Assessing the VIIRS Surface Type EDR Using Global MODIS Data

1
 Assessing the VIIRS Surface Type EDR Using
Global MODIS Data

• K. Jon Ranson1, John R.G. Townshend2 and
• Eric C. Brown de Colstoun3
• 1 Biospheric Sciences Branch, NASA/GSFC
• 2 Dept. of Geography, Univ. of Maryland
• 3 Science Systems and Applications, Inc.
• NPP Science Team Kickoff Meeting, Nov 4-6, 2003

2
Investigator Expertise/Roles

• Jon Ranson Branch Head, Biospheric Sciences
  Branch, NASA/GSFC and Terra Project Scientist.
  Project oversight Assist/review algorithm
  analysis reports, NPP calibration/validation plan
  and science operations concept document
  Facilitate access to MODIS land cover training
  and validation data Provide access to validation
  products/data for Northern Eurasia and China
  projects. Attend two meetings/year.
• John Townshend Chairman and Professor, Dept. of
  Geography, Univ. of Maryland, MODIS Science Team
  Member. Lead University of Maryland Global Land
  Cover Facility (GLCF) activities including
  providing global MODIS training and validation
  data to GSFC Participate in algorithm
  evaluation and validation Assist/review
  algorithm analysis reports, NPP cal/val plan and
  science operations concept document Supervise
  one Faculty Research Assistant (10 FTE) at UMD.
  Attend two meetings/year.
• Eric Brown de Colstoun Senior Support Scientist,
  SSAI. Lead algorithm/code review and evaluation
  effort Develop global metrics from MODIS data
  Test and validate current algorithms Prepare
  algorithm analysis report Contribute relevant
  material to NPP cal/val plan and operations
  concept document Provide relevant information on
  VIIRS land algorithms and ancillary data
  requirements. Interact with VIIRS
  sensor/algorithm contractors as appropriate.
  Attend four meetings/year.

3
EDR/CDR List

• We will focus on evaluating the VIIRS Surface
  Type EDR and in particular
• Quarterly Surface Type Intermediate Product (IP)
  which is equivalent to MODIS/AVHRR global land
  cover products.
• EDRs/IPs that are used by Quarterly Surface Type
  IP
• Vegetation Index EDR/IP
• Monthly Surface Reflectance IP
• Monthly Brightness Temperature IP
• Snow, Fire EDRs??
• Quarterly Continuous Fields IP (Threshold
  requirement).
• Improved MODIS algorithm will require new VIIRS
  approach.

4
Objectives

• 1) Evaluate performance of the current classifier
  OC1 with exact data/methods used to produce
  algorithm spec. in ATBD.
• 2) Evaluate classifier using global, yearly
  temporal metrics derived from monthly MODIS data.
• 3) Validate performance of the classifiers
  against high spatial resolution data and/or
  field-based data.
• 4) Produce at-launch global surface type product
  (1km resolution) for NPP mission.

5
General Approach

• Begin where VIIRS Surface Type ATBD leaves off
• Use Global MODIS data instead of AVHRR.
• Use global training data.

• Produce yearly temporal metrics from monthly
  MODIS data (Max.,Min, Mean, Amplitude).
• Evaluate internal performance and stability
  (e.g. cross-validation).
• Validation against independent data
• MODIS/EDC/UMD training/validation data.
• MODIS Land cover and continuous fields products.
• Validated regional high resolution products
  (MRLC, Northern Eurasia, China, UMD RESAC,etc)
• Produce at-launch Surface Type product.
• Algorithm evolution.

6
Prototype VIIRS 1km Classification
7
Deliverables

• Algorithm analysis report for the VIIRS Surface
  Type EDR including recommendations for
  improvements.
• We will provide support, as appropriate, for
  algorithm reports for other EDRs/CDRs (e.g. VI).
• Contribute as needed to VIIRS cal./val. plan and
  science operations concept documents.
• At-launch surface type product for NPP.

8
Resources Requested/Offered

• No special needs for algorithm evaluation are
  anticipated but

• Help/suggestions with validation data sets is
  always appreciated !

• We look forward to working closely with NASA, IPO
  and other Surface Type and Land EDR evaluators.