Evaluation of geostationary satellite observations and the development of a 1–2 h prediction model for future storm intensity

Abstract

A study was conducted to gain insights into the use of geostationary satellite‐based indicators for characterizing and identifying growing cumulus clouds that evolve into severe weather producing convective storms. Eleven convective initiation (CI), 41 cloud top temperature–effective radius (T‐re), and 9 additional fields were formed for 340 growing cumulus clouds that were manually tracked for 2 h and checked for association with severe weather to 2–3 h into the future. The geostationary satellite data were at 5 min resolution from Meteosat‐8 on six convectively active days in 2010, 2012, and 2013. The study's goals were to determine which satellite fields are useful to forecasting severe storms and to form a simple model for predicting future storm intensity. The CI fields were applied on 3 × 3 pixel regions, and the T‐re fields were analyzed on 9 × 9 and 51 × 51 pixel domains (needed when forming T‐re vertical profiles). Of the 340 growing cumulus clouds examined, 34 were later associated with severe weather (using European Severe Weather Database reports), with the remaining being nonsevere storms. Using a multivariate analysis, transforming predictors into their empirical posterior probability, and maximizing the Peirce skill score, the best predictors were T1451 (51 × 51 pixel T, where re exceeds 14 µm), TG9 (9 × 9 pixel glaciation T surrounding a growing cloud), and ReBRTG51 (51 × 51 pixel re at the breakpoint T in the T‐re profile). Rapid cloud growth prior to severe storm formation leads to delayed particle growth, colder temperatures of the first 14 µm particles, and lower TG values.

The Cloud Top Distribution and Diurnal Variation of Clouds Over East Asia: Preliminary Results From Advanced Himawari Imager

“…With the advent of geostationary satellite, the characterization of clouds has been extensively investigated in previous pioneering studies, including the evolution of convections and the temporal and spatial distribution of clouds (e.g., Bley et al, ; Cintineo et al, ; Mecikalski et al, ). Painemal et al () characterized the diurnal cycle of cloud top height (CTH) and cloud cover over the Southeastern Pacific based on Geostationary Operational Environmental Satellite‐10 images.…”

A New Approach for the Analysis of Deep Convective Events: Thunderstorm Intensity Index

“…Local thunderstorms are usually short-lived and less accurately reported/observed in radar and satellite data [45][46][47][48][49]. Local thunderstorm impact is very localized both in space and time; therefore, in comparison with MCSs or larger systems, it is often rendered insignificant and usually neglected or ignored as part of thunderstorm research.…”

Convective Initiation Proxies for Nowcasting Precipitation Severity Using the MSG-SEVIRI Rapid Scan

“…This size of the box has been chosen to account for different storm spatial extensions (up to the largest one) and to study the evolution stages roughly until maximum expansion. Other works use different box sizes [40], ranging from 3 × 3 to 51 × 51 pixels, depending on the scale of interest. We have found this size is the optimal trade-off fitting the purpose of the study, i.e., to investigate the immediate area near the central pixel close to the maximum peak rainfall.…”

“…The rationale for this work is as follows: we aim to (i) identify a subset of IR indicators-within the wide literature on convective initiation satellite-based proxies (see e.g., [27])-showing distinctive signatures or patterns in the case that convection leads to a hourly cumulative rain greater than 30 mm, as opposed to convective events that only cause less than 10 mm; (ii) show that the spatial extent of the cumuli is, on average, relatively smaller in the case of nonsevere precipitation; (iii) derive tentative thresholds for every indicator, and propose an ideal proxy combination to maximize the predictive skill for the severity of the precipitation. The research discussed here is along the lines of previous works (see e.g., [40][41][42]), similar in spirit but different in both the approach and dataset criteria; we do not use storm-tracking methods but fully characterize the storm features via a static approach by investigating the temporal evolution of relevant IR fields for separate groups of pixels within the area surrounding the storm evolution. The criteria used to design the dataset are peculiar, since we have selected cases where both storm development and subsequent rainfall occurs in the same area (i.e., convective and orographic nature); we do not use the rainfall intensity data, but base our dataset choice on the simultaneous presence of convective initiation and a cumulative hourly rainfall exceeding 30 mm for severe cases (or smaller than 10 mm for nonsevere cases), according to the Italian Radar Network.…”