Download as pdf or txt
Download as pdf or txt
You are on page 1of 68

Flood Simulation using Weather Forecasting and Hydrological

Models

Thesis submitted to the Andhra University, Visakhapatnam in partial fulfilment of


the requirement for the award of Master of Technology in Remote Sensing and GIS.

Submitted By:
Mr. Pratiman Patel

Supervised By:

Dr. Shiv Prasad Aggarwal Dr. Praveen Kumar Thakur


Head Scientist/ Engineer ‘SE’
Water Resources Department, Water Resources Department,
Indian Institute of Remote Sensing Indian Institute of Remote Sensing
Dehradun Dehradun

Indian Institute of Remote Sensing,


Indian Space Research Organnization,
Dept. of Space, Govt. of India,
Dehradun – 248001
Uttarakhand, India

June, 2015
ACKNOWLEDGEMENT

I would first and foremost like to express my deepest sense of gratitude to our respected
director, Dr. A. Senthil Kumar, for his continuous encouraging suggestions and support
during the research which from time to time has led us to rethink and establish our goals and
objectives in a defined way. I would also like to thank Dr. Y.V.N. Krishna Murthy, former
Director, IIRS for his valuable suggestions at the time of synopsis too.

I would like to express my special thanks of gratitude to my supervisors Dr. Shiv Prasad
Aggarwal, Head, Water Resources Department and Dr. Praveen Kumar Thakur,
Scientist/Engineer ‘SE’, Water Resources Department, IIRS, who gave me the golden
opportunity to do this project on the topic “Flood Simulation using Weather forecasting
and Hydrological Models”, which also helped me in doing a lot of Research and I came to
know about so many new things. I am really thankful to them.

I would like to thank to Dr. Bhaskar R. Nikam, Dr. Vaibhav Garg and Mr. Arpit Chouksey
who helped in understanding various subjects during the course work. They have provided
support in various forms. Their suggestions always have been very important to me. I would
like to thank Andhra University for providing Master of Technology Degree in Remote
Sensing and GIS. Along with this, I would also like to thank Dr. S.P.S. Kushwaha, Dean
(Academics), Dr. S.K. Saha former Dean (Academics) and Ms. Shefali Agarwal, M.Tech
Course Director, Indian Institute of Remote Sensing, Dehradun for providing an opportunity
to do this research.

I would like to thank Hydrologic Engineering Center (HEC) and WRF community for
providing the HEC-HMS, HEC-GeoHMS and source code of WRF-ARW, WRF-Hydro,
MET and NCL scripts respectively. Special thanks to Central Water Commission (CWC) for
providing the discharge data of Joshimath and Uttarkashi for model calibration and validation.

I would not be withstanding the course without the help of my friends Vikrant, Surya, Ram,
Raja, Abhishek Saikia, Prakit, Sanjay, Neeraj, K.D., Sukant, Sakshi, Richa, Raunak, Rohit sir
and all others. Their friendship supported me a lot and the time with them can never be
forgotten.

Lastly my moral supporters my family, grandparents, mom, dad & prateek for their utmost
care and concern. They are real motivation to me, without them I would have never thought
of completing this project.

Date: 16 June 2015 Pratiman Patel

II
CERTIFICATE

This is to certify that Mr. Pratiman Patel has carried out the dissertation entitled “Flood
simulation using weather forecasting and hydrological model” in partial fulfilment of the
requirements for the award of Master of Technology in Remote Sensing and GIS. This
work has been carried out under the supervision of Dr. Shiv Prasad Aggarwal, Head, Water
Resources Department and Dr. Praveen Kumar Thakur, Scientist/Engineer ‘SE’, Water
Resources Department, Indian Institute of Remote Sensing, Indian Space Research
Organisation, Dehradun, Uttarakhand, India.

Dr. Shiv Prasad Aggarwal Dr. Praveen Kumar Thakur


Head Scientist/ Engineer ‘SE’
Water Resources Department, Water Resources Department,
Indian Institute of Remote Sensing, Indian Institute of Remote Sensing,
Dehradun Dehradun

Dr. S.P.S. Kushwaha Dr. A. Senthil Kumar


Dean (Academics) Director
Indian Institute of Remote Sensing, Indian Institute of Remote Sensing,
Dehradun Dehradun

III
DECLARATION

I, Pratiman Patel hereby declare that this dissertation entitled “Flood Simulation using
Weather forecasting and Hydrological Models” submitted to Andhra University,
Visakhapatnam in partial fulfilment of the requirements for the award of Master of
Technology in Remote Sensing & GIS, is my own work and that do the best of my
knowledge and belief. It is a record of original research carried out by me under the guidance
and supervision of Dr. Shiv Prasad Aggarwal, Head, Water Resources Department and Dr.
Praveen Kumar Thakur, Scientist/Engineer ‘SE’, Water Resources Department, Indian
Institute of Remote Sensing, Indian Space Research Organisation, Dehradun. It contains no
material previously published or written by another person nor material which to a substantial
extent has been accepted for the award of any other degree or diploma of the university or
other institute of higher learning, except where due acknowledgement has been made in the
text.

Place: Dehradun Mr. Pratiman Patel

Date: 16 June 2015

IV
DEDICATED TO MY FAMILY

V
ABSTRACT
Floods are among one of the most common disaster that happen all around the world, every
year. In mountainous areas flash floods are very common phenomenon due to heavy
precipitation, cloud burst, landslide, or glacier lake outburst. In this research project, flash
floods prediction due to heavy precipitation is being simulated using a weather forecasting
model (WRF-ARW) for precipitation prediction and uncoupled hydrological model for
rainfall-runoff (HEC-HMS) and hydrodynamic (MIKE 11) modelling.

Weather forecasting model (WRF) is parameterized for land use/ land cover, topographical
data and six microphysics and cumulus schemes combinations. These parameters are used for
simulating three past events of cloud bursts i.e. 01 Aug 2012, 13 September 2012 and 13 June
2013. These parameters are validated for accuracy, probability of detection and false alarm
ratio using IMD grids of rainfall and TRMM (3B42 v7 3 hourly product). The most suitable
parameters are MODIS land use/ land cover and WRF Double Moment 6 Class Scheme with
Grell 3D as microphysics and cumulus schemes respectively.

The best output of WRF is selected for hydrological modelling. Hydrological is performed
for uncoupled and coupled models (WRF-Hydro). HEC-HMS is calibrated and validated
before using it for results obtained from WRF. The results of HEC-HMS model for WRF
outputs are compared with TRMM 3B42 v7 3-hourly product. There is a clear under
prediction and time shift of WRF outputs with the TRMM product. This effect is also
observed in the hydrodynamic model. For coupled model, the results are not satisfactory for
the current version of WRF-Hydro. The results obtained are very promising for uncoupled
mode but there is further improvement required for precipitation prediction.

Keywords: Flood forecasting, WRF, rainfall-runoff modelling, hydrodynamic modelling,


uncoupled modelling, coupled modelling.

VI
TABLE OF CONTENTS
Sr. No. Title Page No.

1. Introduction 1

1.1 Research questions 2

1.2 Objective 2

2. Literature Review 3

2.1 Numerical Weather Prediction (NWP) 3

2.2 GIS based Hydrologic and Hydrodynamic Modelling 6

2.3 Digital Elevation Models 7

2.4 Flood 8

3. Study area and data used 10

4. Methodology 16

4.1 Model Overview 16

4.2 Overall Methodology 26

5. Results and Discussions 40

6. Conclusion and Recommendation 52

7. References 53

8. Appendix 1 56

VII
LIST OF FIGURE AND TABLES
List of figures:

Figure 3.1: Map showing the study area with major rivers and important 12
locations.
Figure 3.2: FCC of Landsat 8 for the study area. 12

Figure 3.3: Land Use/ Land Cover of the study area. 13

Figure 3.4: Soil map of the study area. 13

Figure 3.5: CartoDEM of the study area. 14

Figure 3.6: AWiFS land use/ land cover for domain 03 of WRF. 14

Figure 3.7: MODIS land use/ land cover for domain 03 of WRF. 15

Figure 3.8: USGS land use/land cover of domain 03 for WRF. 15

Figure 4.1: ARW n coordinate 16

Figure 4.2: Flow diagram for WRF. 17

Figure 4.3: SCS unit hydrograph. 20

Figure 4.4: Calculation of cross-sectional area for relative roughness not


equal to 1.0. a) Highest water level above maximum specified elevation b) 23
Highest water level below maximum specified elevation

Figure 4.5 Channel section with computational grid 25

Figure 4.6: WRF Hydro architecture showing various components. 26

Figure 4.7: Calling structure of WRF-Hydro 26

Figure 4.8: Flowchart showing overall methodology. 27

Figure 4.9: Domain Configuration of WRF. 28

Figure 4.10: Geogrid description in namelist.wps for use of MODIS 29


LULC.
Figure 4.11: GEOGRID.TBL for SRTM DEM. 30

Figure 4.12: A. CARTOSAT DEM of the area. B. CN Map generated from


LULC & soil Map. C. Impervious Map generated from ISAT-tool. D. Initial 32
Abstraction Map generated from LULC & soil map.

VIII
Figure 4.13: Watershed and River network delineated from HEC- 33
GeoHMS.
Figure 4.14: HEC-HMS model setup. 33

Figure 4.15: Flowchart showing MIKE 11 Methodology. 34

Figure 4.16: Network File of MIKE 11. 35

Figure 4.17: Cross-section file of MIKE 11. 35

Figure 4.18: Channel grid for WRF-Hydro. 36

Figure 4.19: Flow direction grid for WRF-Hydro. 37

Figure 4.20: Land use/ Land Cover for WRF-Hydro. 37

Figure 4.21: Latitude grid for WRF-Hydro. 38

Figure 4.22: Longitude grid for WRF-Hydro. 38

Figure 4.23: Stream order grid for WRF-Hydro. 39

Figure 4.24: Topographic grid for WRF-Hydro. 39

Figure 5.1: USGS LULC comparison with different microphysics and 40


cumulus schemes for 17 June 2013.
Figure 5.2: MODIS LULC comparison with different microphysics and 40
cumulus schemes for 17 June 2013.
Figure 5.3: Difference image of TRMM and WRF results for USGS &
MODIS LULC with different microphysics and cumulus schemes for 13- 41
17 June 2013.
Figure 5.4: WRF model comparison for different microphysics and 42
cumulus schemes for 15 June 2013.
Figure 5.5: IMD and WRF grid difference of different microphysics and 42
cumulus schemes for 17 June 2013.
Figure 5.6: Difference images of TRMM and WRF for different 43
microphysics and cumulus schemes (1-3 Aug 2012).
Figure 5.7: Difference images of TRMM and WRF for different 43
microphysics and cumulus schemes (13-15 Sep 2012).
Figure 5.8: WRF model comparison for different microphysics and 44
cumulus schemes for 03 August 2012.
Figure 5.9: WRF model comparison for different microphysics and 44
cumulus schemes for 15 September 2012.
Figure 5.10: Hydrograph comparison of simulated and observed runoff at 45
Uttarkashi and Joshimath for year 2005.
Figure 5.11: Hydrograph comparison of simulated and observed runoff at 45
Uttarkashi and Joshimath for year 2006.
Figure 5.12: Hydrograph comparison of simulated and observed runoff at 46
Uttarkashi and Joshimath for year 2007.

IX
Figure 5.13: Hydrograph comparison of TRMM & WRF at Uttarkashi. 46

Figure 5.14: Hydrograph comparison of TRMM & WRF at Devprayag. 47

Figure 5.15: Hydrograph comparison of TRMM & WRF at Joshimath. 47

Figure 5.16: Hydrograph comparison of TRMM & WRF at Rudraprayag. 48

Figure 5.17: Hydrograph comparison of TRMM & WRF at Haridwar. 48

Figure 5.18: Hydrograph comparison for different Curve Number Values. 49

Figure 5.19: Stage comparison of WRF and TRMM at Uttarkashi and 50


Devprayag stations.
Figure 5.20: Hydrograph at Haridwar for WRF-Hydro. 51

List of tables:

Table 2.1: Scale definitions of some important atmospheric events. 3

Table 2.2: Major cloud bursts events in Uttarakhand in 2012 8

Table 3.1: Bands in Landsat 8. 11

Table 4.1: Combination for USGS & MODIS land use land cover 28
parameterization simulations.

