Considering the 2008's worldwide data for CO2 emission, Thailand is in 22 ranks. ...... Soil and water assessment tool (SWAT) monitor the five different pools of ...
Climate and Land Use Change Impacts on Streamflow and Water Quality in the Songkhram River Basin, Thailand
by
Binod Bhatta
A thesis submitted in partial fulfillment of the requirements for the degree of Master of Engineering in Water Engineering and Management
Examination Committee:
Nationality: Previous Degree:
Scholarship Donor:
Dr. Sangam Shrestha (Chairperson) Prof. Mukand S. Babel Dr. Duc Hoang Nguyen Dr. Rabin Bhattarai (External Expert)
Nepalese Bachelor in Civil Engineering Nepal Engineering College Bhaktapur, Nepal AIT Fellowship WEM Project
Asian Institute of Technology School of Engineering and Technology Thailand May 2017
i
ACKNOWLEDGEMENTS Foremost, I would like to express my sincere gratitude to my committee chair Dr. Sangam Shrestha for the continuous support for my Master degree research, for his patience, motivation, enthusiasm, and immense knowledge. His guidance helped me during my research and writing of this thesis throughout the study period. He always appreciated my queries and gave comprehensive solution for them. His continuous supervision guided me to complete my research in proposed time. Besides my advisor, I would like to thank the rest of my thesis committee: Prof. Mukand S. Babel, Dr. Duc Hoang Nguyen, and Dr. Rabin Bhattarai, for their encouragement, insightful comments, and questions that were milestones of my research. Their suggestions and comments always promoted me in deep research. Again, I would like to thank Dr. Rabin for his continuous reply to my emails during hydrological modeling. For financial support, I would like to thank Water Engineering and Management department, Asian Institute of Technology and Prof Mukand S. Babel. Without their support, it would not have been possible for me to successfully complete my Master’s Degree from a prestigious institute as AIT. So, my special thanks go to Prof. Mukand S. Babel, for his untiring efforts to manage financial support for my academic credits. My special thanks go, to Senior Research Associate Mr. Pallav Kumar Shrestha for helping me to develop my research skills. I appreciate his help during model calibration and validation, So, I highly acknowledge him. Appreciation is also emblematic to all the Water Engineering and Management (WEM) staffs, students and my colleagues who helped me several times during my research period. I would like to disperse my sincere thanks to WEM department secretary Ms. Pajee Trakanpasakul and Ms. Siriporn Hanmeng who provided a lot of information during the research time. I also would like to thank my dear friend Uttam Ghimire who guided me throughout my academic and research. Last but not the least, I would like to thank my family members for their continuous support, unconditional love, and selfless financial contribution, without which I would not have been able to finish my Masters’ Degree.
ii
ABSTRACT This study examined the combined and separated impacts of climate and landuse change on streamflow, sediment yield and nitrate-nitrogen in Songkhram river basin Thailand. The climate data predicted by ACCESS, CNRM and MPI for three future time periods (2010– 2039, 2040–2069, 2070–2099) were prepared by using linear scaling bias correction method under RCP 4.5 and RCP 8.5. Three calibrated RCM showed that temperature is continuously increasing in the study area, however, future precipitation is highly complex and uncertain, there were significant differences among various RCM under RCP scenario. The future landuse were simulated using the Conservation of Land Use and its Effects at small regional extent (Dyna-CLUE) model by establishing logistic regression model for seven landuse types. The regression model Load Estimator (LOADEST) were applied to developed regression equations and to estimate nutrient and sediment loading at three site of Songkhram River basin from 1990 to 2014. Estimated loading shows increasing trend during past, however, suspended solid showed decreasing trend during historical period. Non-parametric MannKendall test was used to check the significance of trend. Three scenarios (climate change only, landuse change only and combined) were developed and the streamflow, sediment yield, and nitrate yield in future periods under these scenarios was simulated by using Soil and Water Assessment Tools (SWAT) model. Under climate change only annual and seasonal, streamflow, sediment yield, and nitrate yield is in decreasing rate with more in later period. Whereas landuse change increase the flows and sediment yield during all periods and under both scenarios. But LULC change decreased nitrate yield. The results for combined scenario was similar to that of the climate change only scenario. In the future, climate changes tend to affect the hydrological regimes, sediment yield and nitrate yield much more prominently than the land use change, leading significant decreasing in streamflow, sediment yield and nitrate yield. Nevertheless, the role of land use change should not be overlooked, especially if the climate becomes drier in future, as in this case it may magnify the hydrological responses.
iii
TABLE OF CONTENTS
CHAPTER
TITLE
PAGE
TITLE PAGE ACKNOWLEDGEMENT ABSTRACT TABLE OF CONTENTS LIST OF FIGURES LIST OF TABLES LIST OF ABBREVIATIONS 1.
INTRODUCTION 1.1 Background 1.2 Problem statement 1.3 Objectives of the study 1.4 Scope of the study 1.5 Limitation of the study
2.
LITERATURE REVIEW 2.1 Introduction to climate change 2.1.1 Climate change in Asia 2.1.2 Climate change in Thailand 2.2 Climate projections 2.2.1 Introduction of climate models 2.2.2 Global climate model 2.2.3 Regional climate model 2.3 Description of RCPs scenarios 2.4 Bias correction 2.5 Hydrological modeling 2.5.1 Soil and water assessment tools (SWAT) 2.5.2 Model performance evaluation 2.6 Climate change impact on hydrology 2.7 Introduction to landuse 2.7.1 Landuse change modeling 2.7.2 Dyna-CLUE model 2.7.3 Landuse change impact on streamflow 2.8 Introduction to nitrate-nitrogen 2.8.1 Water quality model 2.8.2 Impact of climate and landuse change on water quality 2.9 Load Estimation technique 2.9.1 Load estimation using field data 2.9.2 Load estimation methods used within LOADEST
i ii iii iv vi viii ix 1 1 2 4 4 4
iv
5 5 6 6 7 7 8 9 10 11 12 13 14 14 15 16 17 17 19 20 21 22 22 22
3.
STUDY AREA AND DATA COLLECTION 3.1 Study area 3.1.1 General description of area 3.1.2 Climatology 3.1.3 Topography and geology 3.2 Data collection 3.2.1 Soil data 3.2.2 Landuse data 3.2.3 Historical climate 3.2.4 Hydrological data 3.2.5 Water quality data 3.2.6 Future climate
24 24 24 25 25 26 26 27 28 31 32 33
4.
METHODOLOGY 4.1 Overall research methodology 4.2 Statistical analysis of observed data 4.3 Analysis of future climate 4.4 Rating curve generation 4.5 Load estimation using field data 4.5.1 Regression estimation 4.5.2 Regression model 4.6 SWAT modeling 4.6.1 SWAT hydrology simulation 4.6.2 SWAT sediment simulation 4.6.3 SWAT Nutrient process simulation 4.6.4 Performance evaluation of SWAT model 4.7 Landuse change modeling (Dyna-CLUE)
34 34 34 36 38 39 39 40 41 42 43 45 48 49
5.
RESULT AND DISCUSSION 5.1 Past climate analysis 5.2 Baseline hydrology analysis 5.3 Past landuse analysis 5.4 Past nutrient and sediment analysis 5.4.1 Regression model performance 5.5 Estimation of nutrient and suspended solid loading 5.5.1 Suspended solids loading 5.5.2 Nitrate – Nitrogen (NO3-N) loading 5.6 Future climate scenarios in Songkhram basin 5.7 Future landuse projection 5.8 Hydrological modeling 5.8.1 SWAT model setup 5.8.2 Calibration and validation of hydrological model 5.9 Climate change impact analysis 5.9.1 Climate change impact on streamflow
50 50 53 55 56 57 58 58 60 62 66 69 69 69 72 72
v
6.
5.9.2 Climate change impact on suspended solids 5.9.3 Climate change impact on nitrate-nitrogen (NO3-N) 5.10 Combined impacts of future climate and landuse change
76 79 82
SUMMARY, CONCLUSION AND RECOMMENDATION 6.1 Summary 6.2 Conclusions 6.3 Recommendation
89 89 91 92
REFERENCES APPENDICES
94 104
vi
LIST OF FIGURES Figure
Title
2.1
Concepts of model domain of GCMs and RCMs RCP-scenarios and corresponding additional radioactive forcing value through 21st century Location map of Songkhram River Basin DEM and soil classes of the study area Baseline landuse map and percent area cover by various landuse classes (2009) Network of climate monitoring location in Songkhram Basin Basic Climatology of Upper Catchment Basic Climatology of Lower Catchment Rainfall versus discharge plot at lower catchment KH.55 Monthly Nutrient and Sediment Concentration in Songkhram Basin from 1995 to 2014 Overall methodological framework for climate and landuse change impact analysis on hydrology, sediment, and nitrate-nitrogen Framework for climate change analysis for Songkhram river basin Rating curve for the hydrological station KH.55 (a) Nitrogen Cycle in soil and (b) Partitioning of Nitrogen in SWAT Framework for Dyna-CLUE model for this study Annual average rainfall and temperature trend during 1970 to 2015 Annual rainfall and temperature anomalies Baseline hydrological trend in (a) annual (b) monthly (c) seasonal flow during 1990 to 2014 period Total land area cover by different landuse classes and their rate of change during 2002, 2007, 2009 & 2014 Catchment sub-division to estimate load in different parts (a) Observed versus estimated SS loading (b) Normal probability plot (c) Model residuals (a) Observed versus estimated NO3-N loading (b) Normal probability plot (c) Model residuals Sum of yearly SS loading and area normalized load variation during historical period 1990 to 2014 Estimated seasonal suspended solids load variation at three locations Graph represent (a) annual nitrate load variation (b) area normalized load variation during historical period Box-plot for seasonal variation of nitrate-nitrogen loading during the historical period at three location Change in basin average annual precipitation for three future period relative to the baseline period under RCP 4.5
2.2 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 4.1 4.2 4.3 4.4 4.5 5.1 5.2 5.3 5.4 5.5 5.6 5.7 5.8 5.9 5.10 5.11 5.12
Page No
vii
9 11 25 27 28 30 30 31 32 33 34 38 39 46 49 52 53 54 55 56 58 58 59 60 61 61 62
5.13 5.14 5.15 5.16 5.17 5.18 5.19 5.20 5.21 5.22 5.23 5.24 5.25 5.26
5.27
5.28
2.29
5.30
Change in basin average annual precipitation for three future period relative to the baseline period under RCP 8.5 Change in mean summer season precipitation (%) in three future periods relative to baseline period Change in mean rainy season precipitation (%) in three future periods relative to baseline period Change in mean winter season precipitation (%) in three future periods relative to baseline period Variation of annual Tmax under RCP 4.5 scenario in three future periods relative to the baseline period Variation of annual Tmax under RCP 8.5 scenario in three future periods relative to the baseline period Variation of annual Tmin under RCP 4.5 scenario in three future periods relative to the baseline period Variation of annual Tmin under RCP 8.5 scenario in three future periods relative to the baseline period Future landuse projection (a) by percentage of total area cover by different landuse class (b) relative change in future landuse with respect to baseline map 200 Future landuse map for economic scenario during different period 2020s, 2050s, and 2080s Future landuse projection (a) by percentage of total area cover by different landuse class (b) relative change in future landuse with respect to baseline map 2009 Future landuse map for conservation scenario during different period 2020s, 2050s, and 2080s Watershed delineation of Songkhram River Basin Hydrograph at station kh.55 showing the result of calibration and validation and observed versus simulated plot during calibration period (1990 to 2004) and validation period (2005 to 2014) Comparison of observed versus simulated daily sediment yield during calibration period (2005 to 2011) and validation period (2012 to 2014) Comparison of observed versus simulated daily nitrate-nitrogen yield during calibration period (2005 to 2011) and validation period (2012 to 2014) Change in mean annual streamflow in three future periods relative to the baseline period (1990 to 2009) in Songkhram River Basin under RCP 4.5 Change in mean annual streamflow in three future periods relative to the baseline period (1990 to 2009) in Songkhram River Basin under RCP 8.5
viii
63 63 64 64 65 65 66 66 67 67 68 68 69 71
71
71
73
73
5.31 5.32 5.33 5.34
Relative change in decadal flow with respect to baseline (1990 to 2009) average yearly flow under RCP 4.5 and RCP 8.5
74 75 75 75
5.37
Relative change in summer season streamflow Relative change in rainy season streamflow Relative change in winter season streamflow Change in mean annual sediment yield in three future periods relative to the baseline period (1990 to 2009) under RCP 4.5 Change in mean annual sediment yield in three future periods relative to the baseline period (1990 to 2009) under RCP 8.5 Relative change in decadal sediment yield with respect to baseline period (1990 to 2009) under RCP 4.5 and RCP 8.5
5.38
Change in sediment yield for future years during summer season
78
5.39 5.40
Change in sediment yield for future years during rainy season Change in sediment yield for future years during winter season Change in mean annual nitrate-nitrogen yield in three future periods relative to the baseline period (1990 to 2009) under RCP 4.5 Change in mean annual nitrate-nitrogen yield in three future periods relative to the baseline period (1990 to 2009) under RCP 8.5 Relative change in decadal nitrate-nitrogen yield with respect to baseline period (1990 to 2009) under RCP 4.5 and RCP 8.5
78 78
5.35 5.36
5.41 5.42 5.43
5.44
5.45
4.46
5.47
5.48
5.49
5.50
