Generation of a MODIS Sea Surface Temperature Composite

1
Generation of a MODIS Sea Surface Temperature
Composite Methodology and Validation Stephanie
Haines November 21, 2005
2
Outline

• Motivation to generate a high-resolution SST
  composite
• Methodology used to create several versions of
  MODIS SST composites
• Case Study analysis of May 2004
• Comparison to RTG
• Latency effects
• Validation against buoy data
• Conclusions and future work

3
Motivation

• Coastal NWS offices wanted high resolution SST
  product
• Single pass of MODIS provides limited coverage
• Simple composite generated from multiple passes
  of MODIS data
• Main assumption is that SST does not change
  significantly from one day to the next
• High resolution (1 km), full coverage
• Small scale features in SST gradients retained
• Alternative to the Real-Time Global (RTG) SST
  analysis for model assimilation
• Lower boundary layer forcing important in
  mesoscale models
• RTG is a 0.5 degree daily interpolation of in
  situ and satellite data

4
Methodology

• Use the 1km SST EOS MODIS product to produce a
  detailed and spatially continuous SST field while
  minimizing cloud contamination and latency
  effects
• Retain 5 clear (as determined by a cloud mask)
  SST values for each pixel separately for day and
  night
• Average some of clear values to obtain a more
  consistent field
• Version 2 discards the coldest 2 of the 5 and
  averages the remaining 3 (current operational
  method)
• Version 3 has reduced latency by only considering
  the 3 most recent, to reduce cloud contamination
  the coldest is discarded (experimental method)

5
Case Study Period

• May 2004
• relatively clear and dry period
• General warming trend of SSTs
• Version 2 method used with EOS SST product from
  the DAAC
• EOS cloud mask applied
• Loop of daytime composites shows
• Spatially continuous pattern
• 1 km resolution
• Realistic spatial trends, varying with time

6
Comparison of MODIS to RTG

• Large scale patterns of MODIS match well to RTG
• Strong SST gradient off east coast not in RTG
  product
• RTG is a daily product, therefore no diurnal
  information
• The MODIS composites show diurnal differences
• Operational applications of version 2 revealed
  some latency issues as a result of persistent
  cloud cover reran May 2004 case with version 3

7
Latency Effect on SST
8th May 2004
MODIS SST V2
MODIS SST V3
RTG-SST

• Version 3 developed because of latency issues
  with version 2
• Only small differences seen between the two MODIS
  versions, although there are differences in the
  latency
• Even with several days latency, the composites
  capture both large scale and small scale SST
  gradients

MODIS LATENCY V2
MODIS LATENCY V3
8
Verification Against Buoy Data

• SST composites capture majority of trends in the
  buoy data with high correlation values both day
  and night

Daytime

• Reduced latency improves the composite (average
  latency improvement between versions 2 and 3 is
  1-2 days)

• RTG compares very well to buoy data during the
  day. Both MODIS composites have a warm bias

Nighttime

• RTG too warm at night, with the largest bias of
  0.53

9
Conclusions and Future Work

• Version 2 worked very well for May 2004, but
  persist cloud cover and cooling SSTs revealed
  issues with current compositing method
• Version 3 has reduced latency compared to version
  2 and
• Will be creating a Version 4 algorithm that
  further improves the product
• Add AMSR-E SST data
• Aqua only
• Plan to make AMSR-E SST available in real-time
  2006
• Provides SST values in cloudy conditions

AQUA MODIS AMSR-E MODISAMSR-E