Table 4.2: Combinations for full parameterization. 29

Table 5.1: Curve Number Values for sub-watershed. 49

Table A1.1: Geogrid configuration for domain. 57

Table A1.2: Time information 57

Table A1.3: Domain configuration for namelist.input 57

Table A1.4: Common physics options for all full parameter simulations. 58

Table A1.5: Common physics options for land use land cover 58
simulations.

X
| Flood Simulation Using Weather Forecast & Hydrological Models.

1. INTRODUCTION
Flood is a huge amount of water reaching land surface which is usually not covered with water
in a short period of time. Floods are responsible for loss of crops, damage to property and
even loss of human life. In India, an area of around 40 million hectare has been identified as
flood prone (Singh, 2005). India is having a large number of river systems, which experiences
seasonal floods every year. North and central India is subjected to floods frequently during
the months of monsoon i.e. July, August and September.

Floods can occur due to natural or man-made causes. In India, major cause of flood is due to
intense precipitation, inadequate capacity of riverbanks, siltation of river, landslides causing
obstruction in flow of river, change of river course, poor natural drainage, cyclones, cloud
bursts, glacial lake outbursts and dam break flows.

Losses occurring from floods can be minimized using various flood control structures or using
flood management. Flood management can be categorized into structural or non-structural
measures. Structural measures comprises of multipurpose reservoirs, retarding structures
which can store flood water, increasing the flood carrying capacity of the channels,
embankments and levees. These structures can protect a large area from floods, but these
structures requires a lot of funds and time to build.

It is also known that structural measures are not a permanent protection from floods of all
magnitudes. Also it is neither possible nor feasible to construct these structures due to various
factors like financial constraints, topographic limitation, social issues, etc. Real time flood
forecasting could be a non-structural measures which can be adopted for management of
floods. This process involves the estimation of future stages of vulnerable points along the
river course during floods. This helps in the preparation of evacuation plan in case of flood
emergency. The effectiveness of a flood forecasting system is dependent upon the accurate
prediction of future stages of the incoming flood (Singh, 2005).

The severity of flood depends upon the magnitude, nature and extent of rainfall and
characteristics of the watershed. For example, cloud burst in a hilly catchment can result in
flash flood for a shorter period of time whereas large catchment with heavy rainfall for a
longer duration can generate floods which can sustain for long durations. So, an accurate real
time flood forecast is required in order to provide sufficient time to evacuate people from the
areas which are likely to be affected. The lead time available for flash floods is very short
which makes it even more difficult to evacuate people.

The techniques available for real time flood forecasting can be classified as deterministic,
stochastic and statistical modelling and computational techniques like artificial neural
network (ANN), fuzzy logic, etc. (Singh, 2005)Real time flood forecasting can be divided
into two models atmospheric/weather prediction and hydrological modelling. Atmospheric
model provides the prediction of precipitation and calibrated hydrological model provides the
response of watershed during the predicted precipitation and future stages or flows of the
river.
-1-
| Flood Simulation Using Weather Forecast & Hydrological Models.

Numerical weather prediction is one of the most modern weather prediction technique. It
uses current weather information to predict future weather condition using ocean and
atmospheric mathematical models. This technique requires very high computation facilities
like supercomputers. Even with supercomputers the forecast skill is limited to six days.
Atmospheric and ocean models uses partial differential equation which cannot be solved
exactly and error grows with time. In addition to that model uses parameterizations for solar
radiation, moist process (clouds & precipitation), heat exchange, soil, vegetation, surface
water and effect of terrain.
The developments in remote sensing and geographical information system (GIS) techniques
provides a new platform for spatial visualization of information of natural resources. The
integration of derived formats provides tremendous potential for identification, monitoring
and assessment of floods. Distributed hydrological modelling can be used within a GIS
framework to simulate hydrologic processes from parameter maps derived from geospatial
data. Geospatial data can be used to represent spatial variability of watershed surface
properties that control hydrologic process. This can be the terrain, soil, land use/land cover,
precipitation, channel cross-sections, hydraulic properties and other derived parameters.

In this research project, hydrometeorological modelling is done with numerical weather


prediction and hydrological model. Weather prediction provides the forecasting of the
precipitation and a calibrated hydrological and hydrodynamic models provide the flood stage
or flow for the forecasted precipitation. Hydrological model coupled with atmospheric model
is also used. Upon the completion of this research project answers to following questions can
be given which can be achieved by following objectives.

1.1 Research questions


 Find out flood scenarios from weather forecasting and hydrological process?
 Which parameterization will give the better precipitation forecast for the study area?
 Find out method to integrate forecasted precipitation results and hydrologic process?
 Which hydrologic process is suitable for hydrologic modelling in the study area?

1.2 Objective
 Flood simulation using weather forecasting and hydrologic models.

1.2.1 Sub objectives


 Weather forecasting parameterization for better precipitation forecast in the study
area.
 Integration of weather forecasting and hydrological model using uncoupled and
fully coupled approaches.
 Selection of suitable hydrological process and sensitivity analysis of parameters
for better flood simulation scenarios.

-2-
| Flood Simulation Using Weather Forecast & Hydrological Models.

2. LITERATURE REVIEW
This chapter describes the various studies and some basic concepts related with numerical
weather prediction, weather research and forecasting model (WRF), hydrological modelling
and hydrodynamic modelling.

2.1 Numerical Weather Prediction (NWP)


Numerical weather prediction models are complex and one of the most computationally
expensive applications in the world. An NWP model is consists of an analysis model, which
translates irregularly spaced weather observations into a data format which is used by NWP
model. Secondly, numerical component contains the hydrodynamic equations which is
iterated after every time step. The horizontal and vertical resolutions of the domain must be
defined carefully (Table 2.1 ), for example for a thunderstorm of 20 km model resolution
should be less than 5 km. Model requires a boundary condition to run in a limited area which
is a subset of a global run of the model. Operational NWP models must be three dimensional.
Physical component of the model includes parameterization of physical processes which is
quite complex (Meyer, 1993).

Table 2.1: Scale definitions of some important atmospheric events. (Meyer, 1993)

Scale (km) Atmospheric Events


>10,000 km Planetary atmospheric waves
2,000 to 10,000 km Upper-atmosphere cyclone waves
Fronts & Tropical cyclones
200 to 2,000 km
Mesoscale convective waves
Low-level jets and ocean-land circulations;
downslope mountain windstorms,
20 to 200 km hazardous turbulence, small scale
atmospheric waves (including gravity
waves)

Atmospheric gravity waves caused by combination of gravity and buoyancy forces are most
restricting phenomenon of NWP models. Gravity waves have different speeds for different
phenomena. For example, planetary waves travel at speed of 10 m/s and gravity waves at 300
m/s. This behaviour of gravity waves causes noise in the model. (Meyer, 1993)

The WRF is an update to PSU/NCAR mesoscale model version 5 (MM5) with several
advantages like high resolution time patterns and annual averages. Precipitation is better
captured by WRF than MM5 and TRAMPER (Gsella et al., 2014).

Ferreira et al., (2014) suggests combination of physical parameterization especially


microphysics and cumulus parameterization is very important for the prediction of amount of
precipitation. For high horizontal resolution (<10km) cumulus parameterization schemes
have no effect on model simulations (Pennelly et al., 2014). Domain configuration is also
affects the spatial patterns of the precipitation.

-3-
| Flood Simulation Using Weather Forecast & Hydrological Models.

Simulations carried over east India using WRF-NMM model shows good predictions of
thunderstorms. The model captures rainfall intensity rates with good accuracy (Litta et al.,
2012). The WRF-ARW version model performance evaluated for heavy precipitation events
occurred over Ahmedabad, India, shows that GD and KF schemes underestimates the rainfall
occurred (Deb et al., 2008).

Kumar et al., (2010) found that WRF model is sensitive to the choice of convective scheme.
BMJ cumulus parameterization scheme produces better results than KF and GD for Indian
monsoon region. It can be used to for short range forecasting of precipitation and depression
tracks over Indian monsoon region. The planetary boundary layer option MYJ produces better
precipitation forecast than YSU scheme.

Ravindranath et al., (2010) implements land use land cover classification data from Advanced
Wide Field Sensor (AWiFS) from IRS P6 satellite in the WRF model. The data used over
Indian region is of 2007-2008. The study shows that major effects are over Jammu & Kashmir
region and all other regions have almost no change when compared with USGS land use/ land
cover.

Ochoa et al., (2014) evaluates TRMM 3B42 precipitation product and found that 3B42 V7
shows improvements over the other products of TRMM by reduction in bias. For hydrological
applications, TRMM and NWP estimates differs in precipitation regimes but both can be used
for hydrological modelling in ungauged catchments. The statistical interpretation of NWP
products can be done by accuracy, probability of detection and false alarm ratio (Maini et al.,
2002).

If A, B, C, D are the elements if 2x2 contingency table (Maini et al., 2002).

Forecast Rain Observed Rain


Yes No
Yes A B
No C D

The total number of cases will be M.

Accuracy (ACC) is given by measure of correct forecast.


𝑐𝑜𝑟𝑟𝑒𝑐𝑡 𝑓𝑜𝑟𝑒𝑐𝑎𝑠𝑡 (𝐴+𝐷)
𝐴𝐶𝐶 = = … (2.1)
𝑡𝑜𝑡𝑎𝑙 𝑓𝑜𝑟𝑒𝑐𝑎𝑠𝑡 𝑀

Probability of Detection (PODY) is the fraction of those events when rainfall occurred as
predicted. For perfect forecast value is 1 and for worst it is 0.
𝑐𝑜𝑟𝑟𝑒𝑐𝑡 𝑟𝑎𝑖𝑛 𝑓𝑜𝑟𝑒𝑐𝑎𝑠𝑡 𝐴
𝑃𝑂𝐷𝑌 = 𝑟𝑎𝑖𝑛 𝑜𝑏𝑠𝑒𝑟𝑣𝑎𝑡𝑖𝑜𝑛𝑠
= (𝐴+𝐶) … (2.2)

False Alarm Rate (FAR) is the proportion of forecast rain that fails to materialize. The best
possible is 0 and worst possible is 1.

-4-
| Flood Simulation Using Weather Forecast & Hydrological Models.

𝑓𝑎𝑙𝑠𝑒 𝑎𝑙𝑎𝑟𝑚𝑠 𝐵
𝐹𝐴𝑅 = 𝑟𝑎𝑖𝑛 𝑓𝑜𝑟𝑒𝑐𝑎𝑠𝑡𝑠 = (𝐴+𝐵) … (2.3)

Coupled (1-way) meteorological model MM5 and calibrated hydrological model WaSiM are
run for alpine and orographic complex catchment of the river Mangfall (State of Bavaria,
Germany). The results shown reasonable reproduction of observed runoff but there is a height
dependent bias and an under prediction of about 21% total year precipitation. This type of
modelling approach can be utilized for the prediction of rainfall in missing or poorly gauged
catchments (Kunstmann and Christiane, 2005).

Shih et al., (2014) integrated a hydro-meteorological system to generate flood hydrograph for
specific watersheds. An integrated system of rainfall-runoff and instantaneous data
assimilation from field observations shows that accurate rainfall prediction is the most
important factor in generating hydrological response of the watershed. A calibrated
hydrological model is not enough for flood prediction. Accurate spatial and temporal
distribution of precipitation is also very important. The meteorological model used is WRF in
coupled (1-way) with WASH123D hydrological model.

Yu et al., (1999) tested the ability of meteorological and distributed hydrological model to
simulate single storm events over the Susquehanna River Basin. Penn State – NCAR
Mesoscale Model (MM5) model is used to simulate storm patters and Hydrologic Model
System (HMS) is used for hydrological response. MM5 shows some spatial and temporal
mismatch with the observed precipitation fields. The simulated stream flow is lower than the
observed stream flow. HMS is tested with Curve Number and Green-Ampt methods as loss
methods. Green-Ampt method shows an under prediction while Curve Number matches well
with the observed stream flow.

