Session 2 Statistical Modelling of Rain 2023-2024 Anglais

Download as pdf or txt
Download as pdf or txt
You are on page 1of 41

Session 2 : Rainfall event modelling

Statistical representation of rainfall event. Synthetic


hyetograph

Outline

I. Statistical analysis
I.1 IDF Curves
I.2 Gumbel fitting

II. Synthetic hyetograph


II.1 Keifer methods
II.2 Desbordes methods
II.3 Others

III. Spatialization
I- Statistical characterization of rainfall event :

I.1 IDF curves

For a given event, one seek 𝐼 = max(< 𝐼 > 𝑑 )


Where < 𝐼 > is the average intensity over duration d

Fixed origin
i(mm/h) i(mm/h)
40 40

30 30

20 20
Moving window
10 10 (d=15 mn here)

15 30 45 60 75 100 t (mn) 15 30 45 60 75 100 t (mn)

Averaging with origin kept constant Moving window for the averaging
whatever the duration d for a given duration d
d=15 mn
d=35 mn
I- Statistical characterization of rainfall event :

I.1 IDF curves

𝑗 𝑑𝑢𝑟𝑎𝑡𝑖𝑜𝑛 𝑑𝑗 𝑗
n pluviometric events : 𝐼𝑖 = max(𝐼𝑒𝑣𝑒𝑛𝑡 𝑖 ) (1 < 𝑖 < 𝑛) [1] 𝐼1
𝑗
𝑗
𝐼2
n values of 𝐼𝑖 following def [1] 𝑗
𝐼3
• One keep only p values (p<n).
• Intensities classified in decreasing order : (… )
𝑗
𝐼𝑖
rank i (1<i<p) = rank of the jeme maximum for this duration dj
(… )

𝑗
𝑗
𝐼1,𝑝 Random vector whose values that follow Gumbel law 𝐼𝑝−1
𝑗
𝐼𝑝
𝑗
𝐹′ = 𝑝𝑟𝑜𝑏 𝐼 < 𝐼𝑖
𝑗 𝑗 𝑗 𝑗
𝑗 𝐼1 > 𝐼2 > ⋯ > 𝐼𝑝−1 > 𝐼𝑝
𝐹 = 𝑝𝑟𝑜𝑏(𝐼 > 𝐼𝑖 ) = 1 − 𝐹′
𝑖−𝑎 (Empirical frequency)
approximated by 𝐹 =
𝑝+1−𝑎
a=0.5 (usual choice in France)
I- Statistical characterization of rainfall event :

I.1 IDF curves Duration dj

Rank i associated to frequency


𝑖−𝑎
𝐹𝑖 =
𝑝 + 1 − 2𝑎

And return period 𝑇𝑖

If p is equal to the duration of


observation in year (usually one
value, the max observed, per
year) then :

𝑇𝑖 = 1Τ𝐹𝑖 (years)
By construction

𝑇𝑖1 = 𝑇𝑖2 = ⋯ = 𝑇𝑖𝑚−1 = 𝑇𝑖𝑚


(I assumed that p is the same whatever
the duration of averaging. If not, see
Gumbel analysis to make interpolation in
order to get sample with the same return
period)
I- Statistical characterization of rainfall event :

I.1 IDF curves Duration dj

Intensities associated
to the same return period
𝑇𝑖
for the different duration dj

IDF curve for that return period

𝐼(𝑑𝑗 , 𝑇𝑖 = 𝑇)
I- Statistical characterization of rainfall event :

I.1 IDF curves

Experimental points adjustement (fit) by a « theoretical » law

4,50
T=50 ans
4,00
T=20 ans
3,50
T=10ans
3,00
T=5 ans
2,50
T=10 ans : i(t)=43/(t+10)
2,00

1,50

1,00

0,50

0,00
0 50 100 150
I- Statistical characterization of rainfall event :

I.1 IDF curves

Classical theoretical fit : Montana’s law 𝐼 𝑑, 𝑇 = 𝑎 𝑇 𝑑 𝑏(𝑇) 𝑏 𝑇 <0

• Generally well adapted for rainfall duration ranging from 1h to 72h

• For duration less than 1h, the same formulation can be used but the b coefficient is usually
different than the one for duration >1h.

