Assessing flood forecast uncertainty with fuzzy arithmetic

Testo completo

(1)E3S Web of Conferences 7, 18002 (2016) FLOODrisk 2016 - 3rd European Conference on Flood Risk Management. DOI: 10.1051/ e3sconf/2016 0718002. Assessing flood forecast uncertainty with fuzzy arithmetic 1,a. 2. Bertrand de Bruyn , Laurence Fayet and Vanessya Laborie. 1. 1. CEREMA DtecEMF 134, rue de BEAUVAIS CS 60039 60280 Margny Lès Compiègne CEDEX, France Service de Prévision des Crues Oise-Aisne DREAL Champagne-Ardenne 2, bd Gambetta 60200 Compiègne, France. 2. Abstract. Providing forecasts for flow rates and water levels during floods have to be associated with uncertainty estimates. The forecast sources of uncertainty are plural. For hydrological forecasts (rainfall-runoff) performed using a deterministic hydrological model with basic physics, two main sources can be identified. The first obvious source is the forcing data: rainfall forecast data are supplied in real time by meteorological forecasting services to the Flood Forecasting Service within a range between a lowest and a highest predicted discharge. These two values define an uncertainty interval for the rainfall variable provided on a given watershed. The second source of uncertainty is related to the complexity of the modeled system (the catchment impacted by the hydro-meteorological phenomenon), the number of variables that may describe the problem and their spatial and time variability. The model simplifies the system by reducing the number of variables to a few parameters. Thus it contains an intrinsic uncertainty. This model uncertainty is assessed by comparing simulated and observed rates for a large number of hydro-meteorological events. We propose a method based on fuzzy arithmetic to estimate the possible range of flow rates (and levels) of water making a forecast based on possible rainfalls provided by forcing and uncertainty model. The model uncertainty is here expressed as a range of possible values. Both rainfall and model uncertainties are combined with fuzzy arithmetic. This method allows to evaluate the prediction uncertainty range. The Flood Forecasting Service of

(2) 310 km2. Its response time is about 10 hours. Several hydrological models are calibrated for flood forecasting in this watershed and use the rainfall forecast. This method presents the advantage to be easily implemented. Moreover, it permits to be carried out in real time (for making information available to the user as quickly as possible). It provides end users Flood Forecasting (crisis managers, public) valuable information that allows everyone to better prepare its possible actions (warning, crisis preparation, safekeeping of the furniture, etc.).. 1 Context

(3)

(4)

(5) . ! "#$ % & .

(6) '. ( %( . % % . % %. )*) % & $ % &) ( % . . ! . % & a. " ! + . ( , - ( , (% . ( % . ./0

(7) 1 % % &% & 1#-#1 1 % 1 , %. 23)3 , )4" ( . ( % & . Corresponding author: bertrand.de-bruyn@cerema.fr. © The Authors, published by EDP Sciences. This is an open access article distributed under the terms of the Creative Commons Attribution License 4.0 (http://creativecommons.org/licenses/by/4.0/)..

(8) E3S Web of Conferences 7, 18002 (2016) FLOODrisk 2016 - 3rd European Conference on Flood Risk Management. DOI: 10.1051/ e3sconf/2016 0718002. % " % 2 . , 56 7 6 6 86 96 6 6 6 , $ ( & ( . , ( ,

(9)

(10) # ( :: . .220 " % % :: ( . % :: ( :: $ % . : . % %. % , . A " % % .B0 $ ( ( C: $ 4BD B3D : % )4 % 3). 2 Rainfall-Runoff model. 2.3 Calibration of model parameters. # % $ ! % . % . . % , % . ! ;1 % .)0.230 "7<.*0. % E % % . % ! &% = 2//? ( -. 2//B -. )33? % ' ! , . ' 1

(11) +3 F 33* !, '=3F33*B !- ' F4G > . % H F 333324* I !1 % ' JF F23. 2.1 Program overview ,

(12)

(13) # $ . . &' " + 1. 1

(14) + % 1

(15) +3 . . =3 . . & ( . . & & ( 1 % . 2.4 Operational forecasting operating + ( ,)4 $ % ! , : $ . & 1

(16) + ( !% 2 : % % . . 2.2 Rainfall-runoff model of the river Oise at Hirson . % % " " %

(17) $ 233 &); )23&)> % ?23&@. 2.