Westrick and Mass (2001) simulated a complex rain on snow flood event using
meteorological model as MM5 and hydrological model as Distributed Hydrological-Soil-
Vegetation Model (DHSVM). Results shows that there is under prediction of precipitation
which ultimately results in under prediction of simulated flow with the observed. Sensitivity
simulations of MM5 is conducted by adjusting with the observations and 90% of the total
flow is captured.

In performing a medium flood forecast three different model components are to used, namely
weather forecasting, rainfall runoff and flood routing models. So there could be many
uncertainties linked with each and every model used. Atmospheric models are very limited
due to its high computational requirements and its spatial resolution. Rainfall runoff models
are mainly constrained with sub-grid parameterization of complex and heterogeneous local
flow pathways. Therefore, rainfall runoff model requires some calibration before to be used
with forecasting models. Hydraulic model requires data from high resolution digital elevation
model (DEM), channel width, slope and depth and flood plain roughness value (Pappenberger
et al., 2005).

-5-
| Flood Simulation Using Weather Forecast & Hydrological Models.

2.2 GIS based Hydrologic and Hydrodynamic Modelling


Hydrology, catchment and fluvial systems are closely related with time and space. Chow et
al. (1988) classified hydrological models based on randomness (deterministic and stochastic),
spatial variation (lumped and distributed; space dependent and space independent) and time
variation (steady flow and unsteady flow; time dependent and time independent). This
classification shows a potential role of GIS to handle the dimensions in hydrologic modelling
incudes,

1. Pre-processing of data into suitable form for analysis.


2. Direct support for modelling such as analysis, calibration and prediction can be
carried out by GIS.
3. Post-processing of data through tabulation, reformatting, report generation and
mapping.

These three stages are an overall representation of work that has been widely recognized.
The simulations or predictions is handled by a non-GIS hydrological model that is coupled
to the GIS data input and output either loosely (through the implementation of conversion
protocols for data transfer) or closely (where the GIS and model share a common data
structure and both interact with the same database) (Clark, 1998). These possible
relationships can be used in different applications like,

1. Hydrological assessment to represent hazard or vulnerability.


2. Hydrological parameter determination, like surface slope, land use/ land cover, soil
characteristics, channel length, etc.
3. Hydrological modelling within GIS.
4. Linking GIS and hydrological models to utilize the GIS as an input and real time
process monitoring.

The main constraint in the reliability of the hydrological applications of GIS is data quality.
Data quality is related with precision, accuracy and error. To successfully build a
hydrological model, precision and accuracy are prerequisites. It affects the outcome in the
fluvial system (Clark, 1998). The spatial resolution of for hydraulic modelling affects greatly.
Horritt and Bates (2001) investigates the raster based models from resolution of 1000 to 10
m. The maximum performance of the model is at a resolution of 100 m after which no
improvement is observed by increasing resolution. Projecting water levels into a high
resolution DEM improves more performance but high resolution data is very limited and it
renders some problems like,

1. Increased data volume.


2. Increased processing speed and storage.
3. Noise in adjacent pixel.

Paiva et al., (2011) have shown that large scale hydrologic and hydrodynamic modelling can
be employed using limited data for flood plains and river geometry. GIS algorithms can be
used to extract model parameters from relatively limited data such as SRTM DEM, ASTER

-6-
| Flood Simulation Using Weather Forecast & Hydrological Models.

DEM or CartoDEM. GIS based algorithm includes estimation of river width, depth and river
cross-sections.

Liu and Smedt (2005) have shown a spatially distributed hydrological model working on
hourly time scale. Elevation, soil and land use data are used to predict flood hydrograph and
spatial distribution of hydrological characteristics in a watershed. The model successfully
captures the normal and extreme flood events and is found suitable for hydrological processes
simulations in a complex terrain using GIS and remotely sensed data.

In distributed hydrological modelling approaches, comparison of simulated and observed


measurements is very important. Efficiency criteria are commonly used to provide an
assessment of closeness of the simulated behaviour to the observed measurements. There are
a large number of efficiency criteria to choose. Each criteria reflects different behaviour of
simulated to observed relationship. Nash-Sutcliffe efficiency and index of agreement are
frequently used by hydrologic modelling studies (Krause et al., 2005).

Nash-Sutcliffe Efficiency (E) is defined as one minus the sum of the absolute squared
differences between the predicted (P) and observed (O) values normalized by the variance of
the observed values during the period of investigation. It is calculated as:

∑𝑛
𝑖=1(𝑂𝑖 −𝑃𝑖 )
2
𝐸 =1− 𝑛
∑𝑖=1(𝑂𝑖 −𝑂̅)2
… (2.4)

The range of E lies in between -∞ to 1.0 (perfect fit). An efficiency lower than zero indicates
that the mean value of the observed time series would have been a better predictor than the
model (Krause et al. 2005).

Index of agreement (d) is represented by the ratio of the mean square error and the potential
error and is defined as:

∑𝑛𝑖=1(𝑂𝑖 −𝑃𝑖 )
2
𝑑 = 1 − ∑𝑛 ̅ ̅ 2
… (2.5)
𝑖=1(|𝑃𝑖 −𝑂|+|𝑂𝑖 −𝑂|)

The range of d lies in between 0 to 1 (perfect fit).

2.3 Digital Elevation Models


Digital Elevation Models (DEM) also called digital terrain models provide a 3 D
representation of the real-world topography. DEM creation requires data collection and
processing procedures. Data collection step depends on the areal extent and importance of
the study. They can be constructed by ordinary ground survey when the study area is
relatively small or of minor importance. On the other hand if the study area is large, satellites
can be used for mapping the topography. These maps then have to be processed by remote
sensing imagery to give topographic elevation information. The CartoDEM is generated by
Cartosat-1 (launched on May 5, 2005 by Indian Space Research Organization (ISRO)) data of
2.5 m.

-7-
| Flood Simulation Using Weather Forecast & Hydrological Models.

DEMs can be used in delineating watersheds, stream network and its related analysis. It is
useful in contour generation, quantitative analysis of runoff and soil erosion, volume area
calculations, design of hydraulic structures, flood profiling fly through visualization, etc. The
accuracy of the stream network obtained from DEMs is highly dependent on the resolution
of the DEM.

2.4 Flood
“A flood is an unusually high stage in river, normally the level at which the river overflows
its banks and inundates adjoining area.”(Subramanya, 2009). It can be due to dam break,
sudden discharge of dam into the river, glacier lake outburst, extreme precipitation, etc.

2.4.1 Extreme rainfall events in Alaknanda and Bhagirathi Valley.


The Himalayan region is vulnerable to torrential rainfall in the form of flash flood, glacier
lake outburst or cloud burst. Flash floods and cloud bursts are generally caused by high
intensity rainfall. A study conducted by Joshi and Kumar (2006) over the Himalayan region
indicates that extreme rainfall events are increasing. Every year at least four to five major
events are occurring in the region and causing damage to life and property.

Table 2.2: Major cloud bursts events in Uttarakhand in 2012 (Kumar et al. 2013)

Date State District Effected Villages

05/07/12 Uttarkhand Uttarkashi Assi ganga ghat, Charaghani, Andiyararkal,


Ravada

05/07/12 Uttarkhand Chamoli Beriya

04/08/12 Uttarkhand Uttarkashi Dayara bhugyal, Joshiyada, Gangori bridge

19/08/12 Uttarkhand Uttarkashi Nuranu village, Mori area

14/09/12 Uttarkhand Rudrapryag Timada, Sansari, Giriya, Chunniand, Premnagar,


Juatok villages in Ukhimath area

14/09/12 Uttarkhand Bageshwar Kapkot near Almorah

Some of the main examples of the flash flood in the region are in 1803, in which one third of
Srinagar town was swept away. In 1868, two bridges were swept away and killed 70 pilgrims
who were sleeping on the banks of Alaknanda River at Chamoli. In 1970 due to extreme
rainfall event lake busted out near Gauna village (Chamoli district), causing a sudden
discharge of water and rising the level of Alaknanda river by 50 m at Srinagar, this water
reached Haridwar rising the water level of Ganga river by four meters (Joshi and Kumar,
2006).

-8-
| Flood Simulation Using Weather Forecast & Hydrological Models.

2.4.2 Flood Routing


Also called stream flow routing and channel routing, is one of the classical problems in
applied hydrology. The word routing refers in general to the mathematical procedure of
tracking or following water movement from one place to another; as such, the word also
includes the description of the conversion of precipitation into various subsurface and
surface runoff phenomena. However, flood routing refers specifically to the description of
the behaviour of a flood wave as it moves along in a well-defined open channel (Brutsaert,
2005). For the routing of a flood wave, numerical methods that solve the complete
continuity and momentum equations may be used.

 Fully Dynamic: It is an HD Module which provides fully dynamic solution to the


complete nonlinear Saint Venant equations for an open channel flow.
 Kinematic Models: These models are based on the solution of the continuity
equation and the steady-uniform equation for the dynamic equation. The waves
propagated using these models are called kinematic waves, and routing is called
kinematic routing.
 Diffusion Routing: This is formulated based on the simplified versions of
momentum equation.
 Muskingum-Cunge Method: This method is actually a particular finite-difference
approximation of the kinematic wave equations and present expressions for the
Muskingum coefficients in terms of the physical properties of the channel. And
the coefficients for the method were determined from the observed flood records.

-9-
| Flood Simulation Using Weather Forecast & Hydrological Models.

4. STUDY AREA AND DATA USED


Upper Ganga Basin with outlet at Haridwar is bounded by geographical coordinates of 7800’
E to 80010’ E longitudes and 29050’ N to 31020’N latitudes. The outlet Haridwar is located at
78010’26’’ N longitude and 29057’14’’ E latitude. The drainage area of the watershed is
23,180 km sq. Upper Ganga Basin covers 7 districts of Uttarakhand i.e. Dehradun, Haridwar,
Chamoli, Rudraprayag, Tehri Garhwal, Uttarkashi and Pauri Garhwal. Major rivers are
Bhagirathi and Alaknanda which meet at Devprayag to form Ganga. The minimum elevation
of the watershed is 249 meters above mean sae level (MSL) and maximum elevation is 7626
meters above MSL. Study area with river system and important points is shown in the figure
3.1.

The Upper Ganga basin shows a wide variety of land use/ land cover. It includes evergreen
forests, snow & ice, build up land, crop land, fallow land, deciduous forests, grasslands, mixed
forest and plantation. Land use/ land cover map is shown in the figure 3.3. The land use/ land
cover map is provided by Indian Space Research Organisation Geosphere Biosphere
Programme (ISRO IGBP). The map scale is 1:250,000.

Mostly areas with elevation values more than 4000 are covered with snow and ice which make
up 22% of the study area. Evergreen, deciduous and mixed forests make 42%, 6% and 12%
of fallow & barren land respectively. Grassland, plantation, shrub land and waste land consists
of 2%, 1%, 5% and 2% respectively, while built up and crop land is limited to 1% and 7% of
the total area of the watershed.

Soil distribution in Upper Ganga Basin is very diverse. The soil map is shown in the figure
3.4. The soil map is provided by National Bureau of Soil Survey and Land Use Planning
(NBSS&LUP). The map scale is 1:250,000. It contains sandy, loamy, clay, glaciers, and rock
outcrops. Major soil groups are loamy and loamy skeletal with 50% and 20% respectively.
Other soil groups are sandy, clay, rock outcrops and glaciers with 7%, 1%, 17% and 5%
respectively of the total study area.

The Ganga drainage basin occupies an area of about 1,086,000 km2 of the Indian subcontinent.
It covers 26.2% of total area of India which measures to 862,769 km2 (Nih.ernet.in). Upper
Ganga basin is consists of two major watersheds of Alaknanda and Bhagirathi rivers. The
Alaknanda River basin covers an area of approximately 11.8 x103 km2 and the Bhagirathi
River basin covers approximately 7.8 x 103 km2. These two rivers constitute approximately
2% of the total drainage area of the entire Ganga River basin (cited as in Chakrapani & Saini,
2009). The Bhagirathi River originated from Gomukh (3920m) the snout of Gangotri glacier
whereas the Alaknanda River originates from the snout of Satopanth and the Bhagirathi
kharak glaciers, which rises from Chaukhamba Mountains (cited as in Chakrapani & Saini,
2009). At Devprayag, the Alaknanda and the Bhagirathi Rivers confluence to form the Ganga
River. Drainage Map of the Alaknanda and Bhagirathi Rivers is shown in the figure 3.1.

