Stream Density

COMPARING QUANTILE REGRESSION TECHNIQUE WITH INDEX FLOOD METHOD FOR QUEENSLAND RESEARCH PROPOSAL BY: Jayawickrama chamidu manuranga Abeygunawardana SI: 18098121 SUPERVISORS BY: Prof Atarur rahman Table of Contents ABSTRACT ............................................................................................................................................ 3 INTRODUCTION ................................................................................................................................... 4 LITERATURE REVIEW ........................................................................................................................... 5 MOTIVATION FOR THE RESEARCH .................................................................................................... 11 RESEARCH QUESTIONS ...................................................................................................................... 12 AIMS AND OBJECTIVES ...................................................................................................................... 12 RESEARCH METHODOLOGY ............................................................................................................... 13 PROPOSED TIMELINE ......................................................................................................................... 14 CONCLUSION ..................................................................................................................................... 14 ABSTRACT Floods are one of the world's largest epidemics, affecting people for along period of time that cannot be easily controlled by nature. Flood forecasting and water resources management are among the most important departments that require reliable estimates and statistics. Therefore, the Regional Flood Frequency Analysis (RFFA) isused for this analysis by transferring flood information and features from submerged catchment areas to unprotected catchment areas or simply zoning. The objective of the project isto understand the regional flood frequency analysis methodology for flood assessment planning. This resolution examines RFFA approaches developed by various researchers, applied in the real world, and tested for practical use in the literature review. The research proposal enhances the RFFA approach to better results on RFFA analysis. INTRODUCTION Frequent floods in some places have caused economic hardship over time and such places are unsafe for human habitation. Floods adversely affect land and infrastructure in areas that occur more or less frequently. Individual, private and government infrastructure in flood-affected areas will be destroyed. Infrastructure includes water infrastructure operated by regulators and other infrastructure of high economic importance in the affected areas. Destruction of houses, offices, homes, places of work, crops, livestock, and death of people due to floods leads to people's displacement as they tend to move away from affected regions and search refuge elsewhere. Floods are natural occurrences that cannot be well predicted in places where the severity of floods isnot well known. The floods come unexpectedly or expectedly, depending on the region being investigated. Different places have advantages and uses which have to be utilized by humans due to their economic significance. Since floods in affected regions destroy projects and infrastructure, there isaneed to understand the pattern of floods and put up infrastructures and plans to counter the adverse effects of floods. However, data isthe most important aspect of any planning, depending on deep understanding and analysis of collected data from affected areas. For accurate analysis and design, there must be enough and adequate data to be analysed regarding the floods in affected and regions of interest. This report focuses on Comparing the Quantile regression technique with the index flood method for Queensland, Australia. Due to having avariable climate, the Australian continent experiences a different type of natural disasters and calamities, this includes both droughts and flooding. Flooding isan event where water inundates land that isnormally dry due to heavy and intense rainfall. Australia experiences three categories for flooding, quickset onset flooding, slow onset, and flash flooding. Flood isthe worst natural disaster in Australia in terms of economic damage. For example, in 2013 northern New South Wales was hit with tropical cyclone Oswald that lead to 11 days of heavy rainfall and major flooding. An estimated 41000 people were isolated with peak water levels being at 3.3 meters and 1500 people around the Clarence River were asked to evacuate their properties. Roads and streets were closed, power was cut off and the area was declared adisaster zone. The aftermath of the incident was that 1person went missing, 6people ’sfatality, and 2.52 billion dollars of damage. The 2010-11 floods in Queensland caused 35 deaths and over $5 billion damage. The extreme rainfall in early February 2020 exceeding 390 mm in just four days caused widespread damage across