmerica, Africa, and Asia, respectively. Wefound no appreciable carbon loss in the dataof Baccini et al. for these areas (bottom row ofTable 1), supporting the idea that net change islow or absent in areas of colocated tree coverloss and gain. However, we posit that for other
Hansen et al, Science 363, eaar3629 (2019) 11 January 2019 2 of 4
Fig. 1. Tropical America. Except for some large industrial-scaleclearings, mapped carbon losses within the Amazon Basin are largelyuncorrelated with Landsat-derived tree cover loss. Left: TropicalAmerica tree cover loss from Hansen et al. (2). Right: Aboveground
carbon (AGC) loss from Baccini et al. (1). Top row: Extent of studyarea of Baccini et al. for tropical America. Middle row: 2800 km ×1500 km subset centered on Mato Grosso state, Brazil. Bottomrow: 530 km × 290 km subset of Amazonas state, Brazil.
RESEARCH | TECHNICAL COMMENTon June 16, 2020
http://science.sciencemag.org/
Dow
nloaded from
tree cover loss–dominated areas, carbon lossshould be present. For tropical America, Africa,and Asia, Baccini et al. estimate no carbon lossfor 68%, 76% and 72% of Landsat-derived treecover loss–dominant cells; we identify these aserrors of omission (Figs. 1 to 3) (12, 13).Despite the likely presence of substantial spa-
tial error and claims of capturing carbon lossesfrom disturbances beyond deforestation, thecarbon loss estimates of Baccini et al. are lowerthan those of Tyukavina et al. (14), which reliedon robust and statistically accurate area estima-tion methods. Given the lower overall carbonloss estimate of Baccini et al.—an estimate not
accounted for by colocated gain and loss, andconsisting of considerable commission error—it follows that their estimates of carbon lossesover landscapes experiencing land cover andland use change are underestimated. On theother hand, if we accept that the carbon lossesas depicted are accurate, their results overturnour current understanding of where the globalclimate change science and policy communitiesshould be focused in mitigating emissions fromdeforestation and forest degradation (i.e., notfocused on tropical forest clearing). Accordingto Baccini et al., deforestation is a minoritysource of emissions and is not colocated with
forests experiencing degradation (i.e., sites spreaddiffusely throughout tropical forests, includingforests heretofore thought to be intact) (Table 1and Figs. 1 to 3).We know the limitations of MODIS data in
mapping forest cover, including signal satura-tion within high canopy cover and the impactsof spatial resolution, atmospheric contamination,and sensor degradation (15). Even if these effectscould be accounted for, map characterizationsinevitably contain error and bias. Baccini et al.provide per-pixel estimates of carbon stock changesfrom 1 to >1000 tonnes per hectare—a capabil-ity not previously demonstrated experimentally,
Hansen et al, Science 363, eaar3629 (2019) 11 January 2019 3 of 4
Fig. 2. Tropical Africa. The pattern of smallholder forest clearing alongrural roads in the Congo Basin (12), which includes carbon loss due tothe expansion of agriculture into primary forest, is wholly absent. Left:Tropical Africa tree cover loss from Hansen et al. (2). Right: Aboveground
carbon (AGC) loss from Baccini et al. (1). Top row: Extent of study areaof Baccini et al. for tropical Africa. Middle row: 2800 km × 1500 kmsubset centered on the Congo Basin. Bottom row: 530 km × 290 kmsubset of Tshuapa province, Democratic Republic of Congo.
RESEARCH | TECHNICAL COMMENTon June 16, 2020
http://science.sciencemag.org/
Dow
nloaded from
much less at scale. We believe that the re-sults of their study should be considered withan appropriate level of scientific skepticism,and that other approaches should be used untilcarbon stock change mapping can be moredefinitively demonstrated. We advocate theuse of carbon stock strata—that is, areas ofknown mean carbon and associated uncertaintyderived using probability-based forest inventoryor other data. When such carbon stock data arecombined with accurate estimates of forest/treecover area change, forest carbon change canbe quantified without bias and with knownuncertainty (14).