Cloud free satellite data of Landsat 8 dated 11 Jan 2015 is acquired from
http://earthexplorer.usgs.gov/. False colour composite (FCC) image of Landsat 8 is shown in
the figure 3.2. Landsat 8 satellite is having a repetivity of 16 days. Landsat 8 data is

- 10 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

orthorectified. It is having 2 instruments, Operational Land Imager (OLI) and Thermal


Infrared Sensor (TIRS). Spatial resolution of OLI multispectral bands 1-7, 9 is 30 m and OLI
panchromatic band 8 is 15 m. Data format used is GeoTIFF with cubic convolution
resampling and 16-bit pixel values. World Geodetic System (WGS) 84 is used as datum and
12 m circular error with 90% confidence global accuracy for OLI and 41 m circular error with
90% confidence global accuracy for TIRS. (Landsat.usgs.gov (1))

Table 3.1: Bands in Landsat 8.

Bands Wavelength Resolution


(micrometers) (meters)
Band 1 - Coastal aerosol 0.43 - 0.45 30
Band 2 - Blue 0.45 - 0.51 30
Band 3 - Green 0.53 - 0.59 30
Band 4 - Red 0.64 - 0.67 30
Band 5 - Near Infrared (NIR) 0.85 - 0.88 30
Band 6 - SWIR 1 1.57 - 1.65 30
Band 7 - SWIR 2 2.11 - 2.29 30
Band 8 - Panchromatic 0.50 - 0.68 15
Band 9 - Cirrus 1.36 - 1.38 30
Band 10 - Thermal Infrared (TIRS) 1 10.60 - 11.19 100 * (30)
Band 11 - Thermal Infrared (TIRS) 2 11.50 - 12.51 100 * (30)

* TIRS bands are acquired at 100 meter resolution, but are resampled to 30 meter in
delivered data product. (Landsat.usgs.gov (2))

CartoDEM is acquired from Bhuvan Open Data Archive Portal


(http://bhuvan.nrsc.gov.in/data/download/index.php). CartoDEM is generated using
Augmented Stereo Strip Triangulation (ASST). The data format is GeoTIFF, projection
system and datum is Geographic and WGS84 respectively. Planimetric accuracy is 15m and
vertical accuracy is 8m. It is generated by National Remote Sensing Center (NRSC), ISRO.
CartoDEM of the study area is shown in the figure 3.5.

For weather forecasting Global Forecasting System (GFS) outputs are acquired for dates 01
Aug 2012, 13 Sep 2012 and 14 Jun 2013 from National Center for Environmental Prediction
(NCEP) (http://www.ftp.ncep.noaa.gov/data/nccf/com/gfs/prod/). GFS is a weather forecast
model produced by NCEP. It is a coupled model, composed of four separate models,
atmospheric model, ocean model, land/soil model and a sea ice model. GFS data contains
geopotential height, temperature, relative humidity, U & V component of wind, cloud mixing
ratio, ozone mixing ratio and absolute vorticity for 26 vertical levels of atmosphere.

WRF terrestrial data is downloaded from


http://www2.mmm.ucar.edu/wrf/src/wps_files/geog_complete.tar.bz2. It contains albedo
from NCEP, clay fraction, green fraction, slope, leaf area index, lake depth, land use/ land
cover from USGS and MODIS, soil type, topography, etc. at resolutions of 10, 5, 2 arc minute
- 11 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

and 30 arc seconds. Land use/ land cover maps of the USGS, MODIS and AWiFS for WRF
domain are shown in the figure 3.6-3.8.

Tropical Rainfall Measuring Mission (TRMM) product 3B42 v7 is acquired for 1 Aug 2012
to 4 Aug 2012, 13 Sep 2012 to 17 Sep 2012, 13 June 2013 to 18 June 2013, and monsoon
season of year 2005, 2006 and 2007. 3B42 product is TRMM-adjusted merged infrared
precipitation and root mean square precipitation error estimates. Precipitation obtained from
product is in mm/hr. The temporal resolution of 3B42 is 3-Hour and horizontal resolution of
0.250. The data format used is HDF and NetCDF.

Figure 3.1: Map showing the study area with major rivers and important locations.

Figure 3.2: FCC of Landsat 8 for the study area.


- 12 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

Figure 3.3: Land Use/ Land Cover of the study area.

Figure 3.4: Soil map of the study area.


- 13 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

Figure 3.5: CartoDEM of the study area.

Figure 3.6: AWiFS land use/ land cover for domain 03 of WRF.

- 14 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

Figure 3.7: MODIS land use/ land cover for domain 03 of WRF.

Figure 3.8: USGS land use/land cover of domain 03 for WRF.

- 15 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

4. METHODOLOGY

4.1 Model Overview

4.1.1 WRF Model


The Weather Research and Forecasting (WRF) model is a numerical weather prediction
(NWP) and atmospheric simulation system. The ARW is suitable for use in a broad range of
applications across scales ranging from meters to thousands of kilometers, like real-time
NWP, hurricane research, regional climate research, etc (Skamarock, et. al., 2008).

Governing Equations:

The ARW dynamics solver integrates the compressible, nonhydrostatic Euler equations
(Skamarock, et. al., 2008).

Figure 4.1: ARW n coordinate as shown in (Skamarock, et. al., 2008).

The flux-form Euler equations can be written as

𝜕𝑡𝑈 + (𝛻 · 𝑉𝑢) − 𝜕𝑥(𝑝𝜑𝜂) + 𝜕𝜂(𝑝𝜑𝑥) = 𝐹𝑈 … (4.3)

𝜕𝑡𝑉 + (𝛻 · 𝑉𝑣) − 𝜕𝑦(𝑝𝜑𝜂) + 𝜕𝜂(𝑝𝜑𝑦) = 𝐹𝑉 … (4.4)

𝜕𝑡𝑊 + (𝛻 · 𝑉𝑤) − 𝑔(𝜕𝜂𝑝 − µ) = 𝐹𝑊 … (4.5)

𝜕𝑡𝛩 + (𝛻 · 𝑉𝜃) = 𝐹𝛩 … (4.6)

𝜕𝑡µ + (𝛻 · 𝑉) = 0 … (4.7)

𝜕𝑡𝜑 + µ − 1[(𝑉 · 𝛻𝜑) − 𝑔𝑊] = 0 … (4.8)

along with the diagnostic relation for the inverse density

𝜑 = −𝛼µ … (4.9)

- 16 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

and the equation of state


.…
𝑝 = 𝑝0(𝑅𝑑𝜃/𝑝0𝛼)𝛾 (4.10)

In (4.3)–(4.10), the subscripts x, y and η denote differentiation,

𝛻 · 𝑉𝑎 = 𝜕𝑥(𝑈𝑎) + 𝜕𝑦(𝑉𝑎) + 𝜕(Ω𝑎) … (4.11)

𝑉 · 𝛻𝑎 = 𝑈𝜕𝑥𝑎 + 𝑉𝜕𝑎, 𝑦𝑎 + Ω𝜕𝜂 … (4.12)

where a represents a generic variable. γ = c η=1.4 is the ratio of the heat capacities for dry air,
Rd is the gas constant for dry air, and p0 is a reference pressure (typically 10 Pascal). The
right-hand-side (RHS) terms FU, FV, FW, and FΘ represent forcing terms arising from model
physics, turbulent mixing, spherical projections, and the earth’s rotation. The prognostic
equations (4.3)–(4.8) are in conservative form except for (4.8) which is the material derivative
of the definition of the geopotential (Skamarock, et. al., 2008). The traditional σ coordinate
used in hydrostatic atmospheric models, η varies from a value of 1 at the surface to 0 at the
upper boundary of the model domain (Fig. 4.1).

Use of the WRF Preprocessing System by the ARW:

Figure 4.2: Flow diagram for WRF.

The WPS is a set of programs that takes terrestrial and meteorological data (typically in GriB
format) and transforms them for input to the ARW pre-processor program for real-data cases
(real). Figure 4.2 shows the flow of data into and out of the WPS system. Processes that took
place in WPS are:

 Physical grid definition (including the projection type, location on the globe, number
of grid points, nest locations, and grid distances).
 Interpolation of static fields to the prescribed domain.
 Reformatting of GriB data into an internal binary format.
 Interpolation of meteorological data.

The input to the ARW real-data contains:

- 17 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

 3-dimensional fields (including the surface) of temperature (K), relative humidity


(and the horizontal components of momentum (m/s, already rotated to the model
projection).
 The 2-dimensional static terrestrial fields include albedo, Coriolis parameters, terrain
elevation, vegetation/land-use type, land/water mask, map scale factors, map rotation
angle, soil texture category, vegetation greenness fraction, annual mean temperature,
and latitude/longitude.
 The 2-dimensional time-dependent include: surface pressure and sea-level pressure
(Pa), layers of soil temperature (K) and soil moisture (kg/kg, either total moisture, or
binned into total and liquid content), snow depth (m), skin temperature (K), sea
surface temperature (K), and a sea ice flag. (Skamarock, et. al., 2008)
Physics:

The WRF physics options fall into several categories, each containing several choices. The
physics categories are

 microphysics,
 cumulus parameterization,
 planetary boundary layer (PBL),
 land-surface model and
 radiation. (Skamarock, et. al., 2008)
Microphysics:

Microphysics includes explicitly resolved water vapor, cloud, and precipitation processes.
The model is general enough to accommodate any number of mass mixing-ratio variables,
and other quantities such as number concentrations. Microphysics options available in WRF-
ARW are Kessler, WSM3, WSM5, WSM6, Eta GCP, Thompson, Goddard, Morrison 2
Moment, WDM6, etc. (Skamarock, et. al., 2008)

Cumulus parameterization:

These schemes are responsible for the sub-grid-scale effects of convective and/or shallow
clouds. Cumulus parameterizations are theoretically only valid for coarser grid sizes, (e.g.,
greater than 10 km), where they are necessary to properly release latent heat on a realistic
time scale in the convective columns. Various options available in ARW are Kain-Fritsch,
Betts-Miller-Janjic, Grell-Freitas, Grell-3, etc. (Skamarock, et. al., 2008)

Land-Surface Model:

The land-surface models (LSMs) use atmospheric information from the surface layer scheme,
radiative forcing from the radiation scheme, and precipitation forcing from the microphysics
and convective schemes, together with internal information on the land’s state variables and
land-surface properties, to provide heat and moisture fluxes over land points and sea-ice
points. (Skamarock, et. al., 2008)

- 18 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

4.1.2 HEC HMS:


HEC HMS is the Hydrologic Modelling system computer program developed by the
Hydrologic Engineering Center (HEC).This program simulates precipitation-runoff and
routing process, both natural and controlled. HEC-HMS computes losses using various
methods like initial and constant rate loss model, deficit and constant –rate model, SCS curve
number loss model and Green and Ampt loss model. For runoff computations unit hydrograph
model and kinematic wave model (Feldman, 2000).

SCS curve Number Loss Model:

The Soil Conservation Service (SCS) Curve Number (CN) model estimates precipitation
excess as a function of cumulative precipitation, soil cover, land use and antecedent moisture,
using following equation:
(𝑃−𝐼 )2
𝑃𝑒 = 𝑃−𝐼 𝑎+𝑆 ... (4.13)
𝑎

Where 𝑃𝑒 = accumulated precipitation excess at time t; P = accumulated rainfall depth at time


t; 𝐼𝑎 = the initial abstraction (initial loss); and S = potential maximum retention, a measure of
the ability of a watershed to abstract and retain storm precipitation (Feldman, 2000).

SCS developed an empirical relationship of 𝐼𝑎 and S:

𝐼𝑎 = 0.2𝑆 ... (4.14)

Therefore, the cumulative excess at time t is:


(𝑃−0.2𝑆)2
𝑃𝑐 = 𝑃+0.8𝑆
... (4.15)

The maximum retention S, and watershed characteristics are related through an intermediate
parameter, the curve number (CN) as (Feldman, 2000):
1000−10𝐶𝑁
𝐶𝑁
(𝑓𝑜𝑜𝑡 − 𝑝𝑜𝑢𝑛𝑑 𝑠𝑦𝑠𝑡𝑒𝑚)
𝑆= { 25400−254𝐶𝑁 ... (4.16)
𝐶𝑁
(𝑆𝐼)

Estimating CN:

The CN for a watershed can be estimated by using tables published by SCS.