• Singularity when d→0. Usually several fit on different ranges of duration to hampers
uncertainties at small duration and long duration

• b is weakly dependant on the frequency F (or return period T) : it can be thus


considered as a regional constant

Other formulations :

𝐼 𝑑, 𝑇 = 𝑎(𝑇)Τ(𝑑 + 𝑏 𝑇 ) Talbot

𝐼 𝑑, 𝑇 = 𝑎(𝑇)Τ(𝑑 + 𝑏 𝑇 )𝑐 Keifer and Chu

(…)
I- Statistical characterization of rainfall event :

I.1 IDF curves


Montana’s law : example of the b value evolution
Depending on the rainfall duration

Left : evolution of b (Montan’s law with


the frequency (or return period)
Right : comparison between Montana’s and
Talbot’s adjustment for duration <1h

Source : Christophe Chamoux.


Techniques alternatives en assainissement pluvial. De la théorie à la pratique. Thèse de Doctoral de l’Université Nice
Sophia-Antipolis. 2003
I- Statistical characterization of rainfall event :

I.2 Method using Gumbel fitting

Date Pluie Date Pluie Date Pluie Date Pluie


20/06/1966 80.6 27/11/1975 57.0 03/10/1984 62.80 07/10/1993 87.40
04/11/1967 40.5 09/09/1976 57.1 13/05/1985 40.20 10/01/1994 73.40
03/11/1968 93.1 20/02/1977 60.0 14/11/1986 44.60 18/01/1995 56.20
13/03/1969 61.4 25/02/1978 66.1 05/10/1987 117.80 22/09/1996 89.80
13/11/1970 65.7 19/03/1979 85.0 01/12/1988 62.00 19/12/1997 89.00
12/09/1971 47.7 13/03/1980 40.8 12/04/1989 45.20 30/09/1998 116.60
19/02/1972 61.1 21/12/1981 58.6 16/10/1990 75.60 24/10/1999 108.20
13/10/1973 191.4 09/11/1982 53.4 29/09/1991 84.80
28/09/1974 75.3 18/06/1983 59.0 03/10/1992 104.20

Raw Data. Nice Aeroport


34 years of observations

Observation duration =34 years. P=34

Data classification and empirical


frequency calculation with F=(r-0.5)/p
I- Statistical characterization of rainfall event :

I.2 Method using Gumbel fitting

Gumbel law
𝑥 − 𝑥0

𝐹 =𝑒 −𝑒 −𝑢 𝑢 Reduced variable 𝑢=
𝑔

Methology

• From the empirical Frequency, calculation of : 𝑢 = −ln(− ln 𝐹 ′ )


• The expression of the reduced variable state that : 𝑥 = 𝑥0 + 𝑔𝑢

• Experimental fit to find x0 and g : various strategy

- Linear fit of the data (Least square approximation, « à l’œil ») -


- Moments method
I- Statistical characterization of rainfall event :

I.2 Method using Gumbel fitting

𝑒𝑟𝑟𝑒𝑢𝑟 2 = (𝑝𝑙𝑢𝑖𝑒 − 𝑝𝑙𝑢𝑖𝑒 𝑡ℎ𝑒𝑜)2


𝑥 𝑝𝑙𝑢𝑖𝑒 𝑡ℎ𝑒𝑜 = 𝑥0 + 𝑔𝑢
Random initial guess for 𝑥0 and 𝑔. here x0=30 and g=10
I- Statistical characterization of rainfall event :

I.2 Method using Gumbel fitting

Minimisation of the sum of error2


(Excel solver for instance)
x0=60.06 g=23,52
I- Statistical characterization of rainfall event :

I.2 Method using Gumbel fitting

Methods of moment :

𝑥0 = µ − 0.577𝑔 (0.577 : Euler parameter)