greater Sydney including falling trees, power outage, and flooding of roads. This project proposal reviews the literature on Quntile regional flood frequency analysis by different researchers, identifies the motivation for the research, states the research questions, discusses the aims and objectives of the research, proposes aresearch methodology for investigation including data collection methods to be adopted and analysis to be done, provides atimeline for the research in grant chart and concludes the proposal procedures and activities discussed in the research proposal paper. The research islimited to secondary data due to the difficulty in obtaining real-time data and the time required in collecting the data. LITERATURE REVIEW Research by Haddad and Rahman (2012) on regional flood frequency analysis in eastern Australia compares the Bayesian Generalized Least Squares (BGLS) regression approaches. Afixed and region of influence (ROI) framework isused in the research in order to minimize the Bayesian model error variance. Their research used data from 399 sampled catchment areas ranging from small to medium sizes located in eastern Australia. They predicted from prediction equations developed on average recurrence intervals (QRT) flood quantiles for return periods between 2to 100 years. They employed the moments of log-Pearson type 3(LP3), to the third, also known as the parameter regression technique (PRT) and the Quantile Regression Technique (QRT). Statistical selection procedures and stepwise regressions were adopted in finding optimal regression models for the ROI approach they adopted. The researchers used secondary data and RFFA methods available in an eastern Australian database. According to Haddad and Rahman (2012), the QRT method uses area and design rainfall intensity as significant estimators of aregion's flood quantile. Their research also found out that LP3 distribution uses density rainfall intensity, forest, slope, area, mean annual rainfall, and mean annual evaporation as significant estimators of flood quantile. The BGLS QRT- and PRT-ROI models were better than the fixed region regression approach since they had much smaller regression errors. The QRT-ROI and PRT-ROI models were found to be much similar, with only modest differences hence viable substitutes for each other. The assumptions of ROI were met and well satisfied using QRT-ROI and PRT-ROI models contrary to the fixed region regression models. The research isimportant in RFFA since itprovides for models that try to reduce the estimation errors. The selection and identification of acatchment area with similar and relevant criteria for astudy and its objectives require significant effort and time (Haddad et al. 2010). In their research on regionalization of southeast Australia, they prepared stream flow data for RFFA in Victoria and NSW states. Alarge series of data was obtained from gauging stations available in Victoria and NSW on maximum flood series recorded annually. In collecting the data, alot of uncertainty available in data sources prevents aguarantee for large data sets that increase results accuracy. The study focused on obtaining design flood estimates that would help the Australian Rainfall-Runoff (ARR) through regionalization of Victoria and NSW's states from stream flow data and regional flood frequency analysis. Haddad et al. (2010) reveal that the lack of access to required data for RFFA led to assumptions and estimates from personnel in provision organizations and high dependence on internet sources in order to reduce inappropriate assumptions. In states of Victoria and NSW, 10% to 20% of stations in these regions have over the years since 1990 experienced adownward trend in their annual maximum flood series (Haddad et al. 2010). The variation or change in climate in the past 1990s isthe reason for the decrease in floods in the region, which was highly affected by drought in the 1990s. The regionalization study, which requires data provisions from gauged regions and then transferring the characteristics to ungagged sites, provides adiscussion board between the quantity and quality of the data provided. The compromise extends to the results of the flood quantile estimates and returns periods for estimated areas. The accuracy of flood estimates in aregion isdependent on the available number of stations in the region, depending on the heterogeneity available in the region of interest. Haddad et al. (2010) results are limited to undetected errors in their adopted data sets used in data preparation for the RFFA project in ARR. The RFFA isused in providing estimates of floods in areas where there islittle or no data available. The RFFA thus depends on the regional homogeneity between the gaged and ungagged sites of interest in regions being estimated. Wiltshire (1986) proposed atest for regional homogeneity statistics, which are used in regional flood frequency analysis. The proposed tests work on testing the regional parent of aregion's characteristics and the transfer of region characteristics from one site to another site or its distribution