For a watershed that consists of several soil types and land uses, a composite CN is calculated
as:
∑ 𝐴𝑖 𝐶𝑁𝑖
𝐶𝑁𝑐𝑜𝑚𝑝𝑜𝑠𝑖𝑡𝑒 = ∑ 𝐴𝑖
... (4.17)

In which 𝐶𝑁𝑐𝑜𝑚𝑝𝑜𝑠𝑖𝑡𝑒 = the composite CN used for runoff volume computations with HEC-
HMS; i = an index of watersheds subdivisions of uniform land use and soil type; 𝐶𝑁𝑖 = the
CN for subdivision i; and 𝐴𝑖 = the drainage area of subdivision i. (Feldman, 2000)

Gridded SCS:
- 19 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

The grid based CN modelling option of HEC-HMS can be used. The description of each cell
in the database includes: the location of the cell, travel distance from the watershed outlet, the
cell size, and cell CN. HEC-HMS computes precipitation excess for each cell independently.
(Feldman, 2000)

SCS Unit Hydrograph (UH) Model:

This model is based upon averages of UH (fig.4.3)


derived from gaged rainfall and runoff for a large
number of small agricultural watersheds
throughout the US. (Feldman, 2000)

Basic Concepts and Equations:

SCS UH model is dimensionless, single peaked


UH. This dimensionless UH, expresses the UH

Figure 4.3: SCS unit hydrograph as


shown in (Feldman, 2000).

discharge,𝑈𝑡 , as a ratio to the UH peak discharge, 𝑈𝑝 , for any time t, a fraction of 𝑇𝑝 , the time
to UH peak. (Feldman, 2000)

UH peak and time of UH peak are related by


𝐴
𝑈𝑝 = 𝐶 𝑇 ... (4.18)
𝑝

In which A= watershed area; and C = conservation constant (2.08 in SI and 484 in foot-pound
system). The time of peak (also known as the time of rise) is related to the duration of the unit
of excess precipitation as:
Δ𝑡
𝑇𝑝 = 2
+ 𝑡𝑙𝑎𝑔 ... (4.19)

In which Δ𝑡 = the excess precipitation duration (which is also the computational interval in
HEC-HMS); and 𝑡𝑙𝑎𝑔 = the basin lag, defined as the time difference between the center of
mass of rainfall excess and the peak of the UH. (Feldman, 2000)

Estimating the SCS UH Model Parameters:

For ungagged watersheds, the SCS suggests that the UH lag time may be related to time of
concentration 𝑡𝑐 as:

𝑡𝑙𝑎𝑔 = 0.6𝑡𝑐 ... (4.20)

Time of concentration is a quasi-physically based parameter that can be estimated as:

𝑡𝑐 = 𝑡𝑠ℎ𝑒𝑒𝑡 + 𝑡𝑠ℎ𝑎𝑙𝑙𝑜𝑤 + 𝑡𝑐ℎ𝑎𝑛𝑛𝑒𝑙 ... (4.21)

- 20 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

Where 𝑡𝑠ℎ𝑒𝑒𝑡 = sum of travel time in sheet flow segments over the watershed land surface;
𝑡𝑠ℎ𝑎𝑙𝑙𝑜𝑤 = sum of travel time in shallow flow segments, down streets, in gutters, or in shallow
rills and rivulets; and 𝑡𝑐ℎ𝑎𝑛𝑛𝑒𝑙 = sum of time in channel segments. (Feldman, 2000)

Estimate velocity by Manning’s equation:


𝐶𝑅2/3 𝑆 1/2
𝑉= 𝑛
... (4.21)

Where V= average velocity; R= the hydraulic radius (defined as the ratio of channel cross-
section area to wetted perimeter); S= slope of the energy grade line (often approximated as
channel bed slope); and C = conversion constant (1.00 for SI and 1.49 for foot-pound system).
Values of n, which is commonly known as Manning’s roughness coefficient, can be estimated
from textbooks. (Feldman, 2000)

Channel travel time is computed as:


𝐿
𝑡𝑐ℎ𝑎𝑛𝑛𝑒𝑙 = 𝑉 ... (4.22)

Where L= channel length.

Sheet flow is flow over the watershed land surface, before water reaches a channel. Distances
are short (on the order of 10-100m). The sheet flow travel time can be estimated as:
0.007(𝑁𝐿)0.8
𝑡𝑠ℎ𝑒𝑒𝑡 = ... (4.23)
𝑃20.5 𝑆 0.4

In which N =an overland flow roughness coefficient; L= flow length; 𝑃2 =2 year, 24 hour
rainfall depth, in inches; and S= slope of hydraulic grade line, which may be approximated by
the land slope. Sheet flow usually turns to shallow concentrated flow after 100m. The average
velocity for shallow flow can be estimated as:

16.1345√𝑆, 𝑓𝑜𝑟 𝑢𝑛𝑝𝑎𝑣𝑒𝑑 𝑠𝑢𝑟𝑓𝑎𝑐𝑒


𝑉 = 𝑓(𝑥) = { ... (4.24)
20.3282√𝑆, 𝑓𝑜𝑟 𝑝𝑎𝑣𝑒𝑑 𝑠𝑢𝑟𝑓𝑎𝑐𝑒

From this, the travel time can be estimated. (Feldman, 2000)

Limitations of SCS CN: (Feldman, 2000)

 Predicted values not in accordance with classical unsaturated flow theory.


 Infiltration rate will approach zero during a storm of long duration, rather than
constant rate as expected.
 Developed with data from small agricultural watersheds in Midwestern US, so
applicability elsewhere in uncertain.
 Default initial abstraction (0.2S) does not depend upon storm characteristics or timing
 Rainfall intensity is not considered.

- 21 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

4.1.3 MIKE 11 Hydrodynamic Module:


For unsteady flows in rivers and estuaries MIKE 11 hydrodynamic module uses implicit finite
difference scheme. Depending upon the local flow conditions, module can adopt sub-critical
and super critical flow conditions with the help of numerical scheme. It is possible to include
flow in hydraulic structures as well as the structure operation mechanism. This computational
scheme can be used for vertically homogeneous flow in steep rivers and estuaries. (Water,
2010).

Bed Resistance:

MIKE 11 allows for two different types of bed resistance descriptions (Water, 2010):

1. Chezy
2. Manning

Chezy and Manning:

For the Chezy description, the bed resistance term in the momentum equation is described as:
𝑔𝑄|𝑄|
𝐶 2 𝐴𝑅
... (4.25)

Where, Q is discharge; A is flow area; R is the resistance or hydraulic radius.

For the Manning description, the term is:


𝑔𝑄|𝑄|
𝑀 2 𝐴𝑅4⁄3
... (4.26)

The Manning number, M, is inverse of conventional Manning’s n. The values of n ranges


from 0.01 to 0.1 similarly M ranges from 100 to 10 for smooth channel and thickly vegetated
channel respectively. (Water, 2010).

The Chezy coefficient is related to Manning's n by


1
1
𝑅6
𝐶= 𝑛
= 𝑀𝑅 6 ... (4.27)

C, M or n can be used through model calibration or Manning's n can be found in Chow et al.
(1988). Models can also be calibrated for different topographic characteristics.
Boundary Conditions:

Model branches not connected at a junction i.e. upstream and downstream branches are
required to have external boundary conditions. These boundary conditions can be consists of
the following relationships:

 Constant values of h or Q.
 Time varying values of h or Q.
 Rating curve (Water, 2010).

- 22 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

Discharge Computation:

Figure 4.4: Calculation of cross-sectional area for relative roughness not equal to 1.0. a)
Highest water level above maximum specified elevation b) Highest water level below maximum
specified elevation as shown in (Water, 2010).

Muskingum method is used for hydrologic routing which uses wedge and prism storages for
the calculation of discharge storage relationship. During the advancement of flood, when
inflow exceeds outflow, wedge storage is formed and when outflow exceeds inflow (during
recession), results in negative wedge. The Muskingum method is written:
𝑗+1 𝑗+1 𝑗 𝑗
𝑄𝑖+1 = 𝐶1 𝑄𝑖 + 𝐶2 𝑄𝑖 + 𝐶3 𝑄𝑖+1 + 𝐶4 ... (4.28)

Where indices i and j, respectively, refer to the considered grid point and time level. The
variables, 𝐶1 -𝐶4 are given by
∆𝑡−2𝐾𝑥
𝐶1 = 2𝐾(1−𝑥)+∆𝑡 ... (4.29)

∆𝑡+2𝐾𝑥
𝐶2 = ... (4.30)
2𝐾(1−𝑥)+∆𝑡

2𝐾(1−𝑥)−∆𝑡
𝐶3 = ... (4.31)
2𝐾(1−𝑥)+∆𝑡

𝑙𝑎𝑡 𝑄 ∆𝑡
𝐶4 = 2𝐾(1−𝑥)+∆𝑡 ... (4.32)

K and x are fixed with time and space. So equations (4.29-4.32) has to be evaluated only once.
The input parameter for Muskingum routing is K and x.

The Muskingum-Cunge method uses the same set of equations as Muskingum method but it
neglects local and convective acceleration. It approximates the diffusion of flood wave using
diffusion wave model. The parameters K and x are given by
∆𝑥
𝐾= 𝑐𝑘
... (4.33)

1 𝑄
𝑥 = 2 (1 − 𝐵𝑐 ) ... (4.34)
𝑘 𝑆0 ∆𝑥

𝑑𝑄 𝜕𝑄 𝜕𝑥 𝜕𝑄 𝜕𝑡
𝑐𝑘 = 𝑑𝐴 = 𝜕𝑥 𝜕𝐴
+ 𝜕𝑡 𝜕𝐴
... (4.35)

- 23 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

In Eqs. (4.33) - (4.35), ∆x is the length of the considered Kinematic Routing element and S0
is the bed slope. The variables B, A and Q, respectively, are the width, the cross sectional area
and the discharge, all of which represent the Kinematic Routing element in question. The
variables x and t are space and time variables, respectively (Water, 2010).

Routing:

Governing equations for routing are as follows:

Momentum equation (Water, 2010):

𝑆 = 𝐾𝑄𝑜𝑃 − 𝑇1 𝑄0 ... (4.36)

Continuity equation (Water, 2010):


𝑑𝑆
𝑑𝑡
= 𝑄𝑖 − 𝑄𝑜 ... (4.37)

Where, S is apparent basin storage; Qi is inflow at water way; Qo is runoff at water way; T1 is
time of delay; K, P are constants and T1 is given by (Water, 2010):

𝑄(𝑡) = 𝑄(𝑡 + 𝑇1 ) ... (4.38)

The continuity equation is discretized as:


𝑆𝑡+∆𝑡 −𝑆𝑡 𝑄𝑜,𝑡+∆𝑡 +𝑄𝑜,𝑡
= 𝑄𝑖 − ... (4.39)
∆𝑡 2

And the momentum equation is substituted giving


𝑃 𝑃 −𝑇 𝑄 )
𝐾𝑄𝑜,𝑡+∆𝑡 −𝑇1 𝑄𝑜,𝑡+∆𝑡 −(𝐾𝑄𝑜,𝑡 1 𝑜,𝑡 𝑄𝑜,𝑡+∆𝑡 +𝑄𝑜,𝑡
= 𝑄𝑖 − ... (4.40)
∆𝑡 2

Saint Venant Equations:

The ‘Saint Venant´ equations are based on the following assumptions (Water, 2010):

 The water is homogeneous and incompressible.


 The bottom-slope is small, thus the cosine of the angle it makes with the horizontal
may be taken as 1.
 The wave lengths are large compared to the water depth. This ensures that the flow
everywhere can be regarded as having a direction parallel to the bottom.
 The flow is subcritical.

