Remote Sensing Resolutions:
The goal of this module is to contribute to a common understanding
of remote sensing through a presentation of remote sensing
resolutions.
The concept of resolutions provides a framework for characterizing
remote sensing devices and remotely sensed data.
Moreover, resolution characteristics directly govern the types and
nature of information that can be derived using data from a given
sensor.
A discussion of remote sensing resolutions also provides a convenient
framework for introducing or elaborating on other important
remote sensing concepts that are discussed elsewhere in the RSCC.
These include:
This Module is organized into five main sections as follows:
Spectral resolution refers to the number, spacing,
and width of the sampled wavelength bands
along the electromagnetic spectrum.
Technically, spectral resolution is controlled by the
beam splitting and focusing apparatus
in the sensor, by the design of the detector arrays,
and by the specific sensitivities of the
photon detectors, or photovoltaic elements in
the detector arrays.
It is known and easily observable that different surface objects
have unique, or characteristic spectral signatures.
The higher the spectral resolution, the more completely and precisely the
spectral signature
of each individual
IFOV
will be sampled, and the more readily different scene elements can
be classified
or discriminated based on those signatures. The following figure shows
how a common sensor, Landsat Thematic Mapper, samples the spectra of
three common scene elements.
Spectral band width and band placement are typically chosen so as
to:
The dramatic differential between red (0.63 to 0.69 um on TM) and
near-infrared (0.76 to 0.90 um) reflectance seen in Figure 5 is a
well recognized spectral feature for vegetation.
Most sensors sample in these two wavelength bands. For an example,
go back to red edge.
"Vegetation Indices"
are designed to capitalize on the information contained in reflectance
measurements made in these portions of the electromagnetic spectrum
to indicate the abundance or physiological characteristics of the
vegetation on the surface.
An important consideration in sensor design
is balancing the objectives of keeping the sampled spectral bands
narrow, while maintaining a high signal/noise ratio.
The narrower the wavelength band, the less total radiant energy will be
incident upon the detector element.
This is because a smaller "slice" of the total radiant flux is
being sampled.
As a result, the strength of
the signal will be reduced relative to the magnitude of the
background noise of the sensor.
In order to maintain high image quality, the signal/noise
ratio must be kept large.
Several ways to improve the signal/noise ratio are to:
Each of these options has its drawbacks, and this sets up some
natural trade-offs among the different sensor resolutions.
These tradeoffs will be discussed in more detail at the end of this
module.
Spatial resolution usually
refers to the size of the
instantaneous field of view (IFOV) or
the ground pixel.
Although alternative definitions exist (Mather, 1987),
this is the meaning that will be used here.
The IFOV is the area on the ground that is viewed by the sensor
at a given instant in time.
As such, it usually represents the ground area that is represented by
each pixel in a remotely sensed image.
The IFOV can be indicated as the ground dimensions of each pixel, or
as the instantaneous angular measurement of the sensor field of view.
Spatial resolution is controlled by the scan rate, the dimensions
of the detector array, and the rate at which the analog output, produced by
each detector element as the scanning mirror sweeps across the surface,
is integrated and sampled for conversion to digital output.
Spatial resolution is usually given as "nominal" spatial resolution
which refers to the resolution for a sample obtained from the nadir
viewing position at the specified altitude of the satellite.
For some sensors, the actual resolution can vary considerably across the
scan line. This is especially problematic for the Advanced Very High
Resolution Radiometer (AVHRR) sensors aboard the NOAA series of
meteorological satellites.
Due to the wide swath width of the AVHRR instrument (+/- 56 degrees from
the surface normal or about 3000 km)
AVHRR pixel dimensions become increasingly distorted
away from nadir as view zenith angles increase.
At the extreme edges of the scan, the nominal AVHRR resolution of 1.1 km
is distorted to 2.4 km in the along track direction, and 6.5 km in the
across track direction (Goward et al. 1991).
Spatial resolution has important implications for:
As such, the spatial resolution of the data must be able
to support the objectives of a given remote sensing project.
Clearly, mapping the boundaries of the reservoir shown in Figure 14
cannot be adequately achieved using 1 km resolution data.
Frequently, an important consideration in sensor design
is balancing the objectives of attaining high spatial resolution,
while maintaining a high signal/noise ratio.
