In the real world, precipitation data of automatic weather stations are easily influenced by direct thunderstrokes, instrument ageing, electromagnetic interference, human operation errors and other factors. When close to the observation time, if the missing automatic station data cannot be corrected in a timely fashion, the whole quality of the station data will be affected. Thus, correct handling of the missing precipitation data to maintain their integrity has important significance. In this paper, we propose a “from coarse to fine” (FCTF) neural network to fill out the missing blanks and experiments show that this method to solve the problem of meteorological data shortage is effective.

中文翻译：

---

使用改进的神经网络组成的汽车站降水数据

在现实世界中，自动气象站的降水数据很容易受到直接雷击，仪器老化，电磁干扰，人为操作错误和其他因素的影响。当接近观测时间时，如果不能及时纠正丢失的自动台站数据，则将影响台站数据的整体质量。因此，正确处理丢失的降水数据以保持其完整性具有重要意义。在本文中，我们提出了一种“从粗到细”（FCTF）神经网络来填补缺失的空白，并且实验表明，该方法对于解决气象数据不足的问题是有效的。