The resulting equations are:


𝜕𝑄 𝜕𝐴
𝜕𝑥
+ 𝜕𝑡 = 𝑞 ... (4.41)

𝑄2
𝜕𝑄 𝜕(𝛼 ) 𝜕ℎ 𝑔𝑄|𝑄|
𝐴
+ + 𝑔𝐴 + =0 ... (4.42)
𝜕𝑡 𝜕𝑥 𝜕𝑥 𝐶 2 𝐴𝑅

- 24 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

Where, Q is discharge; A is flow area; q is lateral inflow; h is stage above datum; C is Chezy
resistance coefficient; R is hydraulic or resistance radius; α is momentum distribution
coefficient. (Water, 2010)

Solution Scheme:

Continuity and momentum equations are solved by implicit finite difference scheme. This
scheme is independent of kinematic, diffusive or dynamic wave description. (Water, 2010)

A computational grid of alternating Q (discharge) and h (water level) points is used as


illustrated in fig 4.5. Q-points are placed midway between neighbouring h-points and at
structures, while h- points are located at cross-sections, or at equidistant intervals in between
if the distance between cross-sections is greater than maximum dx. The discharge is defined
by convention as positive in the positive x-direction (increasing chainage). (Water, 2010)

Figure 4.5 Channel section with computational grid as shown in (Water, 2010)

4.1.4 WRF-Hydro
WRF-Hydro is hydrological coupling extension package of WRF. It includes fully distributed
3-D, surface and sub-surface flow model. It is a coupling architecture between weather and
climate models and hydrological models. It can be used either as uncoupled or fully coupled
(atmospheric model) land surface model (Gochis et al., 2013).

Figure 4.6: WRF Hydro architecture showing various components (Gochis et al., 2013).

WRF-Hydro only supports Noah Land surface model with following model physics options:

- 25 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

1. 1-D Land surface parameterization


2. Surface overland flow
3. Saturated subsurface flow
4. Channel routing
5. Reservoir routing
6. Conceptual baseflow

WRF-Hydro receives forcing depending upon the coupled and uncoupled mode. Figure 4.6
shows the basic execution steps which starts with execution of land surface model, grid
disaggregation, subsurface flow routing, overland flow routing, baseflow model, channel &
reservoir routing and grid aggregation (Gochis et al., 2013).

Figure 4.7: Calling structure of WRF-Hydro (Gochis et al., 2013).

WRF-Hydro supports spatially explicit, fully unsteady, diffusive wave with overland flow.
Overland flow is calculated when the depth of water exceeds a specified retention depth. The
diffusive wave equation considers backwater effect and flow on adverse slopes. Manning’s
equation is used for resistance. Overland flow roughness coefficients are obtained from USGS
24-type land cover product. Channel flow routing is done by 1-D diffusive wave through a
gridded channel network. Bucket model is used as baseflow to the stream network (Gochis et
al., 2013).

4.2 Overall Methodology


The methodology is divided into two parts:

 The WRF simulations for forecast.


 Hydrological modelling of WRF forecast.

- 26 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

Figure 4.8: Flowchart showing overall methodology.

4.2.1 WRF Methodology:


In this part, WRF is compiled for real simulations with GCC compilers in a system with 80
cores, 2TB hard disk, 64GB of ram, Cent OS operating system. After the successful
compilation of WRF following executable are created:

 Wrf.exe
 Ndown.exe
 Tc.exe
 Real.exe

WRF Pre-processor (WPS) is compiled which creates following executable:

 Ungrib.exe
 Metgrid.exe
 Geogrid.exe

WRF Domain Wizard is a Java based software for creating and configuring domains of WRF.
This can be initialized by two ways:

 Run runWRFPortal.bat.
 Open cmd in windows, goto WRF Portal folder and type “java -Xmx675m -jar
Portal.jar WRF”.

- 27 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

Figure 4.9: Domain Configuration of WRF.

3 nested domains are created with 27km, 9km and 3km spatial resolution (namelist.wps).
Geogrid.exe is run to get static data for WRF. After that ungrib.exe is run to get atmospheric
data and metgrid.exe is run to interpolate the atmospheric data.

Table 4.1: Combination for USGS & MODIS land use land cover parameterization
simulations.

Sim. No. Microphysics Cumulus


Parameterization

1 Lin et al. scheme Kain-Fritsch Scheme

2 Lin et al. scheme Betts-Miller-Janjic Scheme

3 Eta microphysics Kain-Fritsch Scheme

4 Eta microphysics Betts-Miller-Janjic Scheme

5 WRF Single Moment 6 class Scheme Kain-Fritsch Scheme

6 WRF Single Moment 6 class Scheme Betts-Miller-Janjic Scheme

Table 4.2: Combinations for full parameterization.

Sim. No. Microphysics Cumulus Parameterization

1 Lin et al. scheme Kain-Fritsch Scheme

- 28 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

2 WRF Single Moment 6 class Scheme Kain-Fritsch Scheme

3 WRF Double Moment 6 class Scheme Kain-Fritsch Scheme

4 Lin et al. scheme Grell 3D

5 WRF Single Moment 6 class Scheme Grell 3D

6 WRF Double Moment 6 class Scheme Grell 3D

 Ingestion of LULC in WRF-ARW:

1. USGS and MODIS LULC are standard LULC that comes with WRF-ARW package.
To use USGS in WRF-ARW no need to change in namelist.wps. For MODIS LULC just add
“modis_30s”, “modis_2m”, “modis_5m” or “modis_10m” for 30 arc sec, 2 min, 5 min or 10
min in the namelist.wps (Figure 4.10).

Figure 4.10: Geogrid description in namelist.wps for use of MODIS LULC.

2. AWiFS LULC for WRF can be downloaded from


http://bhuvan.nrsc.gov.in/data/download/index.php. The downloaded zip file is extracted and
files inside the folder are copied to geog/topo_30s (for 30 arc sec dataset) folder. This will
replace the existing USGS dataset.

 Ingestion of DEM:

Download SRTM DEM from http://earthexplorer.usgs.gov/ of the required domain. Open it


in ENVI 5.0. Export it in band interleaved by line (BIL) format. Rename it as “00000.ncols-
00000.nrows”. Create an index file as figure no. …. Create a folder named topo_srtm. Paste
the index file and BIL file in the folder. Update the GEOGRID.TBL as shown in (Figure
3.2.3)To use the SRTM DEM do changes in namelist.wps (Figure 4.11).

- 29 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

Figure 4.11: GEOGRID.TBL for SRTM DEM.

4.2.2 HEC-HMS Methodology:


HEC-HMS is used as a hydrological model. CARTOSAT digital elevation model (DEM) is
used for the pre-processing of the study area. Haridwar is given the outlet of the watershed.
The basic pre-processing includes following operations:

1. Fill
2. Flow Direction
3. Flow Accumulation
4. Stream Definition (200 km2)
5. Stream Segmentation
6. Catchment Grid Delineation
7. Catchment Polygon Processing
8. Drainage Line Processing
9. Adjoint Catchment Processing
10. Slope (in m/m)

Project is generated with Haridwar as an outlet. Following characteristic and parameters are
generated from HEC-GeoHMS.

1. River Length
2. River Slope
3. Basin Slope
4. Longest Flowpath
5. Basin Centroid
- 30 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

6. Centroid Elevation
7. Centroidal Longest Flowpath
8. HMS Process Selection
a. Subbasin – Loss Method: SCS
b. Subbasin – Transform Method: SCS
c. Subbasin – Base flow Method: Constant Monthly
d. River – Rout Method: Muskingum Cunge
9. River Auto Name
10. Basin Auto Name
11. Subbasin Parameters
a. 2 –year rainfall grid
b. Percent Impervious surface grid
c. CN grid
12. Muskingum Cunge Parameters
13. TR55 Flow path segments and parameters
14. CN lag

Export project to HEC-HMS model files.

Import model into HEC-HMS, insert the precipitation gauge values, check for other parameter
values and run the model.

Offline Coupling of WRF-ARW with HEC-HMS:

1. WRF-ARW outputs of domain 3 are exported from netcdf format to raster format
(GeoTIFF format). Using arcpy in python.
2. Zonal stats as table is done for the watersheds created by HEC-GeoHMS.
3. Generated *.dbf files are converted into DSS format using EXCEL macro and
plugin for DSS.
4. Gage file of HEC-HMS is modified according to DSS.
5. Check the model configuration for any inconsistency.
6. Run the model.

Impervious surface grid generation:

4 Download a tool named ISAT from http://coast.noaa.gov/digitalcoast/tools/isat


5 Provide the input watershed shape file and coefficient file for ISAT tool.
6 Run the tool.
7 Export the output to raster format.

- 31 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

A B

C D

Figure 4.12: A. CARTOSAT DEM of the area. B. CN Map generated from LULC & soil
Map. C. Impervious Map generated from ISAT-tool. D. Initial Abstraction Map generated
from LULC & soil map.

- 32 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

Figure 4.13: Watershed and River network delineated from HEC-GeoHMS.

Figure 4.14: HEC-HMS model setup.


- 33 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

4.2.3 MIKE 11 Methodology:

Figure 4.15: Flowchart showing MIKE 11 Methodology.

MIKE 11 hydrodynamic model is used for hydrodynamic modelling. Output from HEC-HMS
is used as an input to MIKE 11. MIKE 11 requires following inputs:

1. Network File
2. Boundary Condition File with Time Series File
3. Hydrodynamic (HD) parameter File
4. Simulation File

Offline Coupling of HEC-HMS with MIKE 11 HD model:

1. Outputs of HEC-HMS are copied into an Excel file according to the boundary
conditions in the MIKE 11 HD model.
2. A time series file (TS file) is created in the MIKE 11 and paste the date from excel
workbook to the time series file.
3. Add the TS file column to the boundary condition file.
4. Run the model.

- 34 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

Figure 4.16: Network File of MIKE 11.

Figure 4.17: Cross-section file of MIKE 11.

4.2.4 WRF-Hydro

Forcing data required for Noah LSM includes:

 Incoming shortwave radiation (W/m2)


 Incoming longwave radiation,

- 35 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

 Specific humidity (kg/kg),


 Air temperature (K),
 Surface pressure (Pa),
 Near surface wind in the u & v components (m/s),
 Liquid water precipitation rate (mm/s)

Using standalone ArcGIS toolset for WRF-Hydro pre-processing, following grids are
generated to be used with WRF-Hydro.

 Channel
 Flow direction
 Groundwater basins
 Lake
 Land Use
 Latitude
 Longitude
 Stream Order
 Roughness factor
 Topography

Figure 4.18: Channel grid for WRF-Hydro.

- 36 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

Figure 4.19: Flow direction grid for WRF-Hydro.

Figure 4.20: Land use/ Land Cover for WRF-Hydro.

- 37 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

Figure 4.21: Latitude grid for WRF-Hydro.

Figure 4.22: Longitude grid for WRF-Hydro.

- 38 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

Figure 4.23: Stream order grid for WRF-Hydro.

Figure 4.24: Topographic grid for WRF-Hydro.

- 39 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

5. RESULTS AND DISCUSSIONS:


WRF Model is evaluated using probability of detection, false alarm ratio and accuracy. The
event, weather rainfall will occur or not is taken into account for 17 June 2013, 03 Aug 2012
and 15 Sep 2012. This analysis is done with USGS and MODIS land use/ land cover (LULC)
dataset as well as six different microphysics and cumulus parameterization schemes. Almost
in every case MODIS LULC has shown better results than USGS LULC.

For microphysics and cumulus schemes used can be found in table 4.1 and table 4.2. Figure
5.3 shows the difference between the TRMM rainfall estimate and WRF predictions, the black
box shows the area of interest.

Figure 5.1: USGS LULC comparison with different microphysics and cumulus schemes for
17 June 2013.

Figure 5.2: MODIS LULC comparison with different microphysics and cumulus schemes
for 17 June 2013.

- 40 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

Figure 5.3: Difference image of TRMM and WRF results for USGS & MODIS LULC with
different microphysics and cumulus schemes for 13-17 June 2013.

With LULC as MODIS and USGS, microphysics and cumulus schemes are tested for
accuracy (ACC), probability of detection (PODY) and false alarm ratio (FAR). The maximum
PODY and ACC is for the combination of WRF Single Moment 6 class Scheme as
microphysics and Grell 3D as cumulus scheme with MODIS as land use/land cover. Since
USGS and AWiFS are similar and found almost no changes in the WRF prediction thus,
MODIS is selected as most suitable land use/ land cover.

