By Narendra Kumar Tiwary (Executive Engineer, Water ... SWAT model for real time flood forecasting...
Transcript of By Narendra Kumar Tiwary (Executive Engineer, Water ... SWAT model for real time flood forecasting...
By
Narendra Kumar Tiwary
(Executive Engineer, Water Resource Department, Bihar, India
Ph.D Scholar, IIT Delhi)
Prof. A.K.Gosain, IIT Delhi
Prof.A.K.Keshari ,IIT Delhi
Civil Engineering Department
Indian Institute of Technology, Delhi
Use of New Technologies & Softwares in Flood sector
Drainage Map of the Bagmati Basin
±
Legend
Drainage lines
Basin boundary20 0 20 40 6010
Km
IntroductionFlood is associated with a serious loss of life, property and damage to utilities.
05-08-2014 3
Event-1 The man-made embankments of river Kosi failed
and Flooded north Bihar, India during 18 August 2008
434 Dead bodies were found until 27 November 2008
05-08-2014 4
Kosi
Burhi
GandakGanda
k
05-08-2014 7
Event-2 In June 2013, a multi-day cloudburst centered on the
North Indian state of Uttarakhand caused devastating floods and landslides in the country's worst natural disaster since the 200 4 tsunami more than 5,700 people were "presumed dead. Destruction of bridges and roads left about 100,000 pilgrims and tourists trapped in the valleys
05-08-2014 8
Event-2
05-08-2014 9
Dehradun -Capital of Uttarakhand
05-08-2014 10
Challenge for Water Resource Engineers and Scientists Today, with modern equipments and radars we should
have been able to predict the floods and expected inundation much earlier and stopped these human disasters
05-08-2014 11
Important Questions to be answered
Where the floodplain and flood-prone areas are?
How often the flood plain will be covered by water?
How long the flood-plain will be covered by water?
At what time of year flooding can be expected?
05-08-2014 13
To modify SWAT model for real time flood forecasting and apply
modified SWAT model for real time flood forecasting of river basins
using web based data,
To enhance the forecast lead time by incorporating the rainfall
forecast issued by IMD and assess its accuracy.
To map expected inundation, and flood zonation by integrating
SWAT outputs into HEC GeoRAS and HEC-RAS models
Objective of the study
Real Time Flood Forecasting and Flood Plane Zoning System
COPONENTS The Real Time Flood Forecasting and Flood Plain Zoning
System usually consists of one or more of the following components:
1. Rainfall Runoff Model.
2. Runoff Routing Model.
3. Error Analysis and Updating Technique
4. Hydrodynamic model for channel routing
(HEC-RAS)
5. Generating TIN and delineating flood plain
05-08-2014 14
METHODOLOGY USED IN PRESENT STUDY1
.
1. Watershed delineation using SRTM
DEM
2. Formation of HRUS
3. Writing input tables
4. Runoff generation using Green Ampt
Method
5. Developing subroutines and
modifying SWAT for real time flood
forecasting application
Models for Forecasting Statistical/ Black box
Correlation
UH Based
Stochastic (ARMAX)
Conceptual (Water Balance)
SWAT
HEC-HMS
Methods for Flood Forecasting
Simple correlations
Use very little data
Only gauge data is sufficient
Work very well for reaches that have very
small contribution from rain
Physically Based ModelsTo use the equations of mass, energy and momentum to
describe the movement of water over the land surface and through the unsaturated and saturated zones.
05-08-2014 18
Ways to enhance the forecasts
Enhanced forecasts require data on
rainfall
To incorporate the contribution of
the catchment
To incorporate the impact of spatial
variability
Forecast lead time increases on
account of catchment lag
Addition to lead time is possible
through precipitation forecasting
Typical flood forecasting situation
Effective Lead Time
Total Forecast Lead Time
Time used by
observation and
forecasting
Flo
od
pro
du
cin
g s
torm
Tim
e e
lapsed in
obser.
