PREDICTION OF SEDIMENT YIELD IN UNGAUGED
CATCHMENTS FOR DAM-RESERVOIR
MANAGEMENT
September 2014
HENG SOKCHHAY
Prediction of sediment yield in ungauged catchments for
dam-reservoir management
ダム管理のための未観測流域における土砂
生産量の予測
A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Engineering
Global Center of Excellence (GCOE)
Special Doctoral Course on Integrated River Basin Management Interdisciplinary Graduate School of Medicine and Engineering
University of Yamanashi
Advisory Committee: Prof. Tadashi Suetsugi (Chairman) Prof. Yasushi Sakamoto
Assoc. Prof. Hiroshi Ishidaira Assoc. Prof. Yutaka Ichikawa
September 2014 Heng Sokchhay
Preface
This dissertation is a result of a three-year Ph.D. program conducted at the Hydraulic Engineering Laboratory, Interdisciplinary Graduate School of Medicine and Engineering, University of Yamanashi, Japan, from October 2011 to September 2014. The study was supported by the University of Yamanashi Global Center of Excellence Program and the Japanese Government (Monbukagakusyo: MEXT).
The content of this dissertation is fully or partially based on works either published or forthcoming publication by Heng Sokchhay. Although these works have co-authors, the substantial contribution (intellectual, research and writing) was that of the first author, and any specific contributions of co-authors have either not been presented in the dissertation or have been duly acknowledged. All peer-reviewed papers accomplished by Heng Sokchhay during the three-year period are:
Heng, S., Suetsugi, T., 2014. Development of a regional model for catchment-scale suspended sediment yield estimation in ungauged rivers of the Lower Mekong Basin. Geoderma 235–236, 334–346. In support of Chapter 3
Heng, S., Suetsugi, T., 2014. Comparison of regionalization approaches in parameterizing sediment rating curve in ungauged catchments for subsequent instantaneous sediment yield prediction. Journal of Hydrology 512, 240–253. In
support of Chapter 3 and 6
Heng, S., Suetsugi, T., 2014. Prediction of sediment yield in an ungauged basin under the impact of cascade dam-reservoirs development. Journal of Japan Society of Civil Engineers, Ser. B1 (Hydraulic Engineering) 70 (4), I_1–I_6. In support of Chapter 3
and 7
Heng, S., Suetsugi, T., 2013. Regionalization of sediment rating curve for sediment yield prediction in ungauged catchments. Hydrology Research (in press). DOI:10.2166/nh.2013.090. In support of Chapter 3
Heng, S., Suetsugi, T., 2013. An approach to the model use for measuring suspended sediment yield in ungauged catchments. American Journal of Environmental Science 9 (4), 367–376. In support of Chapter 4 and 6
Heng, S., Suetsugi, T., 2013. Simulation of suspended sediment load using data-driven models: a comparative study. In: Fukuoka, S., Nakagawa, H., Sumi, T., Zhang, H. (Eds.), Advances in River Sediment Research. CRC Press/Balkema, Leiden, the Netherlands, pp. 91–98.
Heng, S., Suetsugi, T., 2013. Investigation on applicability of data-driven models in ungauged catchments: sediment yield prediction. Earth Resources 1 (2), 37–47. Heng, S., Suetsugi, T., 2013. Coupling singular spectrum analysis with artificial neural
network to improve accuracy of sediment load prediction. Journal of Water Resource and Protection 5 (4), 395–404.
Heng, S., Suetsugi, T., 2013. Using artificial neural network to estimate sediment load in ungauged catchments of the Tonle Sap River Basin, Cambodia. Journal of Water Resource and Protection 5 (2), 111–123.
Heng, S., Suetsugi, T., 2013. Estimating quantiles of annual maximum suspended sediment load in the tributaries of the Lower Mekong River. Journal of Water and Climate Change 4 (1), 63–76. In support of Chapter 5 and 6
Acknowledgements
I would like to express the deepest appreciation to all people who either directly or indirectly involved in the successful completion of this dissertation as well as my Ph.D. study. My high gratitude goes to my advisor, Prof. Tadashi Suetsugi, for his constant guidance, valuable advice and warm encouragement during the entire period of my study. I am also thankful to co-advisors and committee members: Prof. Yasushi Sakamoto, Dr. Hiroshi Ishidaira and Dr. Yutaka Ichikawa for their insightful comments and suggestions.
My sincere appreciation is extended to the University of Yamanashi Global Center of Excellence Program and the Japanese Government (Monbukagakusyo: MEXT) for financial support, and the Mekong River Commission (MRC) for necessary data provision. I owe a very important debt to Mr. Soukaseum Phichit, Database Manager of MRC, for his positive cooperation, generous support and help to applying for the “MRC Non-commercial Data Use License” and obtaining datasets.
I have greatly benefited from the Group Research Seminar – Hydroscience and Hydraulic Engineering group of the study program. Active discussions with all group members during this weekly seminar gave me a lot of ideas to extending scope and improving quality of my research. For this, I am deeply grateful to Prof. Futaba Kazama, Prof. Junko Shindo, Dr. Kei Nishida, Mr. Naoki Miyazawa, Dr. Kazuyoshi Souma, Dr. Jun Magome, Dr. Tetsuya Sano, Dr. Vishnu Prasad Pandey, Dr. Kazuhiro Kakizawa, and many other researchers and colleagues, for their questions, comments and suggestions on my research presentations.
It gives me great pleasure in acknowledging the academic support and administrative assistance from professors and staffs of the University of Yamanashi throughout my Ph.D. study. Many thanks are also presented to my seniors, classmates and juniors in the Global Center of Excellence Program for their sharing, encouragement and kindly help which made my life in Japan comfortable and enjoyable.
Last but not least, I would like to express my profound gratitude to all my family members, relatives and friends for their love, encouragement and support. My special thanks go to my older brother, Mr. Heng Sokheang, for sharing me his personal techniques on writing thesis smoothly and taking responsibility on my affair in the family (in Cambodia) during the three-year period of my study in Japan.
Abstract
Sediment transport information is truly important for sustainable development of water resources system, but observations of this key hydrological variable are lacking for rivers in many parts of the world, especially in developing countries. The Lower Mekong Basin drains part of four Southeast Asian countries: Cambodia, Lao PDR, Thailand and Vietnam. Thailand’s monitoring infrastructure is more developed than the others and therefore, this country is very rich in historical records. On the other hand, the majority of water-related development projects including 96 hydropower dam-reservoirs are planned for the Cambodian, Laotian and Vietnamese portions of the basin, where records of sediment are very poor. There is a need for establishing of a feasible model that allows estimates of sediment yield in such poorly gauged or ungauged areas.