- 41 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

Figure 5.4: WRF model comparison for different microphysics and cumulus schemes for 15
June 2013.

The difference between IMD grids and WRF predictions is shown in figure 5.5, the black box
shows the area of interest. Most of the portion of area of interest is under predicted by WRF
in all the schemes. Figure 5.4 shows that WRF Single Moment 6 class Scheme as
microphysics and Grell 3D as cumulus scheme shows maximum accuracy and probability of
detection for the simulation period of 13 June 2013 to 17 June 2013.

The difference between the TRMM and WRF forecast for the period of 1 August 2012 to 3
August 2012 is shown in the figure 5.6. Similar results are obtained, most of the portion is
under predicated by all the schemes used in the WRF. Figure 5.8 clears that WRF Double
Moment 6 class Scheme as microphysics and Grell 3D as cumulus scheme shows maximum
accuracy and probability of detection.

- 42 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

Figure 5.6: Difference images of TRMM and WRF for different microphysics and cumulus
schemes (1-3 Aug 2012).

Figure 5.7: Difference images of TRMM and WRF for different microphysics and cumulus
schemes (13-15 Sep 2012).

TRMM and WRF difference is shown in figure 5.7 for the simulation period of 13 September
2012 to 15 September 2012. There is under prediction in most of the areas. Figure 5.9 shows
that WRF model performs well with WRF Double Moment 6 class Scheme as microphysics
and Grell 3D as cumulus scheme.

In all the three periods, microphysics scheme WRF Double Moment 6 class scheme has shown
better accuracy and probability of detection in combination with Grell 3D as cumulus scheme.
The false alarm ratio for all the schemes is almost same in all the cases.

- 43 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

Figure 5.8: WRF model comparison for different microphysics and cumulus schemes for 03
August 2012.

Figure 5.9: WRF model comparison for different microphysics and cumulus schemes for 15
September 2012.

So, the best combination is MODIS LULC, WRF Double Moment 6 class scheme for
microphysics and Grell 3D for cumulus scheme. This combination result are used for HEC-
HMS model and MIKE11 for hydrodynamic modelling. The results of WRF are compared
with TRMM at Uttarkashi, Joshimath, Rudraprayag, Devprayag and Haridwar.

The HEC-HMS model is calibrated and validated for year 2005, 2006 and 2007 monsoon
season. The observed and simulated runoff at Uttarkashi and Joshimath are shown in the
figure 5.10-5.12. The model is found to have a coefficient of determination of 0.7, Nash-
Sutcliffe efficiency of -1.32 which shows that model is sensitive with high runoff values, and
index of agreement is 0.77.

The simulated runoff from TRMM data and WRF forecasted precipitation are compared and
is shown in the figure 5.13-5.17.

- 44 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

HEC-HMS is sensitive for curve number. Three different curve numbers are selected for each
sub-watershed. Curve number values are shown in table 5.1 and its response in shown in
figure 5.18.

Figure 5.10: Hydrograph comparison of simulated and observed runoff at Uttarkashi and
Joshimath for year 2005.

Figure 5.11: Hydrograph comparison of simulated and observed runoff at Uttarkashi and
Joshimath for year 2006.

- 45 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

Figure 5.12: Hydrograph comparison of simulated and observed runoff at Uttarkashi and
Joshimath for year 2007.

Figure 5.13: Hydrograph comparison of TRMM & WRF at Uttarkashi.

The hydrograph (figure 5.13) is showing a shift of 5 hours with TRMM at Uttarkashi station.
The peak discharge for TRMM is 3175 m3/s and for WRF is 3046 m3/s. WRF has captured
the rainfall pattern well for this station.

- 46 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

Figure 5.14: Hydrograph comparison of TRMM & WRF at Devprayag.

Figure 5.15: Hydrograph comparison of TRMM & WRF at Joshimath.

Figure 5.14 shows that predicted discharge is almost half of TRMM discharge at Devprayag
station. The cause of this drop is that WRF is unable to capture the full rainfall event in nearby
stations causing an under prediction of rainfall and ultimately affecting the runoff.

Figure 5.15 depicts that Uttarkashi station has been over predicted by WRF which results in
over estimation of discharge with respect to TRMM. Also there is time shift of 5 hours in the
peak of discharge with respect to TRMM.

- 47 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

Figure 5.16: Hydrograph comparison of TRMM & WRF at Rudraprayag.

Figure 5.17: Hydrograph comparison of TRMM & WRF at Haridwar.

Hydrograph comparison at Rudraprayag station (figure 5.16) shows clear under prediction by
WRF with respect to TRMM. The nature of storm event is successfully captured by the WRF.
There is time shift of 5 hours in the peak of WRF.

Figure 5.17 shows that runoff at outlet of study area i.e. Haridwar station is under predicted
by WRF. There is 50% less discharge predicted by WRF when compared with TRMM
estimates. Although nature of the storm is still well captured with a time shift of 5 hours.

HEC-HMS is sensitive to the curve number values. Three cases are selected for sensitivity
analysis (table 5.1). First case is having lower values than original, second case includes
original values and third case includes higher values than original. Figure 5.18 shows that
there is change in the peak of the hydrograph. First case shows, lowest peak of hydrograph
than other two cases, second case shows in between the first and third case and third case
shows the highest peak from the other two cases.

- 48 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

Figure 5.18: Hydrograph comparison for different Curve Number Values.

Table 5.1: Curve Number Values for sub-watershed.

Sub basin CN Case I CN Case II CN Case III


W1100 59.4596 64.63 69.8004
W1090 62.87924 68.347 73.81476
W1080 70.19416 76.298 82.40184
W1070 68.02664 73.942 79.85736
W1060 62.35208 67.774 73.19592
W1050 65.0026 70.655 76.3074
W1040 64.0504 69.62 75.1896
W1030 62.36036 67.783 73.20564
W1020 67.11032 72.946 78.78168
W1010 66.5298 72.315 78.1002
W1000 62.04112 67.436 72.83088
W990 69.86848 75.944 82.01952
W980 68.51608 74.474 80.43192
W970 57.2194 62.195 67.1706
W960 63.80752 69.356 74.90448
W950 66.5988 72.39 78.1812
W940 62.34932 67.771 73.19268
W930 55.39964 60.217 65.03436
W920 55.2598 60.065 64.8702
W910 58.28752 63.356 68.42448
W900 68.59704 74.562 80.52696
- 49 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

W890 55.47048 60.294 65.11752


W880 66.0008 71.74 77.4792
W870 65.0532 70.71 76.3668
W860 63.04116 68.523 74.00484
W850 58.99868 64.129 69.25932
W840 78.982 85.85 92.718
W830 65.94744 71.682 77.41656
W820 61.86264 67.242 72.62136
W810 61.82216 67.198 72.57384
W800 59.24984 64.402 69.55416
W790 68.0708 73.99 79.9092
W780 59.40808 64.574 69.73992
W770 69.62192 75.676 81.73008
W760 62.95836 68.433 73.90764
W750 71.35336 77.558 83.76264
W740 77.70228 84.459 91.21572
W730 81.65736 88.758 95.85864
W720 80.56808 87.574 94.57992
W710 64.12676 69.703 75.27924
W700 71.38648 77.594 83.80152
W690 65.35496 71.038 76.72104
W680 81.85332 88.971 96.08868
W670 57.67296 62.688 67.70304
W660 79.5984 86.52 93.4416
W650 81.39792 88.476 95.55408
W640 77.41984 84.152 90.88416
W630 67.28972 73.141 78.99228
W620 68.24744 74.182 80.11656
W610 78.62136 85.458 92.29464
W600 79.97192 86.926 93.88008
W590 82.80092 90.001 97.20108
W580 69.32108 75.349 81.37692
W570 80.99036 88.033 95.07564
W560 78.5588 85.39 92.2212

- 50 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

Figure 5.19: Stage comparison of WRF and TRMM at Uttarkashi and Devprayag stations.

MIKE 11 HD module is used for hydrodynamic modelling. Outputs from HEC-HMS are used
to drive the hydrodynamic model. Results of TRMM and WRF are compared and is shown in
figure 5.19. WRF under predicts the stages with respect to TRMM estimates. At Uttarkashi
the stage is well captured by WRF but with a shift of 5 hours. At Devprayag, the stage is under
predicted by WRF. There is small variation of stage in WRF at Devprayag.

WRF-Hydro is used as coupled model with WRF. The result obtained from WRF-Hydro is
shown in the figure 5.20. This shows that WRF-Hydro is unable to get the required discharge
at Haridwar. The discharge always keeps on increasing and it is unable to get to the peak as
derived from HEC-HMS.

Figure 5.20: Hydrograph at Haridwar for WRF-Hydro.

- 51 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

6. CONCLUSIONS AND RECOMMENDATION

6.1 Conclusions:
The main objective of this project is flood simulation using weather forecasting and
hydrological models which is completed using WRF-ARW as weather model and HEC-HMS,
MIKE 11 and WRF-Hydro as hydrological models. In order to complete this objective, three
sub-objectives are achieved.

First sub-objective includes the parameterization of weather forecasting model which is


completed by changing the land use/ land cover of AWiFS, MODIS and USGS, topography
to SRTM and six different cumulus and microphysics schemes combinations. Out of these
changes MODIS land use/ land cover with WRF Double Moment Class 6 Scheme as
microphysics scheme and Grell 3D cumulus scheme is selected as most suitable parameters
for weather forecasting. The results of WRF is further validated with IMD rainfall grids which
clearly shows under prediction in most of the areas. These results are used to drive
hydrological models which forms the second sub-objective.

Second sub-objective includes the selection of suitable hydrological processes and its
sensitivity analysis. The HEC-HMS is selected as hydrologic model and MIKE 11 as
hydrodynamic model. The processes selected for HEC-HMS are SCS Curve Number method
for loss estimation and Muskingum-Cunge as channel routing method. For baseflow, constant
monthly method is selected. Loss method SCS curve number is sensitive to the model. So, its
sensitivity analysis is done. The model is calibrated and validated with the observed discharge
at Uttarkashi and Joshimath.

Third sub-objective is the integration of weather forecasting results and hydrological model
using coupled and uncoupled approaches. For uncoupled approach HEC-HMS and MIKE 11
models are used. The precipitation results obtained from WRF are processed for HEC-HMS
and results from HEC-HMS are processed for MIKE 11 as a boundary condition. The results
of HEC-HMS and MIKE 11 are compared with TRMM, which clearly shows the effect of
under prediction in WRF precipitation forecast and time shift in the peak of hydrograph. For
coupled approach WRF-Hydro (version 2.0) model is used. The hydrological data required
by WRF-Hydro is processed and the results obtained are not sufficient.

This overall makes the use of weather forecasting and hydrological model for flood
simulations. Using WRF for forecasting precipitation and HEC-HMS for hydrologic
modelling and MIKE 11 for hydrodynamic modelling. WRF-Hydro is having high
computational requirements and the results are not sufficient.

6.2 Recommendations:
Recommendation for future work is the use of data assimilation techniques for WRF forecast.
Automatic coupling methods to be discovered for the integration of WRF and HEC-HMS.
WF-Hydro needs to be more explored.

