BY

ABHIJIT DAS
BCE(EVENING)-3rd YEAR
ROLL NO: 000810403008

UNDER THE GUIDANCE OF

PROF. KALYAN KUMAR MANDAL

JADAVPUR UNIVERSITY
CIVIL ENGINEERING DEPARTMENT
KOLKATA – 700032
2010-2011
 1
DCPL
[Type the company
address]
[Type the phone number]
[Type the fax number]
[Pick the date]
DCPL

[Year]
STUDY ON FLOOD FREQUENCY ANALYSIS

BY
ADITYA SHANKAR BASU
PTCE – 3rd YEAR
ROLL NO: 000810403019

UNDER THE GUIDENCE OF

PROF. SUDIPTA GHOSH

JADAVPUR UNIVERSITY
CIVIL ENGINEERING DEPARTMENT
KOLKATA – 700 032
2010-2011
 DECLARATION

This seminar paper entitled “STUDY ON FLOOD FREQUENCY ANALYSIS is prepared and
submitted for the partial fulfillment of the continuous assessment of “Bachelor of Civil
Engineering” course of Jadavpur University for the session 2008-2009.

ADITYA SHANKAR BASU

PTCE – 3rd Year
Roll: 000810403019

Date:
Place: Department of Civil Engineering
Jadavpur University
Kolkata – 700 032

1
 ACKNOWLEDGEMENT

I take this opportunity to express my sincere gratitude to Prof. Sudipta ghosh, Professor,
Department of Civil Engineering, Jadavpur University, who has been the guiding spirit behind the
preparation of this seminar report. The report would have never been completed without his able
guidance, constant vigil, careful supervision and inspiration at all stages of the work.

I would also like to express my gratitude to other respected teachers of the department for the
continuous encouragement, support and valuable advices they provided throughout my work.

I would like to thank all the librarians, assistant librarians and library staffs of the Central Library
and the Departmental Library of Jadavpur University for their kind cooperation during preparation
of this thesis by providing me the books and documents as and when necessary.

Last, but not the least, I express my gratitude to my classmates, my friends and my family for the
encouragement and support they provided throughout the work. This thesis would not have been
possible without them.

ADITYA SHANKAR BASU

Roll No: 000810403019

Date :

Place : Department of Civil Engineering

Jadavpur University
Kolkata – 700 032

2
 PREFACE

Flood is phenomena which is haunting the human civilization since time immemorial.

The happening of this incident has destroyed the human civilization, human life, crop, cattles etc.
no of times. As such the rainy season happens to be a terror time through the world in almost all
the countries.Earth quake, Landslide also causes flood on different occations.

In our country INDIA repeated flood in the river Sindhu has destroyed the Mohenjodaro & Horoppa
civilization around 5000 years back. Repeated flood in the river Ganga, Jamuna & Damodar & other
riverlets has caused enormous destruction time & again.In China The river “WHOANGHO” and Niles
In EGYPT have also inundated the civilization on both the banks causing immence lose of life &
property.

In this episode I”ADITYA SHANKAR BASU” shall try to focus some lights on the frequency of
floods & shall evaluate the after effect thereon.

ADITYA SHANKAR BASU

Roll No: 000810403019

Date :

Place : Department of Civil Engineering

Jadavpur University
Kolkata: 700032

CONTENTS
CHAPTER SECTION TOPIC PAGE NO
Acknowledgement 2

3
 Preface 3

Content 4-5

Chapter 1 Introduction
1.0 Introduction 6
1.1 DISCUSSION OF METHODS 7-8
1.2 Analysis Technique: Flood frequency Analysis 8
1.3 Comparative Evaluation Of Four Regional Flood 8
Frequency Methods

Chapter 2 Review of Litarature

1.0 Evaluation Of Design Flood Frequency Methods For 9

IOWA Streams
2.0 Flood Frequency Analysis For U.S Geological Survey 9
Gauging Station.
3.0 Flood Frequency Analysis of The Waimakaririr River 10
4.0 Report on Six Case Studies of Flood Frequency 10
Analysis
5.0 Regional Flood Frequency Analysis For South Africa 10

CHAPTER SECTION TOPIC PAGE NO

Chapter 3 Case Studies 11-23

1.0 Evaluation Of Design Flood Frequency Methods For 11-12

IOWA Streams
2.0 Flood Frequency Analysis For U.S Geological Survey 13-15
Gauging Station
3.0 Flood Frequency Analysis of The Waimakaririr River 15-17

4.0 Report on Six Case Studies of Flood Frequency 18-21

Analysis
5.0 Regional Flood Frequency Analysis For Southern 22-23
Africa

Chapter 4 Conclusion
4.0 Conclusion 24

4
Introduction Chapter 1

1.0 Introduction:

The objective of this study is to develop suitable flood estimation procedures for homogeneous regions
in a country to improve the design and economic appraisal of civil engineering structures.

The components that help to achieve the objective are as follows:

To delineate a region into hydrologically homogeneous regions,

To identify the underlying statistical distributions of flood flows,

To select suitable flood frequency analysis procedures,

To derive regional frequency curves for the delineated homogeneous regions,

To establish a relationship between mean annual flood and catchment

Characteristics for the gauged catchments.

5
The methodology used to carry out regional flood frequency analysis in a region consisted of the
following steps: data screening, delineation of homogeneous regions, selection of appropriate frequency
estimation procedures. All data that are available for the study were screened for inconsistency,
dependence and intersite correlation. The task to delineate a region into homogeneous regions involved
grouping together catchments that are considered to have similar hydrological, climatic and
physiographic characteristics.

