An inventory of benthic marine habitats can be obtained from recent satellite images. During Sub Sea Survey 2008 in Galveston Texas opportunities to preform multi-spectral analysis against satellite imagery were discussed in this presentation.
2. Overview
• Remote sensing using satellite imagery is a cost means
for mapping a large area in a short amount of time.
• Clear water applications ~30m deep
• Different methods:
– Supervised classification
• Uses field survey data (various types)
– Unsupervised classification
– Resolution and accuracy is directly correlated to the
resolution of the original source imagery.
3.
4.
5. Satellite Information
• Different satellites offer multiple options when
ordering/tasking satellites to acquire imagery.
• Resolution of common commercial satellites range from
30 m (LANDSAT) to 60 cm (QuickBird).
• Different bands provided in satellite imagery offer
multiple techniques associated with identifying certain
habitats.
• Normalized Difference Vegetation Index (NDVI) can be
applied to delineate mangroves from surrounding
vegetation. NDVI is a simple formula using two satellite
channels. If one band is in the visible region (VIS) and
one is in the near infrared (NIR), then the NDVI is (NIR -
VIS)/(NIR +VIS). This method is successful for
identifying the spectral signature of mangroves in
imagery where four bands were available.
6. Supervised Classification
• Independent in situ datasets.
– Satellite data
– Field Survey data
• High accuracy GPS unit
• Video
• Depth transducer
• Towed sensor platform (“towfish”)
• Spectral Radiometers
– Correct atmospheric conditions for the time period that
the satellite was collecting imagery.
• Corrections for water column.
• Spectral signatures need to be collected for values
you would like to map.
7. Supervised Classification
If not able to be in the field at the time of imagery
acquisition then you can still ground truth the data.
Disadvantages of this include:
No atmospheric correction (upwelling, reflectance) relative
to the
Areas where there are environmental changes that occur
over time, such as seasonal or yearly trends will
8. Unsupervised Classification
• Satellite imagery
• Classification is done based upon computer clustering of
spectral signature values.
• The user is responsible for distinguishing what habitat is
associated with each cluster.
• Use existing datasets (charts, bathymetry, etc.) to aid in
image interpretation when ground truth data is limited.