6
𝑔= 𝜎 = 0.78𝜎
𝜋

With 𝑝
1
µ= ෍ 𝑥𝑖
𝑝
𝑖=1

𝑝
2
1 2
𝜎 = ෍ 𝑥𝑖 − µ
𝑝
𝑖=1
I- Statistical characterization of rainfall event :

I.2 Method using Gumbel fitting


I- Statistical characterization of rainfall event :

I.2 Method using Gumbel fitting

Direct calculation with the gumbel


fitting function of the rain
corresponding to a given return period

Example direct interpolation from Gumbel law


II- Synthetic hyetograph

II1 – Keifer method (Chicago synthetic hyetograph)


a) Continuous Model

Based on the IDF curves. Rainfall event shape = completly advanced pattern

Continuous model : synthetic hyetograph built using the theoretical ajustement.

1 𝑡
From the definition of the average on time t 𝐼 𝑡, 𝑇 = න 𝑖 𝜏 𝑑𝜏
𝑡 0

𝑑
So : 𝑖 𝜏 = 𝐼(𝑡, 𝑇) × 𝑡
𝑑𝑡

Example with Montana Law

𝐼 𝑡, 𝑇 = 𝑎𝑡 𝑏
𝑑
𝑖 𝑡 = 𝑎𝑡 𝑏+1 = 𝑎(𝑏 + 1)𝑡 𝑏
𝑑𝑡
II- Synthetic hyetograph

II1 – Keifer method (Chicago synthetic hyetograph)


a) Continuous model for an advanced rain pattern

Other example with a 3 parameters law

𝑏
𝐼 𝑡, 𝑇 = 𝑎Τ 𝑡 + 𝑐
𝑏 𝑏−1 Τ 𝑡+𝑐 2𝑏
𝑖(𝑡 = 𝑎 × 𝑡 + 𝑐 −𝑎×𝑏× 𝑡+𝑐

By definition, the black hatched area has the same


surface as the red rectangle of surface I(t,T).t

imax (t, T )

t
II- Synthetic hyetograph
II1 – Keifer method (Chicago synthetic hyetograph)

b) Discrete model for an advanced rain pattern


𝑡
න 𝑖 𝜏 𝑑𝜏 = 𝐼(𝑡, 𝑇)
0
𝑑
𝑖 𝑡 = 𝐼(𝑡, 𝑇) × 𝑡 + 𝐼(𝑡, 𝑇)
𝑑𝑡

The I.D.F curve is discretized on a time step dt.

𝐼 𝑛𝑑𝑡, 𝑇 − 𝐼( 𝑛 − 1 𝑑𝑡, 𝑇)
𝑖 𝑛𝑑𝑡 = × 𝑛𝑑𝑡 + 𝐼( 𝑛 − 1 𝑑𝑡, 𝑇)
𝑑𝑡

𝑖 𝑛𝑑𝑡 = 𝑛 × 𝐼 𝑛𝑑𝑡, 𝑇 − (𝑛 − 1) × 𝐼( 𝑛 − 1 𝑑𝑡, 𝑇)

One can also work with the water depth during time step dt
ℎ 𝑛𝑑𝑡 = 𝑖 𝑛𝑑𝑡 × 𝑑𝑡 = 𝑛 × 𝑑𝑡 × 𝐼 𝑛𝑑𝑡, 𝑇 − 𝑛 − 1 × 𝑑𝑡 × 𝐼( 𝑛 − 1 𝑑𝑡, 𝑇)

With H(ndt)= precipitated rain


ℎ 𝑛𝑑𝑡 = 𝐻 𝑛𝑑𝑡, 𝑇 − 𝐻( 𝑛 − 1 𝑑𝑡, 𝑇) over time n.dt (and the corresponding return
period)
II- Synthetic hyetograph
II1 – Keifer method (Chicago synthetic hyetograph)

b) Discrete model for an advanced rain pattern


With the previous example

hauteur sur le pas de temps


18,00
rang n.Dt(mn) im (mm/mn) ajusté H(nDt) H(nDt)-H((n-1)Dt)
16,00
1 10 1,69 16,86 16,86
2 20 1,32 26,34 9,47 14,00
3 30 1,11 33,29 6,96 12,00