The aim of this research is to develop the Mekong-Feasible Regional Suspended Sediment Yield (Mekong-FReSSY) model for a more efficient management of dam-reservoirs and other water resources system in ungauged areas of the Lower Mekong Basin. The model composes of three modules. Module 1 can be used to estimate monthly specific suspended sediment yield time series (SSYs), with the required input of water
discharge time series. Module 2 and 3 can be respectively used to estimate mean annual/monthly suspended sediment yield (SSYa/SSYm) and quantiles of annual maximum
suspended sediment yield (SSYx*), without requiring any input of hydro-meteorological
information. After development, the Mekong-FReSSY model was coupled with an existing sediment trapping efficiency technique to quantify the potential impact of planned cascade dam-reservoirs on suspended sediment transport in the upper Sre Pok River catchment, southeast of the basin, which is ungauged with respect to sediment data.
In developing Module 1, an ideal regionalization methodology was determined to parameterize sediment rating curves in ungauged catchments for subsequent SSYs
prediction. A comparison of three regionalization approaches (physical similarity, regression and spatial proximity) was conducted in 16 gauged sites. The highest quality results were provided by the physical similarity approach in which a donor catchment was selected in accordance with the minimum catchment similarity index computed by the multidimensional scaling method with catchment descriptors normalized to a specific range. The overall predictive accuracy was further improved through a combined formulation of physical similarity and regression. By applying this ideal methodology, a satisfactory result of SSYs prediction in ungauged catchments was obtained in all 16
modeled catchments. Relied upon this successful validation, the regional SSYs model
(Module 1) was established by integrating the 16 locally calibrated sediment rating curves. Module 2 is an alternative to Module 1 when water discharge data are not available. It is an advanced version of the traditional catchment area-sediment yield model. First, catchment area (A0), the model predictor, was substituted by catchment area with a slope
gradient greater than 15% (A15). By doing so, an analysis based on datasets from 17
gauged sites showed that SSYa in ungauged catchments could be predicted with much
in the case using A15 and A0, respectively. Second, a unit mean annual sedimentograph
(USG) was proposed to monthly distributing SSYa predicted from the regression on A15. A
USG has 12 ordinates representing the ratios of SSYm and SSYa for 12 months. It allows
estimates of SSYm when SSYa is known. The assessment using the same database showed
that SSYm in ungauged catchments could be predicted with fairly good accuracy, showing
65% satisfactorily modeled catchments. Relied upon these validation results, the regional
SSYa/SSYm model (Module 2) was established based on datasets from the 17 gauged sites.
Module 3 is an additional tool to Module 1 or 2. A reservoir planning or management that includes information on extreme sediment transport offered by Module 3 would have more potential to prepare for unexpected events. A logical framework was firstly adopted to extract annual maximum suspended sediment yield (SSYx) from a set of discontinuous
time series. Due to data insufficiency for conducting frequency analysis, a regional linear regression model was then applied to estimate SSYx. After discarding outliers, frequency
analysis was performed on each estimated SSYx series to compute its quantiles (SSYx*) for
return periods between 2 and 100 years. An analysis based on datasets from 24 gauging stations showed that A15 is also a better predictor than A0 in predicting SSYx* in ungauged
catchments. Using A15 as the model predictor, SSYx* could be predicted with reasonably
good accuracy, showing MAPE results between 36.56% and 56.20%. Lastly, relied upon these validation results, the regional SSYx* model for each return period (Module 3) was
established by relating A15 to SSYx* using a database from the 24 gauged sites.
Quantifying the impact of planned cascade dam-reservoirs on suspended sediment transport in the upper Sre Pok River catchment was performed by using the Mekong-FReSSY model (Module 1) to estimate the catchment sediment production and the modified Brune’s equation to compute the sediment trapping efficiency. There are 13 years of water discharge data available to estimate sediment yield at the catchment outlet. The development plan of three dam-reservoirs in this catchment will potentially reduce downstream sediment transport from 1.95 to 0.29 M ton/year. It means that sedimentation may cause a loss in reservoir storage volume about 1.04 million cubic meters every year. For a 100-year period, the cumulative loss is just around 11% of the total reservoir volume. Although this impact is likely not very significant, the 85% decrease in sediment transport may cause many problems to downstream environment such as channel and river bank erosion. This prior information is very essential for water resource managers and different stakeholders in ensuring a sustainability of such development plan.
The Mekong-FReSSY model developed in this research would serve as a central approach in solving ungauged catchment problem with respect to sediment data in the Lower Mekong Basin that is subject to an ambitious development plan of hydropower dam-reservoirs requiring information on sediment transport. The model is important not only for dam-reservoir management but also for many other purposes such as soil and water conservation measure, environmental impact assessment, land use planning and river management. Besides that, the study also contributes a basic framework for developing regional models to estimate other key water parameters of interest. For a similar purpose, this framework can be applied as well in other river basins worldwide.
Table of Contents
Preface ...i
Acknowledgements ... iii
Abstract ...iv
Table of Contents ...vi
List of Tables ... viii
List of Figures ...ix
List of Appendices ... xii
List of Abbreviations ... xiii
List of Symbols ... xv
List of Units ... xviii
Chapter 1 Introduction ... 1
1.1 Background and needs ... 1
1.2 Research objective, scope and limitation ... 10
1.3 Dissertation organization... 12
Chapter 2 Overview of the study area and general research framework ... 15
2.1 Study area: the Lower Mekong Basin ... 15
2.2 General research framework ... 18
Chapter 3 Determination of an ideal regionalization methodology for monthly specific suspended sediment yield time series (SSYs) prediction in ungauged catchments ... 21
3.1 Introduction ... 21
3.2 Materials and methods ... 24
3.3 Results and discussion ... 37
3.4 Concluding remarks ... 56