Once a homogeneous region was identified the next step was to select suitable flood estimation
procedures for the delineated homogeneous regions the assessment of the performance of different
procedures was carried out by applying the predictive ability tests. The predictive ability tests
considered were the bias, the standard error of estimate, the root mean square error and the expected
probability of exceedance. The most robust flood estimation procedures for the regions were selected
from the results of the predictive ability.

Flood frequency analyses are used to predict design floods for sites along a river. The technique
involves using observed annual peak flow discharge data to calculate statistical information such as
mean values, standard deviations, skewness, and recurrence intervals. These statistical data are then used
to construct frequency distributions, which are graphs and tables that tell the likelihood of various
discharges as a function of recurrence interval or exceedence probability. Flood frequency distributions
can take on many forms according to the equations used to carry out the statistical analyses.

Alberta Transportation, Civil Projects Branch (formerly part of Alberta Public Works, Supply and
Services) has identified a problem of consistency regarding flood frequency analyses prepared for water
management projects. With a view to better defining the problem, this document briefly assesses six
cases where inconsistent flood frequency estimates have arisen. The assessments of each case are based
on selected documents that were readily available, and do not necessarily include a review of all past
studies.

 The magnitude of an extreme event is inversely related to its frequency of occurrence, very
severe events occurring less frequently than more moderate events.
 The objective of frequency analysis is to relate the magnitude of extreme events to their
frequency of occurrence through the use of probability distributions.
 Frequency analysis is defined as the investigation of population sample data to estimate
recurrence or probabilities of magnitudes.
 It is one of the earliest and most frequent uses of statistics in hydrology and natural sciences.
 Early applications of frequency analysis were largely in the area of flood flow estimation.
 Today nearly every phase of hydrology and natural sciences is subjected to frequency analyses.

1.1 DISCUSSION OF METHODS:

 Two methods of frequency analysis are described: one is a straightforward plotting technique to
obtain the cumulative distribution and the other uses the frequency factors.
 The cumulative distribution function provides a rapid means of determining the probability of an
event equal to or less than some specified quantity. The inverse is used to obtain the recurrence
intervals.
 The analytical frequency analysis is a simplified technique based on frequency factors depending on
the distributional assumption that is made and of the mean, variance and for some distributions the
coefficient of skew of the data.

6
 1.2 Analysis Techniques: Flood Frequency Analysis
Flood frequency analyses are used to predict design floods for sites along a river. The technique
involves using observed annual peak flow discharge data to calculate statistical information such as
mean values, standard deviations, skewness, and recurrence intervals. These statistical data are then used
to construct frequency distributions, which are graphs and tables that tell the likelihood of various
discharges as a function of recurrence interval or exceedence probability.

Flood frequency distributions can take on many forms according to the equations used to carry out the
statistical analyses. Four of the common forms are:
• Normal Distribution
• Log-Normal Distribution
• Gumbel Distribution
• Log-Pearson Type III Distribution

Each distribution can be used to predict design floods; however, there are advantages and disadvantages
of each technique. According to the U.S. Water Advisory Committee on Water Data (1982), the Log-
Pearson Type III Distribution is the recommended technique for flood frequency analysis.

Normal Distribution
The probability that X is less than or equal to x when X can be evaluated from
x
prob ( X ≤ x ) = p x ( x ) = ∫ (2πσ2 ) −1 / 2 e −( t −µ) / 2σ 2
2
dt
−∞

σ2
The parameters µ (mean) and (variance) are denoted as location and scale parameters,
respectively. The normal distribution is a bell-shaped, continuous and symmetrical distribution (the
coefficient of skew is zero).
GUMBEL DISTRIBUTION:

The extreme value distribution was introduced by gumbel (1941)and is commonly known as gumbel’s
distribution. It is the most widely used probability distribution function for prediction of flood peaks,
maximum rainfall, aximim wind speed etc. Gumbel defined a flood as the largest of the 365 daily flows
and the annual series of flood flow constitute a series of largest values of flows.

xT = x + K T s

 where, K is the frequency factor s is the standard

deviation and x bar is the mean value.

Log-Pearson Type III Distribution:

The Log-Pearson Type III distribution is a statistical technique for fitting frequency distribution data to
predict the design flood for a river at some site. Once the statistical information is calculated for the
river site, a frequency distribution can be constructed. The probabilities of floods of various sizes can be
extracted from the curve. The advantage of this particular technique is that extrapolation can be made of
the values for events with return periods well beyond the observed flood events. This technique is the
standard technique used by Federal Agencies in the United States.

The Log-Pearson Type III distribution is calculated using the general equation:
7
 Logx = logx + Kό(bgx)

where x is the flood discharge value of some specified probability, logx is the average of the logx
discharge values, K is a frequency factor, and ό is the standard deviation of the log x values. The
frequency factor K is a function of the skewness coefficient.

1.3 COMPARATIVE EVALUATION OF FOUR REGIONAL FLOOD-FREQUENCY

ANALYSIS METHODS :

Four popular methods for analyzing regional flood frequency were investigated using Louisiana stream
flow series. The state was divided into four homogeneous regions and all undistorted, long-term stream
gauges were used in the analysis. The generalized extreme value (GEV), two-component extreme value,
and regional log Pearson Type III methods were applied to this data base and compared in terms of
descriptive capabilities. On the basis of several factors, the GEV method was selected as the overall
superior method. The GEV parameters were estimated using the probability-weighted moments
(PWMs). Indexing was accomplished using the first PWM (the mean). A procedure to apply this method
to ungauged watersheds using regression equations and a regional non-dimensional flood distribution
was developed. It was found that the procedure performed well when applied to data not used in the
calibration of the model. The regional GEV procedure was compared with the method of the U.S.
Geological Survey (USGS) and showed significant improvement over the USGS equations in terms of
fit to the observed data. This method is easier to apply and more accurate in terms of descriptive and
probably predictive ability than other feasible methods for Louisiana data.