Tra
nsm
issio
n T
ime
Data
pro
cessin
g
Fore
cast fo
rmula
tion
Tra
nsm
it ion tim
e
Rain
beginsRain
ends
Forecast reaches
userFlood reaches
outflow station
Rain
forecast
05-08-2014 21
Gamma Distribution UH Method
05-08-2014 22
05-08-2014 23
05-08-2014 24
05-08-2014 25
START
Read Input Data
Initialize Parameters
Daily Loop
Yes
End
Simulation
Last hr of Last Day
Simulation
If eventstatus=1 Perform Hourly Simulations
If eventstatus=2 Forecast hourly
Route
Command
Subbasin
Command
Read Sub –daily Precipitation
Write Outputs
to File to be Read by Another SWAT Run
Generate Max/Min Temperature
No
No
Compute and Print Final Basin Statistics
Capture Catchment Characteristics,Save values of Soil Water,Reach Storage,Base flow for next SWAT run
First Hour of First Day Of Simulation and want real time forecast
Hour loop
Import catchment characteristics saved in previous run
Routing Command loop
add
Command
save
Command
Add waterRoute water hourly
through channel
Yes
No
05-08-2014 26
05-08-2014 27
Call
openwth
Call std2
Call simulate
Call finalbal
Call writeaa
Call peatw
Call std3
Call xmon
Call clicon
Call command
Call writed
Call xmon
Call writem
Call tillmix
Call
PmeasCall Tmeas
Call
SmeasCall
HmeasCall
WmeasCall Pgen
Call Weatgn
Call Tgen
Call Clgen
Call Wndgen
Call Slrgen
Call Rhgen
Call Pgen
Aunif/Dstn1
Call
Aunif/Dstn1
Call Tgen
Call Clgen
Call Rhgen
Ee/Aunif/Dstn1Call Wndgen Aunif
Call Aunif/Dstn1
Call
Ee/Aunif/Dstn1Call Aunif
Call Weatgn Aunif/Dstn1
Call Slrgen
Call Subbasin
Call
Route
Call Recmon
Call Recyear
Call Save
Call Recday
Call Reccnst
Call Structu
Call Routres
Call Transfer
Call Addh
Call Saveconc
Call Varinit
Call
AlbedoCall Solt
Call Surface
Call Percmain
Call Etpot
Call Etact
Call Fert
Call Crpmd
Call Graze
Call Nminrl
Call Nitvol
Call Pminrl
Call Gwmod
Call Apply
Call Washp
Call Decay
Call Pestlch
Call Enrsb
Call
PestyCall Orgn
Call Psed
Call Nrain
Call Nlch
Call Solp
Call Subwq
Call Bacteria
Call Urban
Call Pothole
Call Filter
Call Latsed
Call Gwnutr
Call Surfstor
Call
SubstorCall Wetlan
Call Hrupond
Call Irrsub
Call Autoirr
Call Watuse
Call Watbal
Call Sumv
Call Virtual
Call Operatn
SWU (Soil Water
Uptake)
Call Curno
Grow (Plant Growth)
Tstr (Temp Stress)
Nup (Nitrogen Uptake)
Npup (Phosphorus Uptake)
Anfert (Autofert)
Call Nfix Call Nuts
Call Nfix
Call canopyint
Call snom
Call leaf_litter *
Call crackvol
Call volq
Call crackflow
Call surfst_h2o
Call alph
Call pkq
Call tran
Call eiusle
Call ysed
Call percmicro
Call
percmacro
Call Ee
Call Expo
Call Erfc
Call Theta
Call Regres
Call pond
Call Irrigate
Call Irrigate
Call SbsdayCall Sweep
Call impndday
Call Bsbday
Call alph
Call pkq
Call ysed
Call enrsb
Call pesty
Call orgn
Call psed
Call Tair
Call rchinit
Call rtover
Call rtday
Call rtmsk
Call rthourly
Call rthmsk
Call rtsed
Call watqual
Call noqual
Call rtpest
Call rtbact
Call irr_rch
Call rchuse
Call reachout
Call resnut
Call resinit
Call irr_res
Call res
Call lakeq
Call Borwet *
Call Curno
Call Curno
Call Plantop
Call Dormant
Call Harvkillop
Call Harvestop
Call Killop
Call Tillmix
Call impndmon
Call bsbmonCall rchmon
Call
writea
Call sbsmon
Call impndyr
Call rchmyr
Call sbsyr
Call bsbyr
Call
headout
Call header
Call swbl
Call vbl
Call
sbsaaCall
impdaaCall
rchaaCall
bsbaaCall
stdaa
Call
resopmrb
Call
sat_excess
Call
bmp_eff
Call dailycn
Call surq_daycn
Call surq_greenampt
Not
currently
active
Call rchday
Call curno Call ascrv
Call date_and_time
Call initial
Call std1
Call
readwwq
Call readinpt
Call
readsubCall readwgn
Call readswq
Call readrte
Call opensub
Call
readurban
Call readpnd