function. The test isdependent on the size of the region, the choice of the parent distribution function, and the length of the record. The homogeneity test was used in the test to sign the homogeneity of the regions, which depends on the probabilities of flood maxima occurrence inhomogeneous regions. The significance test required data from regions of interest and individual sites that needed to be larger enough to improve the test's power, as claimed by Wiltshire. Wiltshire (1986) conducted two homogeneity tests on homogeneous flood frequency regions with similar probabilities of maximum flood populations and flood frequency. Acoefficient of variation (CV) test on the standardized annual maximum flood series was conducted in gauging the regional variability and homogeneity test. Adistribution-based test of homogeneity test was also conducted on flood frequency in areas of interest to test and verify the availability of homogenous results for regions with homogenous physical and climatic characteristics. In the distribution –based test, a cumulative distribution function (CDF) was adopted by Wiltshire in studying the flood frequency distribution in aset of flood data. The article's homogeneity tests show significant results on test data with results improvement found in larger stations and large data requirements for analysis. The study isimportant as itprovides asolution to most questions on the reliability of regionalization results based on the homogeneity of regions as abasis for estimation of flood frequency in ungagged areas. Acomparison of the two commonly used RFFA methods of the probabilistic rational method (PRM) and generalized least squares method (GLS) based quantile regression technique (QRT) ismade by Rahman et al. (2011). In their research, they used data from asample of 107 catchment areas for the comparison of the PRM and GLS-based QRT methods. The study was conducted in NSW in Australia, where the catchment areas were sampled with all data variables used in prediction and model estimation significantly. For effective results for comparison, the researchers used the same predictor variables in both methods of the RFFA approach in design flood estimation for ungauged catchments. The design flood estimates are very important, as discussed in the hydrologic design. Different split-sample validation and one at atime, tests were used for evaluation of the methods of the RFFA approach for flood estimation in places where estimation isneeded but with little or no data. Rahman et al. (2011) compares the models and finds that the GLS based QRT method performs better than the PRM method of the RFFA. They also establish that the QRT method not only performs better than the PRM but also does not require arunoff coefficient contour map as compared to the PRM, which uses itin its estimation. The QRT model even works better with better results of estimation with the addition of the predictor variables and thus can be integrated into any region of influence. The QRT method isthus more flexible and can provide better results when the analysis isbased on the analysis of the error produced by the results of estimation than the PRM method. The researchers acknowledge the PRM and QRT methods as approximate methods that can provide results for larger catchments and smaller catchments, which are accurate on reasoning with the methods' underlying assumptions. The research information isessential in providing the basis for the development of regional flood estimation techniques for ARR. Pandey and Nguyen's (1999) paper assess regression model performance in flood quantile estimation for select ungagged sites in aparticular region. They look at selecting statistical techniques that can, in the best way possible to determine parameters that link the flood quantiles with physiological characteristics. The accuracy of the predicting regression models isgaged by looking at the performance of different estimators of flood quantiles. The regression models used in flood quantile estimation are tested in the research by varying the predictor variables and checking the performance accuracy of the regression model. The regression models used in the estimation at ungagged sites are used to propose solutions to real-world problems. They compared nine different methods of parameter estimation for the power-form model, which connects flood statistics and physiological characteristics. They conducted research in Quebec, Canada, using the area's physiographic and hydrological data. A 10 to 100 years return period was used in covering abasin area between 3.9 to 86,930 km2 to cater to low and high sides of flood distribution. Ajackknife sampling and simulations were conducted in order to simulate the ungagged site. The model properties, such as model bias, relative error (RE), and root mean squared error (RMSE) were the checks for the overall performance of the nine different models compared in the study. They found that the nonlinear models like the nonlinear least absolute value regression (N_LAV) and nonlinear root mean squared error (N_RMSE) performs better than other