382
M. Menenti
Remote sensing methods applied in support of irrigation management are based
on the type of observations and spectral regions summarized below. Taking into
account the theoretical concepts presented in the preceding chapters of this book,
the description of methods given in this section will be limited to an overview. The
foundation for the statements given here is provided by the material presented in
the section Current Applications.
A. Multispectral classification of crops. The simplest way to discriminate crops
and irrigated area is by relying on reflectance spectra (0.4 through 2.5 Ilffi); the
spectral contrast between crops and soils is large, much less so between different
types of crops. Spectral contrast may be improved by using transforms such the
principal component analysis. In some cases textural measurements might be
helpful because of the relationship of field lay-out with crop type (e.g. orientation
and inter-row distance in vineyards). Agricultural fields comprise a mixture of
crops and bare soil in varying fractions, so that the spectral reflectance of such
mixed targets tends to depend more on the relative amount of soil and foliage than
on the spectral differences between different types of crops. This is an inherent
limitation in the use of automatic classification procedures (see also Chap. 7). In
arid environments, discrimination of cropped irrigated patches becomes simpler
because of the near-absence of green vegetation where no irrigation water is
available. Such simpler classification procedures are generally adequate to obtain
estimates of irrigated area.
B. Multitemporal classification of crops. Cropping patterns often comprise many
crops which differ in terms of sowing date and phenology, which implies that the
temporal evolution of the amount and colour of foliage is a useful feature for crop
discrimination. Both the amount and colour of foliage determine the spectral
reflectance, so multi-temporal sequences of satellite images can be used to
measure spectral reflectance at different phenological (growth) stages. The
simplest option to discriminate crops is to increase the number of attributes by
using simultaneously spectral reflectances on all dates for which images have been
acquired.
A better procedure is by taking phenology explicitly into account. This comprises the following steps:
construct a crop calendar indicating the expected duration of each phenological stage for a given area;
select the dates of acquisition of images on the basis of the crop calendar, i.e.
in such a way that differences in spectral reflectance between crops are largest;
convert the raw at-satellite observations into at-surface reflectances, i.e. taking
into account sensor calibration data and correcting for atmospheric effects.
combine the spectral measurements in spectral indices, so that just one value
is associated with each image pixel on any given date. The temporal sequence
M. Menenti
Remote sensing methods applied in support of irrigation management are based
on the type of observations and spectral regions summarized below. Taking into
account the theoretical concepts presented in the preceding chapters of this book,
the description of methods given in this section will be limited to an overview. The
foundation for the statements given here is provided by the material presented in
the section Current Applications.
A. Multispectral classification of crops. The simplest way to discriminate crops
and irrigated area is by relying on reflectance spectra (0.4 through 2.5 Ilffi); the
spectral contrast between crops and soils is large, much less so between different
types of crops. Spectral contrast may be improved by using transforms such the
principal component analysis. In some cases textural measurements might be
helpful because of the relationship of field lay-out with crop type (e.g. orientation
and inter-row distance in vineyards). Agricultural fields comprise a mixture of
crops and bare soil in varying fractions, so that the spectral reflectance of such
mixed targets tends to depend more on the relative amount of soil and foliage than
on the spectral differences between different types of crops. This is an inherent
limitation in the use of automatic classification procedures (see also Chap. 7). In
arid environments, discrimination of cropped irrigated patches becomes simpler
because of the near-absence of green vegetation where no irrigation water is
available. Such simpler classification procedures are generally adequate to obtain
estimates of irrigated area.
B. Multitemporal classification of crops. Cropping patterns often comprise many
crops which differ in terms of sowing date and phenology, which implies that the
temporal evolution of the amount and colour of foliage is a useful feature for crop
discrimination. Both the amount and colour of foliage determine the spectral
reflectance, so multi-temporal sequences of satellite images can be used to
measure spectral reflectance at different phenological (growth) stages. The
simplest option to discriminate crops is to increase the number of attributes by
using simultaneously spectral reflectances on all dates for which images have been
acquired.
A better procedure is by taking phenology explicitly into account. This comprises the following steps:
construct a crop calendar indicating the expected duration of each phenological stage for a given area;
select the dates of acquisition of images on the basis of the crop calendar, i.e.
in such a way that differences in spectral reflectance between crops are largest;
convert the raw at-satellite observations into at-surface reflectances, i.e. taking
into account sensor calibration data and correcting for atmospheric effects.
combine the spectral measurements in spectral indices, so that just one value
is associated with each image pixel on any given date. The temporal sequence