The smaller the IFOV, the less total radiant energy will be
incident upon the detector element. As a result, the signal/noise
ratio will suffer. This can be counterbalanced by either increasing
the dwell time, or increasing the spectral bandwidth.
Temporal resolution refers
to the temporal frequency with which a
given ground location will be sampled by an individual sensor.
It is controlled by the orbital characteristics of the satellite
and the swath width of the instrument.
Many ground objects have characteristic temporal signatures.
This refers to how the reflectance for a given object in a given waveband,
or band combination, changes through time.
For vegetated surfaces, the temporal signature is related to vegetation
phenology, or the physiological changes of the surface vegetation over
time as it goes through its seasonal cycles.
Because of this, there is potential to discriminate surface types based
on their temporal behavior as it is manifested spectrally and sampled
remotely by satellite.
The higher the temporal resolution,
the more completely and precisely the temporal signatures
of objects are sampled, and the more readily scene elements can be
discriminated based on their temporal behavior.
High temporal resolution, such as that provided by AVHRR,
also improves the likelihood of acquiring
clear images, especially for areas that experience frequent and
persistent cloud cover.
Likewise, high temporal frequency data allow imaging of areas that are
in the process of undergoing rapid change produced, for example, by fire.
Note that, while temporal resolution can
be increased with a wider swath, this results in problems with
spatial resolution as the view angle increases off nadir.
This distortion of the IFOV is illustrated above in
Figure 10.
The highest temporal frequency obtainable from AVHRR (on the order
of 12 hours at the equator) also involves
sampling an area from several different view angles and directions.
This has
considerable implications for radiometric and geometric consistency
when comparing data from the same area across time periods.
A wider swath will also decrease the dwell time, thus reducing the
strength of the signal and, consequently, reducing the signal/noise
ratio.
This can be accomodated (i. e. a high signal/noise ratio can be
maintained) by either enlarging the IFOV, broadening the sampled
wavelength bands, or both; any of which will increase the strength
of the signal.
Radiometric resolution refers
to the dynamic range, or the number of different output levels used
to record the radiant energy for a single measurement.
The number of levels is governed by the byte size, or number of bits used
by the recording medium to represent the signal generated by a single
detector element for a given IFOV, or pixel.
The dynamic range is determined as 2^n, where n
are the number of bits per byte, or pixel.
Note that the detector response to radiant energy is analog and,
therefore, any incremental change in flux will initiate a change
in the detector response. However, this analog response must be
discretized for digital storage, transmission and analysis.
The number of discrete levels is governed by the radiometric
resolution.
The greater the radiometric resolution, the more accurately
the remotely sensed data can represent
variations in surface leaving radiance.
Directional resolution refers to the range of viewing angles from which
the IFOV is, or can be, sampled.
[Link here to a table with many sensors summarized in terms of directional
resolution]
Objects have characteristic directional signatures, or BRDFs.
If adequately sampled, the BRDF can be used to discriminate between objects
and infer surface properties.
The higher the directional resolution, the more completely and precisely
the directional signature, or BRDF, of the surface is sampled.
Directional resolution is governed by: <
Spectral Resolution
Spatial Resolution
Temporal Resolution
Radiometric Resolution
Directional Resolution
Figure 1. Spectral Characteristics of Some Common Sensors
Figure 2. MODIS and its Optical Surfaces
Figure 3. Spectral Signatures for Some Common Materials
Figure 4. Multispectral Image (TM) and Class Spectra
Figure 5. Example Spectral
Signatures With Features Identified
Figure 8. Spatial Resolutions of Some Common Sensors
Figure 9. Nadir and View Geometry
Figure 10. AVHRR: Resolution Dependence on View Angle
The higher (or finer) the spatial resolution, the more completely
and precisely the shapes of objects are sampled, the more readily
objects can be identified based on their shape, and the more accurately
the precise location, extent and area of objects can be determined.
Figure 11. A Multiresolution Image
Figure 12. Resolution Vs. Cover-Type Area Graph
Figure 14. Don Pedro Reservoir at Various
Resolutions
Figure 15. Temporal Resolutions of Some Common Sensors
Figure 16. Temporal Signatures
Figure 17. Radiometric Resolution of Some Common
Sensors
Figure 18. Radiance-In/DN-Out Graph
Figure 19. Digital Image at Various Dynamic Ranges
Figure 20. Something Directional
Figure 22. Something Directional
moody@geog.unc.edu