BIOMASS ESTIMATION USING POLARIMETRIC SAR

1
BIOMASS ESTIMATION USING POLARIMETRIC SAR
Dr. Jakob J. van Zyl RADAR SCIENCE AND
ENGINEERING SECTION JET PROPULSION
LABORATORY CALIFORNIA INSTITUTE OF
TECHNOLOGY 4800 OAK GROVE DRIVE PASADENA, CA
91109
2
OUTLINE

• INTRODUCTION
• SAR RESPONSE TO BIOMASS
• Dynamic Range
• Correlation with Biomass
• ALGORITHMS FOR ESTIMATING BIOMASS
• Dobson et al., 1992
• Ranson and Sun, 1994
• Dobson et al., 1995
• Rignot et al., 1995
• Ranson et al., 1995
• CONCLUSIONS

3
Introduction

• Vegetation biomass is an important geophysical
  parameter to measure since it is a measure of how
  much carbon is stored in the vegetation
• Understanding the total carbon budget, as well as
  its change with time helps to identify sources
  and sinks of carbon
• As important to know as the amount of carbon
  available for release after a forest is cleared,
  is what happens to the cleared area over time.
  If vegetation is allowed to grow back, some of
  the released carbon is absorbed again.
• SAR signals have been shown to be well correlated
  with the amount of woody biomass in vegetation
• Here we review some of the data, as well as
  several approaches published to infer biomass
  from radar data
• The data acquired during the Pacific Rim campaign
  is particularly suited for further research in
  this area

4
The Carbon Cycle
5
Atmospheric Carbon Dioxide Budget
6
Current Estimates of Biomass

• Recent estimates of biomass of tropical forests
  vary between 160 tons/ha (based on forest
  volumes) and 375 tons/ha (based on destructive
  sampling). These estimates come from a mix of
  sources that are often difficult to compare.
• The high estimates assume that all tropical
  forests are undisturbed and productive, and that
  biomass estimates based on direct measurements of
  small areas in a few tropical forest types could
  be extrapolated to all tropical forests.
• According to Food and Agriculture Organization
  (FAO) reports, 56 of open forests are
  unproductive with a correspondingly lower biomass
  density
• Even for closed forests, only 58 are
  undisturbed, and the unproductive forests appear
  to have biomass densities about two-thirds that
  of productive ones
• A study of 1230 one hectare plots along transects
  hundreds of kilometers across the Amazon basin
  showed that more than 50 of the area covered by
  this survey had biomas values less than 200 to
  220 tons/ha

7
Current Estimates of Biomass
8
Deforestation and Atmospheric Carbon Dioxide

• Published rates of deforestation differ by a
  factor of two to three, largely because of
  purpose and definition
• A report released by the FAO erly 1993 showed
  that an average of 15.4 million hectares of
  tropical forests were destroyed per year during
  the 1980s an increase of 40 percent over the
  1970s.
• This report also showed that while 0.6 of the
  worlds rain forests disappear each year, moist
  deciduous and upland forests are disappearing
  even faster
• Skole and Tucker used Landsat TM data to map
  deforestation in the Amazon basin for 1978 and
  1988. Their results show deforestation increased
  by a factor of 2 - 3 in all provinces studied,
  except for Amapa. However, excessive cloud cover
  in this region prevented a complete analysis
• The amount of carbon released by deforestation
  depends on the biomass of the forest cleared and
  that of the ecosystem that replaces the cleared
  forest

9
Deforestation and Atmospheric Carbon Dioxide
10
Net Primary Production
11
Le Toan et al., 1992
12
Isrealsson et al., 1997
13
SAR Signal Dynamic Range
Reference Le Toan et al., 1992
Reference Isrealsson et al., 1997
14
Correlation With Biomass
Reference Le Toan et al., 1992
15
Correlation With Stem Volume VHF
Reference Isrealsson et al., 1997
16
Algorithms Dobson et al., 1992
17
Algorithms Dobson et al., 1992
18
Algorithms Dobson et al., 1992
19
Algorithms Ranson and Sun, 1994
20
Algorithms Ranson and Sun, 1994
21
Algorithms Ranson and Sun, 1994
22
Algorithms Dobson et al., 1995
23
Algorithms Dobson et al., 1995
24
Algorithms Dobson et al., 1995
25
Algorithm Kasischke et al., 1995
26
Algorithm Kasischke et al., 1995
27
Algorithm Kasischke et al., 1995
28
Algorithm Kasischke et al., 1995
29
Algorithm Rignot et al., 1995
30
Algorithm Rignot et al., 1995
31
Algorithm Ranson et al., 1995
32
Algorithm Ranson et al., 1995
33
Conclusions

• SAR backscatter show a high correlation with
  vegetation biomass
• The dynamic range of the SAR signal increases
  with a decrease in frequency
• The largest dynamic range is typically observed
  for the cross-polarized signals
• The single frequency, single polarization SAR
  backscatter saturates for larger biomass values -
  the saturation level is a function of frequency
  and increases with a decrease in frequency
• More accurate results are obtained when using
  multiple polarizations and frequencies
• The biomass range can be increased by using
  ratios of backscatter values measured at
  different frequencies
• The biomass range can also be extended by
  identifying the specific component of the biomass
  that the SAR signals are the most correlated
  with. Allometric equations are then used to
  calculate the total biomass