22
1: Richard Lucas, Aled Rowlands, Olaf Niemann, Ray Merton
7.3.7.3
Temporal Resolution
The temporal resolution refers to the frequency of observation by sensors.
Until the advent of hyperspectral data from spaceborne sensors, observations
were obtained on an 'as needs' basis. Due primarily to limitations of cost, few
multitemporal datasets have been produced. As a result, the complex timeseries of hyperspectral dataset acquisitions targeting geology, soil, and plant
applications constructed primarily of AVIRIS, HyMap, and CHRIS data (1992current) over Jasper Ridge in California (Merton 1999) perhaps represents one
of the few comprehensive datasets available.
The lack of temporal observations of hyper spectral sensors has proved particularly limiting for understanding, mapping or monitoring dynamic environments that experience rapid or marked changes in, for example, seasonal leaf
cover and chemistry (e. g., temperate woodlands), water status (e. g., wetlands)
or snow cover (e. g., high mountains). In many environments, such as arid or
semi-arid zones, temporal changes are less significant which accounts partly
for the success of hyperspectral data in, for example, geological exploration
and mapping.
7.3.7.4
Radiometric Resolution
Radiometric resolution, or quantization, is defined as the sensitivity of a sensor to differences in strength of the electromagnetic radiation (EMR) signal
and determines the smallest difference in intensity of the signal that can be
distinguished. In other words, it is the amount of energy required to increase
a pixel value by one count. In contrast to many MSS, which typically recorded
data up to 8-bit in size, hyperspectral sensors have been optimised to record
data to use at least 10 to 12 bits. As an example, AVIRIS recorded data in 10 bits
prior to the 1993 flight season and 12 bits thereafter (Vane et al. 1993). The
consequence of increasing the quantization has been to increase the sensitivity
of the sensor to variations in the reflected signal, thereby allowing more subtle
reflectance differences from surfaces to be detected and recorded. If the quantization level is too small then these differences can be lost. The acquisition of
data by sensors that support a larger number of quantization levels is therefore
important, especially when dealing with some of the applications discussed
later in this chapter (e. g., determination of vegetation stress and health and
discrimination of minerals).
A fundamental requirement of hyperspectral remote sensing has been the
need to maximise the signal as opposed to the noise; in other words, the
SNR. The SNR is a measure of how the signal from surfaces compares to the
background values (i. e., noise) and is determined typically by estimating the
signal from pixel values averaged over a homogenous target and dividing this
by the noise estimated from the standard deviation of the pixel values. SNR
vary considerably between sensors and spectral regions. Manufacturers of
CASI, for example, state that SNR values are greater than 480:1. Manufacturers
1: Richard Lucas, Aled Rowlands, Olaf Niemann, Ray Merton
7.3.7.3
Temporal Resolution
The temporal resolution refers to the frequency of observation by sensors.
Until the advent of hyperspectral data from spaceborne sensors, observations
were obtained on an 'as needs' basis. Due primarily to limitations of cost, few
multitemporal datasets have been produced. As a result, the complex timeseries of hyperspectral dataset acquisitions targeting geology, soil, and plant
applications constructed primarily of AVIRIS, HyMap, and CHRIS data (1992current) over Jasper Ridge in California (Merton 1999) perhaps represents one
of the few comprehensive datasets available.
The lack of temporal observations of hyper spectral sensors has proved particularly limiting for understanding, mapping or monitoring dynamic environments that experience rapid or marked changes in, for example, seasonal leaf
cover and chemistry (e. g., temperate woodlands), water status (e. g., wetlands)
or snow cover (e. g., high mountains). In many environments, such as arid or
semi-arid zones, temporal changes are less significant which accounts partly
for the success of hyperspectral data in, for example, geological exploration
and mapping.
7.3.7.4
Radiometric Resolution
Radiometric resolution, or quantization, is defined as the sensitivity of a sensor to differences in strength of the electromagnetic radiation (EMR) signal
and determines the smallest difference in intensity of the signal that can be
distinguished. In other words, it is the amount of energy required to increase
a pixel value by one count. In contrast to many MSS, which typically recorded
data up to 8-bit in size, hyperspectral sensors have been optimised to record
data to use at least 10 to 12 bits. As an example, AVIRIS recorded data in 10 bits
prior to the 1993 flight season and 12 bits thereafter (Vane et al. 1993). The
consequence of increasing the quantization has been to increase the sensitivity
of the sensor to variations in the reflected signal, thereby allowing more subtle
reflectance differences from surfaces to be detected and recorded. If the quantization level is too small then these differences can be lost. The acquisition of
data by sensors that support a larger number of quantization levels is therefore
important, especially when dealing with some of the applications discussed
later in this chapter (e. g., determination of vegetation stress and health and
discrimination of minerals).
A fundamental requirement of hyperspectral remote sensing has been the
need to maximise the signal as opposed to the noise; in other words, the
SNR. The SNR is a measure of how the signal from surfaces compares to the
background values (i. e., noise) and is determined typically by estimating the
signal from pixel values averaged over a homogenous target and dividing this
by the noise estimated from the standard deviation of the pixel values. SNR
vary considerably between sensors and spectral regions. Manufacturers of
CASI, for example, state that SNR values are greater than 480:1. Manufacturers