Development of flood control supporting system based on Severe Rainfall prediction system using Artificial Intelligence (SRAI)

本研究的目的是制作防洪保障系统，并将其应用到实际的洪水中。利用作为系统核心的人工智能nee （SRAI）来提高强降雨预报系统的精度是实现强降雨预报的必要条件。SRAI是利用雷达、GMS图像和数值预报资料作为初始值，对小面积、短期阴雨云的爆发和发展进行定性预报的方法。GPV采用RSM （Regional Spectral Model，区域光谱模型）数据作为初始值，难以达到预测精度提升时间。介绍了新一代数值气象模式（5km空间分辨率每小时）NHM（气象研究所/数值预报部非静水模式）的预报数值，并提出了提高预报精度的建议。2 - 3小时以上降雨预报的准确性得到了提高。此外，由于降雨预报精度的提高是可以预期的，所以在研究精度提高的过程中，使用NHM预报的“云水”与“雨水”的混合比作为SRAI的初始值。因此，现阶段的SRAI是一个利用NHM上的全部信息，并对其进行处理，实现短时降水预报的系统。然而，由于大量的时间被用于开发，应用案例无法增加。通过增加未来的应用案例来提高对用户的可靠性将被认为是一个问题。值得自豪的是，通过克服NHM（不能预测100%的降雨量，解译需要很高的气象知识，数据量大）的问题，将数值气象模式预报应用到河流和水库的管理点，是本研究的一大成果。少

The purpose of this research is in making the flood control support system, and the application of this to an actual flood. The accuracy improvement of Severe Rainfall prediction system using Artificial Intellige nee (SRAI) which is the kernel of the system was done became indispensable for that. SRAI is the method to qual itatively forecast the breaking out and development of the rainy cloud of small area and short term, by using the radar, the GMS image, and the numerical prediction data as an initial value.It was difficult for GPV to achieve the forecast accuracy improvement time because it used the data from RSM (Regional Spectral Model) as an initial value. Then, the forecast value of the NHM (Meteorological Research Institute/ Numerical Prediction Division Non Hydrostatic Model), which was the next generation numeric meteorological model (5km space resolution every hour) was introduced, and the improvement of the forecast accuracy was aimed at. The accuracy of rainfall forecast o … More f two-three hours has been improved.In addition, because the improvement of the rainfall forecast accuracy was able to be expected, the mixing ratio of "cloud water" and "rain water" which NHM forecast was made to be used as an initial value of SRAI in the process of the research on the accuracy improvement, has been used. Therefore, SRAI in the present stage is a system which can forecast the rainfall for a short time by taking, and processing all information on NHM. However, because a lot of time had been spared to development, the application case was not able to be increased. It will be assumed to be a problem to improve reliability to the user by increasing the application case in the future.It has pride that making an application for the forecast by a numeric meteorological model to the management site in the river and the reservoir by overcoming the problem of NHM (The rainfall of 100 % cannot be forecasted, necessary of highly meteorological knowledge is necessary for the interpretation, and a large amount of data) by using NHM at the end is a big result in this research. Less