- 52 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

7. REFERENCES

Brutsaert, W., “Hydrology An Introduction”. Cambridge, UK: Cambridge University Press


(2005).
Chakrapani, G. J., and R. K. Saini., "Temporal and spatial variations in water discharge and
sediment load in the Alaknanda and Bhagirathi Rivers in Himalaya, India." Journal
of Asian Earth Sciences 35, no. 6 (2009): 545-553.
Chow, Ven Te, David R. Maidment, and W. Larry., "Applied Hydrology." New York, New
York (1988).
Clark, Michael J., "Putting water in its place: A perspective on GIS in hydrology and water
management." Hydrological Processes 12, no. 6 (1998): 823-834.
Disc.sci.gsfc.nasa.gov,. 'TRMM Online Visualization And Analysis System (TOVAS) — GES
DISC - Goddard Earth Sciences Data And Information Services Center'.
http://disc.sci.gsfc.nasa.gov/precipitation/tovas/
DHI., “MIKE 11 - A Modelling System for Rivers and Channels”.Denmark: DHI (2009).
Deb, S. K., T. P. Srivastava, and C. M. Kishtawal., "The WRF model performance for the
simulation of heavy precipitating events over Ahmedabad during August 2006."
Journal of earth system science 117, no. 5 (2008): 589-602.
Ferreira, J. A., A. C. Carvalho, L. Carvalheiro, A. Rocha, and J. M. Castanheira., "On the
influence of physical parameterisations and domains configuration in the simulation
of an extreme precipitation event." Dynamics of Atmospheres and Oceans 68 (2014):
35-55.
Feldman, Arlen D., “Hydrologic Modeling System HEC-HMS: Technical Reference
Manual.” US Army Corps of Engineers, Hydrologic Engineering Center, 2000.
Gochis, D. J., W. Yu, and D. N. Yates., "The WRF-Hydro model technical description and
user’s guide, version 1.0." NCAR Tech. Doc (2013).
Gsella, A., A. de Meij, A. Kerschbaumer, E. Reimer, P. Thunis, and C. Cuvelier., "Evaluation
of MM5, WRF and TRAMPER meteorology over the complex terrain of the Po Valley,
Italy." Atmospheric Environment 89 (2014): 797-806.
Horritt, M. S., and P. D. Bates., "Effects of spatial resolution on a raster based model of flood
flow." Journal of Hydrology 253, no. 1 (2001): 239-249.
Joshi, Varun, and Kireet Kumar. "Extreme rainfall events and associated natural hazards in
Alaknanda valley, Indian Himalayan region." Journal of Mountain Science 3, no. 3
(2006): 228-236.
Kumar, R. A., Jimy Dudhia, and S. K. Bhowmik., "Evaluation of Physics options of the
Weather Research and Forecasting (WRF) Model to simulate high impact heavy
rainfall events over Indian Monsoon region." Geofizika 27, no. 2 (2010): 101-125.
Kumar, VV Govind, Kamal Jain, and Ajay Gairola. "A Study and Simulation of Cloudburst
event Over Uttarkashi Region using River Tool and Geomatic Techniques."

- 53 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

International Journal of Soft Computing and Engineering (IJSCE), Volume 3, Issue-


1 (2013) :2231-2307.
Kunstmann, Harald, and Christiane Stadler., "High resolution distributed atmospheric-
hydrological modelling for Alpine catchments." Journal of hydrology 314, no. 1
(2005): 105-124.
Krause, P., D. P. Boyle, and F. Bäse., "Comparison of different efficiency criteria for
hydrological model assessment." Advances in Geosciences 5 (2005): 89-97.
Landsat.usgs.gov (1),. 'Landsat 8'. http://landsat.usgs.gov/landsat8.php (accessed on
08/06/2015).
Landsat.usgs.gov (2),. 'What Are The Band Designations For The Landsat Satellites?'.
http://landsat.usgs.gov/band_designations_landsat_satellites.php (accessed on
08/06/2015).
Litta, A. J., U. C. Mohanty, Someshwar Das, and Sumam Mary Idicula., "Numerical
simulation of severe local storms over east India using WRF-NMM mesoscale
model." Atmospheric Research 116 (2012): 161-184.
Liu, Y. B., and F. De Smedt., "Flood modeling for complex terrain using GIS and remote
sensed information." Water Resources Management 19, no. 5 (2005): 605-624.
Maini, Parvinder, Ashok Kumar, S. V. Singh, and L. S. Rathore., "Statistical interpretation
of NWP products in India." Meteorological Applications 9, no. 1 (2002): 21-31.
Meyer, W. Dale., “The Basics of Weather Models.” No. AWS/FM-93/001. AIR WEATHER
SERVICE SCOTT AFB IL (1993).
Nih.ernet.in,. 'About The Ganga Basin.'
http://www.nih.ernet.in/rbis/basin%20maps/ganga_about.htm (accessed June 08,
2015).
Ochoa, A., L. Pineda, P. Crespo, and Patrick Willems., "Evaluation of TRMM 3B42
precipitation estimates and WRF retrospective precipitation simulation over the
Pacific–Andean region of Ecuador and Peru." Hydrology and Earth System Sciences
18, no. 8 (2014): 3179-3193.
Paiva, Rodrigo CD, Walter Collischonn, and Carlos EM Tucci., "Large scale hydrologic and
hydrodynamic modeling using limited data and a GIS based approach." Journal of
Hydrology 406, no. 3 (2011): 170-181.
Pappenberger, Florian, K. J. Beven, N. M. Hunter, P. D. Bates, B. T. Gouweleeuw, J. Thielen,
and A. P. J. De Roo., "Cascading model uncertainty from medium range weather
forecasts (10 days) through a rainfall-runoff model to flood inundation predictions
within the European Flood Forecasting System (EFFS)." Hydrology and Earth
System Sciences Discussions 9, no. 4 (2005): 381-393.

- 54 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

Pennelly, Clark, Gerhard Reuter, and Thomas Flesch., "Verification of the WRF model for
simulating heavy precipitation in Alberta." Atmospheric Research 135 (2014): 172-
192.
Ravindranath, M., R. Ashrit, A.K. Bohra. A report on “Experiment on utilization of AWiFS
LU/LC data in WRF mesoscale model”. NMRF/RR/1/2010, (2010).
Shih, Dong-Sin, Cheng-Hsin Chen, and Gour-Tsyh Yeh., "Improving our understanding of
flood forecasting using earlier hydro-meteorological intelligence." Journal of
Hydrology 512 (2014): 470-481.
Singh, R. D., "Real-time flood forecasting-Indian Experiences." InINTERNATIONAL G-
WADI MODELLING WORKSHOP, vol. 28. (2005).
Skamarock, William C., J. B. Klemp, J. Dudhia, D. O. Gill, D. M. Barker, M. G. Duda, X. Y.
Huang, W. Wang, and W. Wang. Powers., JG., “A description of the Advanced
Research WRF Version 3." NCAR Technical Note (2008).
Subramanya, K. “Engineering Hydrology”, 4e. Tata McGraw-Hill Education, (2009).
Wang, Wei, Cindy Bruyère, Michael Duda, Jimy Dudhia, Dave Gill, and Hui-chuan Lin.
"ARW Version 3 Modelling System User’s Guide." (2009).
Water, D. H. I. "Environment." MIKE 11 Technical Reference (2010).
Westrick, Kenneth J., and Clifford F. Mass., "An evaluation of a high-resolution
hydrometeorological modeling system for prediction of a cool-season flood event in
a coastal mountainous watershed." Journal of Hydrometeorology 2, no. 2 (2001):
161-180.
Yu, Z., M. N. Lakhtakia, B. Yarnal, R. A. White, D. A. Miller, B. Frakes, E. J. Barron, C.
Duffy, and F. W. Schwartz., "Simulating the river-basin response to atmospheric
forcing by linking a mesoscale meteorological model and hydrologic model
system." Journal of Hydrology 218, no. 1 (1999): 72-91.

- 55 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

APPENDIX 1
This portion of thesis describes about the various microphysics and cumulus schemes of
WRF-ARW. The details are taken from ARW technical manual.

 Kessler scheme
This scheme is a simple warm cloud scheme that includes water vapour, cloud water, and
rain. The microphysical processes included are: the production, fall, and evaporation of rain;
the accretion and auto conversion of cloud water; and the production of cloud water from
condensation.

 WSM5 scheme
This scheme take vapor, rain, snow, cloud ice, and cloud water in five different arrays. Thus,
it allows supercooled water to exist, and a gradual melting of snow falling below the melting
layer. This scheme is efficient in intermediate grids between the mesoscale and cloud-
resolving grids.

 WSM6 scheme
The six-class scheme extends the WSM5 scheme to include graupel and its associated
processes. A new method for representing mixed-phase particle fall speeds for the snow and
graupel particles by assigning a single fallspeed to both that is weighted by the mixing ratios,
and applying that fallspeed to both sedimentation and accretion processes.. The behaviour of
the WSM3, WSM5, and WSM6 schemes differ little for coarser mesoscale grids, but they
work much differently on cloud-resolving grids. Of the three WSM schemes, the WSM6
scheme is the most suitable for cloud-resolving grids, considering the efficiency and
theoretical backgrounds.

 Kain-Fritsch scheme
The modified version of the Kain-Fritsch scheme has been modified based on testing within
the Eta model. As with the original KF scheme, it utilizes a simple cloud model with moist
updrafts and downdrafts, including the effects of detrainment, entrainment, and relatively
simple microphysics. It differs from the original KF scheme in the following ways:
1. A minimum entrainment rate is imposed to suppress widespread convection in
marginally
2. unstable, relatively dry environments.
3. Shallow (non precipitating) convection is allowed for any updraft that does not reach
minimum cloud depth for precipitating clouds; this minimum depth varies as a
function of cloud-base temperature.
4. The entrainment rate is allowed to vary as a function of low-level convergence.
5. Downdraft changes:
a. Source layer is the entire 150 – 200 mb deep layer just above cloud base.
b. Mass flux is specified as a fraction of updraft mass flux at cloud base. Fraction
is a function of source layer RH rather than wind shear or other parameters,
i.e., old precipitation efficiency relationship not used.
c. Detrainment is specified to occur in updraft source layer and below.

 Grell-3 scheme
The Grell-3 scheme was first introduced in Version 3.0, and so is new, and not yet well tested
in many situations. It shares a lot in common with the Grell-Devenyi in scheme, being based
on an ensemble mean approach, but the quasi-equilibrium approach is no longer included
- 56 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

among the ensemble members. The scheme is distinguished from other cumulus schemes by
allowing subsidence effects to be spread to neighbouring grid columns, making the method
more suitable to grid sizes less than 10 km, while it can also be used at larger grid sizes where
subsidence occurs within the same grid column as the updraft.

Table A1.1: Geogrid configuration for domain.

parent_id 1,1,2,
parent_grid_ratio 1,3,3,
i_parent_start 1,18,70,
j_parent_start 1,29,23,
e_we 100,124,127,
e_sn 95,133,136,
geog_data_res 'modis_30s+2m','modis_30s+2m','modis_30s+30s',
dx 27000,
dy 27000,
map_proj 'mercator',
ref_lat 31.339,
ref_lon 79.53,
truelat1 31.339,
truelat2 0,
stand_lon 79.53,
ref_x 50.0,
ref_y 47.5,

Table A1.2: Time information

wrf_core 'ARW',
start_date 2013-06-14_00:00:00
end_date 2013-06-17_00:00:00
interval_seconds 10800

Table A1.3: Domain configuration for namelist.input

time_step 120 dy 27000 9000 3000


time_step_fract_num 0 grid_id 1 2 3
time_step_fract_den 1 parent_id 1 1 2
max_dom 3 i_parent_start 1 18 70
e_we 100 124 127 j_parent_start 1 29 23
e_sn 95 133 136 parent_grid_rat 1 3 3
io
e_vert 27 27 27 parent_time_st 1 3 3
ep_ratio
- 57 -
| Flood Simulation Using Weather Forecast & Hydrological Models.

p_top_requested 5000 feedback 1


num_metgrid_levels 27 smooth_option 1

Physics Options in WRF:

Table A1.4: Common physics options for all full parameter simulations.

ra_lw_physics 1 1 1 ifsnow 0
ra_sw_physics 1 1 1 icloud 1
radt 27 9 3 surface_input_source 1
sf_sfclay_physics 1 1 1 num_soil_layers 4
sf_surface_physics 2 2 2 num_land_cat 20
bl_pbl_physics 1 1 1 sf_urban_physics 1 1 1
bldt 0 0 0 maxiens 1
cudt 0 0 0 maxens 3
isfflx 1 maxens2 3
ifsnow 0 maxens3 16
ensdim 144

Table A1.5: Common physics options for land use land cover simulations.

ra_lw_physics 1 1 1 icloud 1
ra_sw_physics 1 1 1 surface_input_sourc 1
e
radt 30 30 30 num_soil_layers 4
sf_sfclay_physics 1 1 1 num_land_cat 20/24
sf_surface_physics 2 2 2 sf_urban_physics 0 0 0
bl_pbl_physics 1 1 1 maxiens 1
bldt 0 0 0 maxens 3
cudt 0 0 0 maxens2 3
isfflx 1 maxens3 16
ifsnow 0 ensdim 144

- 58 -

You might also like