REVIEW OF LITERATURE Chapter

2

REVIEW OF LITERATURE :

1. Bradley Jr, A. A., Wehmeyer L. L. and Chen L.-C. (2009). -

EVALUATION OF DESIGN FLOOD FREQUENCY METHODS FOR IOWA
STREAMS .

Flood frequency analysis for many engineering design problem in IOWA were Needed. Three design
methods were evaluated for estimation of flood frequency. For small IOWA streams. Flood frequency
estimates based on NRCS curves were applied for urban land uses & other approach. However several
methods for improving flood frequency. Estimates were explored by adjusting the runoff co-efficient for
rational method based on nrcs curve.

It was also suggested that time of concentration estimates based on SUDAS procedure would be too
long. However future work should examine the recommendation of SUDAS in designing flood
frequency estimation.

8
2. M. McCarth, Peter., K. Sando Steven (2009). - Flood-Frequency Analyses
for U.S. Geological Survey Gaging Stations Based on Data through
Water Year 2009

The purpose of this research was to determine flood-frequency estimates for USGS gaging stations in
Montana. Flood-frequency estimates were determined by fitting a log-Pearson Type III probability
distribution to the recorded annual peak flows . Research activities in the proposed project were used for
thr improvement of flood-frequency estimates for gaging stations and used to publish the flood-
frequency estimates along with detailed documentation of the specific methods which were used to
develop the frequency estimates.

Objectives

The primary objectives of the proposed project include:

1) Determine flood-frequency estimates (recurrence intervals of 1.5, 2, 2.33, 5, 10, 25, 50, 100, 200,
and 500 years) for more than 660 USGS gaging stations in Montana based on data through water year
2009. These flood-frequency estimates are routinely used by MDT for design purposes.

2) Investigate methods of adjusting flood-frequency estimates for short-term gaging stations with data
restricted to either unusually wet or dry periods (including 1985-present).

3) Investigate the effects of climatic variability on our understanding of flood hydrology in Montana
and potential emerging methods for determining more accurate flood-frequency estimates.

4) Investigate the use of a regional approach for flood-frequency analysis in northweastern Montana,
where unusually large (but infrequent) floods complicate flood-frequency estimate

3. Ware, Robert and Lad, Frank (2003). - Flood Frequency Analysis of

the
Waimakariri River. Journal of Hydrology (NZ),

Flood frequency analysis with particular emphasis on estimating the extreme hydrological
events of group of sites were investigated. The accuracy & prediction were compared. The digital
forecasting procedure were applied to collect data from Waimakariri River in Cantarbary Newzeland
& the appropriateness of frequentist & digital method were scrutinized. It was found that the distribution
computed via the digital method provided much more uniform forecast than the frequentist forecast
method.

4. Neill, C.R. and Watt, W.E. (2001). - Report on Six Case Studies of
Flood Frequency Analysis Alberta Transportation (U.K),

9
Flood frequency analysis results greatly affected the statistical data & flood records available for a
particular period of year. It was a good practice to use the flood record for design if it exceeded the
estimate for the designated return period . Unrecorded period included one or more notable floods or
even highest flood. While developing hydrographs associated with flood frequency estimates for project
purpose due consideration should be given to variability of time to pick & hydrograph shape. Reports on
flood frequency studies included complete data tables, statistical tastes on data & predictions on
distributions etc. were dicussed in this study.

5. Mkhandi, S. H., Kachroo R. K. (1996). - REGIONAL FLOOD

FREQUENCY ANALYSIS FOR SOUTHERN AFRICA, XIIÈMES JOURNÉES
HYDROLOGIQUES DE L’ORSTOM, MONTPELLIER, Tanzania

flood frequency analysis for homogeneous regions in south Africa to improve the design and economy
of the civil engineering was aimed at in this study. The following steps, Data screening, Homogeneous
regions & appropriate frequency estimation etc. were carried out in Flood frequency analysis of south
Africa.

All data that were made available were screened. Once a homogenous region was identified the next
step was to select suitable flood estimation procedure. Different procedure were Carried out by applying
predictive ability test.

The most robust flood estimation procedures for South Africa were selected from the result of the
predictive ability test with Standard Root Mean Square error.

CASE STUDIES Chapter 3

CASE STUDIES :

1. EVALUATION OF DESIGN FLOOD FREQUENCY METHODS FOR IOWA

STREAMS

Investigated by: IIHR–Hydroscience &

A. A. Bradley Jr. (PI) Engineering College of
L. L. Wehmeyer (Research Assistant) Engineering
L.-C. Chen (Research Assistant The University of Iowa
Iowa City

INTRODUCTION

10
Estimates of flood frequencies (e.g., the 25-year return period peak discharge) are needed for many
engineering design problems in Iowa. Three design methods were evaluated for flood frequency
estimation for small Iowa streams, two of which are recommended for use in urban setting by the Iowa
Statewide Urban Design and Specifications (SUDAS) manual. For drainage areas of 200 acres or less,
flood frequency estimates based on the Rational Method were found to significantly underpredict
estimates based on streamgage data. Flood frequency estimates based on the NRCS (The National
Resource Conservation Service) curve number approach are much less biased. Since many engineering
uses of design flood frequency estimates compare pre-developed and post-developed conditions. Future
work should examine the SUDAS recommended time of concentration procedures in design flood
frequency estimation.The following items are highlighted in this case study.

