Short-Term Prediction for Earth-Space Link Failure Due to Rain Using Weather Radar Maps

One of the keys to the future of satellite communications is the Ka and EHF frequency bands. Unfortunately, propagation impairment due to rain of these bands is very severe. Only diversity-based techniques, e.g., site diversity, can mitigate such effect. Although a failure prediction method is essential for the site diversity, its current designs are mostly based on data collected at the site locations. From the real-world experience, this approach can be significantly improved. In this work, a method of a short-term prediction for the link failure due to rain using weather radar maps is proposed. This method is based on the simple digital image processing that can predict several minutes before the link failure. Using available online maps from the Nongchok station, Thailand, with 1-year period, it was found that the performance with 82.8% hit rate and 70.6% precision can be achieved for predicting 5- 30 minutes (25-minute interval) before the link failure. The performance could be traded-off or improved by elaborating the design. These show that the proposed method, with optimized parameters, is promising for the real-time operation of the site diversity systems.