长江流域多尺度时变重力场建模及洪水预警方法研究

发展高精度多尺度区域时变重力场模型，并提升其在洪水监测与预警中的应用能力，是当前大地测量学和水文学这一交叉学科的研究热点与难点。针对当前全球重力场模型的时空尺度过于单一而不能满足洪水监测与预警需求的问题，本项目融合重力卫星沿轨观测数据和区域先验信息，重点研究基于径向基函数的Mascon解算方法、不同时间尺度下的Mascon格网自适应划分方法、基于先验地球物理信息的Kalman滤波方程构建方法，研制并发布年、月、周等多种时间和空间尺度的区域重力场Mascon解；基于多尺度时变重力场模型和多源观测数据，重点分析长江流域各个子流域洪水发生的时延特性，并结合递归神经网络算法这一计算机深度学习技术，构建长江流域的洪水预警方法。预期研究成果不仅为应对长江流域洪水灾害提供数据支撑和技术支持，而且有望提高我国重力卫星数据处理能力，具有重要的科学意义和实用价值。

Developing the regional high-accuracy and multi-scale temporal gravity field model and improving its capacity in flood detecting and warning is still one of the key issues in geodesy and hydrology. Aiming at handling the excessively single temporal and spatial scale of current temporal gravity field models and improving the capacity of flood detecting and warning, we plan to combine the satellite gravimetric data and regional prior observations, build the direct relationship between range rates and mascon solutions via radial basis function, develop the self-adaptive method of mascon grid division for different temporal and spatial scales, develop the Kalman filter method by using the prior geophysical models, and finally develop and publish the regional multi-scale mascon solutions with yearly, monthly and weekly temporal scale and with different spatial scale. Based on the multi-scale mascon solutions and prior geophysical models, we plan to analyze the time lags of flood events over the different sub-basins of the Yangtze River Basin (YRB). Moreover, the recursive neural network, which is one of deep learning technologies, will be used to improve the ability of flood warning over the YRB. The expected achievements will not only provide data support and scientific evidence for dealing with floods over the YRB, but also improve the data processing ability for satellite gravimetry in China.