Nowcasting Thunderstorms with SIGnificant weather Object Oriented

Nowcasting Thunderstorms with SIGnificant weather Object Oriented Nowcasting System Pascal Brovelli, Stéphane Sénési, Etienne Arbogast, Philippe Cau, Sandrine Cazabat, Michel Bouzom, Jérôme Reynaud 5 -9 september 2005 WSN 05 - Toulouse France

SIGOONS principles l Nowcasting significant weather events (0 -4 h) – thunderstorms – later on, fog areas, heavy rain, snow and icing condition, strong wind systems l Based on an object oriented approach – SIGOONS manages Significant Weather Objects (SWO) Notice that usual weather conditions are not describe A hybrid system : man-machine mix l Automated generation of end-user products downstream of supervise database l 5 -9 september 2005 WSN 05 - Toulouse France

SIGOONS: a hybrid system l – Diagnostic of thunderstorm Guesses updates SWO database Production CONO Pre-processing Expertise 5’ Automated processes: Updates Systematic checking Userspecific warnings SWO guesses – SWO systematic check against ground observations – Generation of user-specific warnings • Forecaster expert input: – Arbitrate discrepancies Ground observations 5 -9 september 2005 WSN 05 - Toulouse France

Automated Diagnostic of thunderstorm SWO using CONO (1) l CONO tool analyzes radar data in order to automaticaly detect convective cells and/or systems (Convection Nowcasting Objects, extension of the RDT objects, see also Hering et al. and Autones et al. ) l Detection by adaptative reflectivity thresholding of radar data • • Structure of reflectivity can be complex : smoothing and morphological « closing » operation merges cells and matches the convective system scale Discrimination of convective systems uses lightning data 5 -9 september 2005 WSN 05 - Toulouse France

Automated Diagnostic of thunderstorm SWO using CONO (2) l CONO tool automaticaly tracks convective cells and/or systems l l Tracking by overlapping between a cell detected in the present image and cells detected in the previous image using displacement speed Speed estimate blends move of the cell centroïd and cross-correlation. After tuning, speed diagnostics : – are robust against merges and splits – have smooth variation – improve diagnostic of low group speed on backward regenerating convective systems l CONO intialialize thunderstorm SWO attributes: – horizontal envelope, move speed and lightning activity – rain rate and hail risk 5 -9 september 2005 WSN 05 - Toulouse France

SIGOONS Man-Machine Interaction Challenge: Minimize input by the forecaster l SIGOONS is designed to run automatically l – Merge and check automated SWO from the new « slot » with forecast SWO from the previous « slot » : l l l Tracking allows to propagated forward in all supervise attributes, provided that there is consistency with new observation data Discrepancy send a specific, relevant « disagree signal » to the forecaster Forecaster input is optional – monitors the automated initialization of sensible weather attributes : wind gust, hail risk, rain accumulation l l Choose between different automated diagnostic values Arbitrate discrepancies – Sets a decay/growth tendency on area, duration, attributes … – Creates objects for convective systems or thunderstorm prone areas 5 -9 september 2005 WSN 05 - Toulouse France

Object with mismatch SIGOONS display tool l Fully integrated in the operational Synergie Speed estimate Workstation too high Object with good match 5 -9 september 2005 Significant ground obs without supporting object WSN 05 - Toulouse France

5 -9 september 2005 WSN 05 - Toulouse France

Access to rain rate evolution 5 -9 september 2005 WSN 05 - Toulouse France

Significant rain accumulation Access to ground observations collected over the SWO trajectory 5 -9 september 2005 WSN 05 - Toulouse France

SIGOONS: human expertise l A significant effort on ergonomic: – Real-time experiments and case studies using man-machine interface prototype (since summer 2004) – Ergonomist studies (see Chabaud et al. ) l First study results: – Forecasters feel that the object representation is clear – Forecasters concerns are the workload and their ability to exert expertise: l During the first hour SIGOONS should definitively favour automation l For first to fourth hours ahead, expertise apply to larger scales : thunderstorm systems or thunderstorm-prone areas – Expert input is basically qualitative : “wind gust stronger near the coast” 5 -9 september 2005 WSN 05 - Toulouse France

SIGOONS End-user products 2004 experiment findings : l – – Warning locations are correct and faithfully translate the nowcast database contents The sensible weather diagnostic (wind gust, rain accumulation, hail risk…) is still weak Time consistency and stability must be improved products delivered in “push” mode, like user-specific warnings (e. g. security services) are much promising 5 -9 september 2005 WSN 05 - Toulouse France

SIGOONS : Current status Automated diagnostic and check run in routine mode l Human expertise: l – SWO display tool close to be available for operations to all (7) regional offices forecaster – Experiment on case studies in order to define nowcasting specific tasks l Products – Test with a few customers of a “warning package”, which includes nowcast warnings, and short range forecast – Assessment of the quality of the thunderstorm warnings produced without expert input 5 -9 september 2005 WSN 05 - Toulouse France

SIGOONS Outlook l Automated diagnostics: – Use optimal combination of satellite and radar tracking – Improve the conceptual models for automated convection diagnostics: l l Data fusion with mesoscale analysis Identification of convection organization type Diagnostics for sensible weather attributes Human expertise – Introduce new objects, better suited to human expertise l l thunderstorm-prone area, fog area, surface front ( rain and/or wind )… Products – Introduce uncertainty on phenomena location and intensity – Design graphical and mobile phone products l Extend the OO approach to : – – Tracking objects in Hi. Res NWP simulation (a. s. a. p re. NWP quality) Matching of simulated objects with real objects for real-time trend assessment 5 -9 september 2005 WSN 05 - Toulouse France

Thank you 5 -9 september 2005 WSN 05 - Toulouse France