models. The nonlinear methodologies produce better results for flood quantile estimates for ungagged sites but also limited in many other ways. The nonlinear models do not have statistical tests making them impossible to provide confidence intervals (Pandey and Nguyen 1999). The research has agreat impact and influence in RFFA research as ithas provided abasis of choice between linear and nonlinear models for log and real domain. The RFFA method for arid and semi-arid regions requires stream flow data, which isdifficult to obtain and find records of such data, especially gauged sites. There exists limited hydrological literature for RFAA analysis in most world regions (Zaman, Rahman, and Haddad 2012). Their study looks at the RFFA study for arid and semi-arid areas of Australia using updated data sets and compares itwith world arid region data available in the literature. The study found out that arid and semi-arid catchment areas showed higher losses and smaller than expected runoff coefficients. Partial duration series flood data fitted well with Generalized Pareto distributions as compared to the Exponential distribution. The study also found out homogeneity in developed growth curves for arid and semi-arid regions compared to the world's flood data on arid and semi-arid catchments. Higher ARIs also show decreased mean rainfall with steeper growth curves, while humid regions show flatter curves compared to arid and semi-arid regions in Australia. The study by Zaman, Rahman, and Haddad (2012) shows that design rainfall intensity and catchment area as reasonable estimate predictor variables characteristics in prediction equations of amean annual flood using cross-validation. The study results show that the overall accuracy of developed RFFA methods for arid and semi-arid regions isless than those of Australia's humid regions. Arid and semi-arid areas have higher hydrological variability degree than the humid or coastal regions of Australia. The research developed prediction curves for the arid and semi-arid areas of Australia but also could be used to provide flood quantile estimates for other desired catchments of the world in the RFFA methods for design flood estimation. To improve the accuracy of the RFFA methods of flood design estimates, the researchers proposed an increased establishment of gauging networks streams databases for arid and semi-arid areas of the world. The (Zaman, Rahman, and Haddad 2012) research contributes alot to the RFFA analysis. Itestablishes the need for RFFA methods in the arid and semi-arid areas and methods applicable in such areas. The probabilistic rational method iswidely used in southeast Australia in Australian Rainfall and Runoff for design flood estimation in small to medium sizes of ungauged catchment areas (Rahman 2005). The probabilistic rational method iswidely used since ituses the runoff coefficient as its central component, which varies smoothly over average recurrence intervals (ARIs) and geographical space. Rahman (2005) proposes another method called an Lmoments-based index flood method based on aquantile regression technique. Rahman conducted research using asample of stream flow and catchment characteristics from 88 catchments in southeast Australia. The proposed L moments-index model uses 12 predictor variables of ungauged catchments and transferring those characteristics to homogeneous regions. The model was used to develop prediction equations for 2, 5, 10, 20, 50, and 100 years of ARIs design floods in the quantile regression technique. The Rahman (2005) Lmoments-index model assumed that the explanatory or predictor climatic and physical catchment characteristics selected had aconcrete role in flood generation. The data are obtainable easily. The model requires alot of data, which are time-consuming in gathering for use in the model. The extraction of such data isdifficult and thus the need for amore effective way like an automatic procedure like the use of ageographical information system (GIS). Thus, the paper proposes the need for such facilities in most of southeast Australia's ungauged areas for easier data extraction. The model satisfied the underlying model assumptions, and upon testing with independents, tests confirmed the accuracy of the model's prediction equation for flood estimates. The model isimportant since itdoes not necessarily need the geographical proximity assumption as compared to the probabilistic rational method's runoff coefficients. Cunnane's (1988) paper assesses the regional flood frequency analysis methods and merits in smoothening hydrological data. The researcher discusses twelve different RFFA schemes and their features in flood design estimation and prediction for hydrological projects in the research. The RFFA methods adopted over the period were based on annual maximum flood peaks, and others are based on peaks over the threshold, and others like the U.S. Water Resources Council (USWRC) are Bayesian. Astation year method that does not depend on inter-site uses regional pooling of data where sample statistics are used to reference population data —the