The effects of site diversity techniques on Ku-band rain attenuation are investigated using two kinds of simultaneous BS (Broadcasting Satellite) signal observations: one was conducted among Osaka Electro-Communication University (OECU) in Neyagawa, Kyoto University in Uji, and Shigaraki MU Observatory in Koka for the past ten years, and the other was conducted among the headquarter of OECU in Neyagawa and their other premises in Shijonawate and Moriguchi for the past seven years, respectively. The site diversity effects among these sites with horizontal separations of 3-50 km are found to be largely affected by the passage direction of rain areas characterized by each rain type, such as warm, cold, and stationary fronts or typhoon and shower. The performance of the site diversity primarily depends on the effective distance between the sites projected to the rain area motions. The unavailable time percentages are theoretically shown to be reduced down to about 61-73% of the ITU-R predictions by choosing a pair of the sites aligned closest to the rain area motion in the distance of 3-50 km. Then, we propose three kinds of novel site diversity methods that choose the pair of sites based on such as rain type, rain front motion, or rain area motion at each rainfall event, respectively. As a result, the first method, which statistically accumulates the average passage directions of each rain type from long-term observations, is even useful for practical operations of the site diversity, because unavailable time percentages are reduced down to about 75-85% compared with the theoretical limit of about 61-73%. Also, the third method based on the rain area motion directly obtained from the three-site observations yields the reduction in unavailable time percentages close to this theoretical limit.