Change in nitrate-nitrogen yield for future years under RCP 4.5 and RCP 8.5 compared with baseline years (1990 to 2009) during summer season Change in nitrate-nitrogen yield for future years under RCP 4.5 and RCP 8.5 compared with baseline years (1990 to 2009) during rainy season Change in nitrate-nitrogen yield for future years under RCP 4.5 and RCP 8.5 compared with baseline years (1990 to 2009) during winter season Combined and individual impacts of climate and landuse change on streamflow under RCP 4.5 and RCP 8.5 during 2020s, 2050s, and 2080s in economic scenario Combined and individual impacts of climate and landuse change on sediment yield under RCP 4.5 and RCP 8.5 during 2020s, 2050s, and 2080s in economic scenario Combined and individual impacts of climate and landuse change on nitrate-nitrogen yield under RCP 4.5 and RCP 8.5 during 2020s, 2050s, and 2080s in economic scenario Combined impacts of climate and landuse change on streamflow in three future periods relative to the baseline period (1990 to 2009) in Songkhram river under RCP 4.5 ix
76 77 77
80 80 80
81
82
82
84
85
86
87
5.51
5.52
Combined impacts of climate and landuse change on streamflow in three future periods relative to the baseline period (1990 to 2009) in Songkhram river under RCP 8.5 Combined impacts of climate and landuse change on sediment yield in three future periods relative to the baseline period (1990 to 2009) in Songkhram river under RCP 4.5 and RCP 8.5
x
87
88
LIST OF TABLES Table 2.1 2.2 2.3 3.1 3.2 3.3 3.4 3.5 4.1 4.2 5.1 5.2 5.3 5.4 5.5 5.6 5.7 5.8 5.9 5.10 5.11
Title
Page No
Representative concentration pathways Long term global trend of the forest to agricultural land List of water quality models and their applicability in the various field Data required to research and their sources Thai governmental landuse reclassify criteria Available rainfall stations, with percentage of missing data information Available meteorological stations with its missing data status in percentage Table of observed discharge stations in whole Songkhram River Basin Akaike Information Criteria to Select the LOADEST model Model performance rating criteria of modeling Summary of extreme precipitation indices at various stations from 1975 to 2014 Summary of extreme temperature indices at various station from 1970 to 2014 Tabulation of maximum positive and negative anomalies relative to the average of baseline rainfall in (mm) Tabulation of maximum positive and negative temperature anomalies relative to the average of baseline temperature in (ºC) LOADEST regression performance evaluation Statistical performance of SWAT model for Flow, Sediment, and Nitrate-Nitrogen Hydrological parameter ranges in SWAT-CUP and their optimal value with sensitivity rank Optimal suspended solids parameter in SWAT-CUP with sensitivity rank Optimal Water quality (NO3-N) parameters in SWAT-CUP and their sensitivity rank Percentage change in seasonal nitrate load for different climatic models and scenarios with baseline period (1990 to 2009) for 2020s, 2050s, and 2080s Relative change in suspended solids yield under RCP 4.5 and RCP 8.5
xi
11 19 20 26 27 28 29 32 40 48 50 51 52 53 57 70 72 72 72 81 85
LIST OF ABBREVIATIONS ACCESS AIC AR5 CC CMIP5 CNRM DEM Dyna-CLUE FAO GCM GDP HRU IAMS IPCC IUCN LDD LOADEST LULC MPI NSE PBIAS PCD RCM RCP RID SOTER SRB SRTM SS SUFI2 SWAT SWAT-CUP TMD USGS USLE WMO
Australian Community Climate and Earth-system Simulator Akaike Information Criteria Fifth Assessment Report Climate Change Coupled Model Intercomparison Project Phase 5 Centre National de Research Me ‘te’ Orologiques, France Digital Elevation Model Dynamic Conversion of Land use and its Effects Food and Agricultural Organization Global Climate Model Gross Domestic Product Hydrological Response Unit Integrated Assessment Model Intergovernmental Panel on Climate Change International United Conservation Nature Land Development Department Load Estimator Landuse Landcover Max Plank Institute, Germany Nash-Sutcliffe Efficiency Percent Bias Pollution Control Department Regional Climate Model Representative Concentration Pathways Royal Irrigation Department Soil and Terrain Songkhram River Basin Shuttle Rader Topography Mission Suspended Solid Sequential Uncertainty Fitting Version 2 Soil and Water Assessment Tool SWAT Calibration and Uncertainty Program Thai Meteorological Department United States Geological Survey Universal Soil Loss Equation Word Meteorological Department
xii
CHAPTER 1 INTRODUCTION 1.1
Background
Water resource engineer must know the possible impacts of climate change and global warming, which could further stress in water availability for human use and natural ecosystems (Kim et al., 2013). Also, landuse/landcover (LULC) change such as urbanization, population growth can alter hydrological cycle by affecting evapotranspiration, evaporation, soil infiltration surface and sub-surface flow regimes etc. For long term water resource management and planning, it is very important to better understand the potential effects of landuse/landcover change and climate change on stream water quality and quantity (Kim et al., 2013). There is no any disagreement that climate and landuse change are two environmental active profound factors which affect the watershed hydrology. Intergovernmental panel on climate change (IPCC) fifth assessment report suggested that warming climate has occurred in the past half of century in every region of the world, especially in the mid- latitudes of the northern hemisphere (Ling et al., 2016; Zhang et al., 2016b). In wider contest of the climate Thailand has experienced significant warming trend during the past decades and the trend is expected to continue in future (Parry et al., 2004). No distinct trend of precipitation and augmenting trend of temperature reveal remarkable climate variability, and accelerate the hydrological cycle and swell the frequency of occurrence of extreme events (Budhakooncharoen, 2008; Reda et al., 2013). Global climate change is expected to have serious future effect for the earth’s environment. Water is one of the natural resources that are severely affected by the landuse and climate change (Minville et al., 2008). Climate projection and related records provide sufficient evidence that freshwater resources are vulnerable and have the potential to be strongly impacted by the climate change. This impact will have wide-spread consequences for various ecosystem, human being and water quality (Bates et al., 2008). In many places, climate change will most likely be expressed through change in availability in freshwater. Future climate shifts may further affect different hydrological aspects of the streams such as time to peak, timing of water availability, quantity, and quality of water available etc. Changes in river hydrology will bring risks to many sectors of water resources including irrigation, hydropower generation, navigation, water supply, floods, and droughts (Agarwal et al., 2014). Even though, climate change is a global event, mostly regional communities face its impacts. It is very necessary to assess the impact of climate change on hydrology at basin scale to know the significant change in water resources (Babel et al., 2014). In a developing country many people rely on natural resources (Devkota & Gyawali, 2015; Sayasane et al., 2015). In the recent decade because of climate change people are having the problem of farming. Changes to landuse and climate, and increasing the water demand are the most critical factors that affect water management. Generally, high stream flow is increased and low flow is decreased because of increased in urban area. Many previous studies suggested that climate 1
change has more significant effect than LULC changes (Karlsson et al., 2016). Although, impact of LULC change were minute than those by climate change, LULC changes may enhance the problems of increased seasonal variability in stream flow caused by climate change (Kim et al., 2013; Wilson & Weng, 2011; Zhang et al., 2016). The impact of climate change is more recognized on water quantity rather than quality, but landuse change has more impact on surface water quality. Understanding the water quality is very important for human and aquatic health, and is very significant in terms of cost for water management, regulation, and conservation (Perry et al., 2005). Surface water quality is mainly declining by human activities and climate change (Mujere & Moyce, 2016). As water quality is a subject of ongoing concern during last decade, quality monitoring and regulation in many river basin is needed (Wanielista et al., 1997). In some river basin quality monitoring, have been done, as the result the trend of reliable quality is gradually increased. Modern agriculture has long been recognized as a significant non-point source of water pollution (Carpenter et al., 1998). The control of agriculture pollution has generated much interest regarding both control and evaluation of management options. Water quality is directly related to climate and landuse change phenomenon, there are many research has been done related to water quantity but assessment of water quality is vital to prevent the human life and aquatic ecosystem (Tu, 2009). Assessment of point sources is not a serious issue but Non-point sources is a problem of deterioration of water quality (Basnyat et al., 2000). This study incorporated the both water quality and quantity issues at Songkhram river basin. 1.2
Problem statement
The Songkhram river basin is consider as second largest river basin in northeast Thailand and surface runoff of the basin is suffering from two drivers like direct climatic driver’s temperature, rainfall changes and another factors is non-climatic driving factors such as land cover changes i.e. anthropogenic causes (Khumsri et al., 2009). These factors can alter the surface and subsurface runoff and are very sensitive for surface hydrology. The Songkhram river basin has been experiences the tropical and semi-arid climates (IUCN,2005), so climate and landuse changes has been playing the pivotal role to change the basin flow regime. The study of these drives and its impacts on basin flow is intrinsic to form the adaptation options. Excessive loading of non-point source pollution from different landuse activities cause eutrophication in river and creeks, and are a major concern to water managers (Oeurng et al., 2016). These problems not only include eutrophication that leads to a significant loss of amenity in terms of water supply, fisheries and recreation, but also increases water treatment costs (Shrestha et al., 2008). Basically, two types of pollutant sources are considered first one is point sources pollution that can be track from any point sources like sewage from household, industry, hospital etc. Dealing with point sources pollutant is relatively easy as compare to non-point sources. The non-point source pollutant cannot have tracked back to a single origin or source such as storm-water runoff, water runoff from urban areas or discharge from sudden break down septic systems (Trang et al., 2017). 2
For effective management of water quality, assessment of loading is far important than assessment of concentration. During the last few decades, change in landuse pattern and farming practices, cause by demographic, economy, or cultural alternation, have significant effects on water quantity, water quality and soil erosion (Berka et al., 2001). As a consequences water is becoming polluted by nutrients, increasing the eutrophication hazards. Eutrophication is a continuous problem for the water engineer and managers (Daniel et al., 1998). Point sources are now relatively controlled through governmental plan policy as they are easy to notice. However, even though tertiary treatment has become more widespread in controlling point sources pollution, the limnological problem has not diminished (Shanahan et al., 1998). Attention has now switched to non-point nutrient sources. In contrast to point sources, the non-point sources such as rural and urban stormwater runoff are difficult to control as these are diffusive and chronic in nature (Das et al., 2011). The Songkhram river located in Northeast of the Thailand has played vital role in development of people economy and is the second largest catchment of northeast (Johnston & Kummu, 2012). The rural population of the basin highly depends on natural resources for their livelihood (IUCN,2005). The average income is higher than other seven districts with average US$1474/family/year. Because of changing climate and landuse the quality of water is severely declining in the basin. Meanwhile, the major occupation of the people is agriculture and fisheries, which is entirely depends on water quality and quantity (IUCN,2005). So, declining water quality from non-point sources is the major problem in the basin. Hence proper and detail assessment of the quality and quantity of the stream water is necessary in Songkhram River Basin due to significance influence of climate and landuse change. The following are the research question that are formulated during the research, 1) How the historical trend of climate, nutrient and suspended solid load is changed in study area? 2) What is the future climate and landuse in study area and which method is adopt to project the climate and landuse? 3) Which model is best to represent all these constituents in future? 4) How the flow, suspended solid and nitrate nitrogen behaves in future under different RCPs? These are the most appropriate question from the research point of view. Before start to doing any research, researcher must think about rational of the research. In this study, the above mention questions were incorporated for the best and viable outcome of the research.
3
1.3
Objectives of the study
The overall objective of the research is to evaluate the impacts of climate and landuse change on streamflow and water quality in Songkhram River Basin Thailand. The following specific objectives were achieved. 1. 2. 3. 4.