(mm)
4 40 0,97 38,95 5,65 10,00
5 50 0,88 43,79 4,84 8,00
6 60 0,80 48,07 4,28
6,00
7 70 0,74 51,93 3,86
4,00
8 80 0,69 55,48 3,54
9 90 0,65 58,76 3,28 2,00
10 100 0,62 61,83 3,07 0,00
11 110 0,59 64,72 2,89

10

30

50

70

90

0
12 120 0,56 67,46 2,74

11

13

15

17

19

21

23
13 130 0,54 70,07 2,61
14 140 0,52 72,56 2,49
duree (mn)
15 150 0,50 74,95 2,39
16 160 0,48 77,25 2,30
17 170 0,47 79,46 2,21
18 180 0,45 81,60 2,14
19 190 0,44 83,67 2,07
20 200 0,43 85,67 2,01
21 210 0,42 87,62 1,95
22 220 0,41 89,52 1,90
23 230 0,40 91,36 1,85
24 240 0,39 93,16 1,80

Idf curve h(ndt)


II- Synthetic hyetograph
II1 – Keifer method (Chicago synthetic hyetograph)

c) Random location of the peak


Synthetic hyetograph
Imax(t,T)
a.DT (1-a).DT
IDF curve

a.t (1-a).t

t1 t=t2-t1

DT
t t2

1 𝑡2 DT : total duration of the rain


𝐼 𝑡, 𝑇 = න 𝑖 𝜏 𝑑𝜏
𝑡 𝑡1 t1=a(DT - t)
t2=aDT+(1-a).t
a: rainfall peak intensity position
b) Kiefer method (Chicago synthetic hyetograph)
II- Synthetic hyetograph
II1 – Keifer method (Chicago synthetic hyetograph)

c) Random location of the peak

Procedure to follow for the continous case

1 𝑡2
𝐼 𝑡, 𝑇 = න 𝑖 𝜏 𝑑𝜏
𝑡 𝑡1

𝑑 𝑡2 𝑑
න 𝑖 𝜏 𝑑𝜏 = 𝐼(𝑡, 𝑇) × 𝑡
𝑑𝑡 𝑡1 𝑑𝑡

Leibnitz rule :
𝑑 𝑑 𝑑
𝑡 (𝑡) × 𝑖 𝑡2 (𝑡) − 𝑡 𝑡 × 𝑖 𝑡1 𝑡 = 𝐼(𝑡, 𝑇) × 𝑡
𝑑𝑡 2 𝑑𝑡 1 𝑑𝑡

𝑑
1 − 𝛼 × 𝑖 𝑡2 (𝑡) + 𝛼 × 𝑖 𝑡1 𝑡 = 𝑓(𝑡) with 𝑓 𝑡 = 𝐼(𝑡, 𝑇) × 𝑡
𝑑𝑡
II- Synthetic hyetograph
II1 – Keifer method (Chicago synthetic hyetograph)

c) Random location of the peak

Procedure to to follow for the continous case

With the assumption : 𝑖 𝑡1 𝑡 = 𝑖 𝑡2 𝑡


Finally : 𝑖 𝑡1 (𝑡) = 𝑓(𝑡)

So to build the hyetograph :

Take values of t : 𝑡 ∈ [0, 𝐷𝑇 ]

Calculate : 𝑡1 𝑡 = 𝛼 𝐷𝑇 − 𝑡
𝑡2 𝑡 = 𝛼𝐷𝑡 + 1 − 𝛼 𝑡

𝑖 𝑡1 = 𝑖(𝑡2 ) = 𝑓(𝑡)

Repeat with different values of t to build the rain from a set of points
II- Synthetic hyetograph
II1 – Keifer method (Chicago synthetic hyetograph)

c) Random location of the peak … procedure in the discrete case

Synthetic hietograph
IDF curve

Imax(t,T)

t
DT
t
DT

Non advanced rainfall shape : Discrete case ?

