LibRadtran is a widely used software package for radiative transfer calculations. It allows one to compute (polarized) radiances, irradiance, and actinic fluxes in the solar and thermal spectral regions. libRadtran has been used for various applications, including remote sensing of clouds, aerosols and trace gases in the Earth's atmosphere, climate studies, e.g., for the calculation of radiative forcing due to different atmospheric components, for UV forecasting, the calculation of photolysis frequencies, and for remote sensing of other planets in our solar system. The package has been described in Mayer and Kylling (2005). Since then several new features have been included, for example polarization, Raman scattering, a new molecular gas absorption parameterization, and several new parameterizations of cloud and aerosol optical properties. Furthermore, a graphical user interface is now available, which greatly simplifies the usage of the model, especially for new users. This paper gives an overview of libRadtran version 2.0.1 with a focus on new features. Applications including these new features are provided as examples of use. A complete description of libRadtran and all its input options is given in the user manual included in the libRadtran software package, which is freely available at http://www.libradtran.org.

The role of clouds remains the largest uncertainty in climate projections. They influence solar and thermal radiative transfer and the earth's water cycle. Therefore, there is an urgent need for accurate cloud observations to validate climate models and to monitor climate change. Passive satellite imagers measuring radiation at visible to thermal infrared (IR) wavelengths provide a wealth of information on cloud properties. Among others, the cloud top height (CTH) – a crucial parameter to estimate the thermal cloud radiative forcing – can be retrieved. In this paper we investigate the skill of ten current retrieval algorithms to estimate the CTH using observations from the Spinning Enhanced Visible and InfraRed Imager (SEVIRI) onboard Meteosat Second Generation (MSG). In the first part we compare ten SEVIRI cloud top pressure (CTP) data sets with each other. The SEVIRI algorithms catch the latitudinal variation of the CTP in a similar way. The agreement is better in the extratropics than in the tropics. In the tropics multi-layer clouds and thin cirrus layers complicate the CTP retrieval, whereas a good agreement among the algorithms is found for trade wind cumulus, marine stratocumulus and the optically thick cores of the deep convective system. In the second part of the paper the SEVIRI retrievals are compared to CTH observations from the Cloud–Aerosol LIdar with Orthogonal Polarization (CALIOP) and Cloud Profiling Radar (CPR) instruments. It is important to note that the different measurement techniques cause differences in the retrieved CTH data. SEVIRI measures a radiatively effective CTH, while the CTH of the active instruments is derived from the return time of the emitted radar or lidar signal. Therefore, some systematic differences are expected. On average the CTHs detected by the SEVIRI algorithms are 1.0 to 2.5 km lower than CALIOP observations, and the correlation coefficients between the SEVIRI and the CALIOP data sets range between 0.77 and 0.90. The average CTHs derived by the SEVIRI algorithms are closer to the CPR measurements than to CALIOP measurements. The biases between SEVIRI and CPR retrievals range from −0.8 km to 0.6 km. The correlation coefficients of CPR and SEVIRI observations vary between 0.82 and 0.89. To discuss the origin of the CTH deviation, we investigate three cloud categories: optically thin and thick single layer as well as multi-layer clouds. For optically thick clouds the correlation coefficients between the SEVIRI and the reference data sets are usually above 0.95. For optically thin single layer clouds the correlation coefficients are still above 0.92. For this cloud category the SEVIRI algorithms yield CTHs that are lower than CALIOP and similar to CPR observations. Most challenging are the multi-layer clouds, where the correlation coefficients are for most algorithms between 0.6 and 0.8. Finally, we evaluate the performance of the SEVIRI retrievals for boundary layer clouds. While the CTH retrieval for this cloud type is relatively accurate, there are still considerable differences between the algorithms. These are related to the uncertainties and limited vertical resolution of the assumed temperature profiles in combination with the presence of temperature inversions, which lead to ambiguities in the CTH retrieval. Alternative approaches for the CTH retrieval of low clouds are discussed.

Predictions of thunderstorm‐related hazards are needed in several sectors, including first responders, infrastructure management and aviation. To address this need, we present a deep learning model that can be adapted to different hazard types. The model can utilize multiple data sources; we use data from weather radar, lightning detection, satellite visible/infrared imagery, numerical weather prediction and digital elevation models. We demonstrate the ability of the model to predict lightning, hail and heavy precipitation probabilistically on a 1 km resolution grid, with a temporal resolution of 5 min and lead times up to 60 min. Shapley values quantify the importance of the different data sources, showing that the weather radar products are the most important predictors for all three hazard types.