Chapter 4 Advancement of the traditional catchment area-sediment yield model for mean annual/monthly suspended sediment yield (SSYa/SSYm) prediction in ungauged catchments ... 59
4.1 Introduction ... 59
4.2 Materials and methods ... 62
4.3 Results and discussion ... 66
4.4 Concluding remarks ... 72
Chapter 5 Extraction of annual maximum suspended sediment yield (SSYx) from discontinuous time series for frequency analysis and prediction of its quantiles (SSYx*) in ungauged catchments ... 75
5.1 Introduction ... 75
5.2 Materials and methods ... 78
5.3 Results and discussion ... 85
Chapter 6 Development of the Mekong-Feasible Regional Suspended Sediment Yield (Mekong-FReSSY) model for use in ungauged areas of the Lower Mekong
Basin ... 95
6.1 Introduction ... 95
6.2 Structure of the Mekong-FReSSY model ... 96
6.3 User’s manual of the Mekong-FReSSY model ... 98
6.4 Concluding remarks ... 101
Chapter 7 Coupling the Mekong-FReSSY with a sediment trapping efficiency model to quantify the potential impact of planned cascade dam-reservoirs on suspended sediment transport ... 103
7.1 Introduction ... 103
7.2 Materials and methods ... 105
7.3 Results and discussion ... 107
7.4 Concluding remarks ... 109
Chapter 8 Summary of the research ... 111
8.1 Summary of research results ... 111
8.2 Contributions ... 113
8.3 Future challenges... 113
References ... 115
List of Tables
Table 1.1 Comparison between the traditional and modern regional model. ... 8
Table 3.1 List of catchment characteristics. ... 26
Table 3.2 CUSLE value of different land use/cover types. ... 28
Table 3.3 Fitted sediment rating curves, at-site calibration and Jack-knife validation results. ... 38
Table 3.4 List of donor catchments selected using different CSI measures. ... 43
Table 3.5 Homogeneity test of the SSY annual series at 0.01 significance level. ... 44
Table 3.6 Correlation between model parameters and catchment descriptors. ... 47
Table 3.7 Jack-knife validation results (regionalization using regression approach). . 48
Table 3.8 Statistical performance of refined regionalization methodologies in Level 1 and 2. ... 52
Table 4.1 Summary of catchment information. ... 62
Table 4.2 Statistical characteristics of the alternative SSYa models. ... 69
Table 4.3 Predictive accuracy of SSYm (Jack-knife validation). ... 71
Table 5.1 List of data characteristics. ... 79
Table 5.2 List of distribution models used in this study. ... 83
Table 5.3 Number of stations passing one or several criteria of the SSYx estimation methods. ... 85
Table 5.4 Randomness test of the estimated SSYx series at 0.01 significance level. .... 87
Table 5.5 Statistical characteristics of the best regional SSYx* models. ... 90
Table 6.1 Information for use in Eq. (6.3) to calculate CSI. ... 99
Table 6.2 Ordinates of the double-regional USG model (U). ... 100
Table 6.3 Model equations for estimating SSYx*. ... 101
Table 7.1 Characteristics of the upper Sre Pok River catchment. ... 105
Table 7.2 Characteristics of the planned reservoirs in the upper Sre Pok River catchment. ... 106
List of Figures
Figure 1.1 Schematic representation of river sediment transport. ... 1
Figure 1.2 Diagram of reservoir allocations. ... 3
Figure 1.3 Location map of suspended sediment monitoring stations and river bank erosions occurring in the Lower Mekong Basin. ... 5
Figure 1.4 Schematic diagram of modeling techniques. ... 7
Figure 1.5 Linkage between research objectives in terms of their outputs’ usages. ... 11
Figure 2.1 Map of the study area: the Lower Mekong Basin. ... 17
Figure 2.2 General research framework. ... 19
Figure 3.1 Flow chart of the study in Chapter 3. ... 24
Figure 3.2 Location map of the study catchments in Chapter 3. ... 25
Figure 3.3 Diagram of regionalization methodologies. ... 32
Figure 3.4 Diagram of the regional SSYs model. ... 35
Figure 3.5 Correlation between the SRC model parameters. ... 38
Figure 3.6 Graphical illustration of SRC in Catchment No. 8. ... 39
Figure 3.7 Panel plot of the 17 fitted sediment rating curves. ... 40
Figure 3.8 Graphical comparison between the observed and predicted SSYs in Catchment No. 3. ... 41
Figure 3.9 Distribution of error indicators (NSE, PBIAS and RSR) and rate of satisfactorily modeled catchment corresponding to different regionalization methodologies in the physical similarity approach. Each circular point stands for one modeled catchment. The dash mark indicates the median value. The bar chart shows the rate of satisfactorily modeled catchment. ... 45
Figure 3.10 Distribution of error indicators (NSE, PBIAS and RSR) and rate of satisfactorily modeled catchment corresponding to different regionalization methodologies in the spatial proximity approach. Each circular point stands for one modeled catchment. The dash mark indicates the median value. The bar chart shows the rate of satisfactorily modeled catchment. ... 50
Figure 3.11 Bivariate plot between CSI and D. ... 51 Figure 3.12 Distribution of error indicators (NSE, PBIAS and RSR) and rate of
satisfactorily modeled catchment corresponding to refined regionalization methodologies of physical similarity (PS), regression (RE) and spatial proximity (SP) approach in Level 1, and ensemble (ENS) and combination method (COM) in Level 2. PS: single donor method (CSI4); RE: exponential
km, parameter averaging option); ENS1: PS, RE and SP; ENS2: PS and RE; ENS3: PS and SP; COM: PS and RE. Each circular point stands for one modeled catchment. The dash mark indicates the median value. The bar
chart shows the rate of satisfactorily modeled catchment. ... 52
Figure 3.13 Catchment classification to suit superiority of the physical similarity- and regression-based regionalization approach. ... 54
Figure 3.14 Bivariate plot between model parameters and Sc (left) and between model parameters and KUSLE×CUSLE (right). ... 54
Figure 4.1 Flow chart of the study in Chapter 4. ... 61
Figure 4.2 Location map of the study catchments in Chapter 4. ... 63
Figure 4.3 Extraction of A0, A5, A10, A15, A20 and A30 for Catchment No. 1. ... 64
Figure 4.4 Monthly distribution of suspended sediment transport. ... 67
Figure 4.5 Size of catchment area with different slope characteristics. ... 67
Figure 4.6 Optimization of the alternative SSYa models. ... 68
Figure 4.7 Scatter plot of the observed versus predicted SSYa. ... 69
Figure 4.8 Unit mean annual sedimentograph (USG). ... 71
Figure 4.9 Graphical comparison between the observed and predicted SSYm in Catchment No. 11. ... 72
Figure 5.1 Flow chart of the study in Chapter 5. ... 77
Figure 5.2 Location map of the study sites in Chapter 5. ... 78
Figure 5.3 Framework for extracting SSYx from discontinuous time series. ... 81
Figure 5.4 Frequency of the observed SSYx and Qx per month. ... 85
Figure 5.5 Regional plot of Qx versus SSYx (RAMSRC) and SSYx2 versus SSYx (RLR). ... 86
Figure 5.6 Frequency curves of the estimated SSYx*. ... 88
Figure 5.7 Graphical comparison between the observed and estimated SSYx*. ... 89