Dalrymple's method which uses equal length data records in aregional averaging index flood method. Methods are based on dimensionless moments where regional average values can be used in the estimation of parameters of desired catchment flood frequency estimates. Cunnane (1988) also discusses other methods in the research paper which include: record extension or joint estimation methods, the United States Water Resources Council method that relies mainly on logarithmic skewness, Bayesian methods, amethod based on astandardized probability- weighted moment (PWMs), two-component extreme value (TCEV) method, regional application of Box-Cox transformation, threshold and censored sample methods and the simultaneous at-site and regional parameter estimation. Cunnane finds out that the methods are mainly differentiated on three assumptions; space-time equivalence of station year assumption, Bayesian assumption, and across region averaging assumption. The RFFA methods of regional homogeneity provide better estimates than the at-site approximation methods, which are generally expensive. The paper proves that small regional heterogeneity does not negate the RFFA methods' results, making them much effective flood design and frequency estimation and prediction. The research identifies Wakeby distribution as the best RFFA method for the estimation of regional standardized PMWs. Burn and Goel (2000) developed atechnique for identifying groups for RFFA. The technique uses a clustering algorithm to partition catchments at the beginning of the technique procedure. The K- means algorithm type of clustering algorithm isused in partitioning catchments into regions according to characteristics or features. The study develops aregionalization process that provides effective groups of catchments used in regional flood frequency analysis. The clustering algorithm described by Burn and Goel produces identifiable, homogenous, and sizeable groups from limited physiographic information upon region revision that are effective in regional flood frequency analysis. The clustering algorithm procedure can effectively be used to estimate the flood quantiles for both gauged and ungauged sites of interest (Burn and Goel 2000). More recent research in the design flood estimation has been conducted by Rahman et al. (2019) to estimate peak flows. They have proposed aRegional Flood Frequency Estimation (RFFE) approach that transfers gauged catchment flood frequency characteristics to the ungauged catchment of interest. The proposed approach relies on available and accessible catchment data making the approach simple and easy in design flood estimation. The approach ispegged on four criteria that must be satisfied for the approach to work. They include; national consistency in approach, smooth interfacing between areas at boundaries, use of readily accessible data, and maximum utilization of Australia's stream flow database (Rahman et al. 2019). The research used 853 gauged stations from anational database to develop and test the RFFE technique. The data-driven RFFE technique isan RFFA method that transfers flood characteristics from gauged catchments to catchments where design flood estimates are needed but are ungauged. In the development of the RFFE model by Rahman et al. (2019), the sample of 853 gauged areas was not enough to generalize the diverse and large 7.7 million km2 area of Australia. The model isalso limited to the availability of data for accurate estimation of design flood quantiles. The research could not cover all the variability in catchment areas due to the inadequate data on gauged catchments, more so arid and semi-arid regions. The relative accuracy and reliability of the RFFE technique were checked using the leave one out (LOO) validation. The comparison was made between flood quantile residuals of the RFFE technique and at-site Flood Frequency Analysis. The validation found some catchments with standardized residuals greater or less than three, implying that some of the RFFE confidence intervals could not be relied upon while making reference and conclusions. Rahman et al. (2019) research has agreat contribution to RFFA research as itled to the development of the RFEE software for the analysis. MOTIVATION FOR THE RESEARCH Flood isthe worst natural disaster in Australia. Every year itcauses huge damage to infrastructures such as roads, buildings, and bridges. Flood frequency and severity have increased over the years because of global warming and climate change. Flood has put anegative impact on the Australian agricultural economy destroying thousands of farm animals and agricultural lands. Queensland has faced many distractive floods. For example, In November 2010, aseries of floods have hit Queensland, Australia. Thousands had to be evacuated from cities and towns due to the floods. At least 90 cities and more than 200,000 people were affected. The damage was initially estimated at about $1billion before being raised to $2.38 billion. The estimated decline in Australia's GDP isabout $30 billion. As of March 2012, the 2010-11 Queensland floods had killed 33 people