Close Window

LC-03 Abstract

Radar Remote Sensing of Land Cover and Biomass in the Amazon

Myron Craig Dobson — University of Michigan (US-PI)
Joao Vianei Soares — INPE - Instituto Nacional de Pesquisas Espaciais (SA-PI)

Objectives





This proposal supports the scientific goals of LBA in the following technical theme

areas: (1) land-cover and land-use change, (2) carbon storage and exchange and (3) trace

gas fluxes.











  • The main cross-cutting themes of the proposed effort are: (1) remote sensing, (2) field

    observations and (3) GIS development.


  • Our specific objectives are to use archival orbital imaging radar data (1) to produce

    land-cover classifications of the Amazon basin, (2) to collaborate in mapping of surface

    inundation extent/duration, (3) to estimate aboveground biomass and (4) to collaborate in

    estimation of the biotic carbon pool.










A prototype classifier using orthorectified composites of ERS and JERS

SAR data has already been developed and tested on a number of NSF Long-Term Ecological

Research Sites and a moist tropical site west of Manaus. Simulations (using SIR-C data)

have shown the classifier to be readily adapted to use of the C-band and hh-polarized

channel provided by Radarsat. A large number of polygons will need to be inventoried to

support classifier development and validation.





Prototype biomass algorithms have also been developed and tested using data acquired by

SIR-C, and need refinement and validation for tropical forest types. The orthorectified

results will be expressed as digital thematic maps and placed into the LBA GIS.





The issue of spatial scaling will be explicitly treated within a subsample of primary

and secondary LBA sites via a nested approach using high spatial resolution data (i.e.,

SIR-C/X-SAR or airborne SAR if and when available) and lower resolution orbital data

(i.e., ERS/JERS/Radarsat at full resolution and also degraded to 100-m resolution). The

JERS and Radarsat data were acquired during two sampling periods in 1996. These results

will be cascaded into products generated at 100-m resolution for both the main LBA

transects (during the first three years) and over the full basin (during the ensuing three

years).





The temporal domain will also be explicitly treated. Wet and dry season composites will

be used to ascertain (1) inundation extent and (2) land-cover change over a 6-month period

(nominally for 1996). In addition, it is expected that the future availability of Light

SAR and/or PALSAR data (in the year 2000 to 2002 time frame) may provide the opportunity

for evaluation of decadal-scale changes as well.





The remote sensing effort will be supported by field-level activities to help define a

network of known land-cover polygons for evaluation of the classification and also conduct

biometry of land-cover types (undisturbed forest, regenerating forest, etc.) for

validation of biomass/carbon retrieval algorithms. The project includes remote sensing

scientists, electrical engineers and biologists with collaborative Brazilian

participation.





Research Team











  • Craig Dobson: Land-cover classification and biomass algorithms


  • Fawwaz Ulaby: SAR scattering and textural attributes


  • Leland Pierce: Image processing


  • Robyn Burnham: Tropical forest regrowth, biometry, fieldwork


  • Joćo Soares: Backscattering mechanisms and implementation of final biogeophysical

    processor at INPE


  • Dalton Valeriano: land-cover inventory and biometry


  • Giafracno DiGrandi: ERS SAR


  • Students (Hua Xie and another student from Amazon region), University of Michigan










Activities









Land-Cover Classification:







1. Determination of land-cover classes (1998)


2. Define, locate and label training and testing populations of land-cover polygons

within primary and secondary sites along the transects (1998)


3. Initial classifications at high resolution (SIR-C/X-SAR and ERS/JERS/Radarsat at

full resolution) and at low resolution (100m) (1999)


4. Verification of classifications (1999)


5. Final classification of LBA transects, insertion into LBA/GIS (2000)


6. Derivative products


       -disturbance/change mapping (inundation, clear-cutting, selective logging,

conversion)









Biomass:







1. Define, locate and measure biometry for training and testing populations of various

land-cover classes (stratified sampling). Use archival data where available. (1998)


2. Initial biomass estimation at primary and secondary sites (1999)


3. Verification of aboveground biomass assessments (1999)


4. Apply final algorithms to LBA transects, insert product into LBA/GIS (2000)


5. Creation of derivative products


       - carbon pool assessment



Close Window