Lumped and distributed modelling of suspended solids in a

Report
Lumped and distributed modelling
of suspended solids in a
combined sewer catchment in
Santiago de Compostela (Spain)
R. Hermida, JOSE ANTA, M. Bermúdez, L. Cea, J. Suárez & J.
Puertas
GEAMA Research Team
Universidade da Coruña
1
INTRODUCTION
• Flow and Pollution Modelling in Urban Systems
dust and dirt buildup
washoff
gully-pot processes
sewer erosion - transport
2
OBJETIVES
• Comparison of a lumped and distributed model
for TSS in “El Ensanche” combined sewer
catchment
• Model developed with Infoworks CS 9.x
– Ackers-White equation
– KUL model
• 10 rain event were used for model calibration.
More details presented yesterday:
“Mobilized pollution indicators in a combined sewer
system during rain events” del Río et al.
3
DESCRIPTION OF THE URBAN CATCHMENT
4
MODEL DEVELOPMENT
• Distributed model
(del Río, 2011)
– 316 subcathments:
183 streets, 128
roofs, 5 pervious
areas
– 7 km of pipes
(150 – 1200 mm)
• Lumped model
(Hermida, 2012)
5
BUILDUP
M0 
Ps
K1
 1  e
 N J K1

Model parameters : Ps, K1
WHASOFF
()
()
K a t = C1 × i t
C2
()
- C3 × i t
Model parameters : C1, C2, C3
6
SEWER TRANSPORT MODELS
Ackers & White (1996)
ì
u
æ d50 ö
ï
×í
- K × lc × g × ç
÷ø
R
è
ïî g × s - 1 × R
Model parameter are fixed. Model variables: s, d50
a
b
æ
R ö æ d50 ö
g
Cv = J × ç We × ÷ × ç
×
l
c
A ø è R ÷ø
è
(
)
e
ü
ï
ý
ïþ
m
KUL (Boutelegier and Berlamont, 2002)
æ t - t cr ,erosion ö
qs = a erosion × ç
÷
t
è cr ,erosion ø
berosion
d50
t cr ,erosion = g erosion × g × s -1 × r ×
1000
(
)
Too many model parameters (6 parameters).
Model variables: s, d50
7
SEWER TRANSPORT MODELS
KUL : Shields approach (Shizari and Berlamont, 2010)
g erosion = Shields Number = f (Re*)
Shields number has to be re-evaluated in each time – step
(not allowed in IF)
Ota and Nalluri equation (2003)
 e  0.036
 e  1.67
e  s
gd 50 s  24  e
3
e







(
F = 24× qb - 0.036
)
1.67
KUL equation as function of s, d50
g e = 3g d
8
SENSITIVITY ANALYSIS: POLLUTION MODEL
• InfoWorks doesn’t allow an easy implementation
of formal MC inference
• Sensitivity analysis of the different Infworks
quality subroutines with Matlab.
• Methodology proposed by Kleidorfer (2009):
– Local sensitivity analysis
– Global sensitivity analysis
• Graphical methods
• Hornberger – Spear – Young
9
SENSITIVITY ANALYSIS RESULTS
BUILDUP
Buildup factor is more sensitivity than the decay factor
Model is sensitivity to both parameters
WHASOFF
Model is almost insensitivity to C3 coefficient and can be neglected
C2 is more sensitivity than C1
Model is sensitivity to both parameters
SEDIMENT TRANSPORT MODEL
d50 is more sensitivity than the specific density s
Model is sensitivity to both parameters
10
MODEL CALIBRATION
Hydraulic model calibration
• 11 rainy days: NS=0.85
Pollution model calibration
• Visual calibration: 3 events
• Model validation: 7 events
• Distributed model
– Ackers – White
– KUL (Ota & Nalluri)
• Lumped model
– Ackers – White
– KUL (Ota & Nalluri)
11
MODEL CALIBRATION
12
CONCLUSIONS
•
•
•
•
•
•
Successful application of sensitivity analysis to
determine the most relevant parameters for pollution
modelling in InfoWorks CS
All the sensitivity tests shows similar results
Lumped model works better in terms of NS and EMC
Distributed model works better in terms of Cmax
KUL – Ota & Nalluri approach avoids the
determination of a large number of model
parameters
Ackers – White is more accurate than KUL approach
for lumped model and viceversa.
13
www.geama.org/hidraulica
THANKS FOR YOUR
ATTENTION
[email protected]
14
SENSITIVITY ANALYSIS
Hornberger – Spear – Young Method (Kleidorfer, 2009)
– MC framework
– Comparison of model outputs with a synthetic run with NS
– Analysis of the distance of behavioral (NS>0) and non
behavioral (NS<0) empirical cumulative pdf
NashSutcliffe
Synthetic run
15
BUILDUP
HSY: Ps
HSY: K1
16

similar documents