Shifting the paradigm: Toward real-time tracking of carbon storage on land
-
Upload
centro-de-excelencia-virtual-en-monitoreo-forestal -
Category
Education
-
view
99 -
download
0
Transcript of Shifting the paradigm: Toward real-time tracking of carbon storage on land
Shi%ingtheparadigm:Towardreal-5metrackingofcarbonstorageonland
CONAFOR,Agust2-2016
Overview
§ CarbonStockMonitoring§ Background
§ NewAboveground30mcarbondataset§ Abovegroundcarbonloss
§ Landcoverchange(Hansenetal.2013)combinedwithcarbondataset
§ Abovegroundcarbondensitylossandgain§ “Direct”carbondensitychange
§ Changepointanalysisofbiomass5meseries§ LiDAR(GLAS)§ MODIS
§ Results
• ForestInventories
• Stra%fy&Mul%ply(SM)Approach– Assignanaveragebiomassvaluetolandcover/vegeta5ontypemap
(AsneretAl.2010)
• Combine&Assign(CA)Approach– ExtensionofSM,GISandmul5-layersinforma5on(Gibbsetal.2007)
• EcologicalModels(EM)Approach– Remotesensingtoparameterizethemodel(Hur[etal.2004)
• DirectRemoteSensing(DR)Approach– EmpiricalModelswhereRSdataiscalibratedtofieldes5mates(Baccini
etal.2004,2008,Blackardetal.2008)
Goetz,Baccinietal.2009
Large Area Biomass Estimation
4
Forest
InventoryPlots
Youknowaveragecarbondensity(emissionsfactors)Youknowtheareadeforested(ac5vitydata)
5
DRC= 17 Billion t C
CO2Emissions2000-2010
(tC/ha)
Reduce uncertainty in carbon cycle studies Input for REDD/carbon Market Stock Flow Approach
(ha/year)
Biomass2000 Deforesta5on2000-2010
Distribu5onofforestinventorydatainCentralAfrica
Fieldbiomassmeasurements
MODIS1kmNBAR(RGB2,6,1)
The Geoscience Laser Altimeter System (GLAS)
The figure shows 30% of the GLAS L2A (year 2003) shots after screening procedures (1.3 million observations) Screening for cloud attenuation, topographic slope (SRTM), waveform noise thresholds, etc. 30-40% remain after screening.
Vegeta5onstructurefromLiDAR(GLAS)
70m
Lidarmetricshavebeenextensivelyusedtocharacterizevegeta5onstructure(Sunetal.2008,Lefskyetal.2005,Lefskyetal.1999)
Drakeetal.(2003),Lefskyetal.2005,Drakeetal.2002foundastrongrela5onshipbetweenAGBandLidarmetrics(HOME)
SatelliteInforma5on
Twotypesofinforma5on:PointdataandImagedata
ICESat GLAS (Points)
Biomass=30(t/ha)Biomass=78(t/ha)Biomass=205(t/ha)
Co-locatedFieldMeasurements
Fieldobserva5onnetwork&calibra5on>300loca5ons>30,000treesmeasured
• Columbia• Ecuador• Bolivia• Brazil
• Gabon• DRC• Uganda• Tanzania
• Vietnam• Cambodia• Indonesia
0 100 200 300 400 500
0100
200
300
400
500
Field Biomass (Mg/ha)
GLA
S P
redi
cted
Bio
mas
s (M
g/ha
)
Samples(million) TropicalAmerica
TropicalAfrica
TropicalAsia
Total
Available 13.2 18.2 11.8 43.2A%erScreening 2.3 2.5 0.7 5.5
Percent%used 17.4 13.8 5.9 11.5
Standarderror22.6MgC/haAdjustedR-squared:83.2
Biomass=HOME+H10+H60+CANOPY_ENE+H25
The Geoscience Laser Altimeter System (GLAS)
The figure shows 30% of the GLAS L2A (year 2003) shots after screening procedures (1.3 million observations) Screening for cloud attenuation, topographic slope (SRTM), waveform noise thresholds, etc. 30-40% remain after screening.