1.4
To examine the historical and future rainfall and temperature trend, To analyze the historical trend of nutrient loading, To simulate the hydrology, suspended solids, and nitrate-nitrogen and, To assess the impact of climate and landuse change on hydrology, suspended solids, and nitrate-nitrogen. Scope of the study
This research covered the analysis of individual and combined impacts of climate and land use change on hydrology, suspended solids, and nitrate-nitrogen. The detailed scope of this research includes as follows: • • • • • • • • •
1.5
Collection of hydrological, meteorological, water quality data and others GIS data, RClimDex software was used to analyze extreme climate indices, LOADEST was used to analyze the past trend of Nutrient and sediment loading, Future climate analysis based on selected RCMs, Selection of proper and recommended bias correction technique, Forecast the landuse by using Dyna-CLUE, for different time periods, Simulate the hydrology, suspended solid and nitrate-nitrogen in the basin using SWAT and SWAT-CUP model, Run the SWAT model under different land use and climate change scenarios. Evaluate the impact of climate and landuse change on hydrology, suspended solid, and nitrate-nitrogen under two RCPs (4.5 and 8.5) scenarios. Limitation of the study
• • • • •
Few regional climatic models (RCM) based on the previous research were used, Changes in future climate is projected in the form of rainfall and temperatures only, A single bias correction technique might not be able to correct all the biases, The point sources pollutant is not considered Segregated impact of landuse is not consider
4
CHAPTER 2 LITERATURE REVIEW This chapter summarizes brief introduction of climate change, landuse change, water quality and associated impacts of climate and land use change on streamflow and water quality. Some previous research in diverse areas related climate and landuse change impact on streamflow and water quality are explain in this section. 2.1
Introduction to climate change
The climate of a region or city is its typical or average weather over many years. For example, the climate of Bangkok is sunny and warm. But the climate of California is freezing cold. Earth’s climate is the average of all the word’s regional climates. Climate change is therefore, is a change in the typical or average weather of a region or city. Earth climate is always changing. The short-term variation cannot consider as climate change, if the statistical distribution of long (at least thirty years) term weather pattern has changed is called climate change (Aung et al., 2016). The change is measured with respect to average climate of the regions, where change can be positive and negative. The climate change can be bounded within a region or can happen all around the word. According to IPCC forth assessment report the mean surface temperature (minimum and maximum) of the word is likely to increase but there is no any significant pattern for rainfall, this happen because the increased temperature alters the rainfall timing, magnitude and intensity. The changing climate has direct impacts on hydrological cycle, as a result of which all the natural resources will be disturbed. The hydrological cycle is in constant introduction with climate system, when the hydrological cycle is disturb then world water balance might be change as the consequences of which local and regional community will face the problem in water availability (Jiang et al., 2007). There is some physical evidence for climate change which has been observed due to anthropogenic activities. With an increasing global warming the snow of the Himalayas and glaciers are rapidly melting, causing rise in sea level meanwhile flooding and salt intrusion in the agricultural land and destroying the agriculture. Changes in the meteorological parameter results cyclones, heat waves, droughts, and cold waves, and influence livelihoods, property, and agriculture. Due to this mostly in the developing country population suffers water scarcity, food insecurity, and hunger rate will increase. Because of the poor financial and technical problems, developing country cannot maintain the watersheds but developed country can do it. According to (Boyer et al., 2010) watershed ecosystem has changed because of the climate change ,and drought and flood have an effect on loss of water input by reducing the storage capacity. Water is vital for every sector, energy, food, agriculture, industry, health, and social life of human being as well as all living beings. The impact of climate change on water resources depends not only on change in the volume, timing, and quantity of stream flow and recharge, but also on system characteristics, on the changing pressure on the hydrological systems, on how the management of these systems are developed gradually, and what adaptation to the climate change are implemented (Tu, 2009). 5
The increasing global population and living standard across the world are increasing the demand for fresh water resources (Ouyang et al., 2015; Zhang et al., 2016a). For the longterm effective planning and management of water resources system, future land use pattern, water demand, and water availability should be analyzed before planning any long term effective program. Some of the global changes because of the climate change may not be negative, but should be think and make evaluation as early as possible because of the great importance of water resources for socio-economic value (Jiang et al., 2007). 2.1.1 Climate change in Asia According to (Calvin et al., 2012) almost 60 % of world population are in Asia. It has high contribution for world economic value, about 39% Gross World Product and 44% global energy consumption and emits nearly half of the CO2 emission. So, Asia is very important from any aspect and it has dominant role for world climate change. Carbon dioxide is one of the major GHGs, it has contributed more than half of world emissions. In this Asian region climate change, can observed because of climatic and non-climate drivers. Non-climate drivers such population, urbanization etc has dominant role for climate change. Asia regarded as one of the vulnerable regions with respect to climate change and is already observed sea level rise, heat waves, and others terrible extreme weather events (Manton et al., 2001). Many research has been done in this region regarding climate change and observed annual mean temperature is increased by 0.14°C-0.2°C by 1960s and projected will be limit up to 3°C-6°C till end of 20th century. Precipitation has no defined pattern in this region it varies with time and space. People are facing many problems in region because of frequent flood and drought. Therefore, scientific research can provide vital role for adaptation and mitigation with changing climate (IPCC,2014). 2.1.2 Climate change in Thailand Thailand is one of the largest exporter of rice. Major economic of Thailand is based on agriculture, so called as agricultural based country. From the previous record 49% of population rely on agriculture and more than 10% GDP based on it (Patz et al., 1998). Thailand has more than 3200 kilometers of coastline so tourism and fishery becomes the backbone for economic development of Thailand. Climate change becomes more terrible for all three important sectors of Thailand’s which have vital role for economic enhancements namely tourism, agriculture, and trade (Thanasupsin (B.Sc., 2013). Considering the 2008’s worldwide data for CO2 emission, Thailand is in 22 ranks. But lateral it has great improvement on it, it is estimated that Thailand produce one percentage of total world carbon dioxide emissions. Thailand release 254.8 metric ton of carbon dioxide in the years 2009, and is ranked 23 out of 217 countries. Looking up the past trend of carbon dioxide emission, CO2 emission have decreased by 0.06% by 2008-09. Over the last five years, total CO2 have worsened by 13.05%, and over the last decade the total CO2 has done 6
up by 48.67% (Marks, 2011). Increasing surface temperature, sea level rise, floods, droughts etc. are the major impacts of climate change in Thailand. According to IPCC report increase in sea level are consistent with warming. Global average sea level rose at an average rate of 1.8 mm per year over 1961 to 2003 and at an average rate of about 3.1 mm per year from 1993 to 2003. Many people were suffered from increased sea level. The consequences of the level rise are abundant, millions of people will have subjected to floods, coastal ecosystem will be destroyed, and sea level rises will exacerbate freshwater constraints due to salinization of estuaries and groundwater supplies. IPCC 2007, already warn that the cities like Bangkok located in the mega deltas are more vulnerable due to climate changes. The incidence of the climate change like frequent and intense floods, may disturb the livelihood of people, biodiversity, infrastructure etc. Several assessments across the Thailand has suggested that average annual temperature have been significantly risen by about 0.95°C between 1995 and 2009 (Parkpoom & Harrison, 2008). According to the scientific research in many places over Thailand the projected future temperature will increased between 0.4 and 4°C. Thailand will experience more warm days and less cold days. As Thailand, has observed high temperature it is obvious that numbers of warms days and nights has increased. The precipitation has no any significant trend during recent years, Thailand was flooded in 2011 meanwhile it has experiencing frequent droughts during these years. 2.2
Climate projections
2.2.1 Introduction of climate models Climate model are used to analyze the earth’s climatic system and to simulate the changing pattern of climate in coming future. Climate model to serve to express some properties of Earth’s atmosphere i.e. temperature, precipitation, moisture content, humidity, wind speed by using differential equation based on physics, fluid mechanics, and chemistry as well. To run the climate model Scientists, divide the whole planate into three-dimensional grid before applying the differential equation, and then analyze the results. To predict and understand the behavior of climate system, climate scientist makes a computer based model is called as climate model. It has perfect science; climate models assimilate the physics and chemistry of the atmosphere as well as oceans (Zhang & McFarlane, 1995). The climate model should be run forward in time much faster than the real atmosphere and oceans that is the main challenges for climate model. Scientists makes number of simplifying assumptions, and perform enormous numbers of calculations to do this. There are many types of model used to analyze the climate from different view of climate. The climate model can be, two, or three dimensional and relatively simple. Climate model is used to analyze single physical features of climate relevance, or may contain fully interactive, three
7
dimensional processes in all the three domains like atmosphere ocean and land surface (Zhang & McFarlane, 1995). 2.2.2 Global climate model Global Climate Model (GCMs) are mathematical based models. The models divide the earth, ocean, and atmosphere into number of grids to do the mathematical calculation. Global climate models are very complex and advance climate model among all because the climate systems are described in three dimensions. These are the recently available and upswing tools for simulation the changes in global climate which are the results of increasing greenhouse gases (GHG) concentration, causes of anthropogenic activities and land use changes by various natural phenomenon. The main driving force for GHG are the technological betterment, change in social-economic level, and increasing population. These models represent all the earth system like land surface and sea ice, atmosphere, ocean and able to consider all the climate variability and climate change (Fowler et al., 2007). However, because of the coarser resolution global climate model are unable to resolve significant sub grid scale features i.e. cloud and land use, topography (Grotch & MacCracken, 1991). There is various modeling center all over the world and numbers of GCMs are available from these modeling center. The horizontal grid resolution of model ranges from 250 to 600 km. And it consists of 10-20 vertical layer in the atmosphere and about 30 layers in the oceans. It has relatively high resolution and large number of processes, GCM simulation require a large amount of computer time. According to (Fowler et al., 2007), Global climate model are numerical coupled models representing the various global system such a sea-ice, atmosphere, oceans, land surface and are very advance tools for accessing the climate change and variability. However, they are very course in temporal and spatial resolution. For example, he corroborates that about 0.125º latitude and longitude is needed for hydrological modeling in the Himalayan region but Global Climate Model like HadCM3 model has spatial resolution of 2.5º latitude and 3.75º longitude. Global Climate Model have been highly used to simulate the future stream flow during the recent few decades. Global Climate Model gives much better results on global level than regional level because of its courser resolution. To fill the gap many downscaling techniques have been developed. This technique are very effective tools to forecast future climate in finer resolution. According to the performance of GCMs, some basic consideration can be made (EPRI, 2009). ▪ Global climate Model are more avail to forecasting the future temperature and precipitation. ▪ Global Climate model are more reliable to predict the mean climate and climatic extremes or variability. ▪ Global climate Model is good to predict longer period than shorter one.
8
▪
Global Climate Model has coarser resolution so its outputs are more reliable at the continental scale.
IPCC Fifth Assessment Report (AR5) released some Coupled Model Intercomparison Project (CMIP5) which the framework for a coordinated climate modeling experiment in 2013. About 20 modeling center from all over the world are campaigning the CMIP5 researchers and model data is crowd on the grid system of the earth, which consist of the international data nodes and gateways. The researcher from CMIP5 provides four emission scenarios based on Representation Concentration Pathways, which are classified by the amount of net radioactive forcing inti the earth’s surface by the end of twenty-one centuries. There are numbers of upgraded GCMs in CMIP5. CMIP5 GCM can be download from this website :(http://cmip-pcmdi.llnl.gov/cmip5/terms.html). 2.2.3 Regional climate model To make analysis of Earth’s climate system on regional scale Regional Climate Model is very powerful tools. Global climate model (GCMs) are a type of climate model that employ a mathematical model of the general circulation of planetary atmosphere or ocean (Giorgi & Mearns, 1999). GCMs uses the Naiver-Stokes equations on a rotating sphere with thermodynamics terms for various energy sources (radiation, latent heat). The output from GCMs are gridded in shape (e.g. 100 km) and at particular interval in time (e.g. daily). Regional climate model (RCMs) on the other hand focus only one a certain area of the earth, considering the corresponding atmosphere and ocean region into calculation (refer below Figure 2.1). However, RCMs run with lateral boundary conditions from GCM runs. Thus, all RCMs have parent GCM. Both GCMs and RCMs required high computational resources. The aim of running RCMs as a second step for GCM run is to expand the computational resource for a particular region of interest in order to get finer resolution output (Jones et al., 1995). Thus, RCMs help a great deal in climate change analysis in basin level studies as in the case of this research (Songkhram Basin).According to (WMO, 2014) it is very essential for management unit and policymaker to feigning the impact of climate change at the regional level.
