We present a multi-sensor archive collecting spatial and temporal information about volcanic SO₂ clouds generated by the 11 largest eruptions of this century. The detection and monitoring of volcanic clouds are an important topic for aviation management, climate issues and weather forecasts. Several studies focusing on single eruptive events exist, but no archive available at the moment combines quantitative data from as many instruments. We archived and collocated the SO₂ vertical column density estimations from three different satellite instruments (AIRS, IASI and GOME-2), atmospheric parameters as vertical profiles from the Global Navigation Satellite Systems (GNSS) Radio Occultations (RO), and the cloud-top height and aerosol type from the Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP). Additionally, we provide information about the cloud-top height from three different algorithms and the atmospheric anomaly due to the presence of the cloud. The dataset is gathering 206 d of SO₂ data, collocated with 44 180 backscatter profiles and 64 764 radio occultation profiles. The modular structure of the archive allows an easy collocation of the datasets according to the users' needs, and the cross-comparison of the datasets shows different consistency of the parameters estimated with different sensors and algorithms, according to the sensitivity and resolution of the instruments. The data described here are published with a DOI at https://doi.org/10.5880/fidgeo.2020.016 (Tournigand et al., 2020a).

Precise and reliable information on the tropospheric temperature and water vapor profiles play a key role in weather and climate studies. Among the sensors supporting the observations of the troposphere, one can distinguish the Global Navigation Satellite System Radio Occultation (RO) technique, which provides accurate and high-quality meteorological profiles. However, external knowledge about temperature is essential to estimate other physical atmospheric parameters. To overcome this constraint, I trained and evaluated four different machine learning models comprising Artificial Neural Network (ANN) and Random Forest regression algorithms, where no auxiliary meteorological data is needed. To develop the models, I employed 150,000 globally distributed (45°S–45°N) RO profiles between October 2019 and December 2020. Input vectors consisted of bending angle or refractivity profiles from the Formosa Satellite-7/Constellation Observing System for Meteorology, Ionosphere, and Climate-2 mission together with the month, hour, and latitude of the RO event. While temperature, pressure, and water vapor profiles derived from the modern ERA5 reanalysis and interpolated to the RO location served as the models’ targets. Evaluation on the testing data set revealed a good agreement between all model outputs and ERA5 targets, where slightly better statistics were noted for ANN and refractivity inputs. Vertically averaged root mean square error (RMSE) did not exceed 1.7 K for the temperature and reached around 1.4 hPa and 0.45 hPa for the total and water vapor pressures. Additional validation with 477 co-located radiosonde observations and the operational one-dimensional variational product showed slightly larger discrepancies with the mean RMSE of around 1.9 K, 1.9 hPa, and 0.5 hPa for the temperature, pressure, and water vapor, respectively.Graphical

Tropical cyclones (TC) are natural destructive phenomena, which affect wide tropical and subtropical areas every year. Although the correct prediction of their tracks and intensity has improved over recent years, the knowledge about their structure and development is still insufficient. The Global Navigation Satellite System (GNSS) radio occultation (RO) technique can provide a better understanding of the TC because it enables us to probe the atmospheric vertical structure with high accuracy, high vertical resolution and global coverage in any weather conditions. In this work, we create an archive of co-located TC best tracks and RO profiles covering the period 2001–2018 and providing a complete view of the storms since the pre-cyclone status to the cyclone disappearance. We collected 1822 TC best tracks from the International Best Track Archive for Climate Stewardship and co-located them with 48 313 RO profiles from seven satellite missions processed by the Wegener Center for Climate and Global Change. We provide information about location and intensity of the TC, RO vertical profiles co-located within 3 h and 500 km from the TC eye centre, and exact information about temporal and spatial distance between the TC centre and the RO mean tangent point. A statistical analysis shows how the archive covers all the ocean basins and all the intensity categories well. We finally demonstrate the application of this dataset to investigate the vertical structure for one TC example case. All the data files, separately for each TC, are publicly available in NetCDF format at https://doi.org/10.25364/WEGC/TC-RO1.0:2020.1 (Lasota et al., 2020).

