Abstract Satellite spectral bands have been used to identify patterns associated with thunderstorms such as strong updrafts, deep vertical development, cloud-top cooling, and cloud particle phase. Because of this, satellite data are incorporated into forecasting applications for identification of initiating thunderstorms including the Day Cloud Phase Distinction (DCPD) and Nighttime Microphysics (NtMicro) false color red-green-blue (RGB) composite imagery and machine learning models for short-term forecasting (nowcasting) of thunderstorms and their hazards. However, initiating lightning-producing and non-lightning-producing convective storms can appear similar in satellite imagery, making it difficult to determine which developing storms will become mature thunderstorms. As a consequence, nowcasting models relying solely on satellite imagery, such as LightningCast and ThunderCast, can be imprecise when used to nowcast at local scales. To further verify this and examine the conditions under which non-lightning-producing convective storms appear similar to developing thunderstorms in satellite imagery, four case studies are presented with LightningCast’s and ThunderCast’s predictions. Each case study contained at least one non-lightning-producing convective storm with RGB signatures consistent with cloud-top glaciation. Overall, the case studies demonstrate non-lightning-producing convective storms, appearing similar to developing thunderstorms in satellite imagery, are not limited by latitude and can form in an assortment of meteorological environments including near a sea breeze, in high humidity, and in an environment with key meteorological ingredients for severe weather. While geostationary meteorological satellites provide critical qualitative and quantitative insights for convective weather analysis and nowcasting, these results illustrate their limitations when distinguishing locally between non-lightning-producing and lightning-producing storms.
Abstract The real-time observational system for measuring snowfall generally consists of manual, daily, backwards-looking snow accumulations, at point locations where a volunteer can be recruited. This limited spatial and temporal reporting provides little assistance with warning and mitigation decisions which must be made between observation sites, and before those observations are made. In order to fill this information gap, a snow rate field for the Multi-Radar Multi Sensor (MRMS) has been developed. This MRMS Snow Rate provides a diagnosis of snow accumulation across the CONUS at an hourly cadence, and has been developed with the ability to use six snow-to-liquid ratio (SLR) algorithms. This research is meant to explore which SLR produces the best snow accumulations from the MRMS snow rate product created using MRMS quantitative precipitation estimation (QPE), as there is limited research available to guide this choice of SLR. This product, like any MRMS field, can be accumulated to create multi-hour or multi-day analyses. When 24-hour MRMS snow accumulations are verified against observations, it is not immediately possible to declare an SLR better in all circumstances. Verification statistics depend on observed snowfall amount and verification metric used. The impact of SLRs on decisions on an hourly level is also examined through case studies, illustrating the impact on the spatial and temporal evolution of snow events. The implications of these results are discussed, including the importance of improving all of the components of snow rate diagnoses, including QPE, snow observations, and SLR algorithms.
Abstract Meteorological services face a major challenge in providing reliable snowfall forecasts and detecting solid precipitation, as precise data is needed for weather warnings, damage assessments, hydrological modeling, and many other tasks. This study evaluates the skill of the ALADIN numerical weather prediction model in solid precipitation forecasting at 126 stations of the Czech Hydrometeorological Institute (CHMI) between August 2019 and July 2025. Model performance was assessed against quality-controlled human present weather observations using common verification metrics, including the Probability of Detection (POD), False Alarm Ratio (FAR), Critical Success Index (CSI), and Heidke Skill Score (HSS). Further, the ALADIN_SN weighted index of solid precipitation occurrence was assimilated from ALADIN snow forecasts in hourly intervals using two sets of empirical lead-time-dependent weights. During the snow season, ALADIN_SN showed a mean POD of 0.68, FAR of 0.48, and HSS of 0.53, with an overall accuracy exceeding 83% at all stations. Air temperature is the most important variable affecting the model performance, with better results at temperatures below approximately +2°C. The skill of ALADIN_SN also improved with increasing solid precipitation intensity and duration. Case studies demonstrated that ALADIN_SN successfully captured large-scale snowfall events associated with synoptic systems but struggled with localized or short-lived precipitation and phase transitions at temperatures above freezing, and it often overestimated the spatial extent of solid precipitation zones. Overall, the ALADIN_SN parameter provides a spatially consistent, computationally efficient, and operationally feasible method for identifying solid precipitation occurrence, suitable for supplementing networks of automated measurement systems and human weather observers.
Abstract Storm motion for a right-moving supercell is currently estimated from an environmental sounding as the mean wind in the 0–6-km layer plus a fixed propagation velocity, 7.5 m s −1 , away from the mean wind in a direction that is 90° to the right of the 0–6-km bulk shear vector (after some averaging at the head and tail of this vector). The mean wind and the direction of the bulk shear vector completely determine the predicted storm motion. Although the predicted supercell motion relative to the mean wind is Galilean invariant, it is independent of hodograph size and curvature. There is no provision for upshear propagation, and storm-relative helicity is underestimated. A generally more accurate linear regression equation for predicting supercell motion is proposed as a modification to the commonly used operational supercell motion prediction technique. The statistical model contains multiple unnormalized bulk shears, thus allowing hodograph size and shape to be represented while still retaining Galilean invariance. The updated method removes a statistical bias in estimates of storm-relative helicities.
Abstract This study presents a particle filter–based framework for short-term lightning now-casting using only lightning detection network observations. The method employs DBSCAN clustering to identify lightning-active structures and propagates them probabilistically, yielding useful skill at one- to three-hour lead times, with declining but non-zero skill at four to six hours. The results show that Recall is more robust than precision across increasing lead times, and that performance is sensitive to the ground truth density threshold selection. Despite these limitations, the suggested approach achieves skill scores comparable to more data-intensive machine learning and NWP-based methods, while requiring minimal input data. Importantly, the algorithm can be implemented as a real-time Python application, demonstrating both its scientific value and practical applicability in data-constrained environments.
