6 Precipitation
121
6.3.4 Satellite cloud indexing and life history methods of rainfall estimation
Cloud indexing was the first technique developed to estimate precipitation from
space. It is based on the assumption that the probability of rainfall over a given area
is related to the amount and type of cloudiness present over this area. Hence, one may
postulate that precipitation can be characterised by the structure of the upper surface
of the associated cloudiness.
A cloud structure analysis, either subjectively or objectively performed, is used as
the basis of the definition of a criterion relating cloudiness to a co-efficient (or index)
of precipitation. The characteristic may be, for instance, the number of image pixels
above a given threshold level. Hence, the general approach for cloud indexing
methods involving infra-red observations is to derive a relationship between a
Precipitation Index (PI) and a function of the cloud surface area, S(TBB), associated
with the brightness temperature (TBB) colder than a given threshold To. This
relationship can be generally expressed as follows:
(6.7)
for TBBj
If desired, an additional term related to the visible image can be included on the
right hand side of Equation 6.7. The next step is to associate PI to a physical quantity
related in some way to rain. This is done by adjusting the coefficients A and the
threshold level Toby comparison with independent observations such as raingauge
or radar data (see for example Arkin, 1979).
Richards and Arkin (1981) have shown that all cumuliform clouds with tops colder
than a given threshold temperature T precipitate at a fixed rate G rom h-I. It was found
that T = 235°K and G = 3.3rom h- I are typical of the eastern equatorial Atlantic. Arkin
and Meisner (1987) developed this method into the GOES Precipitation Index (GPI).
One of the problems inherent to this technique is the bias created by the potential
presence of high level non-precipitating clouds such as cirrus. Another limitation
resides in the fact that the satellite measurement represents an instantaneous observation integrated over space, whilst raingauge observations are integrated over time at
a given site.
Life-history methods, as indicated by their name, are based on the observations of
a series of consecutive images obtained from a geostationary satellite. It has been
observed that the amount of precipitation associated with a given cloud is also related
to its stage of development, therefore two clouds presenting the same aspect (from the
VIS-IR images point of view) may produce different quantities of rain depending on
whether they are growing or decaying.
As with the cloud indexing technique, a relationship is derived between a Precipitation Index (PI) and a function of the cloud surface area, S(TBB), associated with a
given brightness temperature (TBB) lying above a given threshold level, To. In
addition the cloud evolution is taken into account and expressed in terms of rate of
change ofS(TBB) between two consecutive observations (see for example, Griffith
et al. (1976), Stout et al. (1979), Scofield (1984».
121
6.3.4 Satellite cloud indexing and life history methods of rainfall estimation
Cloud indexing was the first technique developed to estimate precipitation from
space. It is based on the assumption that the probability of rainfall over a given area
is related to the amount and type of cloudiness present over this area. Hence, one may
postulate that precipitation can be characterised by the structure of the upper surface
of the associated cloudiness.
A cloud structure analysis, either subjectively or objectively performed, is used as
the basis of the definition of a criterion relating cloudiness to a co-efficient (or index)
of precipitation. The characteristic may be, for instance, the number of image pixels
above a given threshold level. Hence, the general approach for cloud indexing
methods involving infra-red observations is to derive a relationship between a
Precipitation Index (PI) and a function of the cloud surface area, S(TBB), associated
with the brightness temperature (TBB) colder than a given threshold To. This
relationship can be generally expressed as follows:
(6.7)
for TBBj
If desired, an additional term related to the visible image can be included on the
right hand side of Equation 6.7. The next step is to associate PI to a physical quantity
related in some way to rain. This is done by adjusting the coefficients A and the
threshold level Toby comparison with independent observations such as raingauge
or radar data (see for example Arkin, 1979).
Richards and Arkin (1981) have shown that all cumuliform clouds with tops colder
than a given threshold temperature T precipitate at a fixed rate G rom h-I. It was found
that T = 235°K and G = 3.3rom h- I are typical of the eastern equatorial Atlantic. Arkin
and Meisner (1987) developed this method into the GOES Precipitation Index (GPI).
One of the problems inherent to this technique is the bias created by the potential
presence of high level non-precipitating clouds such as cirrus. Another limitation
resides in the fact that the satellite measurement represents an instantaneous observation integrated over space, whilst raingauge observations are integrated over time at
a given site.
Life-history methods, as indicated by their name, are based on the observations of
a series of consecutive images obtained from a geostationary satellite. It has been
observed that the amount of precipitation associated with a given cloud is also related
to its stage of development, therefore two clouds presenting the same aspect (from the
VIS-IR images point of view) may produce different quantities of rain depending on
whether they are growing or decaying.
As with the cloud indexing technique, a relationship is derived between a Precipitation Index (PI) and a function of the cloud surface area, S(TBB), associated with a
given brightness temperature (TBB) lying above a given threshold level, To. In
addition the cloud evolution is taken into account and expressed in terms of rate of
change ofS(TBB) between two consecutive observations (see for example, Griffith
et al. (1976), Stout et al. (1979), Scofield (1984».