Figure 5.8 Size of catchment drainage area upstream of each gauging station with different slope characteristics. ... 89
Figure 5.9 Optimization of the alternative SSYx* models. ... 90
Figure 5.10 Graphical illustration of the A15-SSYx* relationships. ... 91
Figure 5.11 Graphical method for estimating SSYx* in ungauged catchments. ... 91
Figure 6.1 Summary of model studies for developing the Mekong-FReSSY model. ... 96
Figure 6.2 Schematic diagram of the Mekong-FReSSY model. ... 97
Figure 7.2 Map of the upper Sre Pok River catchment. ... 106 Figure 7.3 Cumulative impact of the cascade dam-reservoirs on suspended sediment
List of Appendices
Figure A.1 Location map of hydropower dams in the Lower Mekong Basin. ... 124
Figure A.2 Location map of irrigation project headwork in the Lower Mekong Basin. ... 125
Figure A.3 Spatial distribution of mean annual rainfall in the Lower Mekong Basin. 126 Figure A.4 Spatial distribution of mean annual runoff in the Lower Mekong Basin. .. 127
Figure A.5 Rainfall data availability in the Lower Mekong Basin. ... 128
Figure A.6 Water flow data availability in the Lower Mekong Basin. ... 129
Figure A.7 Suspended sediment data availability in the Lower Mekong Basin. ... 130
Figure A.8 Distribution of land use/cover in the Lower Mekong Basin. ... 131
Figure A.9 Distribution of major soil types in the Lower Mekong Basin. ... 132
Figure A.10 Sensitivity of catchment descriptors on computing CSI. ... 133
List of Abbreviations
AMSRC Annual Maximum Sediment Rating Curve ANN Artificial Neural NetworkASCE American Society of Civil Engineers
ASTER Advanced Spaceborne Thermal Emission and Reflection Radiometer C1, 2, 3 … Catchment No. 1, 2, 3 …
CD Catchment Descriptor COM Combination method DEM Digital Elevation Model
EML Exponential Multiple Linear regression function ENS Ensemble method
FAO Food and Agricultural Organization of the United Nations FReSSY Feasible Regional Suspended Sediment Yield model GDEM Global Digital Elevation Model
GEV Generalized Extreme Value probability distribution model GLCC Global Land Cover Classification
GOF Goodness of Fit H0 Null hypothesis
HEC Hydrologic Engineering Center I Input
IACWD Interagency Advisory Committee on Water Data ICEM International Center for Environmental Management IGBP International Geosphere Biosphere Programme LMB Lower Mekong Basin
METI Ministry of Economy, Trade and Industry of Japan ML Multiple Linear regression function
MP Model Parameter
MRC Mekong River Commission
MUSLE Modified Universal Soil Loss Equation N/A Not Applicable
PN Polynomial regression function PS Physical Similarity approach
RAMSRC Regional Annual Maximum Sediment Rating Curve RE Regression approach
RFA Radio Free Asia
RLR Regional Linear Relationship
RUSLE Revised Universal Soil Loss Equation S1, 2, 3 … Station or Site No. 1, 2, 3 …
SAMSRC Site-specific Annual Maximum Sediment Rating Curve SLR Site-specific Linear Relationship
SP Spatial Proximity approach SRC Sediment Rating Curve
SWC Areal coverage of soil and water conservation TDL Total Drainage Length
USA United States of America
USBR United States Bureau of Reclamation USD United States Dollar
USG Unit mean annual sedimentograph USLE Universal Soil Loss Equation WMO World Meteorological Organization
List of Symbols
% Percentage∞ Infinity
°C Degree Celsius °E Degree East °N Degree North
a Rating coefficient of sediment rating curve (power form) A Lower bound of a normalization range
A0 Catchment area
A5 Catchment area with a slope gradient greater than 5%
A10 Catchment area with a slope gradient greater than 10%
A15 Catchment area with a slope gradient greater than 15%
A20 Catchment area with a slope gradient greater than 20%
A30 Catchment area with a slope gradient greater than 30%
A40 Catchment area with a slope gradient greater than 40%
b Rating exponent of sediment rating curve (power form) B Upper bound of a normalization range
CSI Catchment similarity index
CUSLE USLE cover and management factor
D Distance between catchment centroids F Cumulative distribution function
fcl-si Soil erodibility factor related to clay and silt content
fcsand Soil erodibility factor related to coarse sand content
fhisand Soil erodibility factor related to high sand content
forgc Soil erodibility factor related to organic carbon content
K Rank value (positive integer) KS Kolmogorov-Smirnov test KSstat Kolmogorov-Smirnov statistic
KUSLE USLE soil erodibility factor
MAPE Mean absolute percentage error mcl Percentage of clay content
morgc Percentage of organic carbon content
ms Percentage of sand content
msi Percentage of silt content
n, N Sample size
NSE Nash-Sutcliffe efficiency
nSSY Average amount of suspended sediment yield samples per year
O Observed value Oavg Mean observed value
P Predicted value PBIAS Percent bias Q Water discharge
Qa Mean annual water discharge
Qas Mean annual specific water discharge
Qcat Catchment mean annual water discharge
Qlr Mean annual water discharge at the location of the lowest reservoir
Qs Monthly specific water discharge time series
Qx Annual maximum water discharge
R2 Determination coefficient
RFr Empirical cumulative distribution function RMSE Root mean square error
RSR Ratio of the root mean square error to the standard deviation Sc Mean catchment slope gradient
Sr Mean mainstream bed slope gradient
SSCx Annual maximum suspended sediment concentration
SSCx* Quantile of annual maximum suspended sediment concentration
SSY Suspended sediment yield
SSYa Mean annual suspended sediment yield
SSYas Mean annual specific suspended sediment yield
SSYcat Catchment annual suspended sediment yield
SSYm Mean monthly suspended sediment yield
SSYre Annual suspended sediment yield remaining from reservoir trapping
SSYx Annual maximum suspended sediment yield
SSYx* Quantile of annual maximum suspended sediment yield
SSYx1 Annual maximum suspended sediment yield or raw SSYx extracted from
discontinuous time series
SSYx2 Annual maximum suspended sediment yield extracted from discontinuous
time series after excluding SSYx
TE Sediment trapping efficiency
U Ordinate of the unit mean annual sedimentograph V Reservoir storage capacity
WD Distance weighting factor WP Physical weighting factor X Value of catchment descriptor Y Sample
List of Units
G ton/year Giga ton per yearkg/m3 Kilogram per cubic meter km Kilometer
km2 Square kilometer km3 Cubic kilometer m Meter
m3/s Cubic meter per second
m3/s/km2 Cubic meter per second per square kilometer of land area M m3 Million cubic meter
mm/year Millimeter per year M ton Million ton
M ton/year Million ton per year ton/day Ton per day
ton/day/km2 Ton per day per square kilometer of land area ton/month Ton per month
Chapter 1
IntroductionThis chapter firstly explains why this research is important and needed, then lists specific objectives to be performed in subsequent chapters, and finally outlines the dissertation.