EXPERIMENTAL DESIGN

The goal of this study is to evaluate the design approaches for flood frequency estimation and their
application to Iowa streams. To meet this goal, small streams with sufficient streamgage records were
chosen and the flood frequency of each site was estimated. With the aid of a Geographical Information
System (GIS), watershed characteristics were collected and stored for each basin. The rainfall intensities
were then determined. This provided the necessary information for estimating the flood frequency based
on the Rational Method and the NRCS Curve Number approach. Comparisons of the design methods
were made to address the estimation differences and an alternative flood frequency estimation technique
was developed in order to mitigate the differences between the two design methods.

The Selection sites for flood analysis

For site selection, a Midwest region defined by Iowa and its surrounding states was examined. To
evaluate flood frequency design techniques, all streams with a drainage area of 200 acres (0.31 mi 2) or
less and sufficiently long flow records (20 years or greater) were selected. Forty-six sites in the Midwest
region are available, but no Iowa sites met these criteria. However, the 46 sites within the Midwest
region are considered approximately representative of the hydrologic and climatic characteristics of Iowa
streams. Since agricultural land uses are dominate at the 46 Midwestern sites, additional sites were
examined across the United States to find those with significant urban land uses. All streams with a
drainage area of 200 acres (0.31 mi2) or less, a flow record of at least ten years for active gages, and at
least 5% of the area with urban land use, were selected.

Evaluation of design flood frequency techniques

Many of the same parameters are used in the Rational Method and the NRCS curve number approach,
yet the estimates vary greatly. A technique proposed by McCuen and Bondelid (1981) was used to
investigate the discrepancies between the two design methods. The technique involves equating the two
methods and solving for the runoff coefficient in terms of the curve number. Because the NRCS curve
number approach involves the intensity duration frequency (IDF) curve for the region of concern, the
curve number cannot be universally converted to a runoff coefficient. The curve number may be
universally converted for a particular region where a single IDF curve is representative of the rainfall
characteristics throughout the basin. Central Iowa was chosen as the representative IDF curve for the
analysis of the 46 sites located in seven nearby states. Although the rainfall characteristics at each site
may not be identical to central Iowa, the sites were chosen for their similar climatic and hydrologic
regimes, making the central Iowa region an appropriate choice. The outcome of equating the two design
approaches resulted in a new method, named the Iowa-modified Rational Method. This approach takes
advantage of the simplicity of the Rational formula as well as the generally better 57 flood frequency
estimation of the NRCS curve number approach. The explanation of the development and the
implications of this new method are outlined in the following section.
11
Estimate flood frequencies based on the NRCS method

Following the method outlined in Technical Release 55 (TR-55), published by the Soil Conservation
Service (1986), flood frequencies were estimated for the selected sites using NRCS procedures.
Estimates were made for the 2, 5, 10, 25, 50, and 100-year return periods. The SUDAS manual (2004)
recommends 2000 acres as the upper limit for TR-55 methods. TR-55 methods are applicable in small
drainage areas as well, so this analysis included the same set of sites used in the Rational Method
analysis for comparison of the two methods. NRCS curve number procedures were used for runoff
calculations and the NRCS unit hydrograph (time lag approach) was used for flow routing to the basin
outlet.

Estimate flood frequencies based on the Iowa Runoff Chart method

The Iowa Runoff Chart (IRC) estimates flood frequencies using inputs of drainage area, land use, slope
and return period. Land use and slope are grouped together to create a Land Use and Slope Description
(LF) factor. The return period is used in the determinization of the Frequency (FF) factor. The drainage
area is converted to a volume of water that is scaled by FF and LF to provide the design peak discharge
for the given return period.

Comparison of the flood frequency estimates

For each site, a comparison was made between the flood frequency estimates based on the streamgage
data and estimates made using the Rational Method, the NRCS curve number approach and the Iowa
Runoff Chart method. The comparisons provided the information necessary to quantify the uncertainty
in the design flood frequency techniques. There is also uncertainty in the estimates from the streamgage
records, which was accounted for by the contribution of sampling uncertainty (a function of gage record
length) in the comparison. In addition to uncertainty, biases were quantified and systematic errors
(under- or over-estimation) noted.

2. Flood-Frequency Analyses for U.S. Geological Survey Gaging

Stations Based on Data through Water Year 2009
By
Peter M. McCarthy Steven K. Sando
Hydrologist (Eng.) Hydrologist
U.S. Geological Survey U.S. Geological Survey
3162 Bozeman Ave. 3162 Bozeman Ave.
Helena, MT 59601 Helena, MT 59601

U.S. Department of the Interior

U.S. Geological Survey

Background summary

The purpose of this research is to determine flood-frequency estimates for USGS gaging stations in Montana.
Flood-frequency estimates are determined by fitting a log-Pearson Type III probability distribution to the recorded

12
annual peak flows using methods described in Bulletin 17B of the U.S. Interagency Advisory Committee on
Water Data (1982). However, Bulletin 17B analyses using default procedures do not always provide the most
accurate flood-frequency estimates for gaging stations with complicating factors. Research activities in the
proposed project will focus on improving flood-frequency estimates for gaging stations and publishing the flood-
frequency estimates along with detailed documentation of the specific methods used to develop the frequency
estimates. The following items are highlighted in this case study.