Figure 2.1 Concepts of model domain of GCMs and RCMs (IPCC,2013)
9
2.3
Description of RCPs scenarios
Nearly a decade of publishing SRES scenario, IPCC defines RCPs scenario. According to (Moss et al., 2010),all research centers are very eager to developed the new scenarios. To develop the new scenario more minute details is required to run the model for the current generation of climate that was given by foregoing scenario sets. There is rising interest to develop new scenario by various research agencies that explore the different impacts climate change and policies to adopt with it. Such scenario also focuses to analyze the long-term coat and benefit of climate goals (Moss et al., 2010). Nowadays all scientists are more concern with adaptation in more to explore it. To do the research on climate scenario development with prospective integrated information regarding the scenario is needed. IPCC generally request the scientist to develop the new sets of climate scenarios (IPCC 2007). The community developed the new set of climate scenario that incorporate the land-use, emission, and concentration trajectories refer as representation concentration pathways RCPs (Moss et al., 2010). The RCPs is developed to provide the relevant information on feasible development trajectories for the main forcing agents of climate change, predicted by climate models. The climate modeler will handle the time series data for greenhouse gases concentration and emission, land use change from RCPs scenarios to run the climate model to generate new climate scenarios (Vuuren et al., 2011). Meanwhile IAMs (integrated assessment model) explore policy, technological change, and socio-economy that could lead for new pathway and magnitude for climate change. The terms representative address that each RCPs can represent large numbers of scenarios. And the concentration pathways endorse that these RCPs are not final scenarios at all. RCPs are the new scenarios published by IPCC in fifth assessment report (AR5). They invent four pathways (RCP2.6, RCP4.5, RCP6, RCP8.5), based on radioactive forcing reach to earth atmosphere. This radioactive forcing is measure in W/m2. When the incoming solar radiation and outgoing solar radiation is not balance that change the basic composition of earth’s atmosphere. Sometimes incoming radiation is higher than the outgoing radiation this unbalance solar radiation is called as radioactive forcing (Bank, 2012). The major characteristics of RCPs Scenarios are tabulated below in Table 2.1.
10
Figure 2.2 RCP-scenarios and corresponding additional radioactive forcing value through 21st century (IPCC, 2007) Table 2. 1 Representative concentration pathways, Sources:(Vuuren et al., 2011) Name RCP 2.6
CO2 Concentration (p.p.m) 2 Peak at 3 W/m before Peak at 490 before 2100 and then decline 2100 and then decline Radioactive forcing
RCP 4.5
4.5 W/m2 at ̴ 650 (at stabilization stabilization after 2100 after 2100)
RCP 6
6 W/m2 at stabilization ̴ 850 (at stabilization after 2100 after 2100)
RCP 8.5
2.4
>8.5 W/m2 in 2100
>1370 CO2-equiv in 2100
Pathways Peak and decline Stabilization without overshoot Stabilization without overshoot Rising
Temp. anomaly (ºC) 1.5 2.4
3
4.9
Bias correction
The RCM data cannot use directly in the hydrological modeling because RCMs outputs remains relatively course in resolution and is unstable to resolve sub-grid scale features such as landuse, topography and cloud. Thus, the spatial resolution of RCM data generally considers insufficient for representing local variability for climate change impact studies. To some extent all the numerical models suffer from systematic error i.e. the numerical differences between simulated and observed value (Piani et al., 2010). Time independent error component is known as bias. Some form of preprocessing is required to remove the bias present in the simulated climate output, before use for any particular purpose(Sharma et al., 2007). In the simplest formulation of bias correction only the changes in a specific statistical aspect of simulated field is used (Haerter et al., 2011). Before running the hydrological models, the bias of model output should be corrected. Mainly change in variance or mean is employed. The bias can be positive or negative in accordance with the variability so correction methodology may be multiplicative or additive (Teutschbein & Seibert, 2012). 11
Great care is needed while choosing the bias correction methods because it might add more uncertainties in analysis (Chen et al., 2012). They accounted three source of uncertainty as choice of RCM, Choice of future RCPs scenarios, and choice of bias correction parameters which account minor source of uncertainty as compare to other sources. (Haerter et al., 2011) explain the complication for bias correction of climatic parameters which are higher for precipitation as compare to temperature. Depends upon their study area and availability of data, researcher can adopt any bias correction method for analysis, some of which are listed here (Lafon et al., 2013), Delta Change approach, Multiple linear regression, Local intensity scaling, Quantile mapping, Constructed Analogue, Power transformation Method, Variance Scaling Temperature, Gama- Gama Transformation- Distribution Correction Method etc. Before fixing the bias correction method details study about climate, altitude, and other topo-geographic information of the study area is required. Which suggest the appropriate bias correction method for the study site (Christensen et al., 2008). 2.5
Hydrological modeling
To evaluate the availability of water resources under any climate scenario and land use change scenario, hydrological modeling is required. Different climatic variables like temperature, evapotranspiration, precipitation as well as land use change and catchment topography can be used in hydrological modelling to simulate the runoff. (Moradkhani & Sorooshian, 2009) imply that a model is a simplified representation of real word system. Hydrological model are very effective tools to assessment the water balance or water availability given in a different climatic condition and different landuse scenarios. The best model is the one which give results close to reality with the use of parameter and model complexity. Models are mainly used for predicting system behavior and understanding various hydrological processes. In brief hydrological modeling, can be divided into two groups namely physically based distributed-parameter models and lumped models. There are many factors which play vital role while choosing the model, among them propose of study and availability of data are the governing factors (Jiang et al., 2007). In terms of complexity, to prepare the input requirements to set the model, physically based models are more complex but are more accurate in terms of results. So, most of the hydrologist prefer physically based distributed model to analysis the basin hydrology. But in other hands, simple model can also give adequate results at highly reduce cost but the result of these modes has very small range of application. Thus, experience of modeler recognizing in which situations simple model can be used and in which situation complex model must be elect (Middelkoop et al., 2001).
12
To see the special pattern of hydrological response within the basin, process-based distributed-parameter models are needed. Fully distributed models require lots of inputs data and involves many steps in analysis, which might not be possible in most of the cases. Because of this drawback of fully distributed models, semi-distributed model like SWAT, WatBal, SWIM, HYLUX etc. are excessively used. 2.5.1 Soil and water assessment tools (SWAT) The united states departments of Agriculture Research service (USDA-ARS) developed a process based semi-distributed hydrological model which is called as soil and water assessment tools (SWAT). According to (Arnold et al., 1998) SWAT provides a platform for GIS and hydrological model integration. Soil and water assessment tools has been used as very effective and user friendly tools to analyze impact of climate and landuse change on hydrology and water quality (Gassman et al., 2007). SWAT is physically based semi-distributed continuous-time model. This model is operate in daily time steps and able to predict the movement of water in complex watershed with varying land use, management condition, soils over the long periods (Neitsch et al., 2002). In the SWAT model basin is divided into numbers of sub-basin and again further divided into one or more response units based on soil type, topography, landuse etc. These response units are called hydrological response units (HRU). HRUs should have same geomorphological and hydrological properties (i.e. based on same slope, soil type, land use) (Flügel, 1995). The HRUs should have same response to the given inputs likes temperature, precipitation (Ning et al., 2015). There are many methods to measure the surface runoff but in SWAT modified curve number method is mostly used (Shrestha & Htut, 2016). Three methods are more common to estimate the potential evapotranspiration in SWAT model i.e. Hargreaves, Priestley-Taylor and Penman-Monteith (Xiong et al., 2013). To make the input files, SWAT requires climate data, land use and land cover information, topography, Digital elevation model etc. DEM very important for SWAT model. Based on DEM information, modular can delineated the watershed because DEM can be consider as input data for geographic study, climate change, hydrological modeling, environmental impacts etc. (Shrestha & Ye Htut, 2016). The hydrological cycle in the SWAT is based on the water balance equation (Zhang, Y. et al., 2016b). 𝑡
𝑆𝑊𝑡 = 𝑆𝑊𝑜 + ∑𝑖=0(𝑅𝑑𝑎𝑦 − 𝑄𝑠𝑢𝑟𝑓 − 𝐸𝑎 − 𝑊𝑠𝑒𝑒𝑝 − 𝑄𝑔𝑤)
Eq. 2.1
Where, SWt = final moisture content in soil (mm) in t days, SWo = Initial soil water content (mm) on day I, Rday = Amount of precipitation, Ea = Evapotranspiration (mm), Qsurf = Surface runoff (mm), Wseep = Percolation (mm), Qgw = Base Flow (mm).
13
2.5.2 Model performance evaluation The model performance during calibration and validation was carried out by four different statistical parameters namely coefficient of determination (R2), percent bias (PBIAS), RSR and Nash-Sutcliffe efficiency (NS) in this research (Moriasi et al., 2007). The coefficient of determination shows the relationship between observed and forecasted values. The value ranging from 0 to 1, where higher value relies stronger the relation between observed and simulated data. Another important statistical indicator to measure the model performance is Percentage bias, it is a degree of measurement of the average tendency of the simulated data to the observed data. This value can be either positive or negative. Zero is the optimum value that indicate perfect model simulation where as positive value indicates under estimation of model and negative value indicate over estimation. PBIAS is expressed in terms of percentage which is the magnitude of the data being deviated. Other statistical parameter such as percentage of streamflow volume error, prediction error and percent deviation of volume are also used to express the deviation in stream flow (Legates & McCabe, 1999). Nash-Sutcliffe efficiency (NS) is the most widely used performance analysis of the model in which the ratio of mean square error to the variance in observed data is subtracted from unity. The values of NS ranges from 0 to 1 where the higher value indicates the perfect result and the lower value shows the poor result (Moriasi et al., 2007). RMSE-observation standard deviation ratio (RSR) is more powerful indicator to evaluate the model performance. It is the ratio of root mean square error and standard deviation of the given data. The RSR value varies from 0 to large positive number. The lower RMSE gives the lower RSR value, that mean better model simulation performance (Legates & McCabe, 1999). 2.6
Climate change impact on hydrology
Impact of climate change on hydrology has been debatable issue. There are two factors which affects the hydrological cycle, one of them anthropogenic cause has significant impact on water cycle. Because of many human activities, the ratio of greenhouse gas emission is in increasing order (Park et al., 2011). People are more focus on industrialization therefore mean global temperature has been in increasing order. Because of the global warming the water availability in global, regional, or even in basin scale would be changes. Clearly seen that the climate system deeply interacts with hydrological cycle. Such effects may increase the frequency of floods and droughts, fluctuating magnitude of runoff and timing, changes in rainfall pattern, declining the water quality and quantity, frequent extremes events etc. These all phenomena are because of climate change (Dahal et al., 2016). These changes directly influence on water supply system, power generation, irrigation, navigation, sediment transportation and deposition and ecosystem conservation. Without doing proper research nobody can speak about the positive or negative effects. So, to know the impacts of these changes, need to do proper research as early as possible. Sometimes climate changes become so devastated for ecosystem, human beings etc. Water is very important than other natural resources, it has great importance for all living beings, if the 14
water availability is decreases it has serious negative impact for all the sectors (Babel et al., 2014). The water resources are very vulnerable to climate change (Navarro et al., 2016). Many researchers have been evaluated the potential impacts of climate and landuse change on water availability on river basins. For instance,(Shrestha et al.,2016) investigate the impact of climate change on stream flow of upper Bago River Basin, Myanmar. This study was based on discharge simulation by using SWAT model and found that the water availability in Bago Basin is decreasing order during the future periods. (Shrestha et al., 2016) they have conducted research on Indrawati River basin, Nepal. Three CMIP5 models for two RCPs scenarios were used. SWAT was used for simulating future water availability. They found that temperature of the basin has been increasing and there is no any significant trend in annual precipitation. But annual discharge is projected to increase in future for Indrawati River. It is not necessary that temperature should be increase, it can be increase or decrease. Mostly it depends on location of catchments and many other factors has contributed on it. If the rainfall is decrease leading to water crisis. Some areas rely on rainfall for their agriculture, livelihood, and for many other events. If the rainfall pattern is change due to climate change the whole lifestyle of people will be disturbed. Likewise, if flooding occurs on watershed it may disturb all the infrastructure, farmlands, water supply scheme, hydropower production, and creating problem for water storage for the local governments and farmers as well. Therefore, from the above scenarios, both increasing and decreasing from mean pattern is devastating for human livelihood. 2.7
Introduction to landuse
Land cover directly affects biogeochemical cycle, biodiversity,, and peoples livelihood through land surface processes (Trisurat et al., 2009). Studying change in landuse change dynamics, therefore, can help to predict the effect of landuse changes on land degradation, and vulnerability of the places. Among the impacts of landuse changes, the consequences on biodiversity and availability of natural resources are the most sensitive one (Pontius et al., 2008). Landuse changes is the major causes of tropical deforestation (Angelsen & Kaimowitz, 1999). Deforestation results various negative consequences like increase in sediment yield, decrease in rainfall, increase in variability of surface runoff, and induce flood in rainy season and drought in dry season. Moreover, changing the landuse from forest area to agriculture areas will have profound impacts on hydrology and affects the water quality of the regions. Changing landuse and land management practices alter the natural flow of water, and changes its timing and amount. Clearing forest for agriculture can also be linked with shortage of water in the Northern Thailand and has been one of the top debates in the country (Becu et al., 2003). In a developing countries land use change becoming a serious issues regarding the environmental fluctuation such as increasing greenhouse gas emission, soil degradation, decreasing forest product, decreasing agricultural products, soil erosion etc (Fahrig, 2003). 15