Call readwus
Call readres
Call readcrop
Call
readtillCall readpest
Call
readfert
Call readhru
Call readchm
Call readmgt
Call readsol
Call readgw
Call
soil_chemCall soil_phys
Call rteinit
Call h2omgt_init
Call hydroinit
Call impd_init
Call getallo Call
caps
Call
openfile
Call caps
Call readfig Call caps
Call readfile Call
caps
Call
readbsnCall
ascrv
Call
readcod
Call
gcycl
Call caps
Call
readlwqCall ascrv
Call
caps
Call
curno
Call
ttcoef
Call
allocate_parms
Call
Zero0Call
zero1Call
zero2
SWAT FlowChart
05-08-2014 29
05-08-2014 30
05-08-2014 31
05-08-2014 32
05-08-2014 33
05-08-2014 34
05-08-2014 35
Model Set up for study areaData sets used
The SWAT model requires data on terrain, land use, soil, and weather for assessment of inflows and outflows of reaches. Following datasets have been used for setting up of the SWAT model:
(1) DEM – SRTM (90 m resolution)
(2) Landuse – Global USGS (2 M)
(3) Soil – FAO Global soil (5 M)
(4) Rain gauges /Temperature gauges – IMD
(5)Stream Gauges – CWC05-08-2014 36
Study Area
INPUT DEM
05-08-2014 38
05-08-2014 39
05-08-2014 40
05-08-2014 41
Division into subbasins
05-08-2014 42
Soil Details
05-08-2014 43
SEQN SNAM NLAYERS HYDGRP TEXTURE SOL_AWC1 SOL_K1 CLAY1 SILT1 SAND1
3663 3663 2 C LOAM 0.117 35.65 24 35 42
3664 3664 2 C LOAM 0.157 28.52 17 36 47
3682 3682 2 D LOAM 0.175 7.77 24 36 40
3684 3684 2 C LOAM 0.175 13.92 22 38 41
3695 3695 2 C LOAM 0.175 14.96 20 40 40
3743 3743 2 D LOAM 0 6.48 18 44 38
3761 3761 2 C LOAM 0.175 24.73 20 34 46
3808 3808 2 D LOAM 0.175 6.17 21 35 44
3851 3851 2 D CLAY_LOAM 0.137 7.17 27 35 37
Table-6.2 Soil properties
GAUGE STATIONS
05-08-2014 44
05-08-2014 45
05-08-2014 46
05-08-2014 47
05-08-2014 48
05-08-2014 49
05-08-2014 50
y = 108.73x - 4204.1R² = 0.8808
-500.00
0.00
500.00
1000.00
1500.00
2000.00
0.00 20.00 40.00 60.00
dis
ch
arg
e in
m3
/s
Stage in M
discharge-stage for Hayaghat
Series1
Linear (Series1)
05-08-2014 51
y = 4E-09x3 - 1E-05x2 + 0.0182x + 38.219R2 = 0.9605
0.000
10.000
20.000
30.000
40.000
50.000
60.000
0.00 500.00 1000.00 1500.00 2000.00
Sta
ge in
M
Discharge in Cumecs
Stage-discharge for Hayaghat
Series1
05-08-2014 52
y = 4E-07x3 - 0.0002x2 + 0.0567x + 44.834R² = 0.936
44.000
45.000
46.000
47.000
48.000
49.000
50.000
0.000 50.000 100.000 150.000 200.000 250.000 300.000
Sta
ge i
n M
Discharge in Cumecs
Stage-Discharge for Benibad
Results
05-08-2014 53
Calibration - Output
Days
Dis
charg
e m
3/s
Calibration Results Seventeen parameters were selected for sensitivity
analysis, calibration, validation and uncertainty analysis to be carried out by SWAT-CUP4 using SUFI algorithm. On the basis of global sensitivity analysis it can be concluded that most sensitive parameter is CN2.mgt followed by GW_DELAY, CH_N2.rte, ESCO.hru, HRU_SLP.hru, GWQMN.gw, GW_REVAP.gw and SURLAG.bsn. P-factor and r-factor for calibration were found to be 0.74 and is 0.44 respectively, which are very much within the range recommended for a perfect model. Seventy four percent observed and simulated values lie in 95PPU. P-factor and r-factor for validation period are 0.63 and
05-08-2014 55
Validatio for monthly flow
05-08-2014 56
Model Performance Test
The correlation statistics (R2) was determined to measure the linear correlation between the actual and the predicted values and it was found to be 0.93 which indicates the best performance of the model.
R 2 = = 0.96
To estimate the efficiency of the fit, the Nash-Sutcliffe coefficient (Nash) (Nash and Sutcliffe, 1970) was used. It was found to be 0.96 which indicates the best performance of the model.
