These are words to remember for anyone
considering the “dos” and “don’ts” of ordering and using
commercial high-resolution satellite imagery products to map
submerged water features. I learned this important lesson during
the 1960s while working on Texas A&M University’s Spacecraft
Oceanography Project. After analyzing hundreds of color
photographs of ocean and coastal areas taken from NASA’s Gemini
and Apollo spacecraft, I realized that astronauts focused too much
on photographing Earth’s rim. But even the downward-looking images
were often useless due to excessive “sun glitter”—reflected
sunlight from rough water. All of these problems were the result
of undesirable viewing directions. The best pictures came from
unmanned spacecraft taken at nadir. This experience indicates that
what really matters is the combination of target azimuth and
off-nadir angles, as well as the sun’s azimuth and elevation
angles, as they define the critical angle called the Forward
Scattering Angle (FSA).
Imagery Sources Today
Today, there are several U.S-based commercial, high-resolution
Earth imaging systems: DigitalGlobe’s QuickBird (www.digitalglobe.com),
OrbImage’s OrbView 3 (www.orbimage.com)
and Space Imaging’s IKONOS (www.spaceimaging.com).
They all collect multispectral (MS) imagery—blue light (BL), green
light (GL), red light (RL) and near-infrared (NIR) light—with
higher-resolution panchromatic (PAN) imagery. And they all have
azimuth and nadir-angle agility. Selecting imagery wisely is
essential to successfully mapping submerged water features.
For ocean and coastal area mapping purposes, the 2-degree
angular field of view of these commercial cameras minimizes
disparate effects of variations in atmospheric reflectance and
attenuation, refractions at the air-water interface, rough-surface
reflections and attenuation within the water column. This greatly
increases the chances of successfully imaging underwater features.
All of the commercial cameras capture
precise image data over a large dynamic range of digital numbers (DNs).
This also is important for mapping ocean and submerged features,
which are often dark and become darker with water depth.
Using published coefficients and algorithms, DNs from these
imagers can be converted to quantitative estimates of the spectral
radiances at the top of the atmosphere (TOA) and, more
importantly, to apparent reflectances at the TOA or at the surface
(see Author’s Note). This allows for consistent, quantitative
handling of the image data after collection.
Be Careful What You Order
Be careful when ordering an image either as a new collection or from an
image library. Unfortunately, image libraries don’t list all four of the
angles you need to place a proper order. And placing constraints on
target azimuth isn’t normally part of a new collection order. I suggest
you speak with a customer service representative to help resolve these
matters. For example, send an e-mail message with catalog IDs to get
data about sun elevation and azimuth angles. If you have purchased
imagery, target and sun angles are documented in the metadata file (see
Author’s Note).
Be careful about the kinds of standard image products you order. Most
are optimized for land mapping. For example, popular pan-sharpening
products use the high-resolution PAN image, without alteration, to
enhance MS images. Unfortunately, this leads to products that have
brightness attributes that best meet the needs of land mapping, but that
perform poorly for ocean and coastal mapping. The problem lies with the
PAN image, which is dominated by the RL and NIR regions in which water
reflectance is nearly zero. But PAN images can be corrected to reduce
the RL and NR influences so a more appropriate image is made that
matches the spatial properties of submerged objects in the BL and GL
bands.
Physical Considerations
Willebrord van Roijen Snell discovered the Law of Refraction in 1621;
Augustin Fresnel described the Law of Reflection in 1822. From these
quantitative physical optics laws we know that off-vertical angles
change dramatically as radiant energy travels from air to water and from
water to air. In the visible region, the refractive index of water is
about 1.33. Sunlight travels through the atmosphere, across the
air-water interface and into water. The chart below shows the
results of Snell’s Law of Refraction and how the refraction angle (in
the water) varies as a function of the sun elevation angle. A more
vertical pathway—i.e., refraction angle near zero—allows sunlight to
penetrate water better to illuminate submerged objects.
The chart also shows the results of Fresnel’s Law of Reflection and how
the transmittance (100 percent minus percent reflectance) of the
air-water interface changes with sun elevation angle. Technically, there
are two reflectances—one for vertical polarization and another for
horizontal polarization. But sunlight is mostly un-polarized; therefore,
the chart includes average transmittance values.