and left three others missing. To reduce flood damage in Queensland, we need to examine past flood records and develop more accurate flood estimation methods. With the unavailability of enough data within ungauged catchments where projects have to be established, there isaneed to have aconcrete and readily available and applicable model to predict and estimate flood quantiles within such areas. Thus, the research isdriven by the need to provide asecure and more accurate regional flood frequency analysis approach for both gauged and ungauged catchments. The RFFA approach method will inform the engineers and the general public in Australia on the security of their farms, houses, offices, infrastructures, and other amenities that are easily destroyed by floods, and hence early preparations are made towards evicting the epidemic. RESEARCH QUESTIONS 1. carry out the literature review on regional flood frequency analysis? 2. compare regression technique in the index flood method for Queensland? 3. validate test the new regional flood frequency method? 4. to recommend RFFE methods for Queensland? AIMS AND OBJECTIVES The main aim of flood frequency analysis isto understand the nature and magnitude of high discharge in rivers or rather to flood. The research to be conducted after the acceptance and go ahead of this research proposal will aim to understand the whole nature and behaviour of floods in catchment areas with the case study done in Australia. The analysis of the flood pattern will also aim to provide aprediction of flood frequency in both gauged and ungauged catchments for enacting projects. In order to achieve these aims, the proposal proposes the following objectives to be met during the research: 1. To understand the patterns and occurrences of floods in both gauged and ungauged catchments in Australia, itincorporates both humid and arid and semi-arid regions of Australia. 2. Using probability distributions, relate flood magnitudes and characteristics to their frequency of occurring in Australia's catchment regions. 3. To provide and propose working and reliable, easy to use regression models for flood quantile and frequency estimation. 4. Estimate different flood parameters in the proposed model of flood quantile and frequency estimation and prediction. 5. To predict flood frequency in the catchment that needs prediction at the moment of need using the proposed model for project design with the highest accuracy and confidence interval. RESEARCH METHODOLOGY Aresearch plan isan important part of the research and will be well selected in this research. The research plan isthe framework of techniques and methods used by aresearcher to collect and analyze all the variables in the proposed research (Kumar 2019). This allows the researcher to use methods that can solve the research problem perfectly. The research aims to understand the nature and magnitude of river overflow effluent disposal, or simply flood flooding through flood frequency analysis. The analysis will provide dynamic and better models of flood frequency assessment and forecasting in areas of interest for project planning that will emerge even in the event of aflood. Therefore, the research problem proposes aworking and reliable regional flood frequency analysis approach that can predict flood frequency with high accuracy results. Bayesian Generalized Least Squares (BGLS) regression approaches have been suggested and invested in by other research and have been shown to show good results with minimal errors. However, ROI approaches require more step-by-step parameter regression technique (PRT) and quantitative regression technology (QRT) to deliver better output results, which can be tedious, lengthy, and time-consuming. They want many regional features that are not limited to annual evaporation, such as rap rainfall intensity, forest, slope, area, median annual rainfall, and significant flood magnitude estimators. The probabilistic rational method (PRM) has also been suggested, and the implication is that flood frequency isnot abetter estimate model for high accuracy predictions compared to other models that have been shown to be better than the BGLS results. An L-moment-indicator model is also widely used for design flood and flood frequency analysis. The model assumes that the climatic and physical watershed characteristics that explain or predict have apermanent role to play in flood generation and that data are readily available, which they do not do overtime. Proposing Regional Flood Frequency Estimates (RFFE) Approach by Rahuman et al. (2019) have shown excellent results. Accessibility Simplifying and simplifying the procedure for flood assessment depends on available and accessible water supply data. Existing data on the physical and geographical features of the regions of interest have shown that this method gives better results compared to local models in estimating flooding. This model has been integrated into RFFE software for analysis, making iteasy to use and efficient with current technological advances. Therefore, research will focus on analyzing the RFFE approach