2009NBARone8dayscomposite
2009NBARone8dayscomposite
Composite2007-2008
Pixelsmosaicofbestqualityreflectanceovertheperiod2000-2003
KeyInputUsed:NADIR,BRDF-AdjustedReflectance
(Schaafetal.,2002;RSE)
Removes artifacts associated with variable view geometry
NBARcomposi5ng
Atmosphericallycorrectedandcloudcleared
– Spa5alresolu5onat500(nadir)However….Ar5factsduetocloudsresidualsand
shadowsarepresentComposi5ngover5mesuccessfullyremove
ar5facts
Composite2007-2008
Pan-TropicalBiomassMap
• BestqualityMODISmosaic• Mul5pleyears• cloudfree
• ScreenedGLASmetrics• Seriesofmetrics(treeheight,heightofmedianenergy,etc.)
• Co-locatedfieldmeasurements
GLASyear2007
MODIS500mcompositeyear2007-2008
23°N
0°
23°SA.Bausch
CT1
Pantropical Forest Carbon Mapped with Satellite and Field Observations
AmazonBasindetailfromthemap
DRCdetailfromthemapPNGdetailfromthemap
Error25MgCha-1 Error19MgCha-1 Error24MgCha-1
Baccinietal.2012
LandsatBasedAbovegroundBiomass
Landsatcircayear2000RGB:4,5,7(Hansenetal.2013)
GLASbasedbiomassdensityes5mates
BasedonsimilarapproachofBaccinietal.2012
LandsatBasedBiomassDensity(Yr.2000)Baccinietal.2016inprepara1on,Zarinetal.2016
ImprovementinSpa5alResolu5on30m resolution500m resolution
0 2010 Kilometers
0 105 Miles
Aboveground Carbon
tonnes per hectare0 >18012550 1005
LandsatBasedBiomassYear2000
PixelLevelUncertainty(PercentError)
LandsatAnnualBiomassLossfromDeforesta5on
Landsatbiomassdensitycirca2000combinedwithHansenetal.2013deforesta5on
Biomassreferenceyear2000
AnnualGrossEmissionsfromDeforesta5on
!
Zarinetal.2015
Na5onalScale
StateScale
JuntasIntermunicipales
Emissionsfactors
• Statelevelaverage3%
• Municipali5eslevelaverage10.6%
ID=statecode,municipalitycode,andlandcoverclass
IPCCguidelines
Limita5ons
• Grossbiomasslossesfromdeforesta5on– Degrada5on?Verydifficultwhenfocusedonareaextent
• Gains?
• Toolate
Shifting the Paradigm
NoneedtodefineDeforesta5onandDegrada5on
Carbondensitytrajectoriesover5meandspace 2002
2012
• Timeseriesapproachbasedon“changepoint”analysis
• Foreach500mx500mpixelweiden5fythetrajectoryofcarbondensity
2 4 6 8 10 12
050
150
250
Pixel 379226
Year
Biom
ass
(Mg/
ha)
●
●
●
●
●
●
●
●
●
●
2 4 6 8 10 12
050
150
250
Pixel 1800164
Year
Biom
ass
(Mg/
ha)
●
●
●
●●
●
●●
● ●●
2 4 6 8 10 12
050
150
250
Pixel 3762098
Year
Biom
ass
(Mg/
ha)
●
●
●
●●
●●
● ●●
●
Con5nuousspa5allyexplicitcarbondensitychangewithmeasurableuncertainty
High : 128
Low : 1
High : -1
Low : -252
High : 293
Low : 0
Gain
Stable
Loss
Mg/ha
Mg/ha
Mg/ha
190kmx215km
2 4 6 8 10 12
050
150
250
Pixel 3710066
Year
Biom
ass
(Mg/
ha)
●
●●
●●
●● ●
●●
●
Gain=59.2StdEr=24.2P-V=0.041
2 4 6 8 10 12
050
150
250
Pixel 3217595
Year
Biom
ass
(Mg/
ha) ●
● ● ●
● ●●
● ●
● ●
2 4 6 8 10 120
5015
025
0
Pixel 1249653
Year
Biom
ass
(Mg/
ha)
●
● ●●
●
●
●
●
●●
●
●
Loss=-201.2StdEr=8.4P-V=0.003
StdEr=46.1P-V=0.99
High : 128
Low : 1
High : -1
Low : -252
High : 293
Low : 0
Consistentwithdeforesta5onandsensi5veto“degrada5on”?
