现有的全球卫星臭氧总量数据在精度、时空连续性和一致性等方面存在制约，严重影响其后续应用研究。针对这些问题，本项目拟发展全球长时序高精度臭氧数据集重建研究的理论与方法。具体地，提出单传感器臭氧数据的缺失重建模型，利用多时相数据建立局部时空加权回归关系，引入各向异性变差函数进行残差估计，实现卫星臭氧总量数据的无缝重建；发展多传感器臭氧数据的归一化模型，以高精度的数据集为基准，利用多传感器不同数据集的重合时段，订正不同传感器间的相对偏差，实现多传感器数据的时空一致性；引入影响卫星观测精度的多个参量，构建基于广义回归神经网的误差校正模型，提升卫星臭氧总量数据精度。模型通过精度验证后，最终生成一套近42年高精度的全球时空无缝臭氧总量数据集，为全球臭氧总量的时空分布及演变规律研究提供数据支撑。

The existing remotely sensed time series of ozone data often suffer spatio-temporal discontinuous, inconformity and low observation quality, which severely hinder the subsequent application and analysis. To solve these problems, this project intends to develop a set of theory and method for the reconstruction of total ozone data set with global coverage in 1979-2020. Concretely, to achieve spatio-temporally seamless data from different single sensor, this project proposes a reconstruction model, in which locally spatio-temporal weighted regression relationship between multi-temproal ozone data is established, and then residual value is evaluated by introducing anisotropy variogram function. To realize the spatio-temporal consistency of total ozone data from multi-sensor, this project develops a normalized model, with high precision data from OMPS as baseline data, using the overlapping time of different data sets from multiple sensors, relative differences between different data sets from multi-sensor is then eliminated. To improve the accuracy of remotely sensed total ozone data, this project builds an error correction model based on generalized regression neural network (GRNN) by introducing several parameters affecting the observation precision. On the basis of the accuracy verification, a set of spatio-temporally seamless total ozone data set with a high accuracy and global coverage in last 42 years will be generated, which will provides data support for the study of the spatio-temporal distribution and evolution rule of global total ozone.