By construction,
areas : = =

t DT So …
Keiffer synthetic hietrograph
II- Synthetic hyetograph
II1 – Keifer method (Chicago synthetic hyetograph)

c) Random location of the peak … procedure in the discrete case

Advanced
Centered
Discrete hietograph
Discrete hietograph

(…etc)
1 2 3
3 1 2

4 5
5 4
6
7 6
8 7 8
II- Synthetic hyetograph
II1 – Keifer method (Chicago synthetic hyetograph)

c) Random location of the peak … procedure in the discrete case

2 3
Non centered discrete
1 2 3
hietograph
4 5 4 1 5

6 6
7 7 8
8
{
{
b) Kiefer method (Chicago synthetic hyetograph)
II- II- Synthetic
Synthetic hyetograh
hyetograph
II1 – Keifer method (Chicago synthetic hyetograph)

c) Random location of the peak … procedure in the discrete case

rang n.Dt(mn) im (mm/mn) ajustéh(nDt)*nDt h(nDt)-h((n-1)Dt) t=nDt Hk(t-Dt,t)


1 10 1,69 16,86 16,86 10 1,85
2 20 1,32 26,34 9,47 20 1,95
3 30 1,11 33,29 6,96 30 2,07
4 40 0,97 38,95 5,65 40 2,21
5 50 0,88 43,79 4,84 50 2,39
6 60 0,80 48,07 4,28 60 2,61
7 70 0,74 51,93 3,86 70 2,89
8 80 0,69 55,48 3,54 80 3,28
9 90 0,65 58,76 3,28 90 3,86
10 100 0,62 61,83 3,07 100 4,84
11 110 0,59 64,72 2,89 110 6,96
12 120 0,56 67,46 2,74 120 16,86
13 130 0,54 70,07 2,61 130 9,47
14 140 0,52 72,56 2,49 140 5,65
15 150 0,50 74,95 2,39 150 4,28
16 160 0,48 77,25 2,30 160 3,54
17 170 0,47 79,46 2,21 170 3,07
18 180 0,45 81,60 2,14 180 2,74
19 190 0,44 83,67 2,07 190 2,49
20 200 0,43 85,67 2,01 200 2,30
21 210 0,42 87,62 1,95 210 2,14
22 220 0,41 89,52 1,90 220 2,01
23 230 0,40 91,36 1,85 230 1,90
24 240 0,39 93,16 1,80 240 1,80

Example (same IDF curve as the previous examples for the advanced rainfall
event)
II- Synthetic hyetograph
II1 – Keifer method (Chicago synthetic hyetograph)

c) Random location of the peak … procedure in the discrete case

18

16

14

12

10
Hk mm

0
10

20

30

40

50

60

70

80

90

100

110

120

130

140

150

160

170

180

190

200

210

220

230

240
n.dt (mn)

Example (same IDF curve as the previous examples for the advanced rainfall
event)
II- II- Synthetic
Synthetic hyetograh
hyetograph c) Desborde double triangle hyetograph
II2 – Double triangle rainfall event (Desbordes rainfall)
Parameters value: « Guide de Construction et d’utilisation des pluies de
projets », Service Techniques de l’Urbanisme, Paris, Hemain (1986)

Dp DT HT HM1

Example of parameters value. Toulouse-Blagnac (Region 2). Return period T=10 ans for the
duration of the intense period

Same procedure as IDF curves to get the values in the table above.
(which is an example)
For a given set of N events with the same duration DT,
- statistical analysis of the peak events on duration Dp from a
subsample of P events : water depth H(Dp,T) and associated return period
T
- for a given set of P’ subvent whose return period is T for the
peak intensity (so such as H roughly H(Dp,T) on Dp HM1
a) Calculation of the associated water depth on DT
b) Calculation of the return period of the associated
water depth
Dp
tp DT
II- II- Synthetic
Synthetic hyetograh
hyetograph

