| Summary: | 博士 === 國防大學理工學院 === 國防科學研究所 === 97 === This study is to quantitatively estimate precipitation associated with typhoons over ocean using the microwave data of TMI on board TRMM and eleven isolated islands rain gauges near south of Japanese to establish multi-channel linear regression equation by statistical method. The procedure is at first to classify the rain or no-rain areas by using the CC (Combination Check) method, a combination of the SI (Scattering Index) method and TC (Threshold Check) method, and then to distinguish rain types into convective and stratiform rain. That is to estimate convective and stratiform rain rates, respectively. The result shows that overall successful classification of rain and no-rain areas from 2002 to 2004 are 99.4%, 100% and 100% respectively. The coefficient of correlation is 0.74 between quantitatively estimated rain rates and ground truth of rain gauges for oceanic validation. The Root-Mean-Square (RMS) is 3.75 mm/h. Besides, the satellite’s rain rate estimated is overestimated for weak precipitation system. Oppositely, the satellite’s rain rate is underestimated for heavy precipitation system.
This study compares the retrieval rainfall with 2A12 product of GPROF (Goddard Profiling Algorithm) created by physical method through the validation of rain gauges. The result of this study is better than that of the 2A12, It seems that the near surface rainfall of GPROF is only suitable for global scale but insufficient for the local region. Simultaneously, the GPI (GOES Precipitation Index) is applied by combining the microwave and infrared data for calibrating the rainfall threshold of infrared and then continously estimated the rain rate of typhoons by using the high-spatial-resolution of MTSAT-1R (Multi-functional Transport Satellite-1R) data. The total GPI rain rate can be used to indirectly estimate the typhoon’s intensity over sea before affecting the land and becomes an index of typhoon’s rainfall intensity. Results show that for GPI algorithm, combining microwave and infrared data, the correlation between rain rate and FC (Fractional Coverage) for the 1o×1o domain is the best than those for the other domains. But, the spatial resolution seems to be insufficient.
In addition, the new Bayesian approach has been developed to retrieve oceanic rain rate from TMI. The Bayesian approach will create a continuous posterior probability distribution of rain rate by combining a prior distribution of observations and conditional distribution which be derived by a physical model. We utilize an attenuation index with a difference of polarization between rainy and cloud free to estimate rain rate over ocean. We can obtain the advantage of normalized polarization to avoid the saturation of microwave channels happened through the new Bayesian approach. Due to reduction of spatial high-dimension for calculation, the attenuation index can save computer’s times compared with multi-channel brightness temperatures. Retrieved rain rates are validated with measurements of rain gauges located on Japanese islands. To demonstrate the improvement, retrievals are also compared with those from the TRMM/Precipitation Radar (PR), the GPROF, and a multi-channel linear regression statistical method (MLRS). It is found that, qualitatively, all methods retrieve similar horizontal distributions in terms of locations of eyewalls and rain bands of typhoons. Quantitatively, our new Bayesian retrievals have the best linear relationship with rain gauge and the smallest RMS error. The correlation coefficient and RMS of our retrievals are 0.95 and ~2 mm/h, respectively. In particular, at heavy rain rates, our Bayesian retrievals outperform those retrieved from GPROF and MLRS.
|
|---|