1.1 Background and needs
Sediment is the soil particles resulted from land surface erosion. Once eroded, sediment is transferred to the oceans through various transport mechanisms such as river, wave and wind. The transport by rivers is a key pathway contributing more than 90% (15–20 G ton/year) of the global sediment flux and approximately 80% of this input is from Asia (Syvitski et al., 2003; Walling, 2006; Segar, 2009). The Indian Ocean receives the largest amount (about 3.5 G ton/year) which is collectively discharged by four major rivers: the Ganges, Brahmaputra, Irrawaddy and Indus (Segar, 2009). River sediment may be classified by relative particle size or transport mode (Morris & Fan, 1998). By method of transport, the total sediment load mainly consists of bed load and suspended load. Since both components of the sediment load are measured using different samplers, they could be defined simply as the material collected in a bed- and suspended-load sampler. However, bed load is referred to the coarse material moving in contact with the riverbed, while suspended load is the fine particle (silt, clay and sand) moving in suspension in the water column. A schematic representation of the river sediment transport is illustrated in
Figure 1.1. In many streams, the fraction of bed load is very small, while the suspended
portion is predominant and commonly account for around 90% of the total sediment flux (Syvitski et al., 2003; Walling & Fang, 2003; Francke et al., 2008; Zhang et al., 2012).
Source: Morris & Fan (1998)
Figure 1.1 Schematic representation of river sediment transport.
Land surface erosion and sediment transport is a very complex phenomenon varying spatio-temporally. Its main governing factors are hydro-meteorology, topography, land use/cover and soil characteristics (USBR, 2006; Neitsch et al., 2011). Erosive force of runoff from rainfall detaches directly soil particles from its matrix and then carries the
detached materials (sediment) gravitationally to nearby rivers. From erosion sites, the eroded soil particles may require several transporting events to reach the stream network, especially in large catchments. In the river system, the sediment is transported further downstream by streamflow. At the same time, some sediment particles are re-deposited due to a decrease in streamflow velocity. Those reaching the river mouth are deposited and form a delta there. From this viewpoint, sediment variable involves three primary processes: erosion, transport and deposition. Erosion is important in steep-slope terrains or mountainous regions, while deposition predominates in lowland or floodplain areas, although both processes occur simultaneously in all environments.
Sediment plays an important role in sustainable development of water resources system because it controls riverine hydrology, river channel morphology, water quality, aquatic ecology and so on (Morris & Fan, 1998; Walling, 2009; Melesse et al., 2011; Morris, 2014). A rapid population growth led to a high demand of electricity and water for consumptive use. A huge amount of water is of course required for food production, especially in agricultural sector. Consequently, many dam-reservoirs have been developed and planned for storing and regulating this vital resource in order to meet such non-stop increasing need. According to the International Commission on Large Dams, number of the registered large dams (height over 15 m) currently reached 37,641 with their aggregate storage capacity of approximately 15,501 km3. The development of such infrastructures possibly causes a significant impact on river basin environment, e.g. altering the streamflow regime and interrupting the continuity of sediment transport through river systems. In the tributaries of the Lower Mekong River, there are more than 100 dam-reservoirs of hydropower plant (most of them are under planning) and this corresponds to 75.44 km3 of the active storage volume; their development has potential to trap sediment about 25.8 M ton/year(Kummu et al., 2010). Regarding one big project (i.e. Xayaburi Dam) proposed on the mainstream of the Lower Mekong River, sediment issues were a major concern addressed by the riparian countries (PÖYRY, 2012). It is also one among other five key factors included in the design guidance for proposed mainstream dams in this river basin (MRC, 2009a).
Sediment yield is defined as the total sediment load discharged by a catchment or the mass of sediment delivered to a point of interest in the river network over a stated period of time, and the specific sediment yield is the yield per unit of land area (Morris, 2014). Information of sediment yield is essential for planning, design, operation and decommissioning of water resources projects. For a dam-reservoir project, in planning phase, it is needed for the environmental impact assessment as well as the project feasibility analysis; in design phase, an estimate of long term sediment yield is required to estimate reservoir’s life, and to size the sediment storage pool and routing facilities; during the project operation, it is important for an efficient sediment routing operation; and in decommissioning phase, total sediment deposited in the reservoir is a key information for its management. Viglione et al. (2013) stated that “the main control on reservoir sustainability is the sediment input to the reservoir”. Therefore, when developing a dam-reservoir on a river, the project planner has to ensure that sediment will not impair reservoir functions during the useful life of the project, and the project
operation will also not alter the fluvial sediment transport. Failure to do so will lead to many negative impacts on not only the project itself but also the environment downstream, e.g. loss of reservoir storage capacity, shortening of project life, biological damage, river channel degradation, shoreline recession and delta subsidence (Morris, 2014).
In every plan of dam-reservoirs, a certain storage capacity, known as dead storage (Figure 1.2), is basically provided for sedimentation; but, during operation, the provided dead storage is sometimes filled up faster than expected due to ‘extreme events’ of sediment transport, causing the sediment getting deposited for many years of the reservoir’s life. Extreme sediment transports generally occur during large flood events. As a result of heavy rainfall, the extremely high runoff could wash a large amount of particles from the soil surface into the river over a short time. Simultaneously, the river bed sediment is re-suspended due to a sudden increase in flow velocity and water level, initiating more high rates of sediment transport. This issue should be attentively considered for the purposes of a better development and management of water resources (Tramblay et al., 2010). Moreover, during the 58th Conference on Hydraulic Engineering held at Kobe University, Japan, on 4–6 March 2014, extreme events are strongly recommended for consideration in design practices (Takara, 2014). By doing so, a management of dam-reservoirs in the current changing world has more potential to prepare for unexpected events.
Figure 1.2 Diagram of reservoir allocations.
Sediment-related problems have been observed in many river basins around the world. For instance, the projected life of Hao Binh Dam (Da River, Vietnam) was reduced from more than 100 years to about 50 years, as a result of the increased reservoir sedimentation (Imhof, 1998). Because a fraction of sediment flux is trapped in the reservoir behind the dam wall, the sediment-starved water flow or clear water in the river channel downstream of the dam will tend to scour the streambed causing it to coarsen and incise (Morris, 2014). Coarsening of the riverbed can make it unsuitable as ecological habitat and spawning sites for both native and introduced species. Channel incision will accelerate river bank erosion. Nutrients are bound to sediment. Sediment reduction causes a decrease in nutrient supply and thus impacts agricultural productions in floodplains, as well as sustainability of flooded/mangrove forests (rich habitats for fish for feeding and
Dead storage (sediment storage pool)
Non-active storage Active storage Dam
breeding). In a severe condition, sedimentation could lead to a dam failure like the IVEX Dam on the Chagrin River, USA. The collapse of the IVEX Dam (in 1994) was obviously due to a combination of several factors; one of them is the 86% loss of permanent pool capacity because of sedimentation (Evans et al., 2000). A large amount of sediment released to downstream of the project site provoked numerous negative impacts including the change of channel morphology and impairment of water quality. With regard to decommissioning of the San Clemente Dam (Carmel River, USA), it is expected to cost more than USD75 millions, and the largest cost component is associated with management of the 1.9 M m3 of sediment deposited in the reservoir (Morris, 2014).
Sediment issues in the Lower Mekong Basin, draining part of Cambodia, Lao PDR, Thailand and Vietnam, have received considerable attention in recent years while degradation of downstream channel morphology appeared in the basin, which might be due to many factors. One of them could be a reduction of sediment transport which was captured by the upstream constructed dams, together with an over-exploitation of deposited materials (sand) from the riverbed. Many hot spots of severe bank erosion were observed along the mainstream of the Lower Mekong River (Miyazawa et al., 2008), e.g. between Vientiane and Nongkhai reach in Lao PDR (29.1 km long of river bank erosion), at Thuong Phuoc in Vietnam (6 km long of river bank erosion) and at Sa Dec in Vietnam (10 km long of river bank erosion). At Russey Keo (Phnom Penh) in Cambodia, at least 40 houses were damaged by river bank erosion, occurring on 1 April 2008 (Reuters, 2008). At Preah Prasob in Cambodia, another event happening at the night of 18 February 2014 wounded three people and left the other three missing (RFA, 2014). Locations of these hot spots are indicated in Figure 1.3.
The first engineering task in dam-reservoir management with respect to sediment issues is the derivation of sediment yield information. The best estimate of sediment yield is the use of field observed data. Unfavorably, observations of sediment loading are lacking for rivers in many parts of the world, especially in developing and remote regions (Walling, 2009; Isik, 2013; Morris, 2014). Although it is gauged in some areas, the sampling frequency is typically low (monthly or even larger time scale). Only suspended load, in most cases, is measured and measurements of bed load transport are lacking. Such poorly gauged or ungauged sites are usually located in headwater regions. Thus, absence of the historical data records is generally due to inaccessibility, budget and technical constraints (particularly in developing countries), and historical lack of foresight in considering future developments in such areas. For example, the Lower Mekong Basin has a very large drainage area (about 606,000 km2) but contains only 60 monitoring stations of suspended sediment load (Figure 1.3), and a majority of them has only a few samples per month. In North America, the budgetary reasons have caused a decrease in number of suspended sediment monitoring stations over the past three decades (Tramblay et al., 2010).
Figure 1.3 Location map of suspended sediment monitoring stations and river bank erosions occurring in the Lower Mekong Basin.
Considering now the status of sediment observations in the Lower Mekong Basin, the geographic distribution of gauging stations is largely non-uniform (Figure 1.3). Thailand’s monitoring infrastructure is more developed than the others and therefore, this country is very rich in historical records due to its extensive gauging network. Among 46 monitoring stations in the tributary rivers, 83% of them are situated within the Thai sub-catchments, the most promising region for model development. In terms of data quality, observations of sediment transport in Thailand have proven to be the most reliable, complete and easy to access (Fuchs, 2004). On the other hand, a large number of water resources-related development projects including 96 dam-reservoirs (none in Thailand), which require the information on sediment yield, are planned for the Cambodian, Laotian
and Vietnamese portions of the basin (MRC, 2011a). Neighboring Thailand, these three countries all have problems with the extent of monitoring and data availability. Not only sediment but also other related catchment process variables (e.g. water discharge and rainfall intensity) are not available for many areas of the basin. Because sediment yield data are usually limited, it is necessary to use modeling techniques for the estimates.
For a specific river catchment, a model is firstly employed to simulate (calibrate and validate) past sediment transport events, based on their recorded information. After that, the simulated or catchment model can be applied to derive information of other unknown events in that catchment (Figure 1.4a). Such process is known as ‘site-specific modeling’. There are many models that can be used for the simulation task. Anyway, they can be classified as the physically-based (e.g. unit stream power), process-based (e.g. USLE, RUSLE, MUSLE) and data-driven (e.g. sediment rating curve, artificial neural network) (Heng & Suetsugi, 2013a). With regard to site-specific modeling (Figure 1.4a), some historical data of sediment yield are required for calibration and validation before the model can be used for the prediction. This cannot be achieved in river catchments (particularly in less developed areas) having no such data. To overcome this situation, regional models are important.
A regional model is one that was already calibrated and validated for a set of gauged catchments. In this case, the simulation process is so-called ‘regional modeling’. Here, the simulated model or the regional model can be used to estimate sediment yield for other catchments of interest, where no data have been collected. Basically, a regional model is developed using information of some gauged sites located within a specific region (river basin). Its applicability is therefore governed by characteristics of the gauged sites used. In other words, the developed model is applicable only in the study region, not universally. From this viewpoint, a specific regional model has to be build for a specific river basin.
There are several types of regional model. The one that has been seen often in scientific research articles is the relationship between catchment area and sediment yield (Figure 1.4b). Such relationship is hereafter called “traditional regional model” because it cannot be applied to estimate instantaneous sediment yield. Regression analysis (linear or non-linear model) is generally used to determine a relationship between both variables. Catchment area functioning as the predictor in the traditional regional model is time independent variable and hence, the model allows solely a mean or point value prediction, e.g. the mean annual sediment yield (Table 1.1). The single key advantage of the traditional regional model is its broad feasibility because catchment area, the most important variable, is usually known since the early stage of project planning (Tamene et al., 2006). This catchment attribute can also be generated using a global topographic dataset that is easy to access, worldwide, and free of charge. It is well known that the usefulness of mean annual sediment yield, output of such traditional model, is limited. Anyway, it provides enough information for basin or project planning (Haregeweyn et al., 2008) where most of the related data are normally not available. In some cases, seasonal information of sediment transport (or mean monthly sediment yield) is required during
the planning stage and this cannot be provided by the said traditional model. An outstanding challenge in the subject under discussion is advancing the predictive accuracy and capacity of the traditional catchment area-sediment yield model. The research question here is: how to achieve this goal with a condition that the existing advantage of the model (feasibility) is preserved?
Figure 1.4 Schematic diagram of modeling techniques.
In a new scientific society, particularly after launching a scientific program called the IAHS Decade on Predictions in Ungauged Basins (Sivapalan et al., 2003), many research studies have been conducted so as to develop new regional modeling methods which enable time series prediction in ungauged areas. It is worth emphasizing that at least a time series input is required to produce a time series output. Within a region, a model
(a) Site-specific modeling Sediment yield time series Hydro-meteorology Simulation (calibration and validation) Physically-based model Process-based model Data-driven model Catchment model Sediment yield time series? Hydro-meteorology Catchment A (known events) Catchment A (unknown events) Sediment yield time series Hydro-meteorology At-site calibration Physically-based model Process-based model Data-driven model Catchment descriptors Regional validation
Physical similarity approach Regression approach Spatial proximity approach
Catchment models Regional model (modern) Sediment yield time series? Hydro-meteorology Catchment descriptors Gauged catchments Ungauged catchments
(c) Regional modeling (modern) Mean annual
sediment yield
Catchment area
Regional simulation (based on regression analysis) Linear model Non-linear model Regional model (traditional) Mean annual sediment yield? Catchment area Gauged catchments Ungauged catchments
(physically-based, process-based or data-driven) is calibrated on a set of gauged catchments, termed donor catchments; this process is so-called ‘at-site calibration’. After successfully validated on another set of gauged catchments, termed pseudo ungauged catchments, all the locally calibrated models are integrated as a regional model (Figure 1.4c); this process is so-called ‘regional validation’. Such model can be then applied for the prediction in ‘ungauged catchments’ of the region. This kind of modeling tool is hereafter called “modern regional model” because it can be applied to estimate instantaneous sediment yield. Although the modern regional model can produce a time series output which is more advantageous than the point value output produced by the traditional one, its feasibility is narrower because it requires at least a time series input, e.g. water discharge time series of which monitoring stations are also limited worldwide (Table 1.1). This could be evidential from many recent studies on rainfall-runoff modeling in ungauged basins (He et al., 2011; Parajka et al., 2013). As Figure 1.4 showed, this advanced technique is a synthesis of the site-specific modeling and the traditional regional modeling. Besides, for the target ungauged catchments, their model information is transposed from donor catchments. This process is known as ‘regionalization’, and it has been often conducted using the physical similarity, regression and spatial proximity approach. Using the said modern technique, many studies reported in the literature were for developed countries where a dense network of gauging stations is available and most of them focused only on streamflow, not sediment (He et al., 2011; Parajka et al., 2013). A key challenge for this topic is developing a modern regional model that requires a very few time series inputs to possibly estimate instantaneous sediment yield in ungauged catchments. A related question here is that, in a region with a sparse network of basin monitoring like the Lower Mekong Basin, what is the ideal regionalization methodology for continuous sediment yield simulation?
Table 1.1 Comparison between the traditional and modern regional model.
Type of regional model Input (predictor) Output Comparison
Traditional Time independent
(e.g. catchment area)
Mean or point value
(e.g. mean annual sediment yield)
Wider feasibility
Modern Time series
(e.g. water discharge) Time series
Narrower feasibility
With regard to extreme events of sediment transport, there are very few related studies reported in the literature, and they were for developed countries like Canada and the USA (Simon & Klimetz, 2008; Tramblay et al., 2008; Tramblay et al., 2010). The main reason is attributed to the limitation of data. The objective of extreme event analysis is to compute quantiles or magnitude of the variable at different return periods – the basic information needed for many engineering purposes such as planning and design of hydraulic structures (Takara, 2009). Such analysis requires a long term continuous time series data, and this is a major constraint for the case of sediment. Time series data of sediment yield are often discontinuous, making an indistinguishability of annual maxima for the frequency analysis. Here, the traditional regional model could also serve as a
potential tool in solving the problem in ungauged catchments. In this context, a challenge for the Lower Mekong Basin is that the sampling frequency of sediment yield data is relatively low. Hence, the most important question to answer is: how to extract the factual annual maxima from such discontinuous time series?
What are the techniques that have been applied so far to estimate sediment yield in ungauged areas of the Lower Mekong Basin? The traditional regional model (catchment area-sediment yield relationship) has been used in some scientific researches. For instance,
Kummu et al. (2010) employed this kind of modeling tool (developed based on gauged sites in the basin) to estimate the mean annual sediment yield for sub-catchments where no data have been collected, for assessing the basin-wide sediment trapping efficiency of emerging reservoirs. Besides that, a similar technique called the area-reduction factor method was found used in the study report of some dam-reservoir projects located in the basin part of Lao PDR, e.g. in the feasibility study of Xe Pian-Xe Namnoy and Nam Sana hydropower projects, and in the master plan study of hydropower resources in Nam Sum River catchment. The area-reduction factor method uses the mean annual specific sediment yield of one gauged catchment to calculate the mean annual sediment yield for the target ungauged catchments. This estimation strategy is less reliable than the one based on the traditional regional model because only one gauged catchment is present for indexing. There have been no any regional model studies dealing with sediment yield time series and quantiles of extreme sediment yield. Hence, it is important not only to advance to the traditional method being used but also to develop a widely feasible regional model that allows estimates of time series and extreme events in ungauged areas. With respect to sediment issues, the target model will serve as a key tool for sustainable development in this trans-boundary river basin.
The main motivations of this research are summarized as below:
Sediment yield is an important hydrological variable involved in sustainable development of water resources system, but its observed data are usually sparse. In developing countries as located the Lower Mekong Basin, not only sediment data but also its related hydro-meteorological variables like water discharge are not available.
There is an eager need of sediment yield information for a large portion of the Lower Mekong Basin where many water resources-related projects are to be developed.
Regional sediment yield models, particularly for estimating time series of sediment yield and quantiles of extreme sediment yield, have not been established yet for the Lower Mekong Basin.
With regard to prediction of hydrograph in ungauged catchments, most of the research studies reported in the literature focused on streamflow variable, and they were for developed countries.
The traditional catchment area-sediment yield model has been used since long time ago because of its simple structure and broad feasibility in the context of prediction in ungauged catchments, but there have been no many researches attempting to improve its predictive accuracy and capacity.
Time series data of sediment yield are commonly discontinuous, making a difficult selection of annual maxima for extreme event analysis. Relating to prediction of quantiles of extreme sediment yield in ungauged catchments, there are very few studies reported in the literature, and they were for developed countries.
1.2 Research objective, scope and limitation
The aim of this research is to develop a feasible regional model for catchment-scale suspended sediment yield estimations in ungauged rivers of the Lower Mekong Basin, Southeast Asia, for a more efficient management of dam-reservoirs and other water resources system.
The specific objectives of the study are:
(1) To develop a modern regional model for estimating monthly specific suspended sediment yield time series (SSYs) in ungauged catchments
(2) To develop and advance a traditional regional model for estimating mean annual/monthly suspended sediment yield (SSYa/SSYm) in ungauged catchments, an
alternative to the regional SSYs model (Objective 1)
(3) To develop another traditional regional model for estimating quantiles of annual maximum suspended sediment yield (SSYx*) in ungauged catchments, an additional
tool to the regional SSYs model (Objective 1) or the regional SSYa/SSYm model
(Objective 2)
(4) To couple the developed model with a sediment trapping efficiency technique to quantify the potential impact of planned cascade dam-reservoirs on suspended sediment transport
The target feasible regional model to be developed is an integration of three sub-models: the regional SSYs model (Objective 1), the regional SSYa/SSYm model (Objective
2) and the regional SSYx* model (Objective 3). This multi-model is named as
“Mekong-Feasible Regional Suspended Sediment Yield (Mekong-FReSSY) model”. Its schematic diagram is shown in Figure 6.2. ‘Feasibility’ is the principle concept held to build this modeling tool. It means that the developed model should have an optimum capacity to provide information on sediment yield for any catchments of interest even though they are ungauged with respect to some or all catchment process variables.