II2 – Double triangle rainfall event (Desbordes rainfall)

20,00

18,00

16,00

14,00
Hdesborde (mm)

12,00

10,00

8,00

6,00

4,00

2,00

0,00
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
n dt (mn)
II- Synthetic hyetograph

II2 – Double triangle rainfall event (Desbordes rainfall)

Chocat B., Thibault S., Bouyat M. ; "Etude comparative des résultats fournis par la méthode de Caquot et le modèle du
Chocat et al (1981): réservoir linéaire" ; TSM l'eau ; n°7 ; pp 417-424 ; juillet 1981.

1 - (0.1) 120  a  2
t
b +1
i(t2 ) = (0.25.K )b b

0.9  0.1 b

 
Qe

b (0.1) - 1
b
imax = (0.25.K )  120  a  2 b
Qs K
0.9  0.1b
2.d=0.5 K t2=2.25 K t3=2.5.K r=(t3-t2)/t3=0.1
t

imax
t3 • K=time lag : K> 10 mn

• a=a(T) et b=b(T) = coeff de Montana

t et K in mn
i(t2) i(t2) et imax in mm/h

t2 2.d
DT=2.t3
II- Synthetic hyetograph c) Desborde double triangle hyetograph
II2 – Double triangle rainfall event (Desbordes rainfall)

i(t2 ) = (d )b
1 - (0.1) 120  a  2
b +1
b
Canoe Software : symetrical rainfall event

0.9  0.1 b
• DT=10.d (arbitrary) (correspond to Chocat’s and al

imax = (d )

b (0.1) - 1
b

120  a  2b •
results)
t2=(DT-2d )/2
0.9  0.1b

Meaning that the synthetic hyetograph is


made independant on the catchment
response (K is not involved)

imax
• a=a(T) et b=b(T) = coeff de Montana

t et K in mn
i(t2) et imax in mm/h
i(t2)

t2 2.d
DT
II- Synthetic hyetograph

II.3 –Simple One Triangle Rainfaill event (symetrical)

𝑏(𝑇)
𝑖𝑚𝑎𝑥 = 2. 𝑎 𝑇 . 𝐷𝑇
𝑖𝑚𝑎𝑥
This kind of synthetic hietograph is suitable for catchment
with fast response (K below 30 mn).

Above that value, double triangles hyetograph seems


more suitable

tp
DT
. From Chocat et al (1981) : K<10 mn. DT=8.9 C4.23.b . K and DT in mn

C=2.78.IMP.A where IMP=% imperviousness and A surface in ha.


II- Synthetic hyetograph
To summarize

Statistical analysis : IDF Curves 𝐼(𝑡, 𝑇)

Maximum average intensity over duration t

Synthetic hyetograph of a rainfall with duration t (continuous or discrete)

Constant rainfall
Rainfall with a chronogy : i(t). Triangle or double triangle
Fulfill the I.D.F statistic :

i(t) 1 𝑡2
𝐼 𝑡, 𝑇 = න 𝑖 𝜏 𝑑𝜏
𝑡 𝑡1
𝑖 𝑡, 𝑇

t
d

𝑖 𝑡, 𝑇 = 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡𝑒 = 𝐼(𝑡, 𝑇)
advanced
General case
II- Synthetic hyetograph
To summarize

Those synthetic rainfall events are punctual events i(x,y,t) associated to a water depth
H(x,y) over the duration t

For some applications, one are interested in average water depth over the surface
catchment A
1
A 
LM ( A, d ) = H ( x, y, d )dA
A
This require spatialization methods

Constant rainfall
Rainfall with a chronogy : i(t). Triangle or double triangle
Fulfill the I.D.F statistic :

i(t) 1 𝑡2
𝐼 𝑡, 𝑇 = න 𝑖 𝜏 𝑑𝜏
𝑡 𝑡1
𝑖 𝑡, 𝑇

t
d

𝑖 𝑡, 𝑇 = 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡𝑒 = 𝐼(𝑡, 𝑇)
advanced
General case
III- Rainfall spatialization
III- Rainfall spatialization
III-1 Thiessen Polygon’s method (=Voronoi method)

1
LM ( A, d ) = 
A A
H ( x, y, d )dA

1
LM ( A, d ) =  Aj .H j = j w j .H j
A j
t
H j =  i j ( x j , y j , )d
0

Tiré du cours du Pierre F Lemieux, Departement Génie Civil, Université de Sherbrooke


III- Rainfall spatialization

III-2 Spatial damping coefficient

On previous slides : punctual rainfall model.