Nash = 1 - = 0.7505-08-2014 57
Daily Flow Valdation
05-08-2014 58
Flow Hydrograph Daily Simulation
05-08-2014 59
0.0
20.0
40.0
60.0
80.0
100.0
120.0
140.0
160.00
200
400
600
800
1000
1200
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
Dis
cha
rge
in
cu
me
cs
1-July Time in days 31-July
Flow Hydrograph of Hayaghat of July-04
RF at Simra
RF at Kathmandu
RF at Nagarkot
RF at Benibad
RF at Hayaghat
HayaOBS
HayaSim
Valdation for daily flow 2010
05-08-2014 60
Validation for hourly flowNE = 0.72
05-08-2014 61
Validation Graph
05-08-2014 62
ERROR ANALYSIS Time series analysis of errors in hourly
simulations/forecasts has been performed. An ARIMA model developed for forecasting error has been deployed for correcting simulated/forecasted hourly flows. Integration of simulation capability of SWAT model and error forecasting capability by time series analysis has been done for solving the problem of real-time flood forecasting
05-08-2014 63
Error 2004 Simulation
05-08-2014 64
-400
-300
-200
-100
0
100
200
300
400
500
1
16 31 46 61
76 91
106
121
136
151
166
181
196
211
226
241
256
271
286
301
316
331
346
361
376
391
40
6
421
436 451
46
6
48
1
49
6
511
526
541
556
571
586
Dis
cha
rge
in
cu
me
cs
4-July-04 Time in hours 28-July-04
ERROR 4-July-28-July-04
ERROR
Error 2004 hourly simulation
05-08-2014 65
-60
-40
-20
0
20
40
60
80
100
120
1
18 35 52 69
86
103
120
137
154
171
188
205
222
239
256
273
290
307
324
341
358
375
392
40
9
426 44
3
46
0
477
49
4
511
528
545
562
579
596
FIRST DIFF OF ERROR 4-July-28-July on y-axix and time in hrs on X-axix
FIRST DIFF
Error analysis of 2004 flood event
ARIMA Model Parameters for error analysis for hourly simulation of year 2004
ARIMA Parameters
B1(LAG1)= 1.3153
B2(LAG2)= -0.0162
B3(LAG3)=-0.3013
05-08-2014 66
Error Analysis There are 600 observed and simulated values. Four
hundred data have been used for determining ARIMA model parameter and two hundred forecasted errors have been determined. Nash Sutcliffe coefficient before correction for forecasted period of error was 0.64 and after error correction it is 0.795.ARIMA model parameters are given below.
b = 1.3153, -0.0162, -0.3013
Z(t)= 1.3153*Z(t-1)-0.0162*Z(t-2)-0.3013*Z(t-3)
05-08-2014 67
05-08-2014 68
0
100
200
300
400
500
600
700
800
900
1 11 21 31 41 51 61 71 81
91
101
111
121
131
141
151
161
171
181
191
201
211
221
231
241
251
261
271
281
291
301
311
321
331
341
351
Dis
cha
rge
in
cu
me
cs
2-Aug Time in hour 22Aug
Hourly Simulation 2nd August-22nd August-05
OBS
SIMCorrected
SimOld
Hourly 2005 ARIMA Model Parameters for error analysis for hourly
simulation of year 2005
ARIMA Parameters
B1(LAG1)= 0.6056
B2(LAG2)= 0.1792
B3(LAG3)=0.0988
05-08-2014 69
05-08-2014 70
05-08-2014 71
Real time flood forecasting graph
05-08-2014 72
Real time forecasting graphs
05-08-2014 73
0.00E+00
5.00E+00
1.00E+01
1.50E+01
2.00E+01
2.50E+01
3.00E+01
3.50E+01
4.00E+01
4.50E+010
200
400
600
800
1000
1200
1400
1600
1 8 15 22 29 36 43
50 57 64 71 78 85
92
99
106
113
120
127
134
141
148
155
162
169
176
183
190
197
204
211
20 July time in hours 28 July
Real time forecasts issued 0n 20th July and 23 rd July'04
rainfall
Obs
Forecast-205-9
Forecast-202-5
Forecast-202-5
Real time forecast
05-08-2014 74
Forecast generated at Benibad for July 2004 event
05-08-2014 75
Forecast generated at Benibad for July 2004 event
05-08-2014 76
Real time flood forecasting Graph
05-08-2014 77
Real time flood forecasting of Haya
05-08-2014 78
Hourly Discharge
05-08-2014 79
SUMMARY AND CONCLUSIONS
05-08-2014 80
(1)Increase in lead time(cathment response delay+travel time in channel)
(2)Application possible in ungauged cathments also
(3)Sediments,water quality and land use can be modelled on real time basis
simultaneously
(4)Physically based
(5)Easy in adaptability
(6)Capable of modelling snow melt, percolation, lateral flow and base flow
(7)Parsimonious
(8)GIS based
(9)Inundation mapping and flood zonation easily possible
(10)Easy updating of catchment characteristics
(11)Capable to incorporate rain fall forecast of IMD and increase lead time
(12)Increase in accuracy
(13)Capable of generating synthetic sub-hourly hydrographs
CONCLUSIONS
05-08-2014 81
As outcome of the present research work it can be
concluded that Modified SWAT model combined
with time series error forecasting ARIMA model and
HEC-RAS hydraulic model can be used as an
efficient tool for real-time flood forecasting and
inundation mapping. Modified SWAT model
reproduces stream flow hydrographs reasonably
under multiple storm events and can be used as a
real-time flood forecasting tool with enhanced lead
time. Rainfall forecast can further increase the
forecast lead time.
Thank You
END
05-08-2014 82