Tuesday, 9 January 2018: 12:00 AM
Salon H (Hilton) (Austin, Texas)
Identifying existing convection and monitoring its evolution is crucial to minimize severe storms potential damages, specially associated with Mesoscale Convective Systems. Satellite data can provide indirect measurements of instability and moisture of a large area in short periods of time. Lightning location systems provide real time updates of storm severity evolution. Blending information from satellite and lightning data, and numerical weather prediction models, can be even more beneficial for severe weather monitoring and forecasting. Southeastern South America (SSA), a region comprising northeastern Argentina, Paraguay, Uruguay and southern Brazil, is particularly prone to severe weather events, generally associated with Mesoscale Convective Systems (MCS) with intense precipitation, lightning, hail, wind gust and some tornadoes. Lightning activity is intense in the continent throughout the year, but mostly active during the warm season (spring and summer), with maximum in the subtropical region, in this area. There is a limited knowledge in relation to the characteristics of those mesoscale severe storms including frequency, type, morphology, intensity and location of cloud-to-ground lightning associated to the occurrence of mesoscale convective systems. The objective of this study is to better understand the evolution of these storms as they occur in this area, in order to improve our abilities of analysis and forecast of severe weather events and apply directly in an operational environment. This work aims to present an objective methodology for detection and monitoring of convective initiation and the electrification process in regions of deep convection embedded within these MCS, using different techniques and criteria for stratification of a cloud shield into areas of different intensities of convection. Temperature threshold, brightness temperature differences, and temperature trends are used as parameters for the identification of convective initiation and electrification, and then compared with lightning characteristics within those MCS events. Initially METEOSAT data in the water vapor and infrared frequency channels were used in this study. Due to the large amount of information used from the satellite data, a correlation evaluation using principal component analysis was used to reduce the number of variables or predictors. A comparison of lightning observations with collocated satellite information was performed to clusterize the region of interest. A deep learning method was used to select the learning tests and we developed a classifier after comparison of methods such as multi-layer perceptron and radial basis function network. Results indicate a capacity of lightning detection using only satellite data as input for the algorithms in excess of 95%. Further studies are in evaluation to apply this same algorithm with the replacement of EUMETSAT with GOES-16 data. Due to its improved spatial and temporal resolution we expect to be able to have better results which will be presented in this conference.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner