Flood Control? An Evaluation of the Impacts of Flood Control and Drainage Projects in Bangladesh. Final Report

Transcription

1 Flood Control? An Evaluation of the Impacts of Flood Control and Drainage Projects in Bangladesh Final Report Margot Drost Master Thesis of Civil Engineering TU Delft

2 Flood Control? An Evaluation of the Impacts of Flood Control and Drainage Projects in Bangladesh Delft, August 2006 Final report M.Sc. graduation study Author: Margot Drost Graduation Committee: Prof.dr.ir. N.C. van de Giesen Dr.ir. M.W. Ertsen Dr.ir. R.J. Verhaeghe Ir. G. Pichel M.Sc. Thesis Delft University of Technology Faculty of Civil Engineering and Geosciences Section Water Resources Management In collaboration with DHV 2

3 Preface This is the final report of the M.Sc. graduation study of Margot Drost. The study was carried out at the section Water Resources Management of Civil Engineering at Delft University of Technology. In the context of this study I did an internship at DHV in Bangladesh for three months to set up the research. The report consists of four chapters that describe the research and most results are displayed in detail in the appendices. The readers without much time are referred to the first part of the chapter one and the conclusions in chapter four. Those who wish to know more about the hydrodynamic model that was used should read chapter two. For a thorough perception on the results, chapter three and appendix C are recommended. I would like to thank my graduation committee, Prof.dr.ir. N.C. van de Giesen, Dr.ir. M.W. Ertsen and Dr.ir. R.J. Verhaeghe for their assistance during these last months in Delft. During and prior to my stay in Bangladesh, I received much help from Ir. G. Pichel. I would like to thank Gerard for making it possible for me to work and live in Dhaka, where I learned a lot and met many interesting and helpful people. I would also like to thank the Project Director of the Emergency Flood Damage Rehabilitation Project, Mr. Kabir, for his support. He introduced me to many BWDB employees that helped me with my research. Also Dr. Rezaur Rahman and his staff helped me a lot during the difficult first phase of my study. I am also very grateful to all the people of CEGIS, IWM, SPARRSO and the Bangladesh Water Development Board who helped me with finding and getting my data. Upon my return in the Netherlands I had some difficulties with setting up my model. I received much help from the staff of Neele en Schuurmans, where I could stay for a week while building my hydrodynamic model. I would especially like to thank Olivier Hoes for introducing me there and Coen Nengerman for being as patient as he was. Apart from all the help I received concerning the contents of this report, I would also like to thank Mark Huizer for printing it. My final thanks go out to my parents for supporting me during my studies. 3

4 Summary Bangladesh is a floodplain country with an extensive river system. Annual monsoons result in flooding over a significant part of the country. This can vary from twenty percent, to sixty-five percent as in extreme cases such as the disastrous floods of Given Bangladesh s extremely high population density, there is a high pressure on land. This land scarcity forces the government to take flood mitigation measures. In the past this resulted in the implementation of many Flood Control projects with a total investment cost of about US$3 billion. Given this significant sum of money, this research aims to evaluate the effect of these measures has the damage actually been reduced? The Lower Atrai basin is selected for a case study area, in order to evaluate the flood control measures. The area is very flat and counts numerous depressions. The yearly floods have led to the implementation of many Flood Control and Drainage (FCD) Projects. The evaluation of their impacts requires a comparison between the situation both with and without the FCD projects. To make this comparison, simulations of flood events are made. Hydrodynamic models provide a suitable method for the simulation of a flood. The software that was used in order to develop this model is SOBEK Rural. It schematizes the basin by a onedimensional and two-dimensional part. The first section is a system of channels containing the river data i and is linked to the 2D grid containing the surface elevation per cell ii. For the purpose of this research, three scenarios are developed, on which two flood years are imposed to evaluate their performances. In this case, the hydrological conditions of the floods of 1998 and 2004 are used as inputs to the model. First, the existing situation as it occurred in 1998 is simulated; the calibration of the model is based upon this scenario. Subsequently, scenarios with the flood control projects as they were designed and without these measures are simulated. The available data of the 1998 flood was most suitable and this year is therefore used as a basis for the model. Subsequently, the 2004 flood is also simulated for scenario B and C. This year is considered as a moderate flood and is therefore a valuable addition to the simulation of Several factors limit the extent to which the real situation can be met. The main limitations will be discussed here. Firstly, the available data was limited. A coarse grid and missing data on discharges and cross sections complicated the study. A major deficiency of SOBEK is the rainfall input. Although it is possible to introduce precipitation of the grid, this is limited to one station. Consequently, the spatial and temporal variability are not taken into account. For the evaluation of the processes during the monsoon, the temporal variability is however not a determinative factor. Furthermore, the same hydrological input is used for each scenario. For comparison of the different cases, average values therefore satisfy. Another limitation concerns the rainfall-runoff processes. Since the model does not incorporate the groundwater hydrology in its calculations, a large component of groundwater flow is not taken into account. In the middle part of the monsoon, this is, however, acceptable. Since the water levels on the river are high, drainage is impeded anyway. Problems also occurred regarding the schematization. For the purpose of shortening the computation time, the grid was reduced to the area of interest. This caused some complications at the downstream borders, where the water has to leave the grid. This mainly affected the downstream portion of the study area. Therefore, the conclusions are based on the parts of the grid that are not affected by these inaccuracies. Furthermore, the schematizations of river-floodplain connections are merely a rough approximation of reality. Due to a limited amount of information the drainage network is much simplified. The model itself also aggravates the flood by artificially raising the water levels on the Atrai. The cause of this is not exactly clear, but it is very likely that an improvement of the schematization would enhance the model. These limitations have led to the consciousness that caution is bidden for the formulation of the conclusions. i Cross sections, slopes, friction etc. ii A so-called Digital Elevation Model (DEM) was available. 4

5 The study generated several conclusions on how the water system of the Lower Atrai was affected by the FCD measures. From the results of the simulations it can be concluded that the implementation of the FCD polders did not lead to an improvement of the flood situation in The floodplains were more or less equally flooded for the three scenarios. In addition the embankments obstruct the flood ways at the end of the flood season, causing the flood to last longer than it would in natural conditions. From the moderate 2004 flood some additional conclusions can be drawn. The implementation of FCD projects led to a considerable rise of the water levels on the Atrai. These high water levels cause an aggravation of the flooding conditions during August and September. The results show that the flood extent highly depends on the rainfall. Therefore the embanking of the rivers has not proven to be a very effective measure. On the contrary, when it causes a water level rise that impedes the drainage of the floodplain it has a negative impact on the flood parameters. These conclusions apply in case of relatively high water levels downstream. In these cases the embankments are thus merely a hindrance for the drainage of the floodplains. The central part of the Lower Atrai does not cultivate during the monsoon. This already indicates that the projects are not functioning as they should, since they were implemented in order to enable cultivation during the monsoon. From the evaluated floods, it can be concluded that it seems unwise to start cultivating this land in the monsoon. The following issues are suggested for further research. It would be interesting to investigate the effect of the implementation of the projects on the water system during low to moderate flood years. It would also be possible to improve the model itself, in order to enhance the accuracy of the results. Further calibration for the water levels at Chanchkair and a better description of the network can improve the results on the downstream part of the study area. Since the downstream water levels are important for the severity of the flooding, a better simulation of these water levels would be a valuable addition to the research. The adjustments to the schematization can also involve an improvement of the grid; by extending it and by using a smaller cell size. Regarding the method that was used, it can be concluded that it is an applicable method for similar studies that require understanding in the flooding mechanisms of an area. In view of the huge investments that have been made in the past and the low performance of these implemented projects, care should be bidden regarding future measures. Therefore it is highly recommendable to further develop the modeling applications for the flood-prone areas of Bangladesh in order to support the design of flood protection works. 5

6 Table of Contents Abbreviations... 7 Glossary of Terms Introduction The Lower Atrai basin Agriculture Projects Outline of the report Method Hydrodynamic model Data and schematization Scenarios Scenario A: Existing situation Scenario B: Design situation Scenario C: No embankments Input: Hydrological data The 1998 Flood The 2004 Flood Calibration process Results The 1998 Flood Scenario A Scenario B Scenario C The 2004 Flood Sensitivity Analysis Rainfall Floodplain River connections Floodplain roughness Analysis of the results Flood Extent Depth Velocity Duration Agricultural damage Limitations Discussion and Conclusions Conclusions Discussion on results Issues for further research References Appendices A. Input...41 B. Procedure...43 C. Results

7 Abbreviations BBS Bangladesh Bureau of Statistics BUET Bangladesh University of Engineering and Technology BWDB Bangladesh Water Development Board CEGIS Center for Environmental and Geographic Information Services DEM Digital Elevation Map FAP Flood Action Plan FCD Flood Control and Drainage FCDI Flood Control, Drainage and Irrigation FFWC Flood Forecasting and Warning Centre GoB Government of Bangladesh HYV High Yield Variety IWM Institute if Water Modeling WARPO Water Resources Planning Organization 7

8 Glossary of Terms Aus Pre-monsoon rice; grown in kharif 1 B. Aman Deepwater aman rice; grown in kharif 1 and 2 seasons Beel Floodplain lake Boro Rice variety that grows in the dry season. Planted in December/February and harvested in April/June. Khal Canal Kharif Humid period Kharif 1 Pre-monsoon cropping season; March - June Kharif 2 Monsoon cropping season; July - October Rabi Crop growing period; starting at the end of the humid period and lasting until the 250 mm soil moisture is exhausted Rabi crops Crops grown in the Rabi season T. Aman Transplanted aman rice; grown in kharif 2 season Thana Administrative unit also known as upazila; sub-district Zila District 8

9 1 Introduction Bangladesh is a floodplain country with an extensive river system. Every year during the monsoon season, a significant part of the country is flooded which varies from 20% up to 65% in extreme cases as the disastrous flood of These floods cannot be prevented and yearly inundations are even necessary for the country. The wetlands that are hereby created contain a large fish population, which is a direct source of income for 1.3 million fishermen and an important provider of protein for the people of Bangladesh. On top of that, the sediments of the rivers that are deposited on the floodplains during the floods are very important for the fertility of the agricultural land and to maintain the delta of Bangladesh. The region is characterized by a monsoon climate. Over 80% of the annual rainfall occurs during the months May to September. The dry seasons are sunny and relatively cool, with only occasional rainfall. Average annual rainfall ranges from 1600 mm in the North West region to 3200 mm in the eastern regions (FAP 2, 1992). The rivers of Bangladesh mostly originate from outside the country; 93% of their catchments is located in its neighboring countries. The major rivers are the Ganges, Brahmaputra and the Meghna. Together they have enormous catchments and are located in one of the highest rainfall regions in the world. When their discharge peaks coincide severe floods can occur. The most important processes that create an extreme flood in Bangladesh are the higher than normal inflow of the Ganges and Brahmaputra and higher than normal rainfall (Jakobsen, 2005). Bangladesh suffers from all types of floods: rain-fed floods, river floods, storm surges along the coast (cyclones) and even flash floods in the hilly areas. However, serious floods in a densely populated country like Bangladesh can of course cause huge damage. Given the extremely high population density there is a high pressure on land, forcing the government to take flood mitigation measures. In the past this resulted in the implementation of many Flood Control projects. One of the main goals hereby was to protect the crops and thus the food security of the country. Since the 1960s about 617 FCD projects are constructed, together covering 40% of the country! (Choudhury, 2004) The total investment costs are estimated at about US$3 billion. (ADB, 2005) Given this significant sum, this research aims to evaluate the effect of these measures; has the damage actually been reduced? For such an evaluation this research was set up and a case study was done. The Lower Atrai has been selected as a study area and for this basin an evaluation of flood control measures was carried out. The area is very flat and counts numerous depressions. Each year a large part of the basin is flooded and this has led to the implementation of many Flood Control and Drainage (FCD) Projects. It has a very complex water system with large river discharges and is located in one of the highest rainfall areas of the world. Much water passes through the basin and there is a dynamic interaction with the floodplains. A more detailed description of the basin is given in paragraph 1.1. What has been the impact of FCD Projects on the water system in the Lower Atrai? How have the main flood parameters (flooded area, flood depth, duration and velocities on the floodplains) changed due to the implementation of FCD polders. These parameters are the main indicators for agricultural damage! The evaluation of their impacts requires a comparison between the situation both with and without the FCD projects. It appeared to be unrealistic to compare flood events from before the construction of the polders to the situation afterwards; the floods as they occur are too complex to compare. To make this comparison, simulations of flood events are made. Because of the complexity of the basin it was a difficult task to give a good representation of the processes in the Lower Atrai. Another circumstance that complicated this procedure was the limited availability of data. However, due to a Digital Elevation Model with a relative high accuracy iii it is possible to simulate the overland flow on the floodplains. Because this overland flow is a large part of the total water balance, it is important to involve this in the simulation of the basin. A suitable iii The grid of the DEM that was used is coarse (300x300m), but it has a high accuracy regarding the elevation (1cm). 9

10 method that uses the topographical description for the simulation of a flood is to do this by means of a hydrodynamic model. Therefore the development of such a hydrodynamic model seemed the wisest approach to follow in order to find an answer to the research question. For the evaluation of the FCD projects three situations are simulated and analyzed. For this purpose, three scenarios are developed on which two flood years are imposed in order to evaluate their performances. The hydrological conditions of the floods of 1998 and 2004 are used as input to the model. First, the existing situation as it occurred in 1998 is simulated; on this scenario the calibration of the model is based. Subsequently, scenarios with the flood control polders as they were designed and without these measures are simulated. The available data of the 1998 flood was most suitable and this year is therefore used as a basis for the model. Subsequently, the 2004 flood is also simulated for scenario B and C (as shown in figure 1.1). This year is considered as a moderate flood and is therefore a valuable addition to the simulation of Finally, the different results of the simulations are evaluated by means of analyzing the flood parameters. Figure 1.1 illustrates the scheme of the research as it was conducted. Develop a hydrodynamic model in SOBEK Rural Calibration Input: Hydrology of the 1998 and 2004 Floods Scenario A: Existing Situation of 1998 Scenario B: Design Situation Scenario C: No Embankments Evaluation Flood Parameters Evaluation Flood Parameters Evaluation Flood Parameters Compare Results Figure 1.1 Layout Research Plan Conclusions: What has been the effect on the water system? 10

11 1.1 The Lower Atrai basin For this research a case study has been made of the Lower Atrai basin; the location is shown in figure 1.2. This area is suitable for the evaluation of FCD projects, because the area is frequently heavily flooded and therefore many projects were constructed. The Lower Atrai basin is the downstream part of the Atrai, before the river flows into the Jamuna. The basin lies in the North West Region of Bangladesh, which is one of the seven hydrological regions of Bangladesh. The region is bounded by the two major rivers Brahmaputra and Ganges, which come together just after the Atrai joins the Brahmaputra. The total catchment of the Atrai covers an area of km 2 ; twice the size of the selected area iv. The lower part of the basin is characterized by its flatness, the numerous water bodies, its yearly inundations and the many flood control projects that have been constructed since Figure 1.2 Location study area Elevation in the Lower Atrai varies from 5 m above reference v in the southeastern corner to 20 m at the foot of the hills and covers an area of approximately 5000 km2. Atrai river is a tributary to the Brahmaputra that flows into the Jamuna just upstream its confluence with the Ganges. The iv The catchment comprises about 4000 km 2 at Mohadebpur, the most upstream point of the study area. v The national reference level of Bangladesh (+PWD) is the mean sea level at Cox s Bazar. 11

12 study area lies within the districts Naogaon, Bogra, Sirajganj, Pabna, Natore and Rajshahi. Surrounded by two huge rivers and containing numerous depressions, the low-lying area is extremely vulnerable to flooding. The mean monthly river flows of stations on the main rivers are shown in table 1.1. Flooding over three meters regularly occurs over many parts of the Lower Atrai.(GOB Ministry of Water Resources, 1995). This is mainly caused by high water levels on the Brahmaputra that restrict drainage outfall. The area is a network of numerous beels (depressions) and khals (natural drains) and in the monsoon season the bigger part of the region is inundated. Station River Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Year Boral Rly Boral Naldanga Rly Barnai Little Naogaon Jamuna Mohadebpur Atrai Atrai Rly Atrai Nawhata Sib Table 1.1 Mean monthly river flows on the main rivers of the Lower Atrai basin Agriculture The Lower Atrai is a typical rural area where main economic activities take place at a local level. Most people depend on agriculture and the landless generally work as fishermen. Depending on the location of their land and the associated flood risk, farmers can harvest once or twice a year. The cultivable land can be classified vi according to the flood depths of normal inundations. The crops that are cultivated on the land vary with these inundation characteristics; High Yield Varieties of rice (HYV) require more investment and are only grown on higher situated lands. This illustrates that the poorest farmers live on lowlands; where one can only cultivate in the dry season, yields are lower and flood risks are high. The landless are the poorest people in the area and often live on the most flood-prone land; outside the embankments Projects Because the area is vulnerable to floods, many Flood Control and Drainage (FCD) projects have been implemented in the area. The funds for most projects were generally made available after major floods, by international donors. The first projects to be implemented were the Pabna I project vii in 1959 and the Raktadah Lahachura, an Early Implementation Project (EIP), in The Chalan Beel Project (starting in 1974) is situated in the largest depression of the region and is one of the most problematic projects. It consists of four schemes: Chalan Beel A, B, C & D. Difficulties arose from the start due to conflicts between the farmers and fishermen viii and farmers outside the protected areas. Another big project was launched in In order to avoid the difficulties as occurred in Chalan Beel, this project was an integrated study; the Serajgang Integrated Rural Development Project (SIRDP). Other projects worth mentioning here are the projects that started simultaneously around 1984 in the Nagor basin; Nagor Valley on the right bank and Nagor River the left bank of the river. These projects never really worked properly during moderate or high flood years. After completion of the downstream parts, they were extended upstream as sub-projects Upper Nagor Valley and Upper Nagor River. Another relatively new project is the Naogaon Polder. This is a unique project, since it is the only one vi Classification by Harza, vii Construction started after the 1955 flood. viii Fishermen suffered a decreased fishing area due to these projects. 12

13 constructed by higher standards. In other projects return periods are set at 20 years, whereas the Naogaon polder is set at a higher return period. The last one is the Barnai project; the right embankment of the Baranal that was completed in In the table below the main project features of the mentioned projects are summarized, costs are in local currency ix. After the devastating 1987 and 1988 floods, in the context of the Flood Action Plan, a regional study (FAP2) was executed in order to formulate a Regional Water Development Plan. In the FAP 2 report, flood regime planning by assigning flood areas is proposed in the Green River Project. However, no significant changes have been made yet in existing projects. Operation and maintenance are therefore each year a very costly procedure x and funds are generally inadequate. Figure 1.3 The Lower Atrai and its Flood Control and Drainage projects ix Million Thaka. x It is unclear how much is invested in operation & maintenance and rehabilitation over the years, but according to field officials projects are damaged regularly and the yearly O&M funds from the government are never adequate. Major rehabilitation projects are sponsored by international donors after big floods; as was the case in 1987, 1988, 1998 and is still going on from the 2004 flood. 13

14 District Project name Area (ha) FC D Emb. Implementation length (km) Start End Costs (MTh) Naogaon Naogaon Polder / WB Raktadah Lohachura ,3 GoB/WFP Upper Nagor Valley N/A** N/A** 26 88/89 94/95 N/A** EIP Nagor Valley / EIP Bogra Upper Nagor River N/A** N/A** 37 87/88 89/90 77 N/A** Natore Nagor River ,1 EIP Barnai N/A** N/A** 38, /95 834,4 WFP Chalan Beel A , / WFP Chalan Beel B / WFP Chalan Beel C WFP Rahsjahi Chalan Beel D * IDA Serajgonj SIRDP ,7*** ADB Pabna Pabna ,4* N/A** * incl.costs for irrigation scheme ** not known to the writer *** total project costs Table 1.2 FCD project summaries Donor 1.2 Outline of the report In this report the conducted research will be presented. In chapter 2 the approach of the research is presented and the hydrodynamic model will be explained. Afterwards the results of the model are presented in chapter 3. Subsequently the findings of the research will be presented and discussed in chapter 4. 14

15 2 Method To evaluate the constructed FCD projects a model was made. The program used to develop the model is SOBEK Rural. SOBEK is developed by WL Delft hydraulics in partnership with RIZA. In this chapter the background of this software is discussed and its possibilities and limitations for the purpose of this research are mentioned. Subsequently, the approach of the research is discussed. 2.1 Hydrodynamic model SOBEK Rural has a wide range of possible applications; from water quality control to the design of an irrigation system. The program exists of a few modules that can be selected depending on the requirements of the user. For the purpose of this study the module hydrodynamics is used, consisting of a Channel Flow (1D) and an Overland Flow part (2D). They represent the rivers and floodplains respectively. The 1-dimensional section is a system of channels containing the river data (cross sections, slopes, friction etc.) and is linked to a 2D grid containing the surface elevation per cell. This hydrodynamic model is based on the complete Saint-Venant equations; on the conservation of mass and momentum. For Channel Flow these are the following two equations: Continuity Equation: Momentum Equation xi : Q A + = q lat x t 2 Q ( Q / A) h Q Q + + ga + g = 0 2 t x x C RA With: Q = discharge [m 3 /s] A = wet cross section [m 2 ] q lat =lateral discharge in [ h = water level [m above reference] g = gravity acceleration = 9,81[m/s 2 ] C = Chezy coefficient [m 1/2 /s] R = hydraulic radius [m] For the Overland Flow, there are three equations xii : Continuity Equation: ζ ( uh) ( vh) + + = 0 ut x y u u u ζ u V Momentum Equations: + u + v + g + g + au u = 0 2 t x y x C h xi The terms describe respectively the inertia, the convection, the water level gradient and the bed friction; the wind friction is not taken into account. xii The 2D equations do not incorporate the turbulent stress; they are negligible in this context. The wall friction terms have been introduced to account for the added resistance that is caused by vertical obstacles, like houses or trees. 15

16 v v v ζ v V + u + v + g + g + av v = t x y y C h 2 0 With: u = velocity in x-direction [m/s] y = velocity in y- direction [m/s] 2 2 V = velocity = x + y [m/s] ζ = water level in m above reference a = wall friction coefficient [m -1 ] The numerical scheme used in SOBEK solves the Saint-Venant equations by means of a staggered grid. Water levels are defined at the nodes, while discharges are defined in the intermediate reaches. SOBEK reduces the time step automatically when necessary in order to guarantee the numerical stability. The model uses an integrated system of a 1D and 2D scheme. Regarding the momentum balance, the two systems remain strictly separated. Concerning the conservation of mass the volumes of the two systems are combined. Each node in the 1D channel is linked to a grid cell of the digital elevation model; as shown in the figure 2.1. When the 1D channel overflows, the spilled water flows towards the grid cell that is connected to this channel. From these connected cells the water spreads over the rest of the grid. This mechanism also works the other way around; from the connected cells the water flows back into the channel. Figure 2.1 Discretization of river and floodplains 16

17 2.2 Data and schematization Satellite images were available that show the flooding pattern during the 1998 and 2004 flood. Furthermore, information on breaches in the embankments and other project information were available in order to estimate the flood flows. An overview of used hydrological and topographic data is included in appendix A. The mentioned data were incorporated into the schematization of the basin. The river courses of the major rivers were assimilated into the 1D module as channels together with the accompanying cross sections. Where cross section data were not available, assumptions were made. Embankment heights were incorporated into the cross sections. In SOBEK Rural, embankment heights are introduced at the same level at both sides. Since the channel is situated on a grid cell, it is of no use to have different heights; no matter what side the river overflows, the spilled water ends up in the same grid cell. Where the river is only partly embanked, lateral channels are applied to simulate the links with the floodplain. Where breaches occur in the embankments, lateral channels with a weir are applied. Boundaries are set as five incoming discharges upstream and one water level boundary downstream limiting the drainage capacity. The available hydrological data were used as boundaries as well as checks for the calibration. Rainfall on the grid can only be introduced as a homogenous shower. However, the spatial and temporal variation of rainfall is high in the considered area. Therefore, the average precipitation xiii has been determined and used as input. This way the rainfall peaks are reduced. It is not as rainfall occurs in reality, but total volumes are more or less correct. This way the effect of heavy showers is not taken into account in the flooding pattern. 2.3 Scenarios After inserting the above data into the model, the model was calibrated. Stations where discharges and/or observed water levels are available are used as calibration points. The calibration process is described in paragraph 2.5. For the evaluation of the FCD projects three situations are simulated and analyzed. For this purpose, three scenarios are developed on which different conditions are imposed in order to evaluate their performances Scenario A: Existing situation This scenario approaches the situation as the flood occurred during the 1998 monsoon. All breaches from this season are incorporated and the model is calibrated by using the hydrological data of this period. For all breaches and floodplain links, the width and crest heights have been estimated by calibration. Obviously, these are only rough estimations. However, the schematization is physically representative for the 1998 situation. The scenario describes the realtime processes as good as possible and forms a basis for the evaluation of the FCD projects Scenario B: Design situation In this scenario no breaches are taken into account; the projects exist as they were designed. All breaches have been excluded from the calibrated schematization; the remaining features are the same as in scenario A. Naturally, the hydrological input is also the same. It is a purely hypothetical case, the situation never occurred. It is however an interesting reference to the existing situation. xiii The average precipitation was calculated by spatial interpolation, using all relevant and reliable rainfall stations. 17

18 2.3.3 Scenario C: No embankments In this scenario no interventions have taken place in the basin; no Flood Control and Drainage polders are present. In this case, there are no embankments whatsoever. The surface elevation of the DEM is taken as upper limit of the river cross section; as shown in the figure 2.2. This can be considered as the situation as it was before the implementation of the FCD polders. By comparing this scenario to the previous two, the effect of the flood control measures can be evaluated. Figure 2.2 "Assume no embankments" 2.4 Input: Hydrological data For reasons of data availability, the 1998 flood is used as a reference situation; therefore the hydrological data of this year are used. Although this is a flood that has exceeded the design standards of most projects, this season is used as a basis. In order to evaluate the performance of the water system under moderate circumstances, a flood with a much lower frequency has been simulated. For this purpose the flood year of 2004 has been used The 1998 Flood The heavy flood of 1998 was caused by a combination of high water levels on the Brahmaputra and very high rainfall on the Atrai basin. Synchronization of the flood peaks of the Ganges and the Brahmaputra aggravated the flood and high water levels on the Brahmaputra caused the long duration (Monirul Qader Mirza, 2003). The annual rainfall contributed to 47% of the yearly water balance. Heavy rainfall occurred in the beginning of July until the first week of September over the entire catchments of the Ganges and Brahmaputra. In the later part of September, the rainfall was less upstream but still high in Bangladesh. The Brahmaputra and Ganges xiv were above Danger Level for more than a month consecutively. The Brahmaputra recorded its highest ever peak discharge of m 3 /s. The Atrai River also crossed Danger Levels at various places. At Mohadebpur it was above the Dangel Level xv for four weeks, whereas it only crossed the Danger Level briefly at the Atrai Rly station. The discharges of the rivers in the Lower Atrai basin did not exceed the design flood return period of 20 years. However, the downstream water levels were extremely high due to the high flood peak of the Brahmaputra, as was the rainfall. In July the rainfall at Bogra was the highest recorded ever. xiv Known as respectively the Jamuna and Padma rivers in Bangladesh. xv Danger Levels in the Atrai basin differ from the design embankment height about 2 m! This indicates the state of the projects; they are not working at all as they were designed. 18

19 2.4.2 The 2004 Flood The flooding conditions in 2004 were much less disastrous as they were in Rainfall and incoming discharges that occurred can be regarded as moderate. However, due to the synchronization xvi of the flood peaks of the Ganges and the Brahmaputra at the end of July, the downstream water levels were again relatively high. Also in 2004 the contribution of the rainfall was nearly half of the total water balance. In the last week of September and the second week of October the region was hit by flash floods. Breaches in the Brahmaputra Right Embankment caused widespread flooding. 2.5 Calibration process After inserting the rivers with their cross sections and setting the boundaries xvii, the calibration process started. In order to calibrate the model, the calibration points as shown in figure 2.3 were important for verifications and adjustments. Mohadebpur Jote Bazar Atrai Rly Naldanga Singra Chanchkair Figure 2.3 Points that were used for the calibration of the model; at these locations water level or discharge stations exist xvi The synchronization of flood peaks is NOT an unusual event, it happens regularly. In 2004 it only occurred earlier than it normally does. xvii The discharge boundaries were developed by rating curves. 19

20 A first parameter that was unambiguous to adjust was the friction coefficient of the rivers. It is assumed that this is the same for all river stretches. The first simulations already showed that the most upstream boundary of the Atrai was not influenced by the downstream water levels. Since the slope of this stretch is determined by the cross sections, the roughness was adjusted by fitting the simulated water level to the observed water levels at this point. 1D2D links: the river floods Further calibration was more difficult and is open to many interpretations. In this part of the calibration process the spilling of the river towards the floodplains is subject of adjustments. First step in simulating these spills is to interpret the available data and translate this to 1D-2D links in the schematization. Maps, satellite images and project information xviii were used in this context. Where the river is only partly embanked, the links with the floodplain are simulated by lateral channels. Where breaches occur in the embankments, a weir is added to these lateral channels. During the first runs very high water levels and discharges were simulated. After comparing the simulated data with the observed water levels and rating curves the properties of the 1D2D links were adjusted. The main indications that identify which parameter to adjust are described below. - If the simulated water levels and discharges show a peak that is not noticeable in the observed data, it can be concluded that the river is not smoothed out enough. This means either the crest of an existing link should be lowered or an extra link to the polders has to be added. - If the simulated data are merely shifted, but the shapes of the graphs are similar, another option is to alter the width of the crest or the lateral channel. - The crest height and width can be adjusted the other way if the simulation shows that too much water is spilled. E.g. a peak is missing or too much water is spilled along the full hydrograph. Calibration Parameters: Q t Lower (or add) Crest Broaden Spill Figure 2.4 The calibration process: an example of the calibrating the 1D-2D interactions - If a calibration point is behind the confluence of two rivers the different hydrographs can tell which river to spill from. - It is important to study these hydrographs closely, since these describe the behavior of the system; this is what the model needs to simulate correctly! xviii This information is constituted out of many different sources; as found in reports or from field officers and observations in the field. 20

21 Since there were many buttons to push, this is by no means a unique solution. However, the topographical characteristics of the area were incorporated and the physical situation is described as good as possible. The processes in the system are therefore similar to the actually occurred situation. For the downstream portion it is important to simulate the water levels correctly, since the backwater effect has a considerable influence over the major part of the Lower Atrai. This first trajectory of simulating the connections between floodplains and rivers was to determine the spills towards the floodplains. The calibration of this part did not take into account the rainfall. This way, simulations were faster and therefore easier to adjust. For the calibration of this part only the first part of the monsoon season (July) was determinative; while flood water does not yet recede from the floodplains. 1D2D links: the flood recedes For the second part of the monsoon season, the process of return flow is important. As the floods recedes, the water stored on the floodplains finds it way toward the river. As soon as the water levels are low enough, this process starts. In this case the rainfall is an important factor for the calibration, since it contributes a huge amount of water to the storage on the floodplains. In this second stage of the calibration process the drainage connections are subject of adjustments. The major drainage canals of the FCD polders are translated into 1D-2D links in the schematization. Assumptions were made regarding the drainage canals, and orifices represent the (total) regulators of the polders. xix xix At the river reaches that are not entirely embanked, the connections to the floodplains as used for the river spills are representative. 21

22 3 Results In this chapter the results of the model will be described and evaluated. The descriptions of the results for the different scenarios are displayed separately; for the flood of 1998 and For each scenario first the results on the river itself are described (the one-dimensional part), followed by the results on the land (results of the grid). The results of scenario A during the 1998 flood are first compared to the observed values in paragraph 3.1.1, since this was the calibrated scenario. Subsequently scenario B is compared to scenario A in paragraph 3.1.2, followed by a comparison of all three scenarios in paragraph The two scenarios of the 2004 flood are compared in paragraph 3.2. The results of the scenarios are only shortly described, for a more detailed picture of you are referred to appendix C. Subsequently, the sensitivity analysis is described in paragraph 3.3. The outcome of the simulations is then analyzed in paragraph 3.4, where the flood parameters are discussed. Finally, the limitations are discussed. 3.1 The 1998 Flood Scenario A The devastating 1998 flood caused many breaches in the FCD projects. Consequently, the flood peaks were leveled off and much water was stored on the land. Derived from satellite images and project information, the 1998 conditions were simulated in this first scenario. For the location of the stations, see appendix A. In this scenario a reasonable representation of reality has been simulated. A bottleneck is the downstream water level. The model simulates higher water levels at Chanchkair than those actually occurred. This causes a higher backwater effect on the Baranal River as can be seen in the Naldanga graph (appendix C-I). All other calibration points show good fits as in the Atrai Rly station in figure 3.1. In figure 3.2 the flooded area is shown as derived from the satellite images and figure 3.3 shows the flood as it was simulated with SOBEK. It can be seen that on the northern parts, the higher and steeper lands, the model has difficulty with the correct simulation of the flood. The model also encounters difficulties in the downstream part; with the boundaries of the grid. Atrai Rly , ,00 12,00 Discharge (m3/s) ,00 8,00 6,00 4,00 Water Level (+m PWD) 200 2, , Rating curve Scenario A - Q Observed Water Level Scenario A - WL Figure 3.1 Results at Atrai Rly 22

23 Figure 3.2 Satellite Image - August 26th Figure 3.3 Scenario A August 26th 23

24 3.1.2 Scenario B In this simulation the FCD polders exist as they were designed. This means that no breaches occurred and therefore less water was stored on the floodplain. This can directly be seen in the hydrographs of the Atrai river; as shown in appendix C-II. The flood peaks in this simulation (in the middle stretch) are more then twice as high as the actual occurred discharges. The embankments also caused the water levels to rise one to two meters. The absence of breaches in the FCD projects, compared to the first scenario, did not lead to a considerable reduction of the flood extent. This is shown in figure 3.4, where the area that was protected from flooding in scenario B, compared to scenario A, is marked red. It appears that the flood depths are reduced, though only marginally. For a large area this reduction was at most 20 cm. For another significant area the flood depth was reduced between 20 and 50 cm. The precise distribution of this flood depth reduction can be seen in appendix C-II. Figure 3.4 Flood extent scenarios B versus A at September 13 th ; the red area was flooded in scenario A and protected in scenario B. 24

25 3.1.3 Scenario C In this scenario an attempt is made to simulate the natural conditions of the basin; before the interventions. This is done by changing the settings of the Overland Flow Module, so the river starts spilling whenever the water level exceeds the surface elevation of the grid instead of the embankment height. The simulation shows therefore a very dynamic interaction with the floodplains, since there are no banks acting as thresholds anymore. Since it is a dynamic river system, the cross sections can never be the same now as they were before the implementation of the FCD measures. The embankments have surely had their impact on the river. However, since data on the actual cross section before the implementation of the FCD polders are not available, the same river characteristics were used. In reality the river will burst its banks later than in this simulation, because natural rivers normally have elevated watersides. In this case, the surface elevation of the 300x300 m grid is determinative for the moment of overtopping. In this scenario the floodplains store an important part of the flood peaks. The hydrographs are therefore smoothed out and the peak discharges show to be much smaller. A summary of the results of the three scenarios is given in tables 3.1 and 3.2. The first shows the results on the rivers, the other summarizes the results on the floodplains. In figure 3.5 it can be seen that in September the basin is not significantly more flooded than it was for scenario A. The results (appendix C-III) also show that in October the basin is generally less flooded for this scenario; regarding both flood extent and flood depth. Since scenario A and B are at this point more dependent on the schematization xx of the draining conditions, these results are somewhat less reliable. It is however plausible that the draining conditions of the polders are more restricted then a situation without polders. Figure 3.5 Flood extent scenarios C versus A [October 13th]; the red area was flooded in scenario A, but not in scenario C and the green parts vice versa xx The draining conditions are schematized by one main canal per polder. This canal drains on the river through an orifice. 25

26 Station D.L. Emb. L. Max. Q A B C Max. WL Max. Q Max. WL Max. Q Max. WL MOH* 18,59 20, , , ,4 J.B. - 18, , , ,4 BGM - 16, , , ,7 NAL - 15, , , ,6 ATR 13,72 17, , , ,9 SIN - 14, , , ,4 CH - 14, , , ,4 Table 3.1 Results at stations on the river for all three scenarios Figure 3.6 Overview of the FCD polders and stations on the rivers Polder Max. depth (m) 15- jul Scenario A Scenario B Scenario C Flood extent (%) Max. Flood extent (%) Max. Flood extent (%) depth depth (m) (m) aug sep okt jul aug sep okt jul aug sep NAO 3, , , N.R. 3, , , C.B.D 4, , , N.V. 4, , , R.L. 2, , , C.B.C. 4, , , BAR 1, , , C.B.B 3, , , C.B.A 5, , , Total: Table 3.2 Results of all three scenarios on floodplains; the flood depth and flood extent is shown per polder 13- okt 26

27 3.2 The 2004 Flood For the evaluation of a moderate flood year, the hydrological data of 2004 are used as input. From the results in table 3.3 it can be read that the differences between the peak water levels of scenario B and C amount to three meters. The differences between the observed data and B are less, but also reach almost two meters. Concerning the flood conditions on the land, the results imply that the basin does not benefit from the embankments during the middle period of the monsoon. Figure 3.7 illustrates that the basin was flooded more in scenario B; with the implemented FCD projects. As can be seen in appendix C-IV, only the Naogaon polder benefited marginally with a flood depth decrease of less then 20 cm. On July the 15 th, in the beginning of the flood season, the embankment did protect the bigger part of the land. The Naogaon-1 and Chalan Beel C polders benefited the most of the embankment; they led to a decrease of flood depth up to one meter. Figure 3.7 Flood extent scenarios C versus B [September 13 th ]; the red area was flooded in scenario B, but not in scenario C and the green parts vice versa Station D.L. Emb. L. Max. Q Observed B C Max. WL Max. Q Max. WL Max. Q Max. WL J.B. - 18,4 N/A N/A - 16,4-14,2 BGM - 16,5-15,5-15,7-13,6 NAL - 15, , , ,3 ATR 13,72 17, , , ,5 SIN - 14, , , ,9 CH - 14,6-12,6-12,7-11,9 Table 3.3 Observed values and results of scenario B and C at stations on the rivers during the 2004 flood 27

28 3.3 Sensitivity Analysis Naturally, assumptions that are made in the schematization of the model have an influence on the results. The impact of such choices will be tested in the sensitivity analysis Rainfall In the model, the rainfall has been introduced as an average depth over the basin rather than the highly variable rain that actually occurred. In order to estimate the effect this had on the results, a simulation is run with one station - Bogra - as precipitation input. This had a significant effect on the flood extent. The maximum flood depth however, was only marginally affected. The average flood depths in polders as Naogaon, Chalan Beel A and Barnai are 20 to 30 cm higher due to the use of Bogra station. This is caused by the higher annual rainfall of Bogra compared to the average rainfall xxi. Regarding the flood extent, the high variability of the rain is clearly reflected in the flooded area at the evaluated time steps, as shown in table 3.4. Whereas the flood extent lied within a range of 17% when using the averaged values, this range is 76% when using only one station. On the other hand, this variability cannot be seen in the hydrographs of the two scenarios; the difference can hardly be seen. Flood extent (%) Average Bogra 15-Jul Aug Sep Oct Table 3.4 The flood extent for the different rainfall input Floodplain River connections In this case the connections between the river and the floodplains were changed in order to change the effects this imposes on the flooding regime. The thresholds of all connections were raised half a meter, which caused a reduction of spilled water. This led to small changes in the flood extent upstream; the flood extent increased marginally. The flood extent in Chalan Beel A on the other hand decreased. The flood depth also increased for the upstream projects (up to 10 cm), except for Chalan Beel A and B, where the flood depth decreased at some places with almost 30cm. It can be concluded that reduction in the floodplain spills leads to a larger flood extent and higher depth upstream. The downstream projects would most likely suffer to the same extent, but due to the lack of barriers in the grid the model suggests a different effect. xxi The Bogra station registered a total rainfall of 2493 cm in 1998 versus a depth of 1955 cm average over the Lower Atrai basin. 28

29 3.3.3 Floodplain roughness Another parameter that influences the situation on the floodplain is the friction coefficient. In order to test this effect, the manning coefficient of the floodplain was raised from 0.05 to 0.1. This is a high value and can be compared with a forest rather than agricultural area. This value is chosen just to indicate the effect of the change in friction on the behavior of the basin. The rough terrain behaves differently; the water is held on the floodplains. The results show that the flood extent increases by 6% on account of this change. In figure 3.8 it can be seen that this causes an increase of about 50 m 3 /s of the peak discharge at Atrai Rly station Discharge (m3/s) /07/ /07/ /07/ /07/ /07/ /08/ /08/ /08/ /08/ /09/ /09/ /09/ /09/ /09/ /10/ /10/ /10/ /10/1998 Naldanga C=0.1 Naldanga C=0.05 Atrai Rly C=0.1 Atrai Rly C=0.05 Figure 3.8 Results at Naldanga for the different Chezy coefficients 29

30 3.4 Analysis of the results The three scenarios of 1998 have a different development of the storage in time. As can be seen in the hydrographs (fig 3.10) and the storage in the system (fig 3.9), the flood water flows onto the land earlier in scenario C, but recedes sooner as well. Scenario A in its turn stores marginally more than scenario B. The peak storage of the system is highest for scenario A; the current situation. Considering the water levels on the Atrai, it can be read from the hydrographs that the water level difference between the scenarios is about three meters. If the projects would have been working correctly, the water levels would have risen with two meters compared to the water levels as they occurred now. On the other hand, they would be one meter lower if they were not implemented at all. More ambiguous results are the quantifications of the flood parameters. These will be handled separately below. Storage Atrai Rly V o lu m e ( m ^ 3 ) M illio n s A B C D i s c h a r g e (m 3 / s ) W a te r L e v e l (+ m P W D ) Jul 15 -Ju l 29 -Ju l 12 -A ug 26 -A ug 9-Se p 23 -S ep 7-Oct 21 -Oct 15/06/98 29/06/98 13/07/98 27/07/98 10/08/98 24/08/98 07/09/98 21/09/98 05/10/98 19/10/98 Figure 3.9 Total storage in the water system during the 1998 monsoon Scenario C - Q Scenario A - Q Scenario B - Q Scenario C - WL Scenario A - WL Scenario B - WL Figure 3.10 Simulation results at Atrai Rly station for all three scenarios in Flood Extent The total flood extent does not vary significantly for the three scenarios, as shown in table 3.5. Only in October the flooded area is quite a bit lower, because the flood water recedes quicker in scenario C. As could be expected, the Chalan Beel C, Nagor Valley and Raktadah Lohachura Polders had a significantly smaller flooded area in scenario B due to the malfunctioning of the projects as simulated in scenario A. Surprisingly, the Barnai project was less flooded in scenario C than it was in the other two scenarios. Flood extent (%) 15-Jul 14-Aug 13-Sep 13-Oct Scenario A Scenario B Scenario C Table 3.5 The total flood extent of all evaluated polders for

31 3.4.2 Depth Regarding the flood depth, table 3.6 shows the different results of the three scenarios. From the table it can be read that scenario B generally shows lower maximum and average depths then scenario A. Thus, when the project would have functioned this would be an improvement. The maximum flood depths do not vary significantly except for three projects. Nagor Valley has a significant higher max depth (+/-1m) for scenario A, whereas the maximum flood depths of Raktadah Lohachara and Barnai are 1 m higher for scenario C. Furthermore, the average depths prove to be higher without the implementation of FCD polders, except for three projects: Barnai, Chalan Beel B and D. Also flood extent in these cases is larger with implemented projects; they are clearly not an improvement to the situation. Depth A (m) Depth B (m) Depth C (m) Max Mean Max Mean Max Mean Naogaon Nagor River Chalan Beel D Nagor Valley Raktadah L Chalan Beel C Barnai Chalan Beel B Chalan Beel A Table 3.6 Average and maximum flood depths per FCD project for all three scenarios Velocity This flood parameter is only indicative, since the flooding patterns are just a rough approximation of reality. The used grid is too coarse, obstacles are not incorporated into the model and many assumptions have been made regarding breaches and floodplain roughness. Nevertheless, the table shows how the overland velocities have been affected. The maximum velocities are much higher for scenario A and B, of which scenario B shows the highest. For the different scenarios the distribution of the overland velocities were studied, but showed no significant changes. Only locally the velocities were much increased for the scenarios with the implemented FCD projects. This occurs of course at the points in the system where stream flows are concentrated; at breaches in the embankment and at drainage points. However, since these connections are very roughly schematized, no conclusions can be drawn regarding the magnitude of this local change. The overall average values prove not to have changed much. Polder Vel. A (m/s) Vel. B (m/s) Vel. C (m/s) Max Mean Max Mean Max Mean Naogaon Nagor River Chalan Beel A Nagor Valley Raktadah L Chalan Beel C Barnai Chalan Beel B Chalan Beel A Table 3.7 Average and maximum velocities per FCD project for all three scenarios 31

32 3.4.4 Duration The last flood parameter of relevance is the duration of the flood. This variable is evaluated pointwise; so-called history stations are used. These are points on the grid in SOBEK-1D2D that record the water depth as a function of time. Some representative locations are chosen in order to evaluate the duration. The results of 1998 are shown in appendix C-V. In the Naogaon graph (in figure 3.11) it can clearly be seen how the flood water of scenario C recedes quicker and the duration is therefore significantly smaller. This same pattern can be identified for all FCD polders and the left floodplain of downstream Atrai. History Station Naogaon Depth (m) Jun 29-Jun 13-Jul 27-Jul 10-Aug 24-Aug 7-Sep 21-Sep 5-Oct 19-Oct A3 C3 B3 Figure 3.11 Results at a history station on the land of the Naogaon project. The flood depth in time is represented Agricultural damage The flood parameters are related to the agricultural damage. First of all, the agricultural damage depends on the crop types. Each crop has a different value, a different cropping pattern and different reactions towards the flood parameters as depth and duration. The main crop in the area is rice. There are different types of rice in the study area, growing on the different land types. In figure 3.12 the distribution of agricultural land use in the Lower Atrai is shown. The most unambiguous flood parameter is the flooded area; the more flooded land, the greater the damage. More complicated is the relation to flood depth and duration. As an indication, figure 3.13 shows how the depth and duration are related to the damage factor. The effect of velocities is less studied, but farmers indicate that the sudden increase of flood water has a negative impact on the agricultural damage. In the Chalan Beel area xxii - the central part - no crops are grown in the monsoon season. It can therefore already be suspected that the FCD projects are not exactly working as proposed; otherwise farmers would cultivate an extra crop. Only a limited part of the basin is suitable for the cultivation of rice or other crop types during the summer months. For the evaluation of the agricultural damage only these parts are investigated. For the lands where the so-called Aman rice is grown during the monsoon the flood parameters have been assessed. Approximately 15% of the study area cultivates rice between August and November. At the xxii Chalan Beel is the Bangladeshi terminology for low lying area. 32

33 north-western part of the study area there are also some lands that cultivate the Aus variety; this type of rice is harvested in the beginning of the flood season. Figure 3.12 The distribution of agricultural land use; the figure shows that in the central part of the study area no crops are grown during the monsoon. Figure 3.13 Rice damage function 33

34 For the case of 1998, the simulation results are analyzed for those parts that are cultivated during the monsoon months. Regarding the flood extent, it can be concluded that the implementation of the projects led to a negative impact on the flooded area. In table 3.10 it can be seen that the flood extent was the smallest in Scenario C in most cases. Only in September scenario B showed a slight improvement. Compared to the pre-project situation the flood extent increased with 4 to 17%. If the projects would not have breached the flooded area could have been decreased by 1.7% in September. However, the rest of the monsoon the projects would be equally unsuccessful regarding the reduction of the flood extent. As far as the flood depth is concerned, the simulations also show negative impacts. The construction of the FCD polders in all cases led to an increase of the average flood depth. The differences were however only marginal; less than a cm! The assessment of the duration of the flood is done by the observation of history stations. In general the duration of the flood is longer with implemented projects; due to the (already mentioned) quicker withdrawal of flood water. However, the embankments do generally hold back the water three to six days. This causes a delay regarding the time of flooding, which can be important for the harvest of the Aus variety of rice xxiii. 15- Jul Flood extent (%) 14- Aug 13- Sep 13- Oct Scenario A Scenario B Scenario C Table 3.8 The flood extent of 1998 for the land that is cultivated during the monsoon per scenario 15- Jul Average flood depth (m) 14- Aug 13- Sep 13- Oct Scenario A Scenario B Scenario C Table 3.9 The flood depth of 1998 for the land that is cultivated during the monsoon per scenario xxiii The Aus variety of rice is harvested in the beginning of July. 34

35 3.5 Limitations Every model is only an approximation of reality. For this model there are also several factors that limit the extent to what the real situation can be met. First of all, the grid that is used has a big cell size (300x300m). This is favorable for the computation time, but at certain points it causes an unsatisfying accuracy. Another substantial limitation of the model is the fact that the 1D-channels do not form a barrier to the grid. Where the river with its embankments should form a barrier for overland flow, this is not the case in the model! This could have been avoided by integrating the embankment in the gird instead of the cross-sections. But because of the coarse cell size it is not possible to assimilate the embankments into the grid itself. Besides this inaccuracy, it is also laborious. Another major deficiency of SOBEK is the treatment of rainfall. It is possible to introduce precipitation of the grid, the input is however limited to one station. Since the variability of the rain is of importance for the flooding regime, this is a significant restraint. Apart from the constraints of the software, also different problems occurred. Due to time constraints, the computational effort was kept as low as possible; this is of course on the expense of the accuracy. In this case, the grid was reduced to the area of interest. This causes however some complications at the downstream borders of the grid. Also the draining facilities from the polders to the rivers could have been schematized more accurately. The regulators and drainage canals are schematized roughly with a limited amount of information and the drainage network is consequently much simplified. Instead of simulating all the regulators, only a few main connections are chosen. The exact process of this returned flow is difficult to model, but this approximation is satisfying for the purpose of this study. Since the draining of the polders is highly dependent on the Atrai water levels, the schematization is of less importance. One of the major limiting factors in this study is the data. Cross sections of certain river stretches, existing embankment levels and even some incoming discharges are missing. Therefore, assumptions had to be made in these cases. Also the rating curves that were generated from the observed discharges usually deviates from the observed data. For 2004, little observed discharge data were available. Also a large part of water level data was missing for many stations in Regarding the data that are given, some remarks can be made on their reliability. For example, if we look upon the 1998 water level data at Bagmara, a strange detail appears; in July and the beginning of August the water levels at the outflow of the Barnai River on the Fakirni are higher then those upstream. However, the hydrograph of the Barnai River at Nawhata shows a similar pattern to those further downstream at Naldanga. This suggests a more likely backwater effect and leads to the conclusion that the observed data at Bagmara are unreliable. 35

36 4 Discussion and Conclusions In this final chapter the findings of the research are presented. After the conclusions, the method that was used and its limitations are discussed and issues for further research are proposed. 4.1 Conclusions The study that was carried out led to the following ideas on how the water system of the Lower Atrai was affected by the FCD measures. The 1998 flood played a central role in the study. From the results of the simulations it can be concluded that the implementation of the FCD polders did not lead to an improvement of the flood situation. Generally spoken it can be said that the floodplains were more or less equally flooded for the three scenarios. It can even be said that the FCD polders obstructs the water masses at the end of the flood season. They hereby cause the flood to last longer then it would in natural conditions. From the 2004 flood, of which hydrological conditions can be considered moderate, some additional conclusions can be drawn. The implementation of FCD polders led to a considerable rise of the water levels on the Atrai, especially during the peak discharges. These high water levels cause an aggravation of the flooding conditions during August and September. In general the following concluding remark can be made. It seems to be justified to say that rainfall contribution were somewhat underestimated for the design of the FCD projects. Since the flood extent highly depends on the rainfall, the embanking of the rivers is not a very effective measure. On the contrary, when it causes a water level rise that impedes the drainage of the floodplain it has a negative impact on the flood parameters. In case of relatively high water levels downstream, it can be concluded that the FCD projects did not have a positive impact. Due to the confinement of the river, the water levels on the Atrai are raised even more. In these cases the embankments are thus merely a hindrance for the drainage of the floodplains. The main part of the study area only cultivates the boro variety of rice; this crop is not cultivated during the flood season. It already indicates that the projects as they at present exist are not functioning as they should. This can be seen as an opportunity loss; the FCD projects were implemented in order to enable cultivation during the monsoon. However, this currently is not the case for large parts of the study area; accordingly no damage is reduced by the embankments. On the contrary, the land would only benefit from the fertile river sediment that would be deposited on the floodplains. For the evaluated floods, it seems to be unwise to start cultivating the central area (or Chalan Beel area) of the Lower Atrai during the monsoon. Regarding the method that was used, it can be concluded from the finding of this research that it is an applicable method for similar studies that require understanding in the flooding mechanisms of an area. In other parts of Bangladesh it might be interesting to conduct similar studies in order to evaluate the Flood Control measures that have been taken. The method can also be a valuable tool for the assessment of the effects of future measures. However, the limitations of the model have to be taken into account. The approach is not suitable when a precise description of the interaction between river and floodplain is required. 4.2 Discussion on results Based on the findings of this research, some remarks need to be made regarding the results and method of this study. In the preceding chapter the limitations are mentioned of SOBEK, the available data and consequently the schematization. These strong limitations of the used method have led to the consciousness that caution is bidden for the formulation of the conclusions. However, the context of this study is such that a value judgment is more important than quantitative accuracy. Since many limitations apply for all three scenarios, the systematical errors that are caused by the limitations play only a limited role of importance. Some limitations of the method that is used are subject to discussion. One of the main limitations that we have seen is the way rainfall is handled in the software that is used. The spatial and temporal variability cannot be 36

37 incorporated into the model. Instead of using the specific rainfall that is representative for a part of the basin, the averaged rainfall over the whole basin is used. In the preceding chapter it was shown that this spatial and temporal variability highly influences the flood extent; rainfall provides a large contribution to the flooding conditions (flood extent) that are modeled. This issue was mainly a problem for the verification of the model. However, the main interest of the model is to evaluate the reaction of the system in the course of the monsoon; not necessarily the reaction on a specific shower. Furthermore, this same hydrological input is used for each scenario. For comparison of the different cases, average values of rainfall input satisfy. The results of the 2004 flood also made clear that the rainfall-runoff processes are not representative for the actual events that occur during the monsoon. From the many depressions that the basin counts, the floodwater can only recede through the soil. Since the model does not incorporate the groundwater hydrology in its calculations, a large component of groundwater flow is not taken into account. Therefore it is difficult to simulate the basin under low to moderate conditions; especially at the beginning and end of the flood season. In the middle part of the monsoon, the flooding conditions are however acceptable. Since the water levels on the river are high, drainage is impeded anyway. Besides the limitations of the approach that is taken, the schematization that was used has some inaccuracies. First of all, the model aggravates the flood by artificially raising the water levels on the Atrai (at Chanchkair). The cause of this is not exactly clear, but it is very likely that an improvement of the schematization would enhance the model. A better description of the draining conditions would be a valuable addition for simulation of the return flow from the water stored on the floodplains towards the river. It is however questionable to what extent this would matter for the evaluated flood years. For low to moderate flood conditions this would be more important. The downstream boundaries were also a source of errors; regarding the outflow of the floodwater from the grid. This mainly affected the downstream portion of the study area where the grid was kept small in order to reduce the calculation time of the simulation xxiv. Therefore, the conclusions are based on the parts of the grid that is not affected by these inaccuracies. Based on the finding of this research, it is not likely that the main conclusions would alter by solving this boundary problem. Due to a good topographic description (one cm accuracy in height) of the area it was possible to simulate the mechanisms of overland flow. The coarse cell size however, does cause inaccuracies. This was mainly the case for scenario C, where the surface elevation of the DEM acts as the threshold for the river spills. For the surrounding land of the river it would therefore be an improvement to the model if the land elevation would be described more accurately; in other words with a smaller cell size. Due to the coarse grid, obstacles on the floodplains (as for example roads) could not be incorporated into the schematization. However, they do cause retention of water on the floodplains. 4.3 Issues for further research The cases that were studied involved high downstream water levels. It would be interesting to investigate the effect of the implementation of the polders on the water system during low to moderate hydrological conditions. For further research, it is recommendable to extend the scope by including low to moderate flood years. Regarding the simulations that were done, a few adjustments to the schematization can be made in order to improve the accuracy of the results. Further calibration for the water levels at Chanchkair can improve the results on the downstream part of the study area. Since the xxiv The computational effort is highly dependent on the size of the grid. In the schematization that was used the calculation time was approximately fifteen hours; by extending the grid this would increase significantly! 37

38 downstream water levels are important for the severity of the flooding, a better simulation of these water levels would be a valuable addition to the research. The adjustments to the schematization can also involve an improvement of the grid; by extending it and by using a smaller cell size. In view of the huge investments that have been made in the past and the low performance of these implemented projects, care should be bidden regarding future measures. Therefore it is highly recommendable to further develop the modeling applications for the flood-prone areas of Bangladesh in order to support the design of flood protection works. 38

39 5 References ADB AND WORLD BANK, Floods in Bangladesh: Damage and Needs Assessment and Proposed Recovery Program. Dhaka: ADB. ADNAN, S., DALAL-CLAYTON, B. AND HUGHES, R., Floodplains or Flood Plans? A Review of Approaches to Water Management in Bangladesh. Nottingham: Russell Press. AHMAD, Q.K., CHOWDHURY, A.K.A., IMAN, S.H. AND SARKER, M., Perspectives on FLOOD Dhaka: The University Press Limited. BALA, S.K., HAQUE, A., KHAN, S.A., RAHMAN, R. AND SALEHIN, M., Investigation of hydrological aspects of flood 2004 with special emphasis on Dhaka city. Dhaka: BUET. BHUIYAN, M.S., JAKOBSEN, F., HOQUE, A.K.M.Z. AND PAUDYAL, G.N., Evaluation of the Short-Term Processes Forcing the Monsoon River Floods in Bangladesh. Water International, 30 (3), p CHADWICK, M. AND DATTA, A., Water Resource Management in Bangladesh: A Policy Review. Available from: [Accessed ] CHOUDHURY, G.A., In: A.N.H HOSSAIN, ed. National Workshop on Options for Flood Risk and Damage Reduction in Bangladesh, 7 9 September 2004 Dhaka. Dhaka: Masro Printing and Packaging Ltd, xx-xx. CHOWDHURY, O.H., ISLAM, K.M.N. AND BHATTACHARYA, D., Flood 1998: A Rapid Rural Appraisal. Dhaka: ADB. CHOWDHURY, J.U., RAHMAN, M.R. AND SALEHIN, M., Flood control in a floodplain country: experiences in Bangladesh. Dhaka: BUET. DUTTA, D.,HERATH, S. AND MUSIAKE, K., A mathematical model for flood loss estimation. Journal of Hydrology, 277 (2003), p FLOOD ACTION PLAN, North West Regional Study (FAP2) Dhaka: Government of Bangladesh. IFCDR, Performance Evaluation of FCD/FCDI Projects During 1998 Flood. Dhaka: BUET. ISLAM, M.R., BEGUM, S.F., YAMAGUCHI, Y., AND OGAWA, K., The Ganges and Brahmaputra rivers in Bangladesh: basin denudation and sedimentation. Hydrological Processes, 13, ISLAM, M. AND SADO, K., Flood hazard assessment in Bangladesh using NOAA AVHRR data with geographical information system. Hydrological processes, 14,

40 JAKOBSEN, F., ZEAUL HOQUE, A., PAUDYAL,G. AND BHUIYAN, M., Evaluation of the Short-Term Processes Forcing the Monsoon River Floods in Bangladesh. Water International, 30 (3), KHALEQUZZAMAN, Md., Recent floods in Bangladesh: possible causes and solutions. In: Natural Hazards, p MONIRUL QADER MIRZA, M., Three Recent Extreme Floods in Bangladesh: A Hydro- Meteorological Analysis. Natural Hazards, 28, TINGSANCHALI, T. AND KARIM, M.F., Flood hazard and risk analysis in the southwest region of Bangladesh. Hydrological processes, 19, Available from: WERLE, D., MARTIN, T.C. AND HASAN, K., Flood and Coastal Zone Monitoring in Bangladesh with Radarsat ScanSAR: Technical Experience and Institutional Challenges. John Hopkins APL Technical Digest, 21 (1),

41 Appendices A. Input The data that were used as input for the development of the model are described below. First of all, the network had to be schematized. In order to do this, the locations of the rivers were used from the shape file as they were supplied from CEGIS. In these rivers the cross-sections that are known were incorporated. For some river (stretches) no cross-sections are known; for these branches assumptions are made. Figure 1 shows the locations where cross section data are available. In order to represent the 1998 situation the breaches were incorporated into the model. Besides the breaches that allow water to pass to the FCD polders, the floodplains need to be connected with the rivers. The locations of the connections between the 1D-channels and the grid are shown in figure 2. Figure A-1 The locations of the available cross sections Besides the physical data that are necessary for the structure of the network, the hydrological data need to describe the situation regarding rainfall and the rivers. In figure 3 the stations that were used are shown. 41

42 Breaches Floodplain Figure A-2 The Lower Atrai basin and the breaches that were present during the 1998 flood Figure A-3 The hydrological stations in the Lower Atrai that were used 42

43 B. Procedure During this research I encountered some difficulties that influenced the course that I took. In the following pages the choices that were made are explained and the differences between the original research plan and the path that was taken are clarified. Before the field work in Bangladesh, a research plan was proposed. This original plan (figure 1) was kept very broad for reasons of flexibility. During the months in Bangladesh, a more specific focus was chosen. First of all a study area was selected after discussion with local professionals in the water sector. In the original approach the research was aimed on the comparison of flood events; a physical and socio-economical analysis. This scope was however too wide and was soon narrowed down to a mere physical analysis. Four flood years were selected and of these flood years as much data as possible was collected. In the end it proved to be rather complicated to compare the floods as mentioned, due to big differences between the events. Practical knowledge on GIS, image processing and related software Gathering basic information and survey of data availability: geography, demography, hydrology etc. Selection appropriate flood events for case study (first analysis) Damage analysis of flood events Analysis of physical impacts of measures: on floods and inundations Field research Analysis of socio-economic impacts Developing a model: linking physics to socio-economics Appointing evaluation criteria; comparing social impacts Evaluation of impacts Figure B-1 The original research plan at the start of the study 43

44 First, the scheme to describe the water system was aimed to be kept simple. An attempt was made to use a box-model to describe the processes in the basin. In order to do this, the system had to be divided into smaller parts. The separate parts are regarded to as reservoirs, which means that each box is represented by a water level and a volume. The water level of the reservoir - together with those of the connected reservoirs - determines how much water flows from one to another. It also determines in combination with an area-elevation curve how much water is stored at the moment. The virtue of such a model would be that it takes into account the total water balance rather than focus on the river. Due to this different approach, the floodplain flow would play the determinative role as it does in reality. Data at the different points would determine the relations between the reservoirs and consequently it works roughly as a black box. Accordingly the applicability of the method relies on the data availability and was therefore limited. Additionally, another problem occurred while trying to set up the model. There was the dilemma of the size of the boxes. It is questionable to what extent the observed areas can actually be represented by a reservoir. In order to investigate this issue, a small part of the basin was analyzed. Due to limited data availability it proved very complicated to state anything on the error induced by this assumption. The limitation of the available data made it thus impossible to work with this method. The approach that was afterwards taken on was to develop a hydrodynamic model. The advantages of this method over the box-model, is that it places the processes in a physical context. This way the model is not the black box as the reservoir approach in fact is. Another virtue is the fact that the results can be made visible. The limitations of this approach that was finally used are mentioned in the main report. Of course this approach required knowledge on the specialized software. Thanks to help received from the employees of Neele en Schuurmans, I was able to get acquainted with the software easily and could build the model rather quick. The final approach that was used is again summarized in the figure below. Develop a hydrodynamic model in SOBEK Rural Calibration Input: Hydrology of the 1998 and 2004 Floods Scenario A: Existing Situation of 1998 Scenario B: Design Situation Scenario C: No Embankments Evaluation Flood Parameters Evaluation Flood Parameters Evaluation Flood Parameters Compare Results Conclusions: What has been the effect on the water system? Figure B-2 The working method at it was followed for the hydrodynamic model 44

45 C. Results In this chapter the results of the simulations will be presented. For each scenario the results at relevant places on the river will first be analyzed. To start with an indication on the discharges is given for the middle stretch of the river. Then the water levels at the relevant stations are compared. The locations of the stations are shown in figure C-1. Subsequently, the results of the situation on the floodplains are presented. The flood extent and the flood depths at certain time steps are shown. The resulting pictures are then analyzed compared with the satellite images or the other scenarios. First of all, the simulations of 1998 will be shown. Starting with the flood as it occurred, scenario A will be compared with the observations of Subsequently, the results of scenario B and C are presented and compared to the findings of scenario A. Finally, the results of the simulation of the 2004 conditions are run for scenario B and C; with and without FCD polders. Figure C-1 Locations of the relevant stations 45

46 I. Scenario A [1998] This scenario represents the situation as it occurred and therefore the results should match with the flood of In the following pages the model results will be checked with the given data. The discharges and water levels are used for calibration and are therefore compared first. At Mohadebpur the simulated water levels are dependent on the roughness coefficients and the cross sections that are used. As can be seen in the figure the calibration has resulted in a good fit. Mohadebpur Water Level (+m PWD) Jul 15-Jul 29-Jul 12-Aug 26-Aug 9-Sep 23-Sep 7-Oct 21-Oct Observed Water Level Scenario A Figure C-2 The water levels at Mohadebpur show a good fit for scenario A with the observed values Afterwards the model was calibrated by using the Atrai and Naldanga stations. The model also shows a good fit at Atrai Rly as can be seen in figure 3. Water levels as well as discharges correspond with the observed data and the rating curve. At Naldanga the calibration was less successful. Figure 4 shows water level differences up to two meters. This is mainly caused by the backwater effect due to elevated water levels at Chanchkair. 46

47 Atrai Rly Discharge (m3/s) Water Level (+m PWD) Jul 15-Jul 29-Jul 12-Aug 26-Aug 9-Sep 23-Sep 7-Oct 21-Oct Rating curve Scenario A - Q Observed Water Level Scenario A - WL Figure C-3 Observed and simulated water levels and discharges at Atrai Rly; a calibration point Naldanga Discharge (m3/s) Water Level (+m PWD) Jul 19-Jul 2-Aug 16-Aug 30-Aug 13-Sep 27-Sep 11-Oct 25-Oct Rating Curve Scenario A - Q Observed WL Scenario A - WL Figure C-4 Observed and simulated water levels and discharges at Naldanga Rly; a calibration point 47

48 The main bottleneck of the model is the water level at Chanchkair. The water level here has a significant influence upstream. The elevated water levels at this point are caused by the relatively shallow cross sections here. Due to this rise of the bottom level, the water is backed up. As can be seen in graphs 6 to 11, this backwater effect has significant influence upstream. The water levels at Chanchkair clearly have a significant effect on the water levels upstream. The hydrographs of the different stations illustrate that the backwater effect is much stronger in the model then was actually observed. Regarding the Atrai River, Jote Bazar seems to be the upstream limit for this backwater effect in both cases; all stations downstream have a similar shape of the hydrograph as at Chanchkair. In the observed data the backwater seems to affect the Atrai and Naldanga stations as well; though to a lesser extent. The Baral river is not much affected in the observed data at Malanchi, but in the model the backwater effect is clearly present. The fact that the downstream water levels at its confluence with the Atrai are much higher in the model is the reason for this difference. Chanchkair Figure C-5 The sideview of the lower strech of the Atrai shows the bottleneck of the model; at Chanchkair the water levels are artificially raised 48

49 Chanchkair Water Level (+m PWD) Jul 15-Jul 29-Jul 12-Aug 26-Aug 9-Sep 23-Sep 7-Oct 21-Oct Observed Water Level Figure C-6 Observed and simulated water levels at Chanchkair Scenario A Malanchi Water Level (+m PWD) Jun 22-Jun 29-Jun 6-Jul 13-Jul 20-Jul 27-Jul 3-Aug 10-Aug 17-Aug 24-Aug 31-Aug Observed Simulated Figure C-7 Observed and simulated water levels at Malanchi 49

50 Nawhata Water Level (+m PWD) Jun 22-Jun 29-Jun 6-Jul 13-Jul 20-Jul 27-Jul Observed Water Level 3-Aug 10-Aug 17-Aug 24-Aug 31-Aug 7-Sep Figure C-8 Observed and simulated water levels at Nawhata Simulated Water Level 14-Sep 21-Sep 28-Sep 5-Oct 12-Oct 19-Oct 26-Oct Bagmara Water Level (+m PWD) Jun 22-Jun 29-Jun 6-Jul 13-Jul 20-Jul 27-Jul Observed Water Level 3-Aug 10-Aug 17-Aug 24-Aug 31-Aug 7-Sep Figure C-9 Observed and simulated water levels at Bagmara Simulated Water Level 14-Sep 21-Sep 28-Sep 5-Oct 12-Oct 19-Oct 26-Oct 50

51 Simulated Water Level (+m PWD) Jul 15-Jul 29-Jul 12-Aug 26-Aug 9-Sep 23-Sep 7-Oct 21-Oct Atrai Jote Bazar Nawhata Bagmara Naldanga Malanchi Chanchkair Figure C-10 Simulated water levels at seven stations Observed Water Level (+m PWD) Jul 15-Jul 29-Jul 12-Aug 26-Aug 9-Sep 23-Sep 7-Oct 21-Oct Atrai Jote Bazar Nawhata Bagmara Naldanga Malanchi Chanchkair Figure C-11 Observed water levels at seven stations 51

52 Besides the water levels and discharges that are known at the hydro meteorological stations, satellite images are used for the verification of the model. The cell size of these images is much smaller then the cell size of the Digital Elevation Model xxv. This contributes to the differences in pattern between the images and simulation results. The calculation grid is marked as a grey area on the pictures. The general flooding pattern is the same, but there are clearly differences. The boundaries of the grid downstream contribute to the inaccuracies that occur in this part of the basin. These lands are flooded in the model where they were not flooded in the actual situation. This problem is mainly caused by the downstream boundaries of the grid itself xxvi and can be solved by extending the grid. The central part shows the best resemblance with the satellite images. Since the central part of the basin is the low-lying depression, this is the area which is flooded most of the monsoon. Consequently it shows less dynamics and this causes the model to give a reliable output regarding flood extent. Further from the river the results shows less resembling patterns. These steeper parts are less static and more subject to the effect of rainfall and floodplain characteristics. Accordingly the flood extent as simulated deviates more from the satellite images. In figures 14, 17 and 20 the differences between the flood extent as simulated and measured are highlighted. From the satellite images and hydro meteorological stations it can be concluded that the results can be used for the purpose of this study. Main problems are the calculation grid boundaries for the lowest south-western part and the rainfall input for the higher northern lands. The results are therefore not accurate, but the model simulates the flooding mechanism correctly. The results can be used for comparison of the different scenarios, since the mentioned problems are equally Figure C-12 Satellite Image August 26 th xxvxxv 300x300m of the DEM versus 50x50m of the Radar images. This 36 times as small! xxvi The water cannot leave the grid at the borders by itself. Therefore canals are modeled with boundaries and connections to the grid. Despite this measure, the water has trouble leaving the grid. 52

53 Figure C-13 Scenario A results for August 26 th Figure C-14 The area that was flooded in the satellite image of August 26 th, but not in the simulation of scenario A is marked red and the green parts vice versa 53

54 Figure C-15 Satellite Image September 10 th Figure C-16 Scenario A results for September 10 th 54

55 Figure C-17 The area that was flooded in the satellite image of September 10 th, but not in the simulation of scenario A is marked red and the green parts vice versa Figure C-18 Satellite Image September 17 th 55

56 Figure C-19 Scenario A results for September 17 th Figure C-20 The area that was flooded in the satellite image of September 17 th, but not in the simulation of scenario A is marked red and the green parts vice versa 56

57 II. Scenario B In this simulation the FCD polders are incorporated as they were designed; no breaches occurred. The Atrai did not spill her banks in this scenario; the water did not reach the design levels of the embankments. In the following pages the results at the water level and discharge stations on the rivers are compared with the results of scenario A. Afterwards the situation on the land is analyzed. As can be seen in the figure below, the changes imposed on the scenario did not have an effect on the water levels at Mohadebpur. In other words; the polders without breaches did not cause a backwater curve that affects the water levels at the most upstream point. Further downstream at the Atrai Rly station it can be seen how the absence of breaches influences the discharge. The flood peaks in this scenario B are more then twice as high as those in scenario A! The graph shows that there is a huge amount of water stored in the polders in the actual situation. The same counts for the results at the southern branch of the Atrai, as shown in figure.. The differences in water levels for the two stations are two and one meter at Atrai and Naldanga station respectively. Further downstream at Chanchkair the water levels are more or less the same as they were in scenario A; the difference is at most 30 cm. Also on the Baral River (Malanchi), the water levels are more or less the same. On the southern diversion of the Atrai River there were more breaches and therefore the differences between the two scenarios are bigger. At Nawhata and Bagmara the water levels are about one meter higher without the breaches that occurred in scenario A. At Jote Bazar this did influence the water levels; the water levels were raised one meter at the peak levels. Mohadebpur Water Level (+m PWD) Jul 15-Jul 29-Jul 12-Aug 26-Aug 9-Sep 23-Sep 7-Oct 21-Oct Figure C-21 Results 1998 for Mohadebpur Scenario A Scenario B 57

58 Atrai Rly Discharge (m 3/s) Water Level (+m PWD) Jul 15-Jul 29-Jul 12-Aug 26-Aug 9-Sep 23-Sep 7-Oct 21-Oct Figure C-22 Results 1998 for Atrai Rly Scenario A - Q Scenario B - Q Scenario A - WL Scenario B - WL Naldanga Discharge (m3/s) Water Level (+m PWD) Jul 15-Jul 29-Jul 12-Aug 26-Aug 9-Sep 23-Sep 7-Oct 21-Oct Scenario B - Q Scenario A - Q Scenario B - WL Scenario A - WL Figure C-23 Results 1998 for Naldanga 58

59 Chanchkair Water Level (+m PWD) Jul 15-Jul 29-Jul 12-Aug 26-Aug 9-Sep 23-Sep 7-Oct 21-Oct Scenario A Scenario B Figure C-24 Results 1998 for Chanchkair Nawhata Water Level (+m PWD) Jul 15-Jul 29-Jul 12-Aug 26-Aug 9-Sep 23-Sep 7-Oct 21-Oct Scenario A Scenario B Figure C-25 Results 1998 for Nawhata 59

60 Malanchi Water Level (+m PWD) Jul 8-Jul 15-Jul 22-Jul 29-Jul 5-Aug 12-Aug 19-Aug 26-Aug Scenario A Scenario B Figure C-26 Results 1998 for Malanchi Bagmara Water Level (+m PWD) Jul 8-Jul 15-Jul 22-Jul 29-Jul 5-Aug 12-Aug 19-Aug Figure C-27 Results 1998 for Bagmara 26-Aug 2-Sep Scenario A 9-Sep 16-Sep 23-Sep 30-Sep Scenario B 7-Oct 14-Oct 21-Oct 28-Oct 60

61 Jote Bazar Jul 15-Jul 29-Jul 12-Aug 26-Aug 9-Sep 23-Sep 7-Oct 21-Oct Figure C-28 Results 1998 for Jote Bazar Scenario B Scenario A As can be seen in figures 29 to 32, the Lower Atrai basin is also for this scenario heavily flooded. To indicate the differences with scenario A, an analysis is done on the simulated floods between the scenarios. Concerning the flood extent, both scenarios are compared for four time steps xxvii in figure 33 to 35. The area that was flooded in scenario A but protected in scenario B is marked red and the land that was flooded in scenario B where it was not affected for scenario A is colored green. Regarding the water depth, figure 37 to 40 illustrates the differences between scenario A and B at September 13 th. It shows that for a large area the water depths on the land the polders of scenario B reduced by 5 to 20 cm compared to the current situation. For another significant part the depths decreased 20 to 50 cm. Similar patterns were observed at the other time steps. The results of the grid are summarized in table C-1. xxvii Due to the output of SOBEK, other time steps then the satellite images has been chosen for the evaluation of the scenario s. 61

62 Polder Velocity (m/s) Depth (m) Time of wetting (d) Flood extent (%) Max Mean Max Mean >0.5m >1m 15-Jul 14-Aug 13-Sep Naogaon Nagor River Chalan Beel A Nagor Valley Raktadah Lahachura Chalan Beel C Barnai Chalan Beel B Chalan Beel A Table C-1 The results of the flood parameters of scenario B summarized 13- Oct Total: Figure C-29 Simulation results of scenario B on July 15 th 62

63 Figure C-30 Simulation results of scenario B on August 14 th Figure C-31 Simulation results of scenario B on September 13 th 63

64 Figure C-32 Simulation results of scenario B on October 13 th Figure C-33 Flood extent scenario B versus A [July 15 th ]; the red area was flooded in scenario A, but not in scenario B and the green parts vice versa 64

65 Figure C-34 Flood extent scenario B versus A [August 14 th ]; the red area was flooded in scenario A, but not in scenario A and the green parts vice versa Figure C-35 Flood extent scenario B versus A [September 13 th ]; the red area was flooded in scenario A, but not in scenario B and the green parts vice versa 65

66 Figure C-36 Flood extent scenario B versus A [October 13 th ]; the red area was flooded in scenario A, but not in scenario A and the green parts vice versa Figure C-37 Flood depths of scenario B versus A [July 15 th ]; the red area had a deeper flood depth in scenario B, and the red parts were more heavily flooded in scenario A 66

67 Figure C-38 Flood depths of scenario B versus A [August 14 th ]; the red area had a deeper flood depth in scenario B, and the red parts were more heavily flooded in scenario A Figure C-39 Flood depths of scenario B versus A [September 13 th ]; the red area had a deeper flood depth in scenario B, and the red parts were more heavily flooded in scenario A 67

68 Figure C-40 Flood depths of scenario B versus A [October 13 th ]; the red area had a deeper flood depth in scenario B, and the red parts were more heavily flooded in scenario A 68

69 III. Scenario C In this scenario an attempt is made to simulate the natural conditions of the basin; before the interventions. This is done by changing the settings of the Overland Flow Module, so the river starts spilling whenever the water level exceeds the surface elevation of the grid instead of the embankment height. The simulation shows a very dynamic interaction with the floodplains, since there are no banks acting as thresholds anymore. This can also be read on the hydrographs, in the figures below. It can be seen that the graph is even more smoothed out; more water is stored on the floodplains. The water levels differences are highest upstream at Jote Bazar and lowest downstream, at Chanchkair. The differences vary from 0.5m to 4m! The simulation results of the grid are summarized in table 2. Mohadebpur Water Level (+m PWD) Jul 15-Jul 29-Jul 12-Aug 26-Aug 9-Sep 23-Sep 7-Oct 21-Oct Figure C-41 Results 1998 for Mohadebpur Scenario A Scenario B Scenario C 69

70 Atrai Rly Discharge (m3/s) Water Level (+m PWD) Jul 15-Jul 29-Jul 12-Aug 26-Aug 9-Sep 23-Sep 7-Oct 21-Oct Scenario C - Q Scenario A - Q Scenario B - Q Scenario C - WL Scenario A - WL Scenario B - WL Figure C-42 Results 1998 for Atrai Rly Singra Discharge (m3/s) Water Level (+m PWD) Jul 15-Jul 29-Jul 12-Aug 26-Aug 9-Sep 23-Sep 7-Oct 21-Oct Scenario C - Q Scenario B - Q Scenario A - Q Scenario C - WL Scenario B - WL Scenario A - WL Figure C-43 Results 1998 for Singra 70

71 Naldanga Discharge (m3/s) Water Level (+m PWD) Jul 19-Jul 2-Aug 16-Aug 30-Aug 13-Sep 27-Sep Figure C-44 Results 1998 for Naldanga Scenario C - Q Scenario B - Q Scenario A - Q Scenario C - WL Scenario B - WL Scenario A - WL Jote Bazar Water Level (+m PWD) Jul 15-Jul 29-Jul 12-Aug 26-Aug 9-Sep 23-Sep 7-Oct 21-Oct Scenario C Scenario B Scenario A Figure C-45 Results 1998 for Jote Bazar 71

72 Chanchkair Water Level (+m PWD) Jul 15-Jul 29-Jul 12-Aug 26-Aug 9-Sep 23-Sep 7-Oct 21-Oct Figure C-46 Results 1998 for Chanchkair Scenario C Scenario A Scenario B Velocity (m/s) Depth (m) Flood extent (%) Polder Max Mean Max Mean 15-Jul 14-Aug 13-Sep Naogaon Nagor River Chalan Beel A Nagor Valley Raktadah Lahachura Chalan Beel C Barnai Chalan Beel B Chalan Beel A Table C-2 The results of the flood parameters on the grid 13- Oct Total:

73 The simulation results of the flooded area for scenario C are shown in figure 47 to 50. Figure 51 to 54 show the differences regarding the flood extent between scenario A and C. The green parts are flooded in scenario C, but not in scenario A and for the red parts it applies the other way around. Figure 55 to 58 illustrates the effects of scenario C on the water depths on the land. The red parts had higher depths in scenario C, whereas the blue parts were more heavily flooded in scenario A. Figure 58 shows the differences at the end of the flood season (October 13th), where the flood depths are higher for the current situation as simulated in scenario A. The results of October are however less reliable, since the model is in this case more dependent on the schematization xxviii of the draining conditions. It is however plausible that the draining conditions of the polders are more restricted then a situation without polders. Figure C-47 Simulation results of scenario C on July 15 th xxviii The draining conditions are schematized by one main canal per polder. This canal drains on the river through an orifice. 73

74 Figure C-48 Simulation results of scenario C on August 14 th Figure C-49 Simulation results of scenario C on September 13 th 74

75 Figure C-50 Simulation results of scenario C on October 13 th Figure C-51 Flood extent scenario C versus A [July 15 th ]; the red area was flooded in scenario C, but not in scenario A and the green parts vice versa 75

76 Figure C-52 Flood extent scenario C versus A [August 14 th ]; the red area was flooded in scenario C, but not in scenario A and the green parts vice versa Figure C-53 Flood extent scenario C versus A [September 13 th ]; the red area was flooded in scenario C, but not in scenario A and the green parts vice versa 76

77 Figure C-54 Flood extent scenario C versus A [October 13 th ]; the red area was flooded in scenario C, but not in scenario A and the green parts vice versa Figure C-55 Flood depths of scenario C versus A [July 15 th ]; the red area had a deeper flood depth in scenario C, and the red parts were more heavily flooded in scenario A 77

78 Figure C-56 Flood depths of scenario C versus A [August 14 th ]; the red area had a deeper flood depth in scenario C, and the red parts were more heavily flooded in scenario A Figure C-57 Flood depths of scenario C versus A [September 13 th ]; the red area had a deeper flood depth in scenario C, and the red parts were more heavily flooded in scenario A 78

79 Figure C-58 Flood depths of scenario C versus A [October 13 th ]; the red area had a deeper flood depth in scenario C, and the red parts were more heavily flooded in scenario A 79

80 IV. Results 2004 In this scenario the 2004 hydrological conditions are used as input can be considered as a moderate flood year for the study area. Simulations are done for the cases with complete projects and without any projects; scenario B and C. Where observed data are available these are added to the graphs. The graphs of Atrai and Naldanga show that the rivers peaks are very large because the embankments prevent river spills. The peak discharges are doubled and quadrupled for the Atrai and Naldanga respectively! At these points the differences in water level are 4 and 3 meters. At Chanchkair and Jote Bazar the water levels differences are a bit smaller, but still considerable; respectively 1 and 2 meters. Atrai Rly 2, , Discharge (m3/s) 1,600 1,400 1,200 1, Water Level (+m PWD) Jul 15-Jul 29-Jul 12-Aug 26-Aug 9-Sep 23-Sep 7-Oct 21-Oct Rating curve Scenario B - Q Q C Observed Water Level Scenario B - WL Scenario C - Q Figure C-59 Results 2004 for Atrai Rly 80

81 Naldanga Discharge (m3/s) Water Level (+m PWD) Jul 15-Jul 29-Jul 12-Aug 26-Aug 9-Sep 23-Sep 7-Oct 21-Oct Rating Curve Scenario B - Q Scenario C - Q Observed Water Level Scenario B - WL Scenario C - WL Figure C-60 Results 2004 for Nalndaga Chanchkair Water Level (+m PWD) Jul 15-Jul 29-Jul 12-Aug 26-Aug 9-Sep 23-Sep 7-Oct 21-Oct Figure C-61 Results 2004 for Chanchkair Observed Water Level Scenario B Scenario C 81

82 Jote Bazar 17.0 Water Level (+m PWD) Jul 15-Jul 29-Jul 12-Aug 26-Aug 9-Sep 23-Sep 7-Oct Observed Water Level Scenario B Scenario C 21-Oct Figure C-62 Results 2004 for Jote Bazar Concerning the flood extent, the results of scenario B are compared to the available satellite images. Looking at figure 63 to 66, it can directly be seen that the deviations are big. Analysis shows that in July the flooded area as simulated is not much bigger or smaller compared to the images; the flood extent is however differently distributed. For the situation in September, this is a different picture. The simulation shows an overall much larger flooded area. This can be seen in figure 68. The green areas are not flooded in the simulation whereas they were flooded according to the satellite images. For the red areas this is the other way around. Of course some of the areas were flooded due to the breaches. However, this only explains a small part of the green area and does not explain why much of the land is flooded in the simulation whilst it was not in the satellite image. The spatial variation of rainfall that we cannot incorporate in the model can also play a role of importance. This can explain some of the green parts we see in the figures. Another problem that occurs with the hydrodynamic model relates to the rainfallrunoff processes. Water in the depressions of the floodplains normally returns to the rivers through the soil, as a slow component of the runoff. This is not possible in SOBEK, where it is trapped in the low-lying areas. Analysis of the flood depths shows that the basin was more heavily flooded for scenario B at July the 15 th. At August 14 th and September 13 th the area is more seriously flooded for scenario B; the flood depths are up to 1 m higher. Only the Naogaon and Raktadah Lahuachara polders succeeded in protecting their land. Figure 72 shows that the embankment protected the whole central area of the basin at October 13 th. The downstream part was again more seriously flooded for scenario B. 82

83 Figure C-63 Satellite image July 23 rd Figure C-64 Satellite image September 9 th 83

84 Figure C-65 Simulation results of scenario B for 2004, July 23 rd Figure C-66 Simulation results of scenario B for 2004, September 9 th 84

85 Figure C-67 The area that was flooded in the satellite image of July 23 rd, but not in the simulation of scenario A is marked green and the red parts vice versa Figure C-68 The area that was flooded in the satellite image of September 9 th, but not in the simulation of scenario A is marked green and the red parts vice versa 85

86 Figure C-69 Flood depths of scenario C versus B [July 15 th, 2004]; the blue area had a deeper flood depth in scenario C, and the red parts were more heavily flooded in scenario B Figure C-70 Flood depths of scenario C versus B [August 14 th, 2004]; the blue area had a deeper flood depth in scenario C, and the red parts were more heavily flooded in scenario B 86

87 Figure C-71 Flood depths of scenario C versus B [August 14 th, 2004]; the blue area had a deeper flood depth in scenario C, and the red parts were more heavily flooded in scenario B Figure C-72 Flood depths of scenario C versus B [October 13 th, 2004]; the blue area had a deeper flood depth in scenario C, and the red parts were more heavily flooded in scenario B 87

88 V. History stations In the following pages the results at the history stations in 1998 will be discussed. The locations of these history stations are shown in figure 71. The graphs of some history stations were selected on account of the relevance of their results. Figures 72 to 77 show the flood depth as a function of time at history station 928, 919, 944, 940, 908 & 934 respectively; these stations are located on agricultural land that is cultivated during the monsoon. At station 928 the flood depth and duration were decreased by the FCD polders, for 919, 944 & 908 no significant change could be seen regarding the flood depth, but the duration was shorter for scenario C. For station 940 the flood depth was higher than before the implementation of the projects. At this last point, the flood depth stayed below 80cm in scenario C, where it nearly reached 120cm for scenario A and B. At history station 934 the implementation of the FCD projects did not lead to any significant changes at all. For the central area of the study area, graph 78 and 79 illustrate that the implementation of the FCD projects led to some extent to a decrease of the flood depth, but also aggravated the flood situation at the end of the flood season in history station 923 among others. Figure C-73 The locations of the history stations 88