CRAIG SMITH

1
CURRENT MODIS DB PRODUCTS AND FUTURE PLANS AT
ACRES, GEOSCIENCE AUSTRALIA
CRAIG SMITH
03/000
2
ACRES MODIS NRT Online Products

• All Terra Aqua day night Passes acquired
• Available for 7-10 days at no charge, following
  User registration
• Level 1b and Map oriented (Lambert Conformal
  Projection) products available
• Browse JPEG - Bands 1,4,3 as RGB
• Metadata file - 1 for each spatial resolution
  set
• Map oriented products available in ENVI ER
  Mapper format allows for improved access for
  non-expert users
• Approximately 100 Terra Bytes per month of MODIS
  data being downloaded from ACRES ftp site

3
ACRES MODIS Archive Products

• Product specifications similar to the NRT Online
  products
• Post pass ephemeris is used in the generation of
  archive products
• How do we generate additional value from the
  MODIS archive
• The value of ACRES 26 year Landsat archive
  increases every year lessons for the MODIS
  archive!
• Opportunity to optimise the processing of the
  data, and improved information extraction, using
  local conditions and knowledge for improved
  higher level products
• Investigating the generation of higher level
  products, with consideration given to the
  institutional MODIS products available from the
  DAAC

4
ACRES PRODUCTS SERVICES

• Sentinel Project Major operational application
  of NRT MODIS is the mapping of thermal anomalies
  (Hotspots)
• Almost complete raw data archive of MODIS, TERRA
  AQUA, data
• MODIS complements our products and services from
  our archive of Landsat MSS, TM, ETM, ERS,
  RADARSAT JERS data
• Dedicated AVHRR reception facility in Alice
  Springs NRT Online products available from
  ACRES then archived by CSIRO
• Discussions with AVHRR users about potential
  MODIS products
• Investigating the development of a grassland
  curing product
• Geoscience Australia is responsible the Oceania
  ALOS Data Node

5

• Canberra

6
MODIS Data Processing Flow at ACRES Part I
Direct Ingest System
Raw data archive On SDLT
Data acquisition at Alice and Hobart
T0
Tweekly
RT-STPS PDS

• Tape archive at DPF
• Tape transfer to DLI
• Tape transfer to BoM

Sentinel System
T20
IMAPP at DAF
Hot Spot algorithm
7
MODIS Data Processing Flow at ACRES Part II
Direct Ingest System
Data acquisition at Alice and Hobart
T0
RT-STPS PDS
T2hr
WAN
Online products
T2-8hr
IMAPP System
Canberra
8
(No Transcript)
9
MODIS DB RD PROJECTS

• LOCATIONAL ACCURACY ASSESSMENT
• CLOUD COVER MASK (MOD35) ASSESSMENT
• ATMOSPHERIC AND BRDF CORRECTIONS

10
MODIS LOCATIONAL ACCURACY

• Multi-temporal sub-pixel (250 m) registration of
  MODIS AQUA TERRA data are important for most
  applications
• The locational accuracy specifications for TERRA
  MODIS data are about 50 m (1 s) at nadir
• The accuracy specifications for NRT AQUA MODIS
  data are larger as AQUA has no on-board GPS
  capability
• The major determinant of the locational accuracy
  of MODIS products in a NRT environment is the
  accuracy of the ephemeris used in product
  generation
• Post past ephemeris are not available at the time
  of NRT product generation. The locational
  accuracy of ACRES NRT products are approaching
  the theoretical limit

11
MODIS LOCATIONAL ACCURACY

• ACRES is implementing geometric quality assurance
  procedures for our MODIS products
• A sample of MODIS products are automatically
  assessed against a 2002 Landsat ETM mosaic of
  Australia, resampled to 250 metre pixel size
• An intensity based image matching technique is
  used to compare image chips taken from a regular
  grid of MODIS products with the database of
  Landsat image chips

12
MODIS LOCATIONAL ACCURACY

• The mean error of ACRES NRT MODIS Terra and AQUA
  products is 50 and 200 metres respectively with a
  standard deviation of 135 and 160 metres
  respectively
• These figures are derived from a sample of all
  pixels (nadir off nadir)
• Residuals will be used to update the geocoding
  information provide with the ENVI ER Mapper
  formatted products

13
MODIS CLOUD COVER MASK

• The Cloud Cover (CC) mask algorithm (MOD35) is
  being be verified for Australian conditions
• The design of the algorithm is conservative
• There are five groups of spectral tests. If one
  of the five group tests maps a pixel as cloudy
  or probably cloudy then the pixel will be
  labelled accordingly
• Currently investigating of the source of false
  positives
• The spectral band thresholds and ground cover
  types (desert, ocean etc) are the two main
  variables that are used to estimate cloud cover
• Investigating issues of modifying the threshold
  values in the spectral tests and Olsson land
  classification

14
ATMOSPHERIC BRDF CORRECTION

• A research project within the Co-operative
  Research Centre for Spatial Information - lead
  Participants are Geoscience Australia, Western
  Australia Dept of Land Information Cutin
  University
• A major component of this project is validating,
  and applying dynamic corrections for atmospheric
  effects, water vapour MOD05, and aerosols MOD04
• Dynamic inputs being derived from MODIS data with
  supporting information from the Australian Bureau
  of Meteorology
• This work is primarily being undertaken at Dept
  of Land Information and Curtin University, Perth
  
• Dynamic inputs for both-
• DB Version of MOD09
• SMAC code ( Rahman Dedieu )

15
MODIS 09/09/2003 0127UTC 0304UTC
16
Flow chart of atmospheric correction
algorithm H2O vapour is the primary focus of the
current work
Taken from MOD09 ATBD Vermote and Vermeulen
(1999)
17
Radiosonde Locations
18
The algorithms produce a water vapour map over
WA at 1km resolution.
Precipitable Water (kgm-2)
MODIS Terra 0208 UTC 17/12/2004
19
Validation of MODIS H2O algorithms
IMAPP Cloud Mask
DLI Cloud Mask
No Cloud Mask
Sobrino et al.
WVNIR
20
Analysis of data rejected by the cloud mask
DLI Cloud Mask
IMAPP Cloud Mask
Choice of cloud mask may limit good data by up
to 25
21
Grazie
Nov 2003
Apr 2004
NDVI South East Australia