(18) E3S Web of Conferences 7, 18002 (2016) FLOODrisk 2016 - 3rd European Conference on Flood Risk Management. DOI: 10.1051/ e3sconf/2016 0718002. " %. # G & . . 3 Fuzzy arithmetic

(19) :: , . . .220$ ( K .20.L0.40 :: , . % ' 86 666 6. :: % ( % : 2*B *3D ?3D . . $ ( E :: :: $ :: . % . & .G0 ( ( :: % % :: $ ( % & . 4.2 Difficulties encountered in operational forecasting (a few hours before the event) . % % . % % ( . : , E %A . ( % % % % 2 ( 1

(20) + &% & . . % "%. ( . % % % $. ( % % . % ( % . % % : . % .

(21) ) % %)322 " % &1

(22) +: 3 . 4 Results 4.1 test case: the river Oise flood of January 2011 at Hirson L * = )322 " % . C % 233

(23) -. )323 =)322 % % " % % %% , 23 " B3L3 " % 7 % : ' G3 )4 %. L * = )3221 % % ' ! ! " !B M L =**M *=)2'33 ( . . " , . . %. %" % . % % . 4.3- Proposed tool that takes into account the possible levels of easily usable tank and rain forecast % :: : : . & % :: & ( ( :: :: . 24 $ = )322 " ( )322I32I3L 24 ( . 3.

(24) E3S Web of Conferences 7, 18002 (2016) FLOODrisk 2016 - 3rd European Conference on Flood Risk Management. DOI: 10.1051/ e3sconf/2016 0718002. , )322I32I* L . %. )3 43 & %. ?343 2L 2/ )B I ,% % ' ! )323I32I3*L'33 )323I32I3GL' 232B > !)323I32I3GL'33 )323I32I3/L'23 )3 :: . & 1

(25) + , :: . . % ? . 2B3NI "%. % & #

(26) 4 "%. . % 1

(27) + . % % # %. % . .

(28) 4 . % % ( ( ( ( ( . & % & 242IL % % . &.

(29) K, :: :: . % % '3(32(3)( 3?(34( % % 4 %. , :: %

(30) B . 5 conclusion and perspectives. . : . , % ::( % -= )322 " ( K % % . 6 6 86 6 6 : :: : % " ( %. % :: $ &

(31) , :: :: % % .

(32) :: % " (%1

(33) + :: .

(34) :: %" ( %1

(35) + :: . . % % % B3NI . 4.

(36) E3S Web of Conferences 7, 18002 (2016) FLOODrisk 2016 - 3rd European Conference on Flood Risk Management. 6 References 2 7 #( - & O( 2//B

(37) :: ! % . (

(38)

(39) (+#(2//B )7 O()323(P Q R P S C

(40)

(41) ? 7 7( 7 -( ( - -(

(42) O( O )32) TO R E )322 ; , U

(43) . !"#$# $$ %( (2!) P )32) 4 7 7

(44) < )33L' T+ PP , , ' P . O U &

(45) #$"" " )I)33L( 23L!22) B

(46) P ( P )33G ! V R R" ''

(47) (#. )*)B4 L

(48) (2//*(K P P . ' P P C "(#$$ +

(49) '

(50) #$ ,

(51) -" * " ( )33G TO "W-1#7<U .

(52) . $/# GJ#>(; (2/GB($ :: ( 0 .

(53) .

(54) '1(X %!W& /O <("

(55) ( (

(56) O( 77 O )324 -. , J

(57) R % ' "W-1#7<! " ( ( K;+ K ; + ; # 4I)*!)324IBI2 23 )33B( X P P S !P ( S K" 22 Y O #( 2/LB(

(58) :: ( G( ??G!?B?. 5. DOI: 10.1051/ e3sconf/2016 0718002.

(59)