To compute the hydrograph :


- Either one assumes that the rain is uniform all over the catchment
(not always realistic)
- Either one tries to take into account some spatialization

In the later case, spatialization often based on simplified assumption on the rainfall event
structure (epicenter), in particular, continuous decreasing of the rain intensity compared
to the distance with this epicenter

• LM(A,t) = <a> HM (xE,yE,t) avec <a>=<a>(A)

xe, ye : rainfall epicenter


LM : average water depth on surface a for a duration t.
HM : average water depth at the epicenter on the duration t
III- Rainfall spatialization
III- Rainfall spatialization C) Spatial damping coefficient
III-2 Spatial damping coefficient

ie ( xe , ye , t ) Rainfall measured at the « center »

d
H e ( xe , ye ) =  ie ( xe , ye , t )dt Corresponding water depth over duration d
0

d
H ( x, y, t ) =  i( x, y, t )dt Water depth over duration d at an other localation (x,y)
0

i( x, y, t ) = a ( x, y).ie ( x, y, t ) Rainfall intensity assuming spatial damping


III- Rainfall spatialization

III-2 Spatial damping coefficient

1
LM ( A, d ) = 
A A
H ( x, y, d )dA

1  
d
= 
A A 0
  i ( x, y, t )dt dA

1  
d
=    a ( x, y )ie ( xe , ye , t )dt dA
Assumtion :
Spatial damping independant A A 0
on the time. Caution !!!! 
d
1
=
A A a ( x, y )dA  ie ( xe , ye , t )dt
0

LM ( A, d ) = a ( A) H e ( xe , ye , d )

Burkli and Ziegler : <a>(A)=A-e (Model used in Caquot’s method)


III- Rainfall spatialization
C) Spatial damping coefficient
III-2 Spatial damping coefficient

<a>(A) dependance with the rainfall event duration but also the return period

Probabilistic approach (as for the I.D.F curves). Frequency analysis on the quantity L M(A,t) to get
<a>(A,t,T)
III- Rainfall spatialization
C) Spatial damping coefficient
III-2 Spatial damping coefficient

• Rainfall damping depending on the distance to the epicentre : a(r). Can be computed from
the knowledge of <a>

rI Intensité
Bassin I E Pluie à l’épicentre

A Pluie sur le bassin I


rII
Pluie sur le bassin II

B temps
Bassin II

Assumption : concentric damping. <a>(A)=A-e


a(r)=P-e(1-e).r-2.e
Cf demonstration as an exercize
III- Rainfall spatialization

III-2 Spatial damping coefficient

As a conclusion
Simple definition of the spatial damping <a(A)> does not suit well the structure of real spatial
event, as on small time step, fluctutions from an event to another

Probabilistic spatial dumping more satisfying on a conceptual point of view.


But results depend on the method followed to sample event :
In particular, there is no reason that the real event is centered on one of the pluviometer used
to perform statistics (calculation performed on rainfall event already damped)

Mathematical relations are empirical and assume a given « shape » of the spatial variation
(with a continuous decrease) around the center : radar measurements show that this simplified
view is not close to reality.

In urban hydrology :

- For surface catchment lower than 1000 ha : safer to consider a spatially constant
rainfall without damping (rainfall equal to the one considered at the « center »)

- For surface catchment greater than 3000 ha : more precise modelling of the spatial
repartition of rain. Synthetic hyetograph not advised. Better to work with real rainfall event

You might also like