Of course, different sub-models produce outputs with different usage values. The regional SSYs model could provide an estimate of long term sediment yield, valuable
information for not only planning but also design and operation purposes. This sub-model requires hydro-meteorological time series data as the input. Such data may not available in some areas, making the model infeasible. As an alternative, one can employ the regional SSYa/SSYm model but its outputs can be used only for planning, not for design
and operation. Furthermore, to have a robust dam-reservoir management system, information of extreme sediment yield should be additionally considered. Such consideration could be a powerful strategy in reducing vulnerability and enhancing resilience in light of unexpected events, particularly in the current changing world. In this
regard, the regional SSYx* model can be employed as an additional tool. The
Mekong-FReSSY model is able to estimate merely the catchment production, total sediment yield produced by a catchment. To include the impact of dam-reservoirs, Objective 4 was proposed by associating an existing regional tool (sediment trapping efficiency model).
Figure 1.5 portrays a linkage between the four research objectives in terms of usages of their outputs. The regional SSYs model can provide an adequate solution to
dam-reservoir management. If it is used alone (H), the risk in management is higher than it is used in conjunction with the regional SSYx* model (L). When the regional SSYs model is
not applicable due to data constraint, the regional SSYa/SSYm model is the alternative.
Similarly, if it is used alone (H), the risk in planning is higher than it is used together with the regional SSYx* model (L). The higher risk option is definitely more economical than
the lower risk one; hence, the choice between these two options is left to the discretion of decision makers.
Figure 1.5 Linkage between research objectives in terms of their outputs’ usages. Objective 4
Dam-reservoir management
The SSYs model is applicable H Higher risk option Remark:
SSYa is the mean annual suspended sediment yield
SSYm is the mean monthly suspended sediment yield
SSYs is the monthly specific suspended sediment yield time series
SSYx* is the quantiles of annual maximum suspended sediment yield
TE is the sediment trapping efficiency
1
The SSYs model is not applicable
2
Planning Design
Operation Decommissioning
SSYs model SSYa/SSYm model SSYx* model
Objective 1 Objective 2 Objective 3
Mekong-Feasible Regional Suspended Sediment Yield (Mekong-FReSSY) model
TE model H L H L 2 1
Lower risk option
The scope and limitation of this study are listed as below.
The Upper Mekong Basin was not included due to data inaccessibility.
The study was conducted exclusively in the Lower Mekong Basin, a trans-boundary river basin covering part of four Southeast Asian countries (Cambodia, Lao PDR, Thailand and Vietnam), because of two main reasons. First is the viability of the research, meaning that some sediment data are available for analysis. Second is the eager need of sediment information for the ungauged portion of the basin.
This study considered only suspended sediment load because the data of bed load transport are not available.
Only (46) gauging stations in the tributary rivers were considered because data of those (14) in the main Mekong river might be affected by dam operations in the upper basin (in China). Moreover, the catchment drainage areas of the mainstream stations are too large, leading to scaling effects.
Number of modeled catchments for each objective was determined based on the available data obtained from the Mekong River Commission and the minimum requirements of analyses.
For Objective 1, the use of physically- or process-based model for simulation will produce a regional model that is not widely feasible in the target basin because of the mass input data requirement. Therefore, sediment rating curve, a simple data-driven model with only one predictor (independent variable), was selected. Based on a similar concept, the study did not consider much information in the model predictor, for Objective 2 and 3.
Due to data shortage, the sediment trapping efficiency model was not developed and the modified Brune’s equation (Vörösmarty et al., 2003) was used (Objective 4).
A number of global datasets were used in this study: topography or elevation (30-m resolution), downloaded from ASTER GDEM 2, land use/cover (1-km resolution), downloaded from IGBP GLCC 2, and soil type (10-km resolution), downloaded from SOIL-FAO database.
Some software was employed to assist computational processes. They are: ArcGIS 9.3.1 for terrain processing works, DataFit 9 for fitting multi-variable regression equations, EasyFit 5.5 for fitting statistical distributions, IBM SPSS Statistics 20 for statistical non-parametric tests, and Microsoft Office Excel 2007 for fitting single-variable regression equations.
1.3 Dissertation organization
There are, in total, eight chapters present to constitute this dissertation. Each chapter is briefly described as below. Chapters 3 to 7 present the main research results.
Chapter 1 begins with brief introduction of physical processes governing sediment production and transport. Then it describes some facts and figures of sediment-related issues to evidence the important role of this particular hydrological variable
in water environment. After pointing out research needs and challenges in the target river basin, it indicates specific assignments to be performed in this dissertation. Lastly, it outlines the entire works chapter by chapter, such that readers can take things in at a glance.
Chapter 2, in the first part, describes more details of the study area, including hydro-meteorological condition and monitoring stations, physiographic feature and drainage pattern, and land use/cover and soil type. A general framework of the research was presented in the second part.
Chapter 3 focuses on prediction of monthly specific suspended sediment yield time
series (SSYs) in ungauged catchments. Different regionalization methodologies were
setup and applied to parameterize the sediment rating curve models in ungauged catchments for subsequent SSYs prediction. Their relative performances were
evaluated using a set of gauged catchments. The ideal regionalization methodology was finally taken into account to build the regional SSYs model, i.e. the integration of
the locally fitted sediment rating curves.
Chapter 4 focuses on prediction of mean annual/monthly suspended sediment yield (SSYa/SSYm) in ungauged catchments. A new model predictor associating catchment
slope gradient was introduced to improve accuracy of the traditional catchment area-sediment yield model in predicting SSYa. After establishing the regional SSYa model,
it was coupled with a proposed unit mean annual sedimentograph to enable the prediction of SSYm. This coupled approach was called the regional SSYm model.
Chapter 5 focuses on prediction of quantiles of annual maximum suspended
sediment yield (SSYx*) in ungauged catchments. A logical framework was firstly
adopted to extract annual maximum suspended sediment yield (SSYx) from
discontinuous time series for frequency analysis. Sixty-one (61) probability distribution functions were alternatively applied to fit the SSYx series of each study
site. The Kolmogorov-Smirnov goodness of fit test was used to identify the best distribution model for subsequent SSYx* generation. Finally, the regional SSYx* model
was developed, based upon a similar concept used in Chapter 4.
Chapter 6 brings altogether three sub-models: the regional SSYs model from Chapter
3, the regional SSYa/SSYm model from Chapter 4, and the regional SSYx* model from
Chapter 5, to construct a multi-model called the Mekong-Feasible Regional
Suspended Sediment Yield (Mekong-FReSSY) model. A user’s manual of the developed model was also made in this chapter.
Chapter 7 employs the Mekong-FReSSY model in conjunction with a sediment
trapping efficiency technique to quantify the potential impact of planned cascade dam-reservoirs on suspended sediment transport, a case study in the upper Sre Pok River catchment, southeast of the Lower Mekong Basin.
Chapter 8 presents a summary of research results and contributions. It lists a set of remaining challenges for future researches at the end.
