Search

Convective storms and lightning are among the most important weather phenomena that are challenging to forecast. In this study, a novel multi-task learning (MTL) encoder-decoder U-net neural network was developed to forecast convective storms and lightning with lead times for up to 90 min, using GOES-16 geostationary satellite infrared brightness temperatures (IRBTs), lightning flashes from Geostationary Lightning Mapper (GLM), and vertically integrated liquid (VIL) from Next Generation Weather Radar (NEXRAD). To cope with the heavily skewed distribution of lightning data, a spatiotemporal exponent-weighted loss function and log-transformed lightning normalization approach were developed. The effects of MTL, single-task learning (STL), and IRBTs as auxiliary input features on convection and lightning nowcasting were investigated. The results showed that normalizing the heavily skew-distributed lightning data along with a log-transformation dramatically outperforms the min-max normalization method for nowcasting an intense lightning event. The MTL model significantly outperformed the STL model for both lightning nowcasting and VIL nowcasting, particularly for intense lightning events. The MTL also helped delay the lightning forecast performance decay with the lead times. Furthermore, incorporating satellite IRBTs as auxiliary input features substantially improved lightning nowcasting, but produced little difference in VIL forecasting. Finally, the MTL model performed better for forecasting both lightning and the VIL of organized convective storms than for isolated cells. [ABSTRACT FROM AUTHOR]

Based on Multifunctional Transport Satellite data and the infrared window-texture detection algorithm, the level of overshooting top (OT) activity within a tropical cyclone (TC), which is defined as the hourly mean number of OT occurrence, was statistically investigated in the western North Pacific basin for the period 2005-12. Based on the level of OT activity, the samples were divided into OT and non-OT cases or high-activity-OT (HA-OT) and low-activity-OT (LA-OT) cases. The differences in large-scale environmental variables between OT (HA-OT) and non-OT (LA-OT) cases were examined 12 hours prior to the OT occurrence. Statistical analysis showed that environmental differences did exist between the OT and non-OT cases. The OTs were more skewed towards the early stage of the TC life cycle, and mostly concentrated in low latitudes. Meanwhile, a sufficiently deep warm-water layer, large temperature difference between the upper- and lower-level troposphere, large humidity at the middle and upper levels, and large atmospheric instability, were favorable for OT occurrence. The differences in large-scale environmental characteristics between HA-OTs and LA-OTs were not as significant as those between OTs and non-OTs, but the HA-OT samples tended to occur when the vertical shear was weak and the TC intensity was low. Finally, statistical models were designed to predict the OT and HA-OT. When at least three OT (HA-OT) predictor thresholds were satisfied, the Peirce skill score reached a maximum value of 0.49 (0.30). [ABSTRACT FROM AUTHOR]

This paper proposes several quantitative characteristics to study convective systems using observations from Doppler weather radars and geostationary satellites. Specifically, in order to measure the convective intensity of each system, a new index, named the “Convective Intensity Ratio” (CIR), is defined as the ratio between the area of strong radar echoes at the upper level and the size of the convective cell itself. Based on these quantitative characteristics, the evolution of convective cells, surface rainfall intensity, rainfall area and convectively generated anvil clouds can be studied, and the relationships between them can also be analyzed. After testing nine meso- β-scale convective systems over North China during 2006–2007, the results were as follows: (1) the CIR was highly correlated with surface rainfall intensity, and the correlation reached a maximum when the CIR led rainfall intensity by 6–30 mins. The maximum CIR could be at most ∼30 mins before the maximum rainfall intensity. (2) Convective systems with larger maximum CIRs usually had colder cloud-tops. (3) The maximum area of anvil cloud appeared 0.5–1.5 h after rainfall intensity began to weaken. The maximum area of anvil cloud and the time lag between maximum rainfall intensity and the maximum area of anvil cloud both increased with the CIR. [ABSTRACT FROM AUTHOR]