Atmospheric Water Vapor Tomography based on GNSS and Multi-band Remote Sensing Measurements Combination

A three-dimensional water vapor density field has advantages in monitoring atmospheric water vapor, especially for reflecting the vertical motion. The existing multi-source tomography models are around a fusion of single-source remote-sensing signal and Global Navigation Satellite System (GNSS) data. However, different remote-sensing data have advantages and disadvantages regarding spatiotemporal resolution and accuracy. When only single-source remote-sensing data is integrated for tomography, the model's available scenarios are severely limited by weather conditions. Therefore, we construct a tomography model by fusing multi-band spaceborne remote-sensing data and high-precision ground GNSS data, the former includes near-infrared MODIS image, long-wave infrared FengYun-4A image, and morphed integrated microwave image MIMIC. The equations system of the tomographic model is solved based on different strategies of weight determination using the weighted least square algorithm. In addition, to consider the dynamic variations of tropopause height in the research area, the tropopause detection products of Fengyun-4B with high spatial coverage are used to determine the boundary of the tomographic region, and the constraints of model is built by historical GNSS occultation observations. To verify our method, the proposed model is validated by water vapor density from reanalysis and radiosonde data, respectively. The results show that the reasonable prior weights are essential when using multi-source data to perform tightly coupled tomography, the RMSEs of water vapor density are less than 2 g/m³ in most epochs. Compared to the tomographic model based on only GNSS data, the accuracy improvement of the tomographic model fusing multi-band remote sensing data is higher than that of any tomographic model using single-source remote-sensing data. Also, the proposed tomography model can better compensate for the shortcomings of poor time continuity of integrated individual remote-sensing data to expand the application scenarios of the fusion tomographic model.

Acknowledgments: This work was supported by Natural Science Foundation of China (42192534 and 42388102).