375
Global Precipitation Estimation and Applications
with respect to the data pairs selected to fit the power-law curve, resulting in difficulty implementing this single curve for complex cases. Ba and Gruber (2001) proposed the GOES Multispectral Rainfall Algorithm to combine information from five
GOES channels to optimize the identification of raining clouds. Calibration of rain
rate for each indicated raining cloud, referenced by its cloud-top pixel temperature,
was then completed.
15.2.2.2 Cloud Window–Based Algorithms
Cloud local texture−based approaches retrieve pixel rain rates by extending the mapping from one single pixel to a range of the neighborhood pixel coverage. Wu et al.
(1985) used 24 brightness temperature texture features to retrieve rainfall within a
neighborhood size of 20 km × 20 km. Hsu et al. (1997) developed the Precipitation
Estimation from Remotely Sensed Information Using Artificial Neural Networks
(PERSIANN) system, which calculates rain rate at a 0.25° × 0.25° latitude–longitude resolution, based on the brightness temperature variations in a neighboring coverage of 1.25° × 1.25°.
15.2.2.3 Cloud Patch–Based Algorithms
Cloud patch–based approaches estimate rainfall based on information extracted
from the entire cloud coverage determined by various segmentation methods. One
early example of patch-based algorithms is the Griffith–Woodley technique (Griffith
et al. 1978), which segments a cloud patch with an IR temperature threshold of 253 K
and tracks it through its life cycle. The Convective–Stratiform technique (Adler and
Negri 1988) is another example of the cloud patch–based approach. It screens convective cells based on the local minimum of IR temperature and assigns different
rainfall amounts to convective and stratiform components separately. Pixel rain rates
are proportionally distributed starting from the coldest pixel to higher temperature
pixels. Xu et al. (1999) proposed another approach, which determines different
temperature thresholds by separating the rain/no-rain pixels in a cloud patch using
SSM/I MW rainfall estimates.
On the basis of the previous cloud-window-based PERSIANN system, Hong et
al. (2005) developed an automated neural network for cloud patch–based rainfall
estimation. The self-organizing nonlinear output (SONO) model accounts for the
high variability of cloud–rainfall processes at geostationary scales (i.e., 4 km and
every 30 min). Instead of calibrating only one IR–RR function for all clouds, SONO
classifies varied cloud patches into different clusters and then searches a nonlinear
IR–RR mapping function for each cluster. This designed feature enables SONO
to generate various rain rates at a given brightness temperature and variable rain/
no-rain IR thresholds for different cloud types, which overcomes the one-to-one
mapping limitation of a single statistical IR–RR function for the full spectrum of
cloud–rainfall conditions. In addition, the computational and modeling strengths of
neural networks enable SONO to cope with the nonlinearity of cloud–rainfall relationships by fusing multisource datasets. Evaluated at various temporal and spatial
scales, SONO shows improvements of estimation accuracy, both in rain intensity
and in the detection of rain/no-rain pixels. Further examination of SONO adaptability demonstrates its potential as an operational satellite rainfall estimation system
Global Precipitation Estimation and Applications
with respect to the data pairs selected to fit the power-law curve, resulting in difficulty implementing this single curve for complex cases. Ba and Gruber (2001) proposed the GOES Multispectral Rainfall Algorithm to combine information from five
GOES channels to optimize the identification of raining clouds. Calibration of rain
rate for each indicated raining cloud, referenced by its cloud-top pixel temperature,
was then completed.
15.2.2.2 Cloud Window–Based Algorithms
Cloud local texture−based approaches retrieve pixel rain rates by extending the mapping from one single pixel to a range of the neighborhood pixel coverage. Wu et al.
(1985) used 24 brightness temperature texture features to retrieve rainfall within a
neighborhood size of 20 km × 20 km. Hsu et al. (1997) developed the Precipitation
Estimation from Remotely Sensed Information Using Artificial Neural Networks
(PERSIANN) system, which calculates rain rate at a 0.25° × 0.25° latitude–longitude resolution, based on the brightness temperature variations in a neighboring coverage of 1.25° × 1.25°.
15.2.2.3 Cloud Patch–Based Algorithms
Cloud patch–based approaches estimate rainfall based on information extracted
from the entire cloud coverage determined by various segmentation methods. One
early example of patch-based algorithms is the Griffith–Woodley technique (Griffith
et al. 1978), which segments a cloud patch with an IR temperature threshold of 253 K
and tracks it through its life cycle. The Convective–Stratiform technique (Adler and
Negri 1988) is another example of the cloud patch–based approach. It screens convective cells based on the local minimum of IR temperature and assigns different
rainfall amounts to convective and stratiform components separately. Pixel rain rates
are proportionally distributed starting from the coldest pixel to higher temperature
pixels. Xu et al. (1999) proposed another approach, which determines different
temperature thresholds by separating the rain/no-rain pixels in a cloud patch using
SSM/I MW rainfall estimates.
On the basis of the previous cloud-window-based PERSIANN system, Hong et
al. (2005) developed an automated neural network for cloud patch–based rainfall
estimation. The self-organizing nonlinear output (SONO) model accounts for the
high variability of cloud–rainfall processes at geostationary scales (i.e., 4 km and
every 30 min). Instead of calibrating only one IR–RR function for all clouds, SONO
classifies varied cloud patches into different clusters and then searches a nonlinear
IR–RR mapping function for each cluster. This designed feature enables SONO
to generate various rain rates at a given brightness temperature and variable rain/
no-rain IR thresholds for different cloud types, which overcomes the one-to-one
mapping limitation of a single statistical IR–RR function for the full spectrum of
cloud–rainfall conditions. In addition, the computational and modeling strengths of
neural networks enable SONO to cope with the nonlinearity of cloud–rainfall relationships by fusing multisource datasets. Evaluated at various temporal and spatial
scales, SONO shows improvements of estimation accuracy, both in rain intensity
and in the detection of rain/no-rain pixels. Further examination of SONO adaptability demonstrates its potential as an operational satellite rainfall estimation system