Data for Models Jorge E' Pinzon, SSAI

1
Data for ModelsJorge E. Pinzon, SSAI

• Long Time series data involve multiple satellites
  and sensors
• Multiple calibrations and drift
• Sparse and missing data
• A problem extracting invariant signal attributes
  (Morisette), even for the same satellite from
  multiple years

2
VIS/NIR/SWIR Band Comparison
POLDER
VGT
SeaWiFS
AVHRR
MODIS
MISR
3
Warmer Earth Greener North
AVHRR satellite record
4
EMD processing
5
Comparison with other datasets
In northern latitudes Correction does not Change
data significantly.
GIMMS VIe data is most stable of four Datasets in
tropics.
Improvement seen also in southern hemisphere
6
Correction allows for comparisons...
7
(No Transcript)
8
GIMMS Version E NDVI dataset

• Characteristics of the new AVHRR dataset
• Solar zenith angle satellite drift corrected.
• Reduction in uncertainty in tropical areas.
• NOAA 7-14 satellite data integrated with NOAA-16
  data for time series continuity.
• Use of SPOT-Vegetation data to provide bridge
  between historical (NOAA-7 through 14) and
  NOAA-16 data.
• New dynamical range matches MODIS', albeit at 8km
  spatial resolution.
• Many of the advantages of MODIS can be used while
  retaining the historical information.
• More robust method for extracting invariant
  signals

9
Multi-Sensor Comparison

• Use Core sites with ETM scenes and CIMEL Sun
  photometer data through AERONET
  
  
  Data sets
• AVHRR Version E
• SPOT-Vegetation, resampled to 8km Albers
  projections from Vitospot 1km hdf files
  filtered by QA flag
• MODIS-1km NDVI poduct, filtered by QA flag
• ETM, 30m NDVI from Sis-S Atmospherically
  corrected data using water vapor and aerosol
  properties from AERONET