Tropical cyclones (TC) are one of the main producers of clouds in the tropics and subtropics. Hence, most of the clouds in TCs are dense, with large water and ice content, and provide conditions conducive to investigate clouds’ impact on Radio Occultation (RO) measurements. Although the RO technique is considered insensitive to clouds, recent studies show a refractivity positive bias in cloudy conditions. In this study, we analyzed the RO bending angle sensitivity to cloud content during tropical cyclone seasons between 2007 and 2010. Thermodynamic parameters were obtained from the ERA-Interim reanalysis, whereas the water and ice cloud contents were retrieved from the CloudSat profiles. Our experiments confirm the positive mean RO refractivity bias in cloudy conditions that reach up to more than 0.5% at the geometric height of around 7 km. A similar bias but larger and shifted up is visible in bending angle anomaly (1.6%). Our results reveal that the influence of clouds is significant and can exceed the RO bending angle standard deviation for 21 out of 50 (42%) investigated profiles. Mean clouds’ impact is detectable between 9.0 and 10.5 km, while, in the case of single events, clouds in most of the observations are significant between 8 and 14 km. Almost 15% of the detectable clouds reach 16 km height, while the influence of the clouds below 5 km is insignificant. For more than half of the significant cases, the detection range is less than 3 km but for one observation this range spreads to 7–8 km.

Monitoring atmospheric conditions that lead to severe weather events is critical to their timely and accurate prediction and can help prevent of large economic losses. Bulgaria, located in southeastern Europe, has the highest mean number of thunderstorms and hailstorms. These events generally occur between April and September with a peak in July. In this study, both radio occultation (RO) and ground-based observations from the Global Navigation Satellite Systems (GNSS) were used to study two severe hailstorms that occurred in 2014 and 2019. In both storms, a cold upper-air pool was detected in addition to a large specific humidity anomaly between 2 and 6 km. In the hailstorm that occurred in July 2014, there was an RO temperature anomaly between 10 and 14 km as well as a positive specific humidity anomaly between 4 and 6 km. The integrated vapor transport (IVT) reanalysis from ERA5, indicated that the high specific humidity over the Mediterranean could be tracked to an atmospheric river over the North Atlantic, which was connected to a tropical cyclone. In the hailstorm that occurred in May 2019, elevated IVT values were observed before the storm. During this storm, a negative temperature anomaly peak was observed in the RO profile at 11.3 km as well as a positive specific humidity anomaly between 2 and 4.5 km. The WRF Model and the ERA5 dataset could reproduce the temperature profiles for both storms relatively well; however, they tended to underestimate specific humidity. The RO profiles were complemented by ground-based GNSS tropospheric delays with high temporal resolution. The evaluation of the WRF with ground-based GNSS tropospheric products revealed a time delay between the modeled and observed developments of both hailstorms.

Environmental sensor networks play a crucial role in monitoring key parameters essential for understanding Earth’s systems. To ensure the reliability and accuracy of collected data, effective quality control (QC) measures are essential. Conventional QC methods struggle to handle the complexity of environmental data. Conversely, advanced techniques such as neural networks are typically not designed to process data from sensor networks with irregular spatial distribution. In this study, we focus on anomaly detection in environmental sensor networks using graph neural networks, which can represent sensor network structures as graphs. We investigate its performance on two datasets with distinct dynamics and resolution: commercial microwave link (CML) signal levels used for rainfall estimation and SoilNet soil moisture measurements. To evaluate the benefits of incorporating neighboring sensor information for anomaly detection, we compare two models: graph convolution network (GCN) and a graph-less baseline: long short-term memory (LSTM). Our robust evaluation through a five-fold cross validation demonstrates the superiority of the GCN models. For CML, the mean area under receiver operating characteristic curve for the GCN was 0.941 compared to 0.885 for the baseline LSTM, and for SoilNet, it was 0.858 for GCN and 0.816 for the baseline LSTM. Visual inspection of CML time series revealed that the GCN proficiently classified anomalies and remained resilient against rain-induced events often misidentified by the baseline LSTM. However, for SoilNet, the advantage of GCN was less pronounced, likely due to inconsistent and less precise labeling. Through interpretable model analysis, we demonstrate how feature attributions vividly illustrate the significance of neighboring sensor data, particularly in distinguishing between anomalies and expected changes in the signal level in the time series.