There are many driving forces of LULC changes such as technological, biophysical, socioeconomic, geophysical, and combined factors which are more complex (Zheng et al., 1997). To know and analyze the impacts of land use, modeling technique is excessively used and gives better results than other techniques. Modeling technique allows the quantification of past LULC changes, which provides the past trend and challenges theories regarding the LULC change impact on stream flow and makes beliefs about the driving factors of LULC change, and gives better frameworks and results for future (Braimoh & Onishi, 2007). Modeling technique is good to give the answer the question like where and how the LULC is changing, at what rate LULC changes (Qasim et al., 2013). Land use land cover changes is the emerging issues for the environmental planner and manager. The rate of land use land cover is in excessive high. Planner and manager may not concern only in LULC changes; they should think about the location. For example, agricultural lands are changes into urban areas, forest area are degrading in a very devastation way so that planner should think about the location of the changes and should provide better preventive measure. It is being recognized that accounting for spatial variation is essential for understanding varying land use processes (Qasim et al., 2013). Due to improved data handling capacity and increasing access to spatial data, such analysis is becoming more feasible even in the undeveloped areas. This technique (spatially explicit statistical modeling) has been shown to be an effective approach, however different land use land cover data are taken from remote sensing, multivariate statistics can be used to correlate their spatial occurrence with location (Sankhayan et al., 2003). 2.7.1 Landuse change modeling Landuse/ cover changes due to a wide range of highly complex interactions among different drivers of changes. (Lambin et al., 2003) has summarized five fundamental factors of landuse changes: i) increased production pressure due to resources scarcity; ii) changes in market opportunities; iii) outside policy intervention; iv) loss of adoptive capacity and increased vulnerability; and v) changes in social organization. Thus, landuse change = f (Pressure, Opportunities, Policies, Vulnerability, Social organization). The pressure can come from different sectors like population of resources users, labor availability, quality of resources and sensitivity of resources. Opportunity is the function of market prices, production costs transportation costs and technologies. Policies cover subsidies to any dominating classes, taxation system, property rights, infrastructure, and governance. Vulnerability =f (exposure to external perturbations, sensitivity and copying capacity). Social organization = f (resources access, income distribution, household features and urban rural interaction). Land use change (LUC) alter the natural ecosystem, climatic condition, biodiversity, biochemical cycle and brings some unpredictable changes in human activities. Basically, modeling technique is used to quantify the land use change and forecast the future land use change. Modeler can describe the causes by which land use changes have been occurring by using the model. Modeling itself is very complex task and land use modeling is more. To 16
found out some possible trajectories of land use change three main drivers should be keep in mind like socio-economics, biophysical processes, political strategy to manage the land. These factors should be considered for both temporal and spatial aspect. Many techniques can be used while developed the land use change model and some of them are categorized as expert system, statistical, cellular, systems dynamics, based on mathematical equation and hybrid (Parker et al., 2003). There are two types of modes used for study of land use change. One is spatially explicit models and another is non-spatially based models. Many researchers prefer spatially explicit models because it gives how the pattern of land use change with respect to driving forces in the spatial and temporal sense. Some example of spatial explicit, none-economic models of land use change are listed here (Irwin & Geoghegan, 2001),Cellular automata (CA),Simulation models of urban growth, Empirical models of land use and Hybrid models of land use change (DynaCLUE). 2.7.2 Dyna-CLUE model Dyna-CLUE includes both spatial and non-spatial consideration and is flexible on determining the drivers of the landuse change. it is applicable to a wide-range of conditions to developed future landuse scenarios based on the management policies. It is capable of producing spatial variation in landuse, which is required in any research work to analyze the effects to both landuse and climate change using SWAT model with the interface of ArcGIS. Dyna-CLUE has two modules the first one is non-spatial (demand) module and another is spatial allocation module (Zhang et al., 2016). The non-spatial module determines demand of various landuse through past landuse trends or scenario-based landuse and then translates the demand of land used by the spatial allocation module (Verburg & Overmars, 2009). Data required for Dyna-CLUE model are past landuse map to set up and validate the model, digital elevation model, soil properties, rainfall, and demography. Dyna-CLUE can simulate the time series map, the simulated map should be compare with the observed one to validated the model. The validation error is quantified using Kappa statistical analysis (K). Value of K ranges from 0 to 1; 1 represent perfect simulation match of image and 0 meaning complete disagreement of model set up. 2.7.3 Landuse change impact on streamflow Many research has been done to estimate how much fresh water is available in the earth. Researcher got result about 2.5% water is available as a fresh water and two third of its volume is occupied by ice and snow. Hence only 0.77% is freely available as a fresh water (Postel et al., 1996). The important sources of fresh water are ground water withdrawal and rainfall. The ground water is considered as non-renewable sources of fresh water like an oil, so only the rainfall is renewable sources of freshwater. The rainfall pattern and frequency is varying from place to place, it has great influence on human livelihood because most of the people highly depends on water availability from rainfall directly or indirectly. Therefore, 17
any observable change in rainfall pattern and frequency will have great effect on human activities and progress of the region. The quality and quantity of water available in term of runoff from the rainfall are highly depends on climatic condition and how land use changes in area. Land use changes such as decreasing agricultural area, increasing urbanization, decrease in forest etc. will have dominant impact on water quality and quantity of the region. (Thanapakpawin et al., 2007) conducted a research in Mae Cham River basin in Thailand, research found that most of the crop lands are expanded on the forest area in the lower and upper part of the basin. This gives profound impact on stream flow and hydrology of the catchment. The study makes four different scenarios and use DHSVM model to simulate runoff, result gives 4% increase in dry season flow hence runoff increase due to various human activities (Githui et al., 2009). In the developing countries climate change and land use change are two major factors which influences the stream flow. (Sayasane et al., 2015) used logistic regression method to predict the future land use and SWAT was used to simulate the future stream flow. The study found that most of the wood and shrub land was converted into agricultural land in the middle part of Nam Xong Watershed. The stream flow will be decrease due to climate change and land use change over the 20 years, stream flow will decrease by 11.7-12.2%. Land Use change will happen because of the human activities. People would like to shift to urban area rather than rural area, so urbanization is the major driver for LUC. Wolf Bay Watershed was dramatically converted to urban area. By 2030 the percentage of urbanization will be double. Most of the forest land were converted into urban area. LULC was very less responsible to change the stream flow in basin (Wang et al., 2014). Water Resource Manager should importantly consider landuse change for development of tenable water resources system, because future land use change may have impact on hydrology individually and also combined with climate change in future (DeFries & Eshleman, 2004). In a tropical area, conversion of natural landscape to agriculture one is particularly evident. The estimates in a global extent, the conversion over last three centuries shows that the forest clearing for cropland has increase significantly since 1950. The following table shows the long term global trend of the forest land to agricultural land. A report publishes by ITTO (2002) estimates 850 Mha. (Million hector) of forest land has been degraded. In addition, many tropical regions are under rapid urbanization. The global population living in cities has increased from10% in the 1990s to more than 50% in 2008. It is further expected to reach to 60% by 2030s (Altintas et al., 2011). The majority of this increase is occurring in the tropical region of Asia (UNPD,2006). These changes in landcover changes will have direct and indirect impacts to integrates streams and watershed.
18
Table 2. 2 Long term global trend of the forest to agricultural land (Source: Mayers, 1996) Land area (x106ha) Year Forest Grassland Cropland 1700 6215 6860 265 1850 5965 6837 537 1920 5678 6748 913 1950 5389 6780 1170 1980 5053 6780 1500 2000 3800 1360 2.8
Introduction to nitrate-nitrogen
The term “quality” is very important in water resources because it has pivotal role to impact on aquatic life and river health. These days’ the amount of nitrate nitrogen in surface and subsurface water bodies are amplifying (Altman & Parizek, 1995; Sebilo et al., 2003). Many studies have been done in diverse regions of the world and they found that more nitrate nitrogen exported from watershed soil i.e. mainly from agriculture and more than 60% is eliminated by denitrification, before reaching to the water bodies (Mayer et al., 2002). During the recent decades, the amount, of nutrients reach to the ocean is amplified by three times. The whole nitrogen cycle is altered because the nexus between food, energy and water has been changed because of the increasing population. There are abundant sources of nitrate, among them pesticide and fertilizer application into agricultural land, effluent from waste water treatment plant and burning processes are the major (Oeurng et al., 2016). The process by which a body of water becomes enriched in dissolved nutrients (as nitrogen) that stimulate the growth of aquatic plant life usually resulting in the depletion of dissolved oxygen called eutrophication. High concentration of nitrate-nitrogen in water bodies and aquifers are major concern for the water and environmental engineer because river health is very important for human lifecycle and aquatic ecosystem (Cambardella et al., 1999). Hydrological cycle is altered by the climate change so the amount, of nutrients and its characteristics is change. Climate is the strong driver which has strong influences to the biological, physical as well as chemical process of nitrogen cycle (Worrall et al., 2009). While the natural flow is reduced the retain, transport and transform of nutrient load of upstream region would be alter. To have a better understanding of nutrient loading from watershed, analysis of extreme event are important because during flooding period the mobility of the nutrient load from catchment to river will be high. (Oeurng et al., 2016) suggested that nutrient transport will be temporally varies according to the seasons, even within season the load is strongly vary because of not uniform flow frequency. (Zweimüller et al., 2008) had one research in Austrian Danube river to analyze the climate change impacts on nutrient loading and the finding was the load will be high during winter and low during summer as a consequences of temperature change. Climate change can have both positive and negative impacts on nutrient loading and the effect is different for diverse region. So the important task to researcher is to find the variation in riverain nutrient load on a regional and catchment basis (Oeurng et al., 2016). 19
2.8.1 Water quality model Various water quality models have been developed to simulate the nitrogen yield, transformation processes, and enrichment in basin or regional scale. These models are able to study the point and non-point source dynamics of nutrient loading. The Water Quality Analysis Simulation Program Version 4 (WASP4) is a dynamic compartment-modeling system that can be used to analyze a variety of water-quality problems in a diverse set of water bodies. The WASP4 model can simulate the nutrient loading from various water bodies like pond, stream, reservoir, lake etc. And hydrodynamics, eutrophication-dissolved oxygen kinetics, conservative mass transport, and toxic chemical- sediment are the four principle components of WASP4 model. So, this model is considering very effective in the era of water quality modeling (Beck, 1987). In the field of water quality modeling there are so many water quality models are working on. Among them hydrological simulation program Fortran (HSPF), Agricultural Non-point source pollution Model (AGNPS), soil and water assessment tools (SWAT), water quality analysis simulation program ((WASP), Better assessment science integrating point and nonpoint sources (BASINS), integrated nitrogen in catchments (INCA), simulation for water resources in rural (SWRRB), The chemicals runoff and erosion from agricultural management systems (CREAMS), Ground water loading effects of agricultural management systems (GLEAMS), Storm water management model (SWMM), Erosion prediction impact calculation (EPIC)(Beasley et al., 1980; Di Toro et al., 1983; Duda et al., 2003; Gironás et al., 2010) are the commonly used water quality models. Table 2. 3 List of water quality models and their applicability in the various field Model Name
Field Scale, Agricultural watershed, urban watershed
1
AGNPS/AnnAGNPS
Agricultural
Empirical
Event/Continuous
2
ANSWERS
Agricultural
Physically based
Event
3
BASIN(SWAT)
Agricultural
Semi-Empirical
Continuous
4
DWSM
Agricultural
Physically based
Event
5
HSPF
Agricultural and Urban
Physically based
Continuous
6
GLEAMS
Field
Empirical
Continuous
7
PRMS
Agricultural
Physically based
Both
8
KINEROS
Agricultural and Urban
Physically based
Event
SN
Types
Temporal Scale
9
MIKE-SHE
Agricultural
Physically based
Both
10
SWAT
Agricultural and Forest
Semi-Empirical
Continuous
11
SWMM
Urban
Physically based
Both
20
2.8.2 Impact of climate and landuse change on water quality Linking the impact of climate and landuse change on water quality is very complex because water quality is highly depending on water quantity. The water quality is fluctuating by both climate and landuse change. There have been many studies related to stream water quality with changing landuse and climate, by using many modeling technique like SWAT, correlation, regression, QUAL2K etc. (Mehdi et al., 2015) conduct the research to evaluate the combined impact of climate and landuse change on stream water quality and quantity in Bavaria Germany. The SWAT model was used to evaluate impacts of CC in addition with LUC on stream quality and quantity. Researcher analysis only nitrates and phosphorous as water quality parameter and founded that 3 folds increase in nitrates and 8 folds increased in phosphorous due combine impact of CC and LU change as compare to CC alone. (Khoury et al., 2015) analyze the water quality and quantity in Canadian river basin by using SWAT model, result shows that climate change will drive the monthly streamflow, nitrate, and organic phosphorus loads up, while decreasing organic nitrogen and nitrate loads, while land use changes will drive the same variables in the same direction as climate change, except for organic nitrogen, for which effects of the two stressors are opposite. The magnitude of the impacts is not same are vary according to the water quality or quantity parameters, changes in streamflow was highly driven by climate change whereas changes in water quality parameter driven by landuse changes (Trang et al., 2017). (Fan & Shibata, 2015) study impact of climate and landuse change on both stream quality and quantity in Teshio River Basin japan. Result shows that the trend of precipitation is in rising trend due to climate change which reflects increase in surface flow, baseflow and ground water. Agriculture activities is increasing in the basin, farmer rely on fertilizer application result of which N and P yields also increased. Study found climate change has more significant impacts on both quality and quantity of water in basin. There was very strong relationship between hydraulic processes and fertilizer application, especially in sediment and nutrient yields during the rainy seasons. The loads P, N and sediments were mainly derived by the landuse change. Another similar study by Li et al., (2008) using regression and correlation analysis, shows the facts that percentage of urban was the predictor for water quality, if the sewage and waste from urban area is not manage properly the stream water quality will decrease. The evaporation from Rivers, Lakes, and stream would increase in Vietnam under the climate and landuse change scenarios in future. Thus, the water quality and quantity will decrease in near and far future (Truong, 2008). There was many research in which authors shows that certain land uses which is mainly have positive effects on environment also have a negative impact on water bodies (Mehdi et al., 2015).
21
2.9
Load Estimation technique
2.9.1 Load estimation using field data It is very desirable to know the amount of suspended sediment and chemical constituents that comes into the river from watershed. To measure the daily concentration of these constituent is not possible so generally those constitutes are measure in every four-fivemonth interval. This data is not well enough to analysis the past trend of the water quality and suspended sediment. So, US geological survey department developed a FORTRAN based program which is called Load Estimator. LOADEST is very powerful tool to calculate the daily load form discrete monthly based concentration. The load (weight of the material transported during specific time period) of water quality constituent in a stream is the function of the concentration of the constituent by the stream discharge(Shrestha et al., 2008). The total mass of the loading over any time can be calculated from following relationship: 𝑡
L= [∫0 𝐶𝑄𝑑𝑡]
.
Eq. 2.2
Where L is the total load, C is the concentration of constituents and Q is the stream discharge. Continuous measure of Q and C are rarely available so equation (2.2) is problematic for discrete dataset. So in practice this equation cannot be use directly to calculate the pollutant loading, as the sampling period for discharge and concentration is longer than the period over which concentration and discharge are invariant(Runkel et al., 2004). So, following equation is more commonly used to calculate the sediment and nutrient loading. 𝑁𝑃
𝐿 = 𝛥𝑡 ∑𝑛=1(𝐶𝑖𝑄𝑖)
Eq. 2.3
Where NP is the number of discrete points in time and 𝛥𝑡 is the time interval represented by the instantons load. Development of load from this equation is the three steps process i.e. Model formulation, Model Calibration, and Load Estimation. 2.9.2 Load estimation methods used within LOADEST The load estimation processes will be complete by retransforming the bias of the data, censoring of data and checking the nonnormality. According to (Bruce & Ferguson, 1986) the rating curve estimates of instantaneous load are bias, estimates may be underestimate the true load by much as 50 percent load. Data can be sensor by comparing the measure value in the site and in the laboratory. If there are no differences in the concentration between those two measures, the load calculated from these data will be accurate. Therefor a more rigorous treatment of the data is required for perfect analysis on loading. A final complication is the assumption of ordinary least square(OLS) regression model that the model residuals are normally distributed. There are so many methods to calculate the model coefficients, but other methods are applicable when the model residuals do not follow the normal distribution. 22
Adjacent maximum likelihood Estimation (AMLE) is used in this study to calculate the sediment and nutrient loading (Runkel et al., 2004). 𝑚
𝐿 = 𝑒𝑥𝑝(𝑎0 + ∑
(𝑎𝑗 𝑋𝑗 ))𝐻(𝑎, 𝑏, 𝑆2, 𝛾, 𝑘)
Eq. 2.4
𝑛=1
The Akaike information criteria (AIC) is a measure of the relative quality of statistical models for a given set of data. AIC= 2k-2ln(L)
Eq.2.5
Akaike Information Criterion (AIC) is a model selection tool. If a model is estimated on a particular data set (training set), AIC score gives an estimate of the model performance on a new, fresh data set (testing set). It can be shown that, for Gaussian models with known residual variance, AIC is equivalent to an estimate of the in-sample error of the estimated model (true prediction error on the training data set). The lower AIC score signals a better model. To use AIC for model selection, we simply chose the model giving smallest AIC over the whole set of candidates. AIC attempts to mitigate the risk of over-fitting by introducing the penalty term 2 * d, which grows with the number of parameters. This allows us to filter out unnecessarily complicated models, which have too many parameters to be estimated accurately on a given data set of size N. AIC has preference for more complex models compared to Bayesian Information Criterion (BIC). In some textbooks and software packages an alternative version of AIC is used, where the formula above is divided by the sample size N. Such definition makes it easier to compare models estimated on different data sets of varying sizes.
23
CHAPTER 3 STUDY AREA AND DATA COLLECTION Songkhram river basin is selected for the research. The basin is located in northeast part of Thailand. This chapter covers climatic condition, water quality, geographic, topography, landuse, and hydrological condition around the basin. 3.1
Study area
3.1.1 General description of area The Songkhram river has the second largest catchment of any basin in northeast Thailand after the Mun-Chi Bain and is one of the largest Mekong river tributary. Sonogram river brings the discharge about 350m3/sec, which is about 2% of the Mekong river (IUCN,2005). The catchment area of the Songkhram river basin is 12,880 km2 and the total length of the river is 420 km. At the end of the catchment there is no any hydrological station so outlet of the basin is considered at Kh.55 which is laying 10 km upstream of Mekong confluence point and catchment is 12,430 km2. This river flows through four provinces i.e. Udon Thani, Nong Khai, Sakon Nakhon and Nakon Phanom and collectively call as Sakhon Nakhon Basin (Judprasong et al., 2006). According to the past trend flood frequency is increased in rainy season and people are facing the water storage problem in dry season. To enhance the agriculture product people, use fertilizer so water quality is also decrease. Most of its lower reaches the river menders over an excessive floodplain at an altitude of between 145 m to 160 m from mean sea level, with gentle gradient of about 1: 30,000. Songkhram river supports an estimated population of 1.7 million people (Blake & Pitakthepsombut.,2006). Oon and Yam are two major tributaries of the Songkhram river. Songkhram river covers the 33 district out of which 12 districts cover by lower Songkhram basin. One study was conducted on Songkhram river basic, on the basis of which 54% of Songkhram basin is wetland and out of which 38% was comprises of wet rice fields (Blake & Pitakthepsombut., 2006). As per (HALLS, 2010) Songkhram river is in natural flow condition, no dam was constructed along the mainstream. High flow can be observed in rainy season. According to Mekong Wetland Biodiversity Conservation and Sustainable Use Program (MWBP), Lower Songkhram covers 3000 km2 and the density of population is very high as compare to other part of basin. Approximately 130 villages are in the lower Songkhram with an approximate population of 108000. Most these population are closely linked with the natural resources (Blake & Pitakthepsombut.,2006). Even today, many families rely on fishing and trading fish for their livelihood.
24
Figure 3. 1 Location map of Songkhram River Basin 3.1.2 Climatology The Songkhram river basin is experiences a tropical and semi-arid climate with three distinct climate seasons namely summer (Jan-Apr), Monsoon (May-Oct) and Winter (Nov- Dec) (Blake & Pitakthepsombut.,2006). About 90% of annual rainfall can be observed in rainy season. The annual rainfall has vast deviation in between southern and northern part. Approx. less than 1,200 mm per annum in southern and more than 2,100 mm per annum in northern part of the basin has been observed. The annual evaporation rate is varying between 1,5582,054 mm/year. The relative humidity is ranges from 65 to 78% according to seasons. The mean annual air temperature of the basin is 31.5ºC. Minimum temperature rarely declines below 10ºC in winter season but in the summer temperature can goes up to 40ºC. 3.1.3 Topography and geology The SRB between lies in between 17º 20’ to 18º 10’ N and 103º 30’ to 104º 30’ E and it can drain water from 12,700 km2 of area. Majority parts of Songkhram River Basin consist with flat marshland where the elevation ranges from 135 m to 675 m from the MSL with gradient of about 1: 30,000 (Blake & Pitakthepsombut.,2006). SRB is a natural floodplain grassland wetland with a creek system with levees, scrubs, savannah, and herbaceous swamp. In the lower parts of the Songkhram, many tributaries like Mao, Oon Yam Rivers from South, and 25
Hi Rivers from North join to the mainstream. All this tributary and Lower Songkhram river creates the lowland flood plain. It is drained by Songkhram river which enters the Mekong River in Tha Uthen district of Nakonphanom province, about 40 km north of Nakonphanom city. There was a survey conducted by a project and they found that only 23.56% of soils is potentially irrigated and were not saline, but the rest of the agricultural being saline or with the potential to be saline. The topography of the study area is extracted from digital elevation model (DEM) with 30 resolution, which is downloaded from ASTER website. This DEM is used to set up the hydrological model SWAT and to extract the basic information about the study area. The DEM is processed for the study area with the help of arc hydro tools. The projected coordinated system of the study area is WGS1984 UTM_ Zone 48N. The processed DEM of the study area is shown in figure below. 3.2
Data collection
This section provides the information regarding the required data to conduct the proposed research and some preliminary analysis on data. Mainly Spatial data, water quality, hydrological and meteorological data are required to pursue the research. The detail information of data is presented below. Table 3. 1 Data required to research and their sources SN
Data type
Time
Frequency
1
DEM(Topography)
30m X 30m
Raster
2
Soil data
Raster
Raster
3
Land Cover data
Raster
Year
4
Meteorological data (Tmax, Tmin, Rain)
1975-2014
Daily
5
Hydrological data (Q) 1990-2014
Daily
6
Water quality data (Nutrient, Sediment)
Daily
1996-2014
Department
ASTER 30 Land development Department, Thailand (LDD) Land development Department, Thailand (LDD) Meteorological Department Thailand (TMD) Royal Irrigation Department Thailand (RID) Pollution Control department Thailand (PCD)
3.2.1 Soil data The soil data required to conduct the research is collected from land development department Thailand. The assess of the soil property was analysis through ArcGIS tools. The map is available in the shapefile format so which can directly use to set up the hydrological model. There are seven types of soil are present and all analysis are done based on these soil types. And the digital elevation model was downloaded from ASTER website with the 30m grid resolution. 26
Figure 3. 2 DEM and soil classes of the study area 3.2.2 Landuse data Past landuse map were collected from land development department (LDD) Thailand. Four land use map were available and past analysis was conducted based on these available landuse map. The spatial resolution of all these available map is 1:25,000. The landuse map of 2002,2007 and 2009 consists of 13 classification whereas 2014 landuse map consists of 31 classification. For the analysis landuse type of collected map were reclassified into 7 classes as defined by the LDD. The LDD landuse classification guidance is shown in given Table 3.2. Table 3. 2 Thai governmental landuse reclassify criteria Code
Aggregated class
0
Crop
1
Built-up land
2
Planted trees
3
Paddy
Thai landuse classification Field crop, Horticulture, Integrated farm/ Diversified farm, Orchard/Horticulture, Village, Transportation, Communication and Utility, Industrial land, City, Town, Commercial, Village/Orchard, Other built-up land, Pasture and farm house, Institutional land Perennial, Orchard, Perennial/Orchard, Forest Plantation, Field crop/Perennial, Field crop/Orchard Paddy field
4
Water Body
Reservoir, Marsh and Swamp, Natural water body, Aquaculture land
5
Miscellaneous
Other miscellaneous land, Mine, pit, Rangeland
6
Forest
Evergreen forest, Deciduous forest
27
Figure 3. 3 Landuse map and percent area cover by various landuse classes (2009) 3.2.3 Historical climate In this study past and future climate change is measure in terms of rainfall, minimum and maximum temperatures. Daily data was collected from Thai Meteorological Department (TMD). There are thirty rainfall and six temperature stations. Geographical situation of all these rainfall and precipitation are presented in below figure. Here historical climate data are used for historical trend analysis and to developed the hydrological model. To developed the SWAT model missing data was filled by -99 and for historical climate trend analysis few stations were selected depending on the data availability and the completeness of the time series, only six stations were selected for historical trend analysis. APHRODITE gridded datasets were used to fill the missing data for those selected stations. There are 30 rainfall stations and out of which seven station were used to analyze the past trend of rainfall. Trend was analyzed by using RClimDex, there is no any significant trend in rainfall for the whole basin. Rainfall data are available from 1975 to 2015. All station was used for hydrological model set up, because missing data can interpolate by SWAT model itself. Table 3. 3 Available rainfall stations, with percentage of missing data information Station Name
St ID
Lat_N
Long_E
Alt(m)
Missing (%)
A. Tha Uthen A. Si Songkhram A. Ban Phaeng A. Bung Kan Phon Phisai
357002 357004 357005 352007 352009
17.5 17.6 17.9 18.3 18.2
104.6 104.2 104.2 103.6 103.2
151 172 152 160 153
1 3.6 5.4 17.8 28
28
Station Name A. So Phisai A. Phon Charoen A. Sawang Daen Din A. Phanna Nikhom A. Waritchaphum Sang Kho Highway A. Wanon Niwat A. Akat Amnuai Phu Phan National A. Kusuman A. Ban Muang A. Phang Khon A. Song Dao A. Kut Bak A. Nong Han A. Ban Dung A. Chai Wan 352201 354201 356201 356301 357201 357301
St ID 352006 352010 356008 356001 356009 356012 356002 356003 356011 356004 356005 356006 356007 356014 354004 354005 354020 352201 354201 356201 356301 357201 357301
Lat_N 18.1 18 17.4 17.3 17.2 16.8 17.6 17.6 17.2 17.3 17.8 17.3 17.3 17.7 17.3 17.7 17.3 17.8 17.3 17.1 17.1 17.4 17.4
Long_E 103.4 103.7 103.4 103.8 103.6 103.8 103.7 103.9 103.7 104.3 103.5 103.7 103.4 103.8 103.1 103.2 103.2 102.7 102.8 104.1 104.6 104.7 104.7
Alt (m) 178 165 176 171 177 395 170 156 316 170 152 169 189 194 168 163 182 173 177 171 190 145 153
Missing (%)
43.6 45.1 4.6 25.6 11.6 31.7 17.3 10.5 3.3 2.8 1.6 14 24.9 33.3 17.4 9.5 32.5 5.3 4.9 4.8 4.4 2.7 24.4
Table 3. 4 Available meteorological stations with its missing data status in percentage Station Tmax Tmin RH Wind SR 352201 1.4 0.5 1.5 24.0 354201 1.5 7.4 1.5 26.2 356201 1.3 7.6 1.5 24.0 2.4 356301 3.4 9.4 6.8 78.4 3.1 357201 1.4 7.5 1.5 24.2 4.9 357301 33.9 35.0 31.4 78.4 30.6
29
Figure 3. 4 Network of climate monitoring location in Songkhram Basin
40
350
35
300
30
250
25
200
20 150
15 10
100
5
50
0
Rainfall (mm/month)
Temperature (ºC)
The monthly average temperature and rainfall for upper and lower part of the basin are represented in following figures.
