Abstract |
- The physical, biological, chemical, and optical
processes of the ocean operate on a wide
variety of spatial and temporal scales, from
seconds to decades and from micrometers to
thousands of kilometers (Dickey et al., this
issue; Dickey, 1991). These processes drive
the accumulation and loss of living and nonliving
mass constituents in the water column
(e.g., nutrients, phytoplankton, detritus, sediments).
These mass constituents frequently
have unique optical characteristics that alter
the clarity and color of the water column
(e.g., Preisendorfer, 1976). This alteration
of the ocean color, or more specifically the
change in the spectral “water-leaving radiance,”
L𝓌(λ), has led to the development of
optical techniques to sample and study the
change in biological and chemical constituents
(Schofield et al., this issue). Thus, these
optical techniques provide a mechanism to
study the effects of underlying biogeochemical
processes. In addition, because time- and
space-dependent changes in L𝓌(λ) may be
measured remotely, optical oceanography
provides a way to sample ecological interactions
over a wide range of spatial and temporal
scales.
The question often posed by scientists
trying to resolve problems involving the
temporal and spatial variation of oceanic
properties is: “What is the optimal time/
space sampling frequency?” The obvious answer
is that the sampling frequency should
be one half the frequency of the variation
(i.e., Nyquist frequency) of the property of
interest. However, therein lies the rub for
the oceanographer: the range of the relevant
scales is large, and the range of available
resources and/or actual engineering capabilities
to sample all relevant scales is often
small. Hence, the decisions affecting resource
allocation become critical in order to
maximize the total data information in both
quantity and quality. While these scientific
resource decisions are rarely discussed in
explicit terms of cost-benefit analysis, such
discussions should be integral parts of the
scientific design of instruments, platforms,
and experiments aimed at resolving oceanic
processes.
The practical examples of this problem in
remote sensing include: “What is the optimal
repeat coverage frequency?” and “What is the
optimal Ground Sample Distance (GSD) or
pixel size of the data?” For the optical oceanographer,
there is also the issue of optimal
spectral coverage needed to resolve the optical
constituents of interest (Chang et al., this
issue). The sum of these considerations feed
into the sensor, deployment platform, and
deployment schedule decisions. For polarorbiting
and geo-stationary satellites that
cost hundreds of millions of dollars, as well
as airborne sensors that have smaller upfront
costs but higher deployment costs, the decision
of sampling frequency directly impacts
the scientific use of the data stream, and
what processes may be addressed with data
streams collected by these sensors. These
scientific cost-benefit analyses extend beyond
the cost in dollars because the typical
lifetime and replacement cycle of these sensors
is on the order of years to decades, and
a poorly designed sensor package is very difficult to replace.
In 2001, the Office of Naval Research
(ONR) sponsored the Hyperspectral Coastal
Ocean Dynamics Experiment (HyCODE)
(Dickey et al., this issue), which presented
the opportunity to study the question of
scales of variability in remote-sensing data.
Hyperspectral airborne sensors were deployed
on several platforms at various altitudes.
This coverage was supplemented
by numerous space-borne, remote-sensing
satellites. The airborne instruments included
two versions of the Portable Hyperspectral
Imager for Low-Light Spectroscopy (PHILLS
1 and PHILLS 2) (Davis et al., 2002) operating
at an altitude of less than 10,000
feet and 30,000 feet, respectively, as well as
the NASA Airborne Visible/Infrared Imaging
Spectrometer (AVIRIS) sensor operating
at 60,000 feet. These sensors provided
hyperspectral data at 2 m, 9 m, and 20 m
GSDs, respectively. The satellite data collected
included the multi-spectral images
from Sea-viewing Wide Field-of-view Sensor
(SeaWiFS), Moderate Resolution Imaging
Spectroradiometer (MODIS), Fengyun 1
C (FY1-C), Oceansat as well as the multispectral
polarimeter Multiangle Imaging
SpectroRadiometer (MISR) sensor and sea
surface temperature (SST) sensor Advanced
Very High Resolution Radiometer (AVHRR).
These collections provided a wealth of remote-
sensing and field data during a spatially
and temporally intense oceanographic
field campaign, and they offered the ability
to begin to address the issue of optimal sampling
scales for the coastal ocean.
The use of these multiple remote-sensing
data streams requires the calibration, validation,
and atmospheric correction of the sensor signals to retrieve estimates of L𝓌(λ),
or “remote sensing reflectance,” Rᵣₛ(λ), a
normalized measure of the L𝓌(λ). Our goals
in this paper are to illuminate some of the
issues of remote sensing spatial scaling in
the nearshore environment and attempt to
derive some understanding of appropriate sampling scales in the nearshore environment.
We will focus on the data collected by
a single sensor (PHILLS 2) to reduce uncertainties
in the analysis that may result from
the different data processing techniques applied
to each of the individual sensors’ data.
|
Citation |
- Bissett, W. P., Arnone, R. A., Davis, C. O., Dickey, T. D., Dye, D., Kohler, D. D. R., & Gould Jr., R. W. (2004, June). From Meters to Kilometers: A Look at Ocean-Color Scales of Variability, Spatial Coherence, and the Need for Fine-Scale Remote Sensing in Coastal Ocean Optics. Oceanography, 17(2), 32-43. doi:10.5670/oceanog.2004.45
|