In order to aid feature selection in thunderstorm nowcasting, we present an analysis of the utility of various sources of data for machine-learning-based nowcasting of hazards related to thunderstorms. We considered ground-based radar data, satellite-based imagery and lightning observations, forecast data from numerical weather prediction (NWP) and the topography from a digital elevation model (DEM), ending up with 106 different predictive variables. We evaluated machine-learning models to nowcast storm track radar reflectivity (representing precipitation), lightning occurrence, and the 45 dBZ radar echo top height that can be used as an indicator of hail, producing predictions for lead times of up to 60 min. The study was carried out in an area in the Northeastern United States for which observations from the Geostationary Operational Environmental Satellite-16 are available and can be used as a proxy for the upcoming Meteosat Third Generation capabilities in Europe. The benefits of the data sources were evaluated using two complementary approaches: using feature importance reported by the machine learning model based on gradient-boosted trees, and by repeating the analysis using all possible combinations of the data sources. The two approaches sometimes yielded seemingly contradictory results, as the feature importance reported by the gradient-boosting algorithm sometimes disregards certain features that are still useful in the absence of more powerful predictors, while, at times, it overstates the importance of other features. We found that the radar data is the most important predictor overall. The satellite imagery is beneficial for all of the studied predictands, and therefore offers a viable alternative in regions where radar data are unavailable, such as over the oceans and in less-developed ares. The lightning data are very useful for nowcasting lightning but are of limited use for the other hazards. While the feature importance ranks NWP data as an important input, the omission of NWP data can be well compensated for by using information in the observational data over the nowcast period. Finally, we did not find evidence that the nowcast benefits from the DEM data.

The complex problem of quantitative precipitation estimation in the Alpine region is tackled from four different points of view: (1) the modern MeteoSwiss network of automatic telemetered rain gauges (GAUGE); (2) the recently upgraded MeteoSwiss dual-polarization Doppler, ground-based weather radar network (RADAR); (3) a real-time merging of GAUGE and RADAR, implemented at MeteoSwiss, in which a technique based on co-kriging with external drift (CombiPrecip) is used; (4) spaceborne observations, acquired by the dual-wavelength precipitation radar on board the Global Precipitation Measuring (GPM) core satellite. There are obviously large differences in these sampling modes, which we have tried to minimize by integrating synchronous observations taken during the first 2 years of the GPM mission. The data comprises 327 “wet” overpasses of Switzerland, taken after the launch of GPM in February 2014. By comparing the GPM radar estimates with the MeteoSwiss products, a similar performance was found in terms of bias. On average (whole country, all days and seasons, both solid and liquid phases), underestimation is as large as −3.0 (−3.4) dB with respect to RADAR (GAUGE). GPM is not suitable for assessing what product is the best in terms of average precipitation over the Alps. GPM can nevertheless be used to evaluate the dispersion of the error around the mean, which is a measure of the geographical distribution of the error inside the country. Using 221 rain-gauge sites, the result is clear both in terms of correlation and in terms of scatter (a robust, weighted measure of the dispersion of the multiplicative error around the mean). The best agreement was observed between GPM and CombiPrecip, and, next, between GPM and RADAR, whereas a larger disagreement was found between GPM and GAUGE. Hence, GPM confirms that, for precipitation mapping in the Alpine region, the best results are obtained by combining ground-based radar with rain-gauge measurements using a geostatistical approach. The GPM mission is adding significant new coverage to mountainous areas, especially in poorly instrumented parts of the world and at latitudes not previously covered by the Tropical Rainfall Measuring Mission (TRMM). According to this study, one could expect an underestimation of the precipitation product by the dual-frequency precipitation radar (DPR) also in other mountainous areas of the world.

Deep generative modeling is able to generate highly realistic atmospheric fields, one prominent example being precipitation. So far, almost all studies have used generative adversarial networks (GANs) for this purpose, but recent progress in machine learning research has had a new class of methods called diffusion models replace GANs in many applications. Diffusion models have been often shown to be able to generate a wider variety of samples than GANs, suggesting that they might be able to better capture uncertainty in applications such as weather and climate where quantifying it is important.In this presentation, we describe our research on using diffusion models for short-term prediction (nowcasting) of precipitation fields. We adapt the latent diffusion model used by Stable Diffusion (Rombach et al. 2022) to the this problem, predicting precipitation up to 3 hours ahead to the future at 5-min temporal resolution and 1-km horizontal resolution. Predictions can be produced as an ensemble where each member represent a possible future evolution of the precipitation field.We show that our model:generates highly realistic precipitation fields that are consistent with the past precipitation used as input. outperforms the state-of-the-art GAN-based Deep Generative Models of Rainfall (DGMR) model by most relevant metrics. performs particularly well at representing the uncertainty of its own predictions, as shown by uncertainty quantification methods developed for ensemble forecast verification. Therefore, it appears that diffusion models are indeed suitable for generative modeling of precipitation fields with highly realistic representation of uncertainty. Our model architecture also permits multiple inputs data sources to be combined, in particular allowing seamless generative predictions to be made by exploiting observations and numerical weather predictions.