The transmittance is better than 94 percent when the sun elevation angle
is larger than 30 degrees. But sunlight has to cross this interface
twice—from the sun to the water and again from the water to the imager.
A worse case is when the imager’s off-nadir angle is 30 degrees and the
sun elevation angle is 30 degrees. In this case, the two-way
transmittance would still be better than 92 percent. When the sun
elevation angle is 30 degrees, the refraction angle is only 40.6
degrees.
Consider the 2.2 percent to 5.9 percent
reflectance that occurs at the air-water interface. This amount of
reflection may seem insignificant. If the interface is flat, then the
reflected radiant energy travels in a direction not likely to be seen by
the imager. But if the interface is rough, which is typically the case,
then the reflected radiant energy is dispersed over a range of upward
traveling angles as a function of the FSA. If the camera looks too
closely to the sun’s reflection center off the water’s surface, the
apparent reflectance will be much higher than 2 percent to 6 percent.
To successfully map submerged water features, I’ve found that the FSA
should be greater than 40 degrees. For mapping oil slicks and surface
features, however, the FSA should be less than 40 degrees. In clear
water, BL penetrates water better than GL. But in more turbid water, GL
penetrates better. In either case, neither NIR nor RL imagery will
penetrate water well enough to show submerged water features.
Preferred Image Products
Image providers need to perform well-chosen post-collection processes
that maximize the likelihood of the successful observation and
quantification of underwater features. The dynamic ranges of brightness
in image DNs over water areas often are limited to only a few hundred
integers—out of perhaps thousands for the dynamic range over land.
Here are some guidelines for imagery intended for mapping submerged
features:
• Purchase either MS data or an MS/PAN bundle as 16-bit integers—8-bit
products suffer from the scale reductions that force water features into
a low range for the overall range of 1 to 255. The darkest and
brightness features will be truncated to fit the DNs into the smaller 1
to 255 range.
• Don’t order Dynamic Range Adjustment (DRA) products.
• Don’t buy pan-sharpened products. The NIR and RL parts of the spectrum
dominate the PAN band. In most pan-sharpening processes, the overall
intensity of the image is controlled by the intensity of the PAN image.
For water, PAN intensities tend to be very dark. Thus, the PAN band will
suppress spectral information in the BL and GL bands during sharpening.
It would be far better to modify the PAN band by subtracting NIR and RL
components before using the modified PAN image for sharpening
purposes—probably for sharpening a natural color image.
Putting It All Together
The color composite image in the "Imagery Sources Today" section
above illustrates some of the
complexities inherent with imagery selection. The FSA for the image was
only 19.8 degrees, so it wasn’t expected to be ideal for viewing
submerged features. Depending on wind speed and fetch—the distance over
which the wind works on the water to create roughness—sun glitter is
likely to obstruct the viewing of submerged features. This is certainly
the case for most of the bay north of the mainland. The water features
that are visible are all surface features—primarily boat wakes
highlighted by sun glitter. In the ocean south of the mainland, the high
contrast between submerged sand and submerged benthic grass tends to
overcome sun glitter interference. Look carefully at the shoreline next
to the northern edges of the landmass, and you can see places where the
water is smoother due to the short fetch of the wind.
The lesson to be learned here is that if you want to do the best job
possible when mapping underwater features with high-resolution
commercial satellite imagery, you should select a pair of target
off-nadir and target azimuth angles that reduce or eliminate sun glitter
reflections from rough water surfaces. In addition, you need to
calculate and consider the FSA (see Author’s Note below). Then order an
MS bundle (MS + PAN) or just MS data that have the full range of
brightness (16-bit integer products). If you really need to pan-sharpen
the submerged water features, consider working with an expert who does
this correctly. In most cases, just working with MS data, with an
emphasis on the BL and GL bands, will yield great results.
Author’s Note:
I wrote a conversion script, SRFI.SML, for TNTmips image processing
software from MicroImages Inc. (microimages.com).
SRFI stands for Standardized Reflectance Factor Index. Also, I wrote an
Excel spreadsheet that calculates FSA as a function of the target’s
off-nadir angle, the target azimuth angle, the sun’s elevation angle and
the sun’s azimuth angle. FSA is critically important when ordering
satellite imagery for mapping submerged or floating features on water.
For more information, contact me at the e-mail address listed at the
beginning of this article.