to its improvement in order to obtain better results. The RFFE approach isdata-based, which extracts flood information from surveying sites and transfers itto insecure sites. The technology isbased on two main steps; Establishment of zones and development of regional estimation equations. The RFFE approach combines logical methodology, reaction-based techniques, and indicator flood methods to accurately estimate regional flood frequency estimates and provide better results. PROPOSED TIMELINE CONCLUSION Regional flood frequency analysis (RFFA) isan essential aspect among hydrologists, which creates a need for research. The research proposal on regional flood frequency analysis discussed above has described the floods in the introduction part with the need for the flood frequency prediction and estimation. The introduction has provided importance to the RFFA on water management, flood forecasters, resource planning, flood insurance studies, river ecological studies, flood plain management, and hydraulic structures. The proposal has also provided background information and reviewed different publications on RFFA, stating the RFFA approaches adopted in those publications, assumptions of the approaches, their contributions to the RFFA research area, and the drawbacks of the procedures. The proposal has also provided the proposed methodologies and research designs that will be used in the research towards obtaining the project objectives. The RFFA approach that the investigation will adopt has been outlined, and the methods of data collection and analysis are described. The proposed timeline for the project to ensure that the projected ispresented in time has also been established with the help of aGantt chart. The proposal's acceptance and go ahead will ensure that the projected iscompleted and published for reference on regional flood frequency analysis. REFERENCES Aaltonen, K. (2011). Project stakeholder analysis as an environmental interpretation process. International journal of project management, 29(2), pp.165-183. Baird, M.E., Suthers, I.M., Griffin, D.A., Hollings, B., Pattiaratchi, C., Everett, J.D., Roughan, M., Oubelkheir, K. and Doblin, M. (2011). The effect of surface flooding on the physical –biogeochemical dynamics of awarm-core eddy off southeast Australia. Deep Sea Research Part II: Topical Studies in Oceanography, 58(5), pp.592-605. Ball, J.E., Babister, M.K., Nathan, R., Weinmann, P.E., Weeks, W., Retallick, M. and Testoni, I.(2016). Australian Rainfall and Runoff-A guide to flood estimation. Burn, D.H. and Goel, N.K. (2000). The formation of groups for regional flood frequency analysis. Hydrological Sciences Journal, 45(1), pp.97-112. Cunnane, C. (1988). Methods and merits of regional flood frequency analysis. Journal of hydrology, 100(1-3), pp.269-290. Haddad, K. and Rahman, A. (2012). Regional flood frequency analysis in eastern Australia: Bayesian GLS regression-based methods within fixed region and ROI framework –Quantile Regression vs. Parameter Regression Technique. Journal of Hydrology, 430, pp.142-161. Haddad, K., Rahman, A., Weinmann, P.E., Kuczera, G. and Ball, J.(2010). Streamflow data preparation for regional flood frequency analysis: Lessons from southeast Australia. Australasian Journal of Water Resources, 14(1), pp.17-32. Hamed, K. and Rao, A.R. eds. (2019). Flood frequency analysis. CRC press. Kumar, R. (2019). Research methodology: Astep-by-step guide for beginners. Sage Publications Limited. López, J.and Franc és,F. (2013). Non-stationary flood frequency analysis in continental Spanish rivers, using climate and reservoir indices as external covariates. Hydrology and Earth System Sciences, 17, pp.3189-3203. Pandey, G.R. and Nguyen, V.T.V. (1999). Acomparative study of regression based methods in regional flood frequency analysis. Journal of Hydrology, 225(1-2), pp.92-101. Rahman, A. (2005). Aquantile regression technique to estimate design floods for ungauged catchments in south-east Australia. Australasian Journal of Water Resources, 9(1), pp.81-89. Rahman, A., Haddad, K., Kuczera, G. and Weinmann, E. (2019). Regional flood methods. Australian Rainfall and Runoff: AGuide To Flood Estimation. Book 3, Peak Flow Estimation, pp.105-146. Rahman, A., Haddad, K., Zaman, M., Kuczera, G. and Weinmann, P.E. (2011). Design flood estimation in ungauged catchments: acomparison between the probabilistic rational method and quantile regression technique for NSW. Australasian Journal of Water Resources, 14(2), pp.127-139. Sharp, N., Enzi, J., Fountoulakis, E., Lam, B. and Rabbior, M.C., International Business Machines Corp (2012). System and method for displaying gantt charts with other project management tools. U.S. Patent 8,245,153. Verma, J.P. (2012). Data analysis in management with SPSS software. Springer Science &Business Media. Wiltshire, S.E. (1986). Regional flood frequency analysis I:Homogeneity statistics. Hydrological Sciences Journal, 31(3), pp.321-333. Zaman, M.A., Rahman, A. and Haddad, K. (2012). Regional flood frequency analysis in arid regions: A case study for Australia. Journal of Hydrology, 475, pp.74-83.