Recent Progress: Automated Retrieval of Surface Reflectance from Landsat

1
GSFC Carbon Theme
Recent Progress Automated Retrieval of Surface
Reflectance from Landsat Dr. Jeffrey Masek,
GSFCDr. Forrest Hall, GSFC/UMBCDr. Robert Wolfe
GSFC/RSCDr. Nazmi Saleous GSFC/RSTX
2
Development Status

• Capitalizes on MODAPS hardware/software
  environment for automated global processing
• saves
• consistent processing with MODIS
• Surface Reflectance (SR) processing stream
  prototyped
• atmospheric correction algorithm from MODIS
  (Vermote, 1997)
• aerosol retrieval uses dark vegetation approach
  of Kaufman et al. (1997)
• surface reflectance retrieval uses 6S code
• currently uses climatological water vapor, will
  use NCEP historical
• testing on Boreal study set of 18 scenes
• Validation studies underway
• comparisons with MODIS (8-day composite, daily
  reflectance)
• comparison with in-situ reflectance retrievals
  (BOREAS MMR)
• comparison of retrieved aerosols with AERONET
  data
• evaluation of stability of invariant targets
  in interannual data

3
Landsat Preprocessing and Analysis
2000, 1990 GeoCover TM/ETM
1975 GeoCover MSS
Other Landsat Data
MODIS Aerosols

• Orthorectification
• Precision Correction
• Invariant Target
• Normalization or Atm. Correction

• Calibration
• Atm. Correction
• Cloud//Snow mask

• IT Normalization

Preprocessing
Radiometrically Consistent Surface Reflectance
Dataset (1975-2000)

• Disturbance Rate, Type
• Land Cover Conversion Rate
• Spectral Unmixing for Fractional Change

Analysis
Aggregation
QA/ Validation
Disturbance/Recovery Products for Carbon
Assessments
4
Surface Reflectance (SR) Product
TM, ETM Nadir looking (within 7.5 degrees)
surface reflectance No BRDF, phenology
correction Corrected for - Rayleigh
scattering - aerosols - water vapor - ozone
Topographic correction possible MSS Soil
Line Rectified Implicitly corrects for
atmosphere, soil BRDF effects No correction for
canopy BRDF, phenology
5
Landsat-5 Atmospheric Correction
1990s Landsat-5 mosaic
TOA reflectance
Surface reflectance
BOREAS Study Region
100 km
6
August 2000 MODIS 500m 8-day surface reflectance
composite
7
L-7 vs. MODIS 09A Daily Comparison
Landsat-MODIS surface reflectances agree within
1 SR for the visible and within 2 SR for the
VNIR-SWIR. Dr Goal lt0.4 visible, lt1
VNIR-SWIR
August, 3 2000
Dr () rETM rMODIS
Landsat ETM Band
8
Effect of Atmospheric Correction

• Aerosol, Rayleigh correction reduces path
  radiance (VIS-NIR)
• Water vapor corrects for absorption (NIR-SWIR)

Dr ()
9
BOREAS Barnes MMR Data

• BOREAS flew Barnes MMR (Modular Multiband
  Radiometer) on helicopter, with sun photometer,
  in 1994
• Processing of simultaneous Landsat-5 data (/- 5
  days) allows validation of reflectance algorithm

10
MMR vs. Landsat Reflectance
Old Jack Pine (NSA T7Q8T) June 9, 1994
Fen Flux Tower (SSA F0L9T) Sept. 18, 1994
Reflectance 100
Landsat Band
Helicopter MMR data Landsat-5 TM Surface
Reflect. Error bars /- 1 std dev within 3x3
pixel window of center
Old Black Spruce (SSA G6K8S) Sept. 18, 1994
11
Example Mapping Disturbance
August 1989
Consistent surface reflectance data allow the
mapping of disturbance and recovery using known
spectral-temporal relationships.
August 1999
Change 1989-1999
Burned (Biomass loss)
Regrowth (Biomass gain)
No change
12
Mapping Disturbance at Regional Scales from
Geocover Dataset
1990 Landsat
2000 Landsat

• 1990-2000
• 1.03 Mha burnt, logged (5.8 of area)
• 0.89 Mha in early regrowth (4.9 of area)
• Note these results are illustrative and not
  validated

Change
13
Processing System Performance
Time to process 1 scene, 1 CPU

• MODAPS allows rapid processing of large data
  volumes
• Software/hardware reuse reduces costs

Wall-clock time (hr)
Time to process N scenes, 16 CPUs
Decadal NAM
One-time NAM
Processing a 30-year North American surface
reflectance dataset should take lt 4 days
14
Conclusions

• Results from prototyping are promising
• Surface reflectance retrieval now accurate to
  1-2 SR units
• Planned improvements will increase accuracy and
  utility
• new 6S code
• NCEP historical water vapor
• improved dark-object selection criteria
• extension to 1975 MSS (invariant targets
  approach)
• MODAPS system performance can scale to NAM
  dataset
• Beginning increased focus on disturbance mapping
  algorithms