Objectives

The primary objectives of the proposed project include:

1) Determine flood-frequency estimates (recurrence intervals of 1.5, 2, 2.33, 5, 10, 25, 50, 100, 200, and 500
years) for more than 660 USGS gaging stations in Montana based on data through water year 2009. These
flood-frequency estimates are routinely used by MDT for design purposes.
2) Investigate methods of adjusting flood-frequency estimates for short-term gaging stations with data
restricted to either unusually wet or dry periods (including 1985-present).
3) Investigate the effects of climatic variability on our understanding of flood hydrology in Montana and
potential emerging methods for determining more accurate flood-frequency estimates.
4) Investigate the use of a regional approach for flood-frequency analysis in northwestern Montana, where
unusually large (but infrequent) floods complicate flood-frequency estimate

INFORMATION SUPPORTING THE NEED FOR RESEARCH ON PERIOD-OF-RECORD

Since the early to mid-1980’s large parts of Montana have experienced two severe droughts. Occurrence of these
dry periods during a relatively short time frame has the potential to substantially affect flood-frequency estimates
for gaging stations with relatively short periods of record. Further, recent climatic patterns, and uncertainties
concerning the relative contributions of natural and anthropogenic factors as driving forces, have led some
researchers to question some of the basic premises of flood-frequency methods typically used for structure design
(Milly and others, 2008).

Comparison between flood-frequency curves based on 1985-2008 data and flood-frequency curves based on long-
term period-of-record data (figs. 3 and 4) shows effects of recent climatic patterns on flood-frequency estimates.

Methods

Fitting the LPIII distribution to systematic records of annual peak flows is the currently accepted methodology for
estimating flood frequency for design purposes. However, if default Bulletin 17b methods are applied to gaging
stations with anomalous periods-of-record (including the period 1985-2008 in some areas) there is potential for
introducing substantial inconsistency between gaging stations.

Where applicable, record-extension methods will provide improved flood-frequency estimates for some short-
term gaging stations. Further, consistency in flood-frequency estimates between closely-located gaging stations
with different periods of record will be improved. Thus, application of currently accepted flood-frequency
methods will be improved. However, uncertainties in the driving forces of recent climatic

patterns and whether the recent patterns will actually prevail on a long-term basis might have substantial effects
on what are considered appropriate flood-frequency methods in the future.

13
To remain current with emerging climatic and hydrologic research, initiation of systematic tracking of stationarity
in peak-flow records for Montana is proposed. Temporal patterns in statistics of peak-flow records (that is, the
mean, standard deviation, and skew) will be investigated for long-term stations. Spatial variability in any apparent
temporal trends also will be investigated. Further, a literature review and consultation with USGS researchers will
be conducted on nonstationary probabilistic models and the potential for application to flood-frequency analyses
in Montana.

Benefits

Up-to-date flood-frequency estimates will meet critical needs for numerous structure-design and floodplain-
management issues in Montana, especially for MDT. Further, the proposed research will provide valuable
information for evaluating the effects of climatic patterns on flood-frequency estimates and investigating
emerging flood-frequency methods that address climatic uncertainty.

o Economical hydraulic design

o Establishment of road grades and low beam elevations

o Defending against lawsuits

o Securing floodplain permits.

o Stream restoration and fish passage.

o Preliminary-engineering and planning process

Detailed discussion

Flood-frequency estimates for USGS gaging stations generally will be updated using standard procedures
presented in Bulletin 17B. However, if default Bulletin 17B methods are applied to gaging stations with short or
unusually wet or dry data-collection periods (including the period 1985-2008 in some areas) there is potential for
introducing substantial inconsistency between gaging stations. Specific methods that will be investigated for
adjusting flood-frequency estimates to compensate for period-of-record effects are presented.

Recent climatic patterns, and uncertainties concerning the relative contributions of natural and anthropogenic
factors as driving forces, have led some researchers to question some of the basic premises of typical flood-
frequency methods, including stationarity of peak-flow records. There is a substantial need to begin systematic
tracking of stationarity of peak-flow records in Montana. Specific methods that will be investigated to remain
current with emerging climatic and hydrologic research, and initiate systematic tracking of stationarity of peak-
flow records for Montana are presented.

14
Flood-frequency estimation for some gaging stations in northwestern Montana is complicated by the presence of a
small number of unusually large floods during the period of record. In previous USGS flood-frequency reports, a
mixed-population flood-frequency analysis was performed on a station-by-station basis to improve flood-
frequency estimates for these gaging stations. However, a regional mixed-population analysis might provide
further improvement to flood-frequency estimates in northwestern Montana, and also potentially have application
to other parts of Montana. Specific methods concerning application of the regional mixed-population analysis are
presented.

3. Flood Frequency Analysis of the Waimakariri River

Robert Ware and Frank Lad

Approximation of Posterior Means and Variances
of the Digitised Normal Distribution using Continuous Normal
Approximation.
Technical Report UCDMS2003/16, Department of Mathematics and
Statistics,
Univeristy of Canterbury, Christchurch, N.Z.

INTRODUCTION

A common problem in many areas of environmental engineering is that of estimating the return period of rare
geophysical events, such as extreme floods, for a site or group of sites. A large volume of work considering the
estimation of flood risk has appeared in the last 20 years. Modern flood frequency theory is typical of much
conventional statistical theory, in that most effort is expended in determining an appropriate form to model the
“underlying distribution” of floods, and then estimating the parameters of thisunderlying
distribution.Conventional estimates of flood exceedance quantiles arehighly dependent on the form of the portion
of the underlying flood frequency distribution(the right tail) which is most difficult to estimate from observed
data.Currently there is no compelling theory on which to base the distributional form ofthe right hand tail
The following items are highlighted in this case study.

The Waimakariri River

The Waimakariri River is located in the Canterbury region of the South Islandof New Zealand. It is classified as a
“Main Divide” river, meaning that it hasa catchment which drains from the ranges east of the Main Divide of the
SouthIsland. It has a catchment area of 3210km2, the largest of any river in Canterbury.The Waimakariri River
flows through the northern outskirts of Christchurch. Floodprotection works have been constructed to protect most
of urban Christchurch andKaiapoi.

The daily flow of the Waimakariri River has been recorded since 1930 at thesite of the Old State Highway One
bridge. The site is 5.4km from the mouth ofthe river, and consequently the water level is effected by the tide. The
Waimakariri River is also measured at a site in the Waimakariri Gorge. When records of river levels do not exist,

15
for example if the mechanical recorder is broken or if the flow records have been lost, flow records are estimated
based on water levels at the Waimakariri Gorge recording site. Studies of the relationship between recorded flow
levels at these two sites show this to be a reasonable resolution to the “missing data problem”. Recent
Waimakariri River flows can be viewed on the Canterbury Regional Council’s internet site.

Between 1930 and 1966 water levels at the Old State Highway One bridge site were recorded visually at irregular
intervals. The AMS series for this period was calculated retrospectively using slope-area calculations and records
from the Waimakariri Gorge site. “Slope-area gaugings” are post-flood event measurements.

They take into account the highest water level mark, the associated cross-sectional area of the channel, slope of
the channel bed, and likely velocities of the flow as read from a standard water depth-flow velocity relationship
table. It is widely recognized that slope-area gaugings are less precise than records obtained via conventionally
recorded flow levels. In 1966 a mechanical recorder was installed to record the water level every 15 minutes.
Flow measurements for years 1960 through 1966 were complicated by the constantchange of the river-bed
profile due to large amounts of shingle being removed AMS recorded for the Waimakariri River.

from the river bed for use in construction of Christchurch’s Northern motorway. During the period on record there
has been a small (< 10 cumecs) amount of upstream water diversion. This is an insignificant amount considering
the level of measurement precision.

The highest mechanically measured discharge for the Waimakariri River is 3070m3/sec (1979), a value exceeded
twice in the recorded period (1940 and 1957). AMS values range between 710m3/sec (1939) and 3990m3/sec
(1957) with mean1485m3/sec, standard deviation 632m3/sec and skewness 1.77. The upper panel of

At-Site Flood Frequency Analysis

The simplest flood estimation methods involve collecting AMS data for a site, and using this data to estimate the
characterising parameters of the underlying distribution, the functional form of which is assumed known. Popular
estimation techniques include the method of moments and maximum likelihood estimation. The method of
moments is notoriously unreliable for fitting extreme value distributions due to the poor sampling properties of
second and higher order sample moments. The method of maximum likelihood has been used when dealing with
extreme values, however it doesn’t work well when the sample size is small to moderate. Moreover its
computational aspects are based on iterative procedures which require reasonable starting 8 values. For the
parameter estimation undertaken in this Report we shall use the method of L-moments, a linear extension of the
conventional method of moments. L-moments have been widely used in recent studies of extreme phenomenon.
For a taste of the breadth of current research see Kjeldsen et al. (2002), Kroll and Vogel(2002) and Park et al.
(2001) to see L-moments applied in studies in South Africa, the United States of America and South Korea. The
theory of L-moments is introduced in Section 3.3. Three popular candidates for underlying flood distribution are
introduced

Regional Flood Frequency Analysis

Recent research into flood frequency estimation has focused on developing and evaluating regionally derived
flood frequency estimates. In regional flood frequency analysis it is assumed that the data from all gauged sites in
a region can be combined in such a way as to produce a single regional flood frequency curve. This curve is
applicable, after appropriate rescaling, anywhere in that region. Regionalisation allows us to pool data from m
sites. Each site has ni years of recorded measures, where ni can be of any length. Conventional regionalisation
techniques identify a fixed set of recording sites which adjoin each other. Each region is identified by considering
which sites are ‘close’ to each other. Proximity can be assessed using statistical measures (e.g. coefficient of
variation (CV ) or ratio of mean flow to drainage area) or spatial measures (e.g. longitude and latitude of each
site). The biggest advantage of regional estimation is seen to be the increase in record length. A regional approach
is necessary when estimating floods at sites with no observed data. Many studies (e.g. Lettenmaier et al., 1987;
Hosking, 1990) have shown that flood estimates based on regional information are more accurate (have less

16
absolute error) and are more stable (have less variance) than those based solely on at-site records. The most
commonly used regionalisation techniques are based on the index flood approach.

Summary:

This Report has focussed on flood frequency analysis, and in particular on the estimationof flood quantile levels.
Our problem of characterising extreme floods wasintroduced in Section 2. We described river flow measurement
procedures, payingparticular interest to the Waimakariri River. In Section 3, conventional frequentist estimates
were calculated for the Waimakariri River. A Monte Carlo procedure wasused to compare different estimation
methods in terms of accuracy and precision, foran experimental data set. In Section 4 a procedure for scoring
sequential forecastsusing digitised mass functions was developed. This procedure is based on the work developed
in Ware and Lad (2003). Different scoring rules were examined usinga sequence of annual maximum river flows
from the Waimakariri River. Finally, we compared the appropriateness of the frequentist and digital procedures
using both subjective and objective techniques. The scores of conditional expectations of both procedures were
similar, but the score of the conditional variance was much better for the updated mixture forecasts. When
objectivist measures were considered, the mixture distributions computed via the discrete digital method provide
forecasts with lower root mean-squared-error. This is despite the fact that when the coefficient of variation is
small the digital methods are more biased than the frequentist methods. As the coefficient of variation increases,
the accuracy of the digital methods improves rapidly, even reducing the bias.

4. Report on Six Case Studies of Flood Frequency Analys

by
C.R. Neill, Northwest Hydraulic Consultants, Edmonton, Alberta
and
W.E. Watt, Queen’s University, Kingston, Ontario

INTRODUCTION

The cases considered are as follows:

1. Willow Creek near Claresholm, related to the Pine Coulee Dam project.
2. Highwood River at High River, related to the Little Bow River Dam project and to floodplain

17
 delineation .
3. Red Deer River at Drumheller, related to a hospital project at Drumheller and to floodplain
delineation.
4. Sawridge Creek at Slave Lake, related to urban flood damages.
5. Bow River at Calgary, related to floodplain delineation.

6. Oldman River near Brocket, related to Oldman River Dam project.

has been used in the preparation of a National Water Plan. The Generalized
Extreme Value (GEV) distribution is being used in the Flood Action Plan.

Discussion
Various difficulties arise over conducting a flood frequency analysis for the Bow River at Calgary.
Some relevant points are as follows:

1. The amount of live storage in the basin was increased at intervals over the period 1912 to 1955, to a
total of about 9000 cms-days. Logically, all recorded flood peaks should be naturalized before attempting a
frequency analysis.

2. More important than the storage effect is a large statistical discrepancy between the earlier and later
records, apparently caused by a major shift in rainstorm patterns after 1932. The cause of this shift and its
possible periodicity are unknown. It is generally considered unjustified in statistical practice to lump together
two series from clearly different source populations and fit a single probability distribution. One study used a
method of combining two component distributions into a single dual-slope frequency curve. In such cases,
the most rational method of selecting a project flood may be to adopt an arbitrary value or a specific
historical flood. Both these approaches have been used at different times. Construction of a composite
frequency curve from identified marginal distributions also appears to be a legitimate approach.

HIGHWOOD RIVER ABOVE HIGH RIVER

Flood Record

Flood frequencies above High River town were required in connection with the Little Bow River Dam
project. The main source of data is the downstream gauging station near Aldersyde. By including a station
below the Little Bow Canal that began in 1986, a nearly continuous record of annual maximum discharges
can be constructed from 1912 to 1996. Instantaneous maximum values are reported for most years since
1957, and for all but one (1932) of the higher floods before then. The drainage basin is basically unregulated
with respect to floods.

18
Alberta Environment (1991) adjusted recorded values for occasional spills to only) and for generally trivial
canal withdrawals, to provide a "naturalized" instantaneous series upstream of High River town. If the
subsequent record flood of 1995 is added, the seven highest events in the series rank as follows:

The 1995 estimate used here was arrived at by adding an arbitrary allowance of 27 m 3/s to the recorded peak
for overspill below Little Bow Canal. The origin of the 1932 estimate is explained in Section 2.3 below. A
plot of the entire 1912-96 sequence is shown in Figure

Rank Year Peak flow m3/s

1 1995 830 (estimate)
2 1932 818 (estimate)
3 1952 782
4 1923 702
5 1929 593
6 1953 510
7 1963 453

19
OLDMAN RIVER NEAR BROCKET
Background and Flood Record

Flood frequencies near Brocket were investigated in connection with the Oldman River Dam project in the
mid-1980s. The primary source of data was the Environment Canada gauging station a short distance
downstream, designated as Oldman River near Brocket (# 05AA024, drainage area = 4400 km2). This has an
annual maximum discharge record extending from 1966 to present - except that the "Extremes Report" has
no entries for 1995, when the largest flood since 1966 occurred. Since around 1990, flows at the station have
been affected by the Oldman River Dam.
In flood frequency studies conducted for the project (Hydrocon 1985), the short 19-year record then
available for Brocket (1966-84) was extended by correlating and interpolating earlier records for upstream
and downstream stations, to create a 73-year table of maximum daily discharges covering the period 1910-84
except for a 2-year gap in 1932-33. (The events of 1932 and 1933 seem to have been unremarkable at
Lethbridge, so their omission is probably of little significance.)
Within Hydrocon's extended 1910-84 series, the three highest daily maxima are as follows:

1923: 1865 m3/s (synthesized)

1975: 1220 (recorded)
1948: 1045 (synthesized)

A historical flood of 1908 is thought to have been higher at Lethbridge than the flood of 1923. Because of a
lack of information from other locations, no 1908 estimate was developed for Brocket.
As noted above, data for the 1995 flood are missing in the Brocket station record. However, estimates of
hourly inflows to the Oldman River Reservoir a short distance upstream were synthesized by Alberta
Environment on the basis of reservoir levels and outflows (Magowan and Ruttan 1997, Magowan 2001).
Estimated daily and instantaneous maxima are approximately 2050 and 3490 m3/s. Assuming the synthesis to
be reasonably reliable, the peak/daily ratio is then 1.70. The peak inflow appears to have occurred at
midnight on 6/7 June, a timing that tends to maximize the ratio.

Relationship of PMF to Flood Frequency Estimates

A separate project study by Alberta Environment (1985) estimated the Probable Maximum Flood using
standard hydrometeorological methods. Their PMF had an instantaneous peak of approximately 9500 m3/s,

20
 approximately 5 times Hydrocon's 100-year value. This ratio, although very high, is not incompatible with
some other recent studies.

The PMF hydrograph was of markedly shorter duration than the flood frequency equivalents. The rise time
was only 24 hours instead of 36, the duration up to the 15% recession point was less than 3 days instead of 7,
and runoff was virtually complete at 5 days. The PMF and 10,000-year hydrographs are compared.

Reconstructed 1995 inflow hydrograph compared with pre-project 10,000

year and PM

5. REGIONAL FLOOD FREQUENCY ANALYSIS FOR SOUTHERN AFRICA

S. H. Mkhandi, R. K. Kachroo

Water Resources Engineering Programme

University of Dar es Salaam
P.O.Box 35131
Dar es Salaam
Tanzania

21
INTRODUCTION

The data used in the study was contributed to the Southern Africa FRIEND data base by the
participating countries. This data base is currently situated at the Water Resources Engineering
Programme at the University of Dar es Salaam, Tanzania. Information on the
number of sites, their corresponding record lengths. Data from a total of 754 gauging stations
with average record length of 24 years from eleven countries in Southern Africa was used
as the basis for carrying out this study. Annual maximum instantaneous discharge series
were used for most of the countries. For those countries where annual maximuminstantaneous flows
were not available, annual maximum mean daily discharge series were used.
The following items are highlighted in this case study.

Predictive ability tests to select suitable frequency procedure for the

analysis of flood flows

The selection of a distribution for flood frequency analysis goes in hand with the selection of the method
of parameter estimation. The most commonly used parameter estimation methods are the methods of
Moments (MOM), Maximum Likelihood (ML) and the Probability Weighted Moments (PWM).
Parameters estimated by any of these methods are subject to sampling errors. While a method may be
efficient for one distribution it is not necessarily true for other distributions.

In this study the predictive ability tests were used to evaluate the performance of different procedures to
estimate the flood magnitudes in the Southern Africa region. Predictive ability tests investigate how well
a distribution and its associated method of parameter estimation can estimate the Q-T relationship or
the frequency of future events when the population distribution is not identical to that of the proposed
distribution. The results of the predictive ability tests were consequently used to select the most robust
flood estimation procedure for the regions.

Methodology to select the most robust flood estimation procedure.

1. Estimate population parameters from weighted regional average Standardized statistics (Qbar, Cv
& Cs)

2. Obtain the exact Q-T relationship for T = 100, 200 and 500 using the estimated population
parameters

3. Perform Monte Carlo experiments to generate random samples of size 15, 30 and 50.

4. Estimate flood quantiles, T = 100, 200 and 500 using each of the 14 estimation procedures.

5. Repeat the procedure (step iv and v ) 1000 times.

6. Evaluate performance of each procedure by computing test indicators for the bias, se, rmse, pe
from 1000 quantile values estimated for a given sample size and return period.

7. Obtain the average over the five values determined from simulations using 5 parent distributions for
each test indicator and estimation procedure.

8. Rank the performance of the different estimation procedures for a given return period and sample
size.

22
 Analysis of results from predictive ability tests
The results are obtained from the simulation study involving homogeneous regions from 11 countries in
the Southern Africa. The results presents the best overall procedure which gave the lowest test
indicator values for the bias, the standard error, the root mean square error and the expected probability
of exceedance for the given return period (T=100, 200 and 500) and sample size (n=15, 30 and 50).
The indicator values for the probability of exceedance are presented as percentage of the difference in
the estimated return period in years over the specified return period.

From the results obtained from the predictive ability tests, different procedures emerged as robust when
considering the different test indicator estimators. However in this study the procedures which were
indicated by lower bias and smaller difference in the expected probability of exceedance were preferred
over those indicated by lower standard error of estimate but poor fitting ability. In this regard, the
procedures which were indicated by lower bias i.e. P3/PWM and LP3/MOM were adopted for deriving
regional flood frequency curves for the regions in Southern Africa.

The estimated regional parameters of the selected procedures for all the regions are considered in the
study. The model parameters estimated for a given region were used to compute standardised quantile
estimates for the return periods (T=2, 5, 10, 20, 25, 50, 100, 200 and 500). The resulting quantile
estimates were then used to construct the regional frequency curves.

The investigation on the slopes of the derived frequency curves for the different countries indicated that
the extent of the slopes seem to be governed by the variability of the flood regimes. For instance if the
coefficient of variation (Cv) of flood flows for a given region is high the frequency curve for that
particular region should be expected to be steep and vice versa.

Summary
One of the objectives of the study was to delineate the Southern Africa region in homogeneous regions
which would form the basic units to develop frequency curves with smaller standard errors. Despite
many attempts made by the research hydrologists to delineate homogeneous regions no general
methodology is accepted universally. This is due to the complexity in understanding precisely the
factors that have effect on the generation of floods that would provide guidelines to delineate regions
with similar flood characteristics. The procedure to delineate flexible homogeneous regions based on
available information on topography, mean annual rainfall, drainage pattern and maximum flood peak
regions proved useful. The majority of the proposed regions satisfied the homogeneity test applied in
the study. Predictive ability tests have been applied to identify robust flood estimation procedures for
the regions in Southern Africa. Criteria of bias, standard error, root mean square error and expected
probability of exceedance were used to evaluate the robustness of various flood estimation methods
considered in the study. The performance of a given estimation procedure was judged through the
magnitude of each of the test indicator applied in the study.

CONCLUTION Chapter 4

23
CONCLUTION:
• One of the objective of the study is to delineate the method which would form the basis to develop the
flood frequency curves with smaller standard errors. Despite many attempts made by research hydrologist
no general methodology is accepted universally. This is due to the complexcity is understanding precisely
the factors that have effect on the generation of floods & that would provide the guideline.

• To Delinate regions with similar flood characteristics. The procedure to delinate flexible homogeneous
regions based on the available information on topography mean annual rainfall, drainage pattern & max.
flood regions proved useful.

• The results obtained from the study show that different procedure immerged as robust flood estimation
procedures when different predictive tests were applied to evaluate the performance of the various
procedures. The selection of one procedure from among the best four procedures in the study will depend
on the emphasis put in the predicting model.

24