Update of the National Biomass and Carbon Dataset 2000 using ALOS
PALSAR L-band data
Update of the National Biomass and Carbon Dataset 2000 using ALOS
PALSAR L-band data
Josef Kellndorfer, Wayne Walker, Oliver Cartus
The Woods Hole Research Center
Josef Kellndorfer, Wayne Walker, Oliver Cartus
The Woods Hole Research Center
NACP Meeting – February 2013
Continental-Global Biomass Maps (500-1000m)
Continental-Global Biomass Maps (500-1000m)
Santoro et al., 2011, RSE
Baccini et al., 2012, Nature Clim. Change
OutlineOutline National-scale mapping of canopy height and biomass at hectare-
scale (scale of disturbance): using inventory, Lidar and optical/radar remote sensing
using radar (without inventory for calibration)
The National Biomass and Carbon Data Set for the U.S.
(Collaborators: Elizabeth LaPoint, Mike Hoppus, James Westfall, USDA FS)
Multi-temporal L-band Radar (ALOS PALSAR ALOS-2, DESDynI-R)
(Collaborators: Maurizio Santoro, GAMMA RS)
National-scale mapping of canopy height and biomass at hectare-scale (scale of disturbance): using inventory, Lidar and optical/radar remote sensing
using radar (without inventory for calibration)
The National Biomass and Carbon Data Set for the U.S.
(Collaborators: Elizabeth LaPoint, Mike Hoppus, James Westfall, USDA FS)
Multi-temporal L-band Radar (ALOS PALSAR ALOS-2, DESDynI-R)
(Collaborators: Maurizio Santoro, GAMMA RS)
Pilot Studies for SRTM Height Retrieval: Georgia
Pilot Studies for SRTM Height Retrieval: Georgia
Kellndorfer, J.M., W.S. Walker and L.E. Pierce, M.C Dobson, J. Fites, C. Hunsaker, J. Vona, M. Clutter, "Vegetation height derivation from Shuttle Radar Topography Mission and National Elevation data sets." Remote Sensing of Environment, Vol. 93, No. 3, 339-358, 2004.
Kellndorfer, J.M., W.S. Walker and L.E. Pierce, M.C Dobson, J. Fites, C. Hunsaker, J. Vona, M. Clutter, "Vegetation height derivation from Shuttle Radar Topography Mission and National Elevation data sets." Remote Sensing of Environment, Vol. 93, No. 3, 339-358, 2004.
In the US a unique opportunity (for 2000) to combine several 30m national EO and ancillary data sets to extend plot level FIA height and biomass to wall to wall maps:
Shuttle Radar and Topography Mission (SRTM)
National Elevation Dataset (NED) Compiled from Topographic Survey data Cohesive processing for the first time around 2000 “Bald Earth” model
National Land Cover Database 2001 Provides Landcover, Treecover, Imperviousness
MRLC Landsat ETM+ Datasets 1999-2002
In the US a unique opportunity (for 2000) to combine several 30m national EO and ancillary data sets to extend plot level FIA height and biomass to wall to wall maps:
Shuttle Radar and Topography Mission (SRTM)
National Elevation Dataset (NED) Compiled from Topographic Survey data Cohesive processing for the first time around 2000 “Bald Earth” model
National Land Cover Database 2001 Provides Landcover, Treecover, Imperviousness
MRLC Landsat ETM+ Datasets 1999-2002
Modeling ApproachModeling ApproachHypotheses:Hypotheses:
Height = f (SRTM-NED, canopy density, cover type, …)Height = f (SRTM-NED, canopy density, cover type, …)
Biomass = f (height, density, cover type, …)Biomass = f (height, density, cover type, …)
Used a statistical approach (ensemble regression-tree Used a statistical approach (ensemble regression-tree algorithm) to model the relationships between remote sensing algorithm) to model the relationships between remote sensing measurements and FIA plot datameasurements and FIA plot data
Assess at different spatial scales how well height (LoreyAssess at different spatial scales how well height (Lorey’’s) and s) and biomass could be predictedbiomass could be predicted
Radar Optical
7
In National Geographic!
http://www.whrc.org/mapping/nbcd/index.html
http://daac.ornl.gov/NACP/guides/NBCD_2000.html
Plot Level Accuracy via Bootstrap Validation
Plot Level Accuracy via Bootstrap Validation
Biomass estimates less accurate than height as biomass integrates vertical and horizontal structure
NBCD Blackard et al.
NBCD multi-scale comparison with Blackard et al. and FIA biomass estimates
NBCD multi-scale comparison with Blackard et al. and FIA biomass estimates
Hexagon ScaleSize ~ 650 km2~ 25 FIA plots
NBCD represents a unique product because several 30 m remote sensing products were available for the same time frame
Continuity of global L-band observations warranted with ALOS-2 and DESDynI-R
Update of NBCD with ALOS PALSAR?Update of NBCD with ALOS PALSAR?
USDA project: Towards Spatially Explicit Quantification of Carbon Flux (2000-2007) in
Northeastern U.S. Forests Linking Remote Sensing with Forest Inventory Data
Investigators: Kellndorfer, J., Cartus, O., Houghton, R. A., Walker, W. S.
Collaboration: Maurizio Santoro, GAMMA RS
USDA project: Towards Spatially Explicit Quantification of Carbon Flux (2000-2007) in
Northeastern U.S. Forests Linking Remote Sensing with Forest Inventory Data
Investigators: Kellndorfer, J., Cartus, O., Houghton, R. A., Walker, W. S.
Collaboration: Maurizio Santoro, GAMMA RS
655 PALSAR FBD images for 2007/08
Multi-temporal coverage: 1-5
BIOMASAR Algorithm (Santoro et al.,
2011)
BIOMASAR Algorithm (Santoro et al.,
2011)
Semi-empirical Model:
Automated Model Calibration and Inversion for each image (or pixel
level)
Automated Model Calibration and Inversion for each image (or pixel
level)
ALOS biomass map for 2007from 655 HH/HV acquisitions
30m pixel posting
When aggregating to county scale …When aggregating to county scale …
Pixel-level Comparison with NBCDPixel-level Comparison with NBCD
NLCD 2001/06 Change product used to mask areas of forest cover change
Importance of multi-temporal dataImportance of multi-temporal data
Pronounced weather (rain) effects on single image retrieval
> 4 images per year (available only locally from ALOS PALSAR)
SummarySummary
•L-band radar: continuity warranted with ALOS-2 and DESDynI-RMulti-temporal acquisition strategy is key!
•Two different methods and data sets:Good agreement at spatial scales >500m
•To be investigated: Better understanding of uncertainty in the mapsDisturbance/Change, Map difference vs. signal
change