october 2018 flood event in the Aude River basin France CharpentierNoyer M Nicolle P Payrastre O Gaume E Bouttier F and Marchal H New shortrange rainfall forecast ID: 1036432
Download Presentation The PPT/PDF document "Relevance of using ensemble forecasts of..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
1. Relevance of using ensemble forecasts of flash-flood impacts for an emergency service: an evaluation for the october 2018 flood event in the Aude River basin, FranceCharpentier-Noyer M., Nicolle P., Payrastre O., Gaume E., Bouttier F., and Marchal H.
2. New short-range rainfall forecast productsA first evaluation based on discharge threshold anticipationObjectivesPrinciples of the multi-agent modelResultsConclusionsnew experimental short range ensemble rainfall forecast products2Experimental ensemble rainfall products based on the convection permitting Arome EPS and Arome NWP model
3. A first evaluation based on the discharge thresholds anticipation3Zero future rainfallpepi ensemble product - 75% percentile What is the best forecast for an user ? Charpentier-Noyer et al., 2022, NHESSd, https://doi.org/10.5194/nhess-2022-182New short-range rainfall forecast productsA first evaluation based on discharge threshold anticipationObjectivesPrinciples of the multi-agent modelResultsConclusionsAnticipation capacity of T=10 years discharge threshold (Example of the Aude River basin: 1174 sub-basins)
4. Flash floods in the Aude River basin 14-16 October 2018New short-range rainfall forecast productsA first evaluation based on discharge threshold anticipationObjectivesPrinciples of the multi-agent modelResultsConclusionsObjective: evaluate the forecasts based on the capacity to take efficient decisionsCase study: Aude October 2018 floodThe agents of a fire service are rescuing teamDecisions based on impacts forecasts32 municipalities with rescue operationsInterventions triggered by reports of observed impacts15 fatalties200 M€213 mm in 6 hours4
5. Overall principles of the multi-agent model New short-range rainfall forecast productsA first evaluation based on discharge threshold anticipationObjectivesPrinciples of the multi-agent modelResultsConclusionsForecast of flooded buildings, at the river reach scale and 15 min time resolution5Maximum number of flooded buildings in oct. 2018A decision threshold (number of inundated buildings) triggers the decision of sending a rescue team
6. Overall principles of the multi-agent model New short-range rainfall forecast productsA first evaluation based on discharge threshold anticipationObjectivesPrinciples of the multi-agent modelResultsConclusionsPrinciple of the impact forecasts used to take decisions6A discharge-impacts relation is built on each river reach (7 flood scenarios, T=5 to 1000 years)Ensemble discharge forecasts can be direcly translated in terms of impacts
7. Overall principles of the multi-agent model Assumptions & parameters7Intervention threshold: minimum number of flooded buildings for sending a rescue team for a river sectionIntervention capacity: number of buildings treated by a rescue teamTeam displacement: displacement matrix with an average speed of 30 km/h92 rescue teams (42 firemen centers)New short-range rainfall forecast productsA first evaluation based on discharge threshold anticipationObjectivesPrinciples of the multi-agent modelResultsConclusions
8. Examples of modelling resultsLocation (and number) of rescue teams in operation8New short-range rainfall forecast productsA first evaluation based on discharge threshold anticipationObjectivesPrinciples of the multi-agent modelResultsConclusionsEvolution of the number of buildings really rescuedEvolution of the number of rescue teams mobilized
9. Measurement of the efficiency of decisions9Computation of a cost functionCumulative number of buildings for which no rescue team was present during the floodBefore the begining of the flood cost = 0 buildingsLess than 1h after the flood begining cost = 10 % flooded buildingsBetween 1h and 3h after the begining of the flood cost = 50 % flooded buildingsMore than 3h after the begining of the flood cost = 100 % flooded buildingsNew short-range rainfall forecast productsA first evaluation based on discharge threshold anticipationObjectivesPrinciples of the multi-agent modelResultsConclusions
10. Comparison of the efficiency of forecasts10Reference simulation based on observed rainfall: RS (used as reference, decisions based on field observations)AROME-EPSNew short-range rainfall forecast productsA first evaluation based on discharge threshold anticipationObjectivesPrinciples of the multi-agent modelResultsConclusionsForecasts based on zero future rainfall: RF0Forecasts with the 3-ensemble rainfall forecast products (75% percentile):pepipertDpepi
11. An ideal case with unlimited number of rescue teams 11Number of buildings really rescuedIntervention threshold: 10 buildingsTeam number: unlimitedIntervention capacity: 100 buildings/team New short-range rainfall forecast productsA first evaluation based on discharge threshold anticipationObjectivesPrinciples of the multi-agent modelResultsConclusions
12. An ideal case with unlimited number of rescue teams 12Number of rescue teams mobilized Intervention threshold: 10 buildingsTeam number: unlimitedIntervention capacity: 100 buildings/team New short-range rainfall forecast productsA first evaluation based on discharge threshold anticipationObjectivesPrinciples of the multi-agent modelResultsConclusions
13. An ideal case with unlimited number of rescue teams 13Cost function: buildings not rescuedIntervention threshold: 10 buildingsTeam number: unlimitedIntervention capacity: 100 buildings/team New short-range rainfall forecast productsA first evaluation based on discharge threshold anticipationObjectivesPrinciples of the multi-agent modelResultsConclusions
14. A more realistic case with 92 rescue teams14Intervention threshold: 10 buildingsTeam number: 92Intervention capacity: 100 buildings/team Number of buildings really rescuedNew short-range rainfall forecast productsA first evaluation based on discharge threshold anticipationObjectivesPrinciples of the multi-agent modelResultsConclusions
15. A more realistic case with 92 rescue teams15Intervention threshold: 10 buildingsTeam number: 92Intervention capacity: 100 buildings/team Number of rescue teams mobilized New short-range rainfall forecast productsA first evaluation based on discharge threshold anticipationObjectivesPrinciples of the multi-agent modelResultsConclusions
16. A more realistic case with 92 rescue teams16Intervention threshold: 10 buildingsTeam number: 92Intervention capacity: 100 buildings/team Cost function: buildings not rescuedNew short-range rainfall forecast productsA first evaluation based on discharge threshold anticipationObjectivesPrinciples of the multi-agent modelResultsConclusions
17. Comparison of the efficiency of forecasts17Performance of the different tested rainfall forecastsImpacts model based on affected population rather than buidingsNew short-range rainfall forecast productsA first evaluation based on discharge threshold anticipationObjectivesPrinciples of the multi-agent modelResultsConclusionsRealism of the agent-based model - improvementsTravel timesCost function (case of late arrival of rescue) The added value of forecasts can be identified, particularly in case if unlimited meansForecasts are largely affected by false alarms in case of limited means. Decisions based on zero future rainfall may be better in this case.Decisions based on forecasts never worse than decisions based on field observationsDuration of rescue operations and reaffectation of rescue teams