Deforesta5onLandsatbased(30mresolu5on)Hansenetal.2014
Gain
Stable
Loss
Mg/ha
Mg/ha
Mg/ha
190kmx215km
2 4 6 8 10 12
050
150
250
Pixel 3710066
Year
Biom
ass
(Mg/
ha)
●
●●
●●
●● ●
●●
●
Gain=59.2StdEr=24.2P-V=0.041
2 4 6 8 10 12
050
150
250
Pixel 3217595
Year
Biom
ass
(Mg/
ha) ●
● ● ●
● ●●
● ●
● ●
2 4 6 8 10 120
5015
025
0
Pixel 1249653
Year
Biom
ass
(Mg/
ha)
●
● ●●
●
●
●
●
●●
●
●
Loss=-201.2StdEr=8.4P-V=0.003
StdEr=46.1P-V=0.99
SouthEastAsiaBiomasschange2003-2014
0 105 Kilometers
0 14070 Kilometers
-1 < -500
0 105 Kilometers
SouthEastAsiaBiomasschange2003-2014
0 14070 Kilometers
-1 < -500
Deforesta5onLandsatbased(30mresolu5on)Hansenetal.2014
UncertaintyAssociatedtoChange
“Preliminary result based on GLAS/MODIS
44
MexicoAnnualCarbonNetLossandGainBaccinietal.2016.Inreview
<-200
10204060
80
-40-70-100-150
-10-20
Mg/ha
RegionalBiomassLoss
47
XinguBiomassLoss
48
Democ.Rep.CongoAnnualCarbonNetLossandGain
●
●
●●
● ●
●
●
●●
●
−100
−50
050
100
Congo, DRC
Year
Car
bon
(Tg)
03−0
4
04−0
5
05−0
6
06−0
7
07−0
8
08−0
9
09−1
0
10−1
1
11−1
2
12−1
3
13−1
4
●
●
●
●
● ● ●
●
●
●
●
Net LossNetNet Gain
●
●
●
●
● ●
●
●
●●
●
Baccinietal.2016.Inreview
Valida5on
50
Issueswithexis5ngfielddata• Size• Measurements• Dateofmeasurements• Geoloca5on
512004 2006 2008 2010 2012
020
4060
8010
0
Sum Delta Biomass Change
Year
Biom
ass
(Mg/
ha)
Parcel B−RSParcel B−FParcel C−RSParcel C−F
Brandoetal.2014
Cumula5
veBiomassloss(Mg/ha)
Valida5on/Assessment?
Valida5on
52
2004 2006 2008 2010 2012
020
4060
8010
0
Year
Cum
ulat
ive
Bio
mas
s C
hang
e (M
g/ha
) RS estimatesField estimates
Valida5onBiomassGrowth
53
Planta5onsmaps(WRI)Stapeetal.2010Biomassincrements16to22Mgha-1y-1
2004 2008 2012
050
150
250
Pixel 4307245
year
Bioma
ss (M
g/ha)
●●
● ●●
●●
●
● ●
●
●
Titulodelapresentación
Geospatial Data Integration for Forest Carbon Monitoring:
Southeast Mexico and Beyond
Wayne Walker1, Alessandro Baccini1, Curtis Woodcock2, Luis Carvalho2, Alicia Peduzzi3, Fabio Gonçalves4, Javier Corral Rivas5, Carlos Lopez Sanchez5
Direct Measurement of Aboveground Carbon Dynamics In Support of Large Area CMS Development
2001-2012 2001-2012 2001-2012
Applications and Impact:
4,100 km2 of LiDAR
1000’s of field plots
Thank you!
Wayne Walker Mary Farina Luis Carvalho Tina Cormier Damien Sulla-Menashe Skee Houghton