|
The retrieval of MTSAT multi-spectral satellite image rainfall intensity field was studied, with which the "Unit-Feature
Spatial Classification (UFSC) method" was proposed to become the foremost basis of the possibility of continuous
observation of real-time precipitation from geostationary satellite. In this method, MTSAT multi-spectral satellite
measured value and measured precipitation rate from high density ground stations of plum rain season in east china
(Jiangsu Province, Zhejiang Province and Anhui Province) in 2007 are combined to conduct the cooperative analysis,
and therefore the distribution features of the level of each precipitation probability and each precipitation intensity are
well established on different two-dimensional and three-dimensional spectral feature spaces. On the basis, the
discrimination matrices, correspondingly, are established for precipitation probability and precipitation intensity of
different spectral combinations. Different spectral combinations are used for the construction of the discrimination
matrices of the day and the night, respectively. For the day, IR1 (11µm), IR3 (6.7μm), VIS (0.7m), IR12 (TIR2-IR1) and
IR13 (TIR3-IR1) are available, among which IR1, VIS and IR3 (or IR13) are mainly used ; for the night, IR1, IR3, IR4
(3.7μm), IR12, IR13, IR14 (TIR4-IR1)and IR24 (TIR4-IR2) are available and IR1, IR3 and IR24 (or IR14) are mainly used.
The contrast test between the observed data of precipitation and the retrieval results based on precipitation data from
basic stations and reference stations in China in 2007 shows that, 30% precipitation probability can ideally distinguish
precipitation area from non-precipitation area; and the analysis of precipitation intensity category also matches well with
the fact. It is well known that the observation of satellite is instantaneous one time per hour while the rain gauge
observation is an accumulative process during an hour. The error study further suggests that the difference between the
two observation methods is the vital cause of the relative error.
|