REFERENCES
1. A. Baccini et al., Science 358, 230–234 (2017).2. M. C. Hansen et al., Science 342, 850–853 (2013).3. M. C. Hansen et al., Earth Interact. 7, 10 (2003).4. M. C. Hansen, P. Potapov, S. Turubanova, in Global Forest
Monitoring from Earth Observation, F. Achard, M. Hansen, Eds.(Taylor and Francis, 2012), pp. 93–110.
5. P. Olofsson et al., Remote Sens. Environ. 148, 42–57(2014).
6. S. V. Stehman, Int. J. Remote Sens. 35, 4923–4939 (2014).7. Intergovernmental Panel on Climate Change, Good Practice
Guidance for Land Use, Land-Use Change and Forestry (IPCCNational Greenhouse Gas Inventories Programme, Hayama,Japan, 2003).
8. United Nations Food and Agriculture Organization,Integration of Remote-Sensing and Ground-BasedObservations for Estimation of Emissions and Removals
of Greenhouse Gases in Forests: Methods and Guidancefrom the Global Forest Observations Initiative(ed. 2.0, 2016).
9. P. Potapov et al., Sci. Adv. 3, e1600821 (2017).10. A. Tyukavina, M. C. Hansen, P. V. Potapov, A. M. Krylov,
S. J. Goetz, Glob. Ecol. Biogeogr. 25, 151–163 (2016).11. M. C. Hansen, S. V. Stehman, P. V. Potapov, Proc. Natl. Acad.
Sci. U.S.A. 107, 8650–8655 (2010).12. G. Molinario et al., Environ. Res. Lett. 12, 104001 (2017).13. M. Broich, M. Hansen, P. Potapov, M. Wimberly, Int. J. Remote
Sens. 34, 5748–5760 (2013).14. A. Tyukavina et al., Environ. Res. Lett. 10, 074002 (2015).15. Y. Zhang, C. Song, L. E. Band, G. Sun, J. Li, Remote Sens.
Environ. 191, 145–155 (2017).
30 October 2017; accepted 8 November 201810.1126/science.aar3629
Hansen et al, Science 363, eaar3629 (2019) 11 January 2019 4 of 4
Fig. 3. Tropical Asia. The documented difference in logging within theprimary forests of the Malaysia/Indonesia border (13) is wholly absent in thecarbon loss map. Left: Tropical Asia tree cover loss from Hansen et al. (2)Right: Aboveground carbon (AGC) loss from Baccini et al. (1). Top row:
Extent of study area of Baccini et al. for tropical Asia (east and west edgestruncated). Middle row: 2800 km × 1500 km subset centered the islandsof Sumatra and Borneo. Bottom row: 530 km × 290 km subset of centralBorneo along the Malaysia/Indonesia border.
RESEARCH | TECHNICAL COMMENTon June 16, 2020
http://science.sciencemag.org/
Dow
nloaded from
of gain and loss''Comment on ''Tropical forests are a net carbon source based on aboveground measurements
Matthew C. Hansen, Peter Potapov and Alexandra Tyukavina
DOI: 10.1126/science.aar3629 (6423), eaar3629.363Science
ARTICLE TOOLS http://science.sciencemag.org/content/363/6423/eaar3629
CONTENTRELATED
http://science.sciencemag.org/content/sci/363/6423/eaat1205.fullhttp://science.sciencemag.org/content/sci/358/6360/230.full
REFERENCES
http://science.sciencemag.org/content/363/6423/eaar3629#BIBLThis article cites 12 articles, 4 of which you can access for free
PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions
Terms of ServiceUse of this article is subject to the
is a registered trademark of AAAS.ScienceScience, 1200 New York Avenue NW, Washington, DC 20005. The title (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement ofScience
Copyright © 2019, American Association for the Advancement of Science
on June 16, 2020
http://science.sciencemag.org/
Dow
nloaded from
