Precipitable water vapor (PWV), as an important representative parameter of atmospheric water vapor contents, can be obtained by means of Global Navigation Satellite Systems (GNSS) using both ground-based and space-borne observation techniques. However, the PWV prediction models currently accessible tend to be simplistic combinations or individual models. In this study, we develop a WOA-CNN-LSTM-Attention model to predict PWV, which takes the sixteen GNSS PWV values near the HKKP station as characteristic parameters and the spatial relationship between the point of interest and its neighboring GNSS stations into consideration. An optimal model via the whale optimization algorithm (WOA) is investigated by using a wavelet analysis to separate noises, through combining convolutional neural network (CNN), long short-term memory neural network (LSTM) and attention mechanism. Results show that considerable improvement in the prediction accuracy has been achieved through a comparison between CNN-LSTM-Attention and the conventional LSTM and CNN-LSTM models. In terms of long-term predictability, CNN-LSTM-Attention is proven to be a superior model when 8 features are incorporated. The model’s root mean square error (RMSE) is 2.30 mm which is reduced by 20.42 % than in the case of 0 feature is used. As a further analysis, we also examine the prediction performance of various models for hourly PWV using 7, 15, 30, 60 and 90 days of data as different lengths of training. The results show that CNN-LSTM-Attention has a better prediction effect when the training length is 30 days, the RMSE is 0.74 mm and the Nash-Sutcliffe efficiency coefficient (NSE) is 0.98.

Small object detection in large-scale remote sensing images (RSIs) is crucial for military and civil applications, but it remains challenging. Since small objects occupy few pixels, their features are easily interfered with by complex backgrounds and large objects. In addition, they are susceptible to localization offsets, which are prone to false or missed detections as there are few predicted bounding boxes matching the ground truth. To overcome these issues, this article proposes a filter progressive small object detection (FPSOD) model that is based on the progressive mechanism. With the proposed attention-based soft-threshold filtering module, FPSOD significantly filters out redundant information in high-level feature maps thus enhancing the semantic features of small objects. Furthermore, a progressive regression loss (PR-Loss) function is proposed to facilitate the precise localization, which mitigates predicted bounding box drift by limiting the fluctuated range of the gradients. The experimental results show that the proposed model substantially improves the precision and recall of small objects, effectively reduces missed detections, and improves detection performance.

Layer precipitable water (LPW), a water vapor product similar to precipitable water vapor (PWV), reports partial moisture content within a specified vertical range. Compared with PWV data, the latest LPW products can describe more refined distributions and variations in water vapor in the troposphere. Global Navigation Satellite Systems (GNSSs), as a powerful water vapor sensing tool, only provide the opportunity to retrieve all-weather PWV, not LPW products. To this end, we develop the first deep-learning-assisted, global water vapor stratification (GWVS) model to estimate the GNSS LPW within any given vertical range. The proposed model is trained and tested using the global radiosonde data, with the training and testing root mean square error (RMSE) of 0.94 and 1.10 mm for radiosonde LPW, indicating the excellent generalization of the GWVS model. Furthermore, the model is comprehensively validated using the data from the two regional GNSS networks and one global network. The RMSEs of the predicted GNSS LPW from the three GNSS networks compared with the co-located radiosonde LPW are 1.52, 1.80, and 1.54 mm, respectively. To study potential applications, we use the model-derived GNSS LPW products to calibrate Geostationary Operational Environmental Satellite-16 (GOES-16) LPW products and improve the GNSS water vapor tomography technique. Results show that the accuracy of three GOES-16 LPW products is improved by 31.3%, 23.3%, and 17.9%, respectively, and the RMSE of the tomography results is reduced from 2.28 to$1.67~\text {g}/\text {m}^{3}$. Both validation and application results highlight that the GWVS model retrieves the required GNSS LPW products and provides additional value for water-vapor-related studies.

The Global Navigation Satellite Systems (GNSS) water vapor tomography technique has been successfully used as a promising tool for sensing atmospheric water vapor and applied to weather forecasting in recent years. In most tomography models, the single grid resolution, i.e., the same horizontal resolution, is widely adopted to divide the three-dimensional (3D) tomographic domain into many small voxels. However, the single resolution GNSS tomography (SRGT) method implements the grid-based parametrization of the physical domain and does not follow the vertical spatial heterogeneity of atmospheric water vapor. To this end, we develop a new multi-resolution GNSS tomography (MRGT) method that incorporates the vertical decline tendency of water vapor. The MRGT method generates different resolution tomographic water vapor products in the lower, middle, and upper domains of the troposphere. Besides, a new indicator, known as the integrated water vapor (IWV) lapse rate, is introduced to determine the appropriate non-uniform stratification strategy. Eight tomography schemes were analyzed to compare the tomography results obtained from different tomography models based on the GNSS data in Hong Kong region during June and July 2015. The results show that with respect to radiosonde data, the MGRT method reconstructs a more accurate water vapor distribution than the SRGT approach, with the average root mean square error of tomographic results reduced by 12%. Moreover, in rainfall conditions, the tomographic water vapor profiles from the MGRT model agree well with the radiosonde profiles, which highlights the potential of multi-resolution tomographic water vapor products for rainfall-related studies.

High spatiotemporal resolution atmospheric water vapor can be retrieved using the Global Navigation Satellite System (GNSS) tomography technique, in which the remained ill-posed problem of the tomography system resulting from the acquisition geometry is a vital issue to be addressed. Remote sensing (RS) water vapor data, with high-resolution and global coverage, show great potential for retrieval of slant water vapor (SWV) observations to improve the tomographic geometrical distribution. In this article, we develop a GNSS-RS (GNSS combining RS) tomography model to fully exploit the value of observation signals from GNSS and RS measurements. The two key factors of retrieving the RS SWV are performed by calibrating the original precipitable water vapor (PWV) images and adding the tropospheric horizontal gradients. The results reveal that when introducing the RS SWV observations into the tomography model, the acquisition geometry is significantly improved, with the average rate of voxels crossed by rays from 62% to 95% and the mean number of observation signals from 395 to 508 during the tomographic periods. Independent radiosonde data are used to validate the tomographic water vapor fields. The mean root-mean-square error (RMSE) and bias of the water vapor profiles derived from GNSS-RS solutions are decreased by 28% and 45% with respect to the GNSS-only results, respectively. Such improvements highlight that GNSS-RS troposphere tomography has significant potential to improve the reconstruction of the atmospheric water vapor fields.