0 Jan Feb Mar Apr May Jun
Jul Aug Sep Oct Nov Dec
Monthly Avg rainfall
Tmax
Tmin
Figure 3. 5 Basic Climatology of Upper Catchment There is not much variation in rainfall from lower and upper catchment, but by some amount lower catchment has more rainfall. As explain in above section Thailand has three distinct seasons and as compare to rainy season other two seasons has very less rainfall. Both lower and upper part of basin experience high amount of rainfall in the month of August. Meanwhile, both station observed less rainfall in the month of January and December. The minimum and maximum monthly rainfall is 5mm and 320 mm for the upper catchment respectively. Likewise, lower catchment station measure 3 mm and 360 mm as a minimum and maximum monthly rainfall.
30
40
400
35
350
30
300
25
250
20
200
15
150
10
100
5
50
0
Rainfall (mm/month)
Temperatrure (ºC)
Similarly, almost resembling temperature has been observed in both lower and upper station. By comparing average annual temperature of these two station, it’s about 2ºC more temperature is observed in upper station. The maximum and minimum of maximum temperature at upper station is found 36ºC and 28ºC, and maximum and minimum of minimum temperature is 25ºC and 16ºC respectively. In spite of this, 34ºC and 28.5ºC is maximum and minimum of maximum temperature at lower station. The minimum temperature at lower station ranges in between 15ºC to 25ºC according to the season.
0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Monthly Avg rainfall
Tmax
Tmin
Figure 3. 6 Basic Climatology of Lower Catchment 3.2.4 Hydrological data The hydrological data were collected from royal irrigation department (RID) Thailand. There are nine gauging stations from which both gauge and discharge are measured. The quality of hydrological data is very low. So, rating curve was generated for outlet from 1990 to 2014. Due to lack of data other station are not considered. Yam and Oon are the two-major tributary of the Songkhram river. Hydrological data was used to calibrated and validated the hydrological model. For this purpose, only one station was used because of the lack of data for other stations. The rating curve was generated for KH.55 station and quality of data was checked by simple plot of discharge and rainfall. From the plot below, the response of hydrology is very good but the rainfall is not smooth enough. More than 90% of annual rainfall is observed in the rainy season so flow in Songkhram basin is highly seasonal. As compare to rainy flow, flow in other two seasons is very low. The peak discharge usually sense in August and September, while minimum flow occurs between February and April. The hydrological situation of lower part of the basin is very complicated because as occasional backwater effect from Mekong river has been occurring during the rainy seasons, and can have impact about 126 km upstream from the confluences point.
31
Figure 3. 7 Rainfall versus discharge plot at lower catchment KH.55 Table 3. 5 Table of observed discharge stations in whole Songkhram River Basin No Station ID Name Latitude Longitude 1 Kh.20 Phanna Nikhom 103.8 17.3 2 Kh.20B Ban Khok Sa-at 103.7 17.3 3 Kh.54 Ban Na Wa 104.1 17.4 4 Kh.55 Ban Pak Un 104.2 17.6 5 Kh.60 Ban Dong Yen 103.4 17.5 6 Kh.71 Ban Khon Sai 103.7 17.5 7 Kh.74 Ban Tha Huai Lua 103.3 17.8 8 Kh.93 Ban Kok Kham Lai 103.3 17.5 9 Kh.98 Ban Thakoh Daeng 103.7 17.8 3.2.5 Water quality data The required water quality data were collected from pollution control department (PCD), Thailand. There are nine water quality stations, five are in Songkhram river and four are in Oon river. SO1, SO2.SO3, SO4, and SO5 are located on Songkhram river and are located from 1 km, 4 km, 88.8 km, 118.4 km, and 189.4 km from the river mouth along the Songkhram river respectively. Nutrient and sediment data were used to calculate the past loading in the mouth of the river. The load generated from LOADEST was used to calibrated the hydrological model on daily basis. For past Nitrogen (in terms of nitrate, nitrite, and ammonia), phosphorus, suspended sediment and total dissolved solid were collected. Pollution control department measure only concentration of these constituents at every five/six months’ intervals. The concentration data were available from 1994 to 2014 in every 5/6 months’ intervals.
32
Figure 3. 8 Monthly Nutrient and Sediment Concentration in Songkhram Basin from 1995 to 2014 3.2.6 Future climate The RCMs being used for climate change analysis in Songkhram River Basin (SRB) are enlisted in given Table 3.6 with brief details. Although RCM run output comprise of large range of climate variables, the main influences once are rain and temperature (minimum and maximum). In this, study climate change analysis will cover the study of these three variables in SRB. Representative concentration pathways (RCPs) are four greenhouse gas concentration trajectories (“scenarios”) adopted by IPCC for its fifth assessment Report (AR5) in 2014. The pathways describe four possible climate futures depending on how much greenhouse gases are emitted in the years to come. But in this study only two medium and high concentration scenarios i.e. RCPs 4.5 and RCPS 8.5 are considered. Table 3.6 Regional Circulation Model (RCM) use in study Serial Resolution Coordinate RCM No (degree) System
Developer
1
ACCESS1
Rectangular
Collaboration for Australian Weather and Climate Research, Australian Government
2
CNRN-CM5CSIRO-CCAM
Rectangular
National Meteorological Research Centre (NMRC)
3
ICHEC-ECEARTH-SMHIRCA4
4
MIP-ESM-LR
0.5 Curvilinear/ Rotated Rectangular
33
Swedish Meteorological and Hydrological Institute (SMHI) European Network for Earth system Modeling (ENES)
CHAPTER 4 METHODOLOGY 4.1
Overall research methodology
This study will integrate the relationship between climate and landuse change impacts on watershed hydrology, sediment yield and water quality. The following methodological framework is used to accomplish the research, tailspin principle has been used to complete the probe examination of climate, landuse/landcover change and its solely and syndicate impacts on flow, sediment yield and water quality.
Figure 4. 1 Overall methodological framework for climate and landuse change impact analysis on hydrology, sediment, and nitrate-nitrogen 4.2
Statistical analysis of observed data
Tests for the detection of significant trends in climatologic time series can be classified as parametric and non-parametric methods. Parametric trend tests require data to be independent and normally distributed, while non-parametric trend tests require only that the data be independent. In this study, two non-parametric methods (Mann-Kendall and Sen's slope estimator) were used to detect the meteorological variables' trends. Time series data have been collected from the period of 1975 to 2014. For future climatology analysis, past analysis is very essential. Only visual assessment of the trend is not enough to say increasing 34
or decreasing trend. So, to strengthen the validity of result Mann-Kendall Non-Parametric test is used. Generally, temperature is increasing in trend and precipitation has no any define trend. Mann-Kendall trend test In 1945 Mann develop the test for significance of Kendall’s tau where the X variable is the time as a test for trend. This was directly analogues to regression, where the test for significance of the correlation coefficient r is also the significant test for the simple linear regression. The Mann Kendall test can be stated most generally as a test for whether Y values trend to increase or decrease with monotonic change. let x1, x2……. xn, be sequence of time series data over the time. Mann propose the test of null hypothesis of population sample mean Ho, where random variables are independent and identically distributed. Following is, the Kendall statistics S, 𝑛 S= ∑𝑛−1 𝑖=1 . ∑𝑗=𝑖+1 𝑠𝑔𝑛(𝑋𝑗 − 𝑋𝑖)
Eq. 4.1
Where n is the total number of data points xi and xj are the data value time series in i and j (j>i), respectively and sgn(xj-xi) is the sign function and define as,
+1,if x j -x s gn(x j -x i )= 0,if x j -x +1,if x j -x
Eq. 4.2
Kendall (1973) explain that S is the normality distribution and provide the mean and variance of S, for which there are tied in the x values as, var(S)=
1 m (n(n-1)(2n+5) - i=1 t i (t i -1)(2t i -1)(2t i +5) 18
Eq. 4.3
where n is the number of data points, m is the number of tied groups and ti denotes the number of ties of extent i. A tied group is a set of sample data having the same value. In cases where the sample size n >10, the standard normal test statistic ZS is computed using following equation.
s-1 var(S) ,if S>0 Zs = 0 , if S=0 s+1 ,if S Z1−α /2, the null 35
hypothesis is rejected and a significant trend exists in the time series. Z1−α/2 is obtained from the standard normal distribution table. In this study, significance levels α=0.01 and α=0.05 were used. At the 5% significance level, the null hypothesis of no trend is rejected if |ZS|>1.96 and rejected if |ZS|>2.576 at the 1% significance level. The Mann-Kendall statistical test has been frequently used to quantify the significance of trends in hydrometeorological time series analysis. RClimDex (1.0) RClimDex is the statistical software for computing the climatic indices for analysis and monitoring the climate change. It can be run by in the interface of R-Programme. It was developed by Byron Gleason at national Climate Data Centre (NCDC) of NOAA, and has been used in CCI/CLIVAR workshop on climate indices from 2001. There are total core climatic indices and out of them six temperature and eight rainfall indices were analyzed in this study. The main objective of using RClimDex in this research is to analyze the past climatic trend in terms of rainfall and temperature. To run this software data quality should be good. So, in this study quality of data is checked and prepare input format for the software. Missing data were fill by -99.9. RClimDex manual can be refer for detail study of the software. 4.3
Analysis of future climate
In this study, RCM with two RCPs scenarios (medium and high) were used for future climate change analysis. The study compare the qualitative relationship between the observed value of meteorological data and RCPs Data. As a conclusion from many past studies there must be some bias between these two values. So, the bias is corrected using linear scaling approach. There are many sites available to download the RCPs data, but in this study future climate data is download from https://pcmdi.llnl.gov/projects/esgf-llnl/ website of Earth System Grid Federation (ESGF) under CMIP5. As mention earlier, RCMs helps to achieve climate variables at finer resolution for a particular region of interest. However, the RCM data used in this study were downloaded from regional climate research web portal which have larger domain compare to SRB in this study. This is also evident from the fact that out of hundreds of grids of RCM output, only 12 grids intersect with the boundary of SRB. The large domain on which the RCMs are run make it highly unlikely that the grids over the SRB sufficiently posses with the spatialtemporal variation observed over the basin. Also, the output may possess some bias for the SRB region due to lack of fine tuning of climate model for sub region or priority of calibration for other sub regions of the RCM domain. In order to check this, the grid values are compared with the corrected for improvement (if any bias is present) using the measurement data. The correction parameters so calculated are applied to the future projections of RCMs. This process is known as bias correction
36
Linear scaling is simplest bias correction technique applied in various studies(Teutschbein & Seibert, 2012;Vuuren et al., 2011) and is used to adjust the RCMs mean value. The differences between the monthly mean observed and model data is applied to the model data to obtain bias corrected climate data. Data analysis relied on linear scaling bias correction (V.1.0) (Shrestha, 2015) was used for the application of this method. In the linear scaling bias correction method, the daily precipitation obtained from the RCM are transferred to P* such that P* = a.P, by using the linear scaling conversion factor “a”. Where “a “is the ratio of monthly observed and RCM precipitation for that grid point. Here, the monthly scaling factor is applied to each of uncorrected daily observations of that months, generated the corrected daily time series data. This method is adopted by many researcher because of its simplicity and less data requirements, only monthly climatological information is required to calculate the monthly correction factor (Lafon et al., 2013). Simply precipitation is corrected be multiplying be factor. In case of temperature, correction factor is calculated by subtracting long term monthly mean of historical value from observed temperature. And the temperature is corrected by adding this correction factor to the respected historical and simulated temperature. Temperature Correction: T*his (d) = This (d) + [µm (Tobs (d) - µm (This (d))]
Eq. 4.5
T*scen (d) = Tscen (d) + [µm (Tobs (d) - µm (This (d))]
Eq. 4.6
Precipitation Correction: P*his (d) = Phis (d) x a
Eq. 4.7
P*scen (d) = Pscen (d) x a
Eq. 4.8
μ (P ) a= m obs(d) μ m (Phis(d) ) Where, “*” = after bias correction his = Historical scen = Simulated obs = Observed (T, P, d) = (Temperature, Precipitation, Daily) µm= long term monthly mean
Eq. 4.9
The complete methodology for climate change analysis of SRB is as described in given Figure 4.2. The data worked with included-1) The measured data, 2) Historical/ Hindcast RCM data and 3) Projected/forecast RCM data. As mention, earlier the climatic variables taken in account for the study are daily values of rain, maximum and minimum temperature. 37
As the first step, the Hindcast RCM data are corrected using measurement values as references via Linear Scaling Method.
Figure 4. 2 Framework for climate change analysis for Songkhram river basin 4.4
Rating curve generation
In the basin, there are nine hydrological station but the measure discharge is for very short duration. This short duration discharge is not well enough to analysis the hydrology of the basin. At the outlet of the basin we had stage height for fifteen years so rating curve was generated for the discharge at outlet. Two years’ measure discharge is available for that station so with the help of this discharge we generate the rating curve for the whole-time series. the development of the stage-discharge relationship which forms the first step is very important. Once the stage discharge (G-Q) relationship is established, the subsequent procedure consists of measuring the gauge (G) and read the discharge (Q) from the (G-Q) relationship. The gauge height was taken from royal department of Thailand, they measure the stage of the river by using any mechanical instrument. By using any mechanical instrument like current meter, we can create the relationship between discharge and stage which is commonly known as rating curve. The measure value of discharge when plotted against the corresponding stage give the relationship that represent the integrated effect of wide range of channel and flow parameters. The compile effects of these two parameters is termed as control. In this research two (2013 and 2014) years discharge was used to developed the rating curve for fifteen years.
38
Two rating curve was generated based of the height of water level from mean sea level. The height greater than 136m from mean sea level has one rating curve and below 136m has another rating curve. After generation of taring curve was validated with short term available discharge and found very well performance. Following two are the rating curve equation that has been used to calculating the discharge at outlet of the basin (KH.55). The first equation is applicable only for gauge height more than 136m msl. Q = -0.0405Y5 + 28.5Y4 – 8024 Y3 +(1E+06Y2) – (8E +07Y) +2E+09
Eq.4 10
Rating curve for gauge height less than 136 msl. Q = 0.3003Y4 – 160.23 Y3 +32058Y2 – (3E +06Y) +(1E+08)
Eq.4.11
Where Q is the discharge in m3/sec and Y is the gauge height from mean sea level in m.
Figure 4. 3 Rating curve for the hydrological station KH.55 4.5
Load estimation using field data
4.5.1 Regression estimation In this Load Estimator was used as a regression model which developed the relationship between flow and concentration based on samples taken from river, usually mean daily flow and concentration are the input for regression equation. The most commonly used regression equation is the log-log rating curve. Log10 = a + b.log10 (Q)
Eq.4.12
When the daily discharge and discrete time interval concentration are fitted to that regression then it develops daily load. The product of the daily mean discharge and concentration gives the daily load. So, the LOADEST regression model first generate the daily concentration from the given input and then it gives daily load as output of model. 39
L= i=1 Ci .Qi n
Eq.4.13
The regression model does not required excessive data but the quality of predictand is totally depends on input data quality(Smith & Croke, 2005). Based on the research (Koch & Smillie, 1986) the load can be over and under estimates if the input data quality is low. This concept is applicable for all nutrients as well as sediments load. Load estimator can calculate the load by using three methods based on the quality and nonnormality of data. Maximum Likelihood Estimation (MLE), Adjusted Maximum Likelihood Estimation (AMLE) and Least Absolute Deviation (LAD) are the methods used in load estimator (Runkel et al., 2004). When the datasets are censored (measured in both lab and stream) application of maximum Likelihood Estimation method is better, if there is no any bias between these two, measure data. If some bias occurs while censoring this method cannot be used to estimate the time series load. In such case Adjacent Maximum likelihood Estimation is the best method (Runkel et al., 2004). AMLE load estimation method is apply in this research.
LALME = exp a 0 +
m j=1
a j x j H(a,b,s 2 ,α,k)
Eq.4.14
Where, LALME = is the instantons load, a and b are the function of explanatory variables, α and k are parameter of gamma distribution, and s2 is the residual variance(Cohn et al., 1992). In the above equation, the secondary term H (a, b, s2, α, k) is the bias correction factor. This equation (4.12) develops the instantaneous time series load for all observation and the mean load for various time periods are calculate by using equation (4.13). 4.5.2 Regression model The following regression based model LOADEST (Runkel et al., 2004)was used to estimate the pollutant loads. There are twelve automated model selection option, out of which model on nine was selected for this study. Which gives the better statistical result as compare to others model. In this study Akaike information is consider for model selection, for all the constituent, model performance was found moderate. The AIC criteria for model selection is tabulated below. Lowest AIC value is better to select the model. AIC can be calculated by using AIC= 2k-2ln(L), relationship, where k is the number of parameter used in regression model and L is the maximum value of the likelihood function for the estimated model. Table 4. 1 Akaike Information Criteria to Select the LOADEST model S No AIC Value Selection Criteria 1 ≤2 Substantial evidence for the model evidence 2 3>AIC≤ 7 Less support for the model 3
>10
the model is unlikely
40
ln(Li )=a o +a1lnQ+a 2lnQ2 +a 3sin(2Πdtime)+a 4cos(2Πdtime)+a 5dtime+a 6dtime2
Eq.4.15
Seven parameter model was used, where, ln=natural logarithm, L=pollutant load in (kg/day), ao= regression constant, a2, a3, a4, a5, a6 are regression coefficient, Q is daily mean streamflow in ft3/sec, dtime is time parameter in decimal years. LOADEST analysis showed that data from all stations generally fit the model well. The best model was selected atomically based on AIC value. Model was developed for upper, middle, and lower catchment of the basin. Adjacent maximum likelihood estimation (AMLE) and calibration option was selected on loudest as residuals approximated a normal distribution. Despite of AIC criteria other statically parameters were used to check the model performance. The regression model which has more than ±25% bias with observed load cannot be used to estimate the load. Partial load ratio is also model performance evaluation statistics, PLR>1, indicates overestimation and PLR< 1 indicates under estimation.
PLR=
Bp +100
Eq.4.16
100
E Nash Sutcliffe Efficiency Index E ranges from -infinity to 1.0, E = 1; a perfect fit to observed data. E = 0; model estimates are as accurate as the mean of observed data E < 0; the observed mean is a better estimate than the model estimates. 4.6
SWAT modeling
Soil and Water Assessment Tool (SWAT) excessively used word-wide in the watershed of various scales(Devkota & Gyawali, 2015; Easton et al., 2010; Mango et al., 2011; Mengistu & Sorteberg, 2012) is adopted for hydrological modeling of the Songkhram River Basin. The model developed by United State Department of agriculture- Agricultural Research Services (USDA-ARS) and Texas A & M Agrilife Research can simulate hydrology, sediment as well as water quality. It can simulate both surface and subsurface hydrology. It not only helps analyze present status of hydrology but also predict impacts of climate, landuse change and management intervention on water, sediment, and agricultural chemical yield. The model is physically based, semi-distributed, and computationally efficient and uses readily available inputs. SWAT model is based on the water balance of a river basin. It analyzes hydrological cycle components, sediment, and nutrient loading to the basin and routs the movements of these elements through the channel networks in the basin. The main components of hydrological processes in SWAT are precipitation, surface runoff, infiltration, evapotranspiration, groundwater flow, and soil water content. This study uses Arc-SWAT as Graphical User Interface (GUI) of SWAT model. Arc-SWAT is a public domain software based on ARCGIS, a popularly used Geographic Information System tool.
41
4.6.1 SWAT hydrology simulation SWAT can simulate the hydrology of the basin in two phases i.e. land and water phase. The land phase of the hydrological cycle controls the amount of water, sediments, pesticide, and nutrients yields to the main stream in each sub basin, while in water phase controls the movements of water, nutrients, pesticide, and sediments through the channel network of the watershed to the outlet. According to (Babel et al; 2011) in land phase SWAT simulates nine primary components like weather, sediments, soil temperature, bacteria, pesticide, crop growth, nitrates, and land management. The basic water balance equation of SWAT in Land phase is explain below. SWt = SWo + ∑it (Rday – Qsurf – Ea –Wseep –Qgw)
Eq.4.17
Where, SWt = final soil water content (mm), SWo = initial soil water content on day i (mm), t = time (days), Rday = the amount of precipitation on day i(mm), Qsurf = amount of surface runoff on day i(mm), Ea = amount of evaporation on day i(mm), Wseep = seepage water (mm), Qgw = amount of return flow on day i(mm), Surface Runoff will estimate by using modified curve number method of the soil conservation service (SCS). The SCS curve number method depends on permeability, landuse and soil moisture content (Neitsch et al., 2009). The SCS equation is given below. (𝑅𝑑𝑎𝑦−𝐼𝑎)2
𝑄𝑠𝑢𝑟𝑓 = (𝑅𝑑𝑎𝑦−𝐼𝑎+𝑆) 𝑆 = 25.4(
1000 𝐶𝑁
Eq. 4.18
− 10)
Eq. 4.19
Ia = 0.2 x S C.i.A q peak = 3.6
Eq. 4.20 Eq.4.21
Where, Qsurf = total excess rainfall (mm), Rday = rainfall depth (mm), Ia = initial abstraction (mm), S =retention (mm), CN = Curve Number, qpeak = peak runoff (m3/sec), i= rainfall intensity (mm/hr), A= area (km2) The main input parameter for SWAT is the Digital Elevation Model (DEM), which provide bunch of information on topography, landscape, elevation, slope, and other aspects of the study area. DEM can explain the direction of flow and the other physical characteristics of the basin. LULC map, meteorological data, and soil type are other input data required for SWAT. Model was calibrated and validated by discharge and water quality data. Landuse and soil type have significant impact on water balance of the study area.
42
4.6.2 SWAT sediment simulation Modified universal soil loss equation (MUSLE) is used to calculate the erosion and sediment yield from the basin developed by (Arnold et al., 1995). MUSLE is the modified form of USLE equation developed by Wischmeier and smith. USLE predict the volume of considering the rainfall factor, but in MUSLE rainfall energy factor is replace by runoff factor. Sediment yield is more depends on runoff factor rather than rainfall factor. MUSLE estimate the precise sediment yield because rainfall has no relationship with moisture content of soil but runoff is the function of antecedent moisture content of soil. The modified USLE, MUSLE is given by: Sed=11.8.(Qsurf .q peak .area hru )0.56 .K USLE .CUSLE .PUSLE .LSUSLE.CFRG
Eq.4.22
Where Sed= sediment yield in metric ton, Qsurf = surface runoff volume in mm, qpeak= peak flow runoff (m3/sec), areahru= area of hru in hector (ha), KUSLE = soil erodibility factor (0.013 metric ton m2 hr /(m3-metric ton cm), CUSLE = cover and management factor, PUSLE = USLE support factor, LUUSLE = USLE topographic factor and CFRG is the coarse fragment factor. Soil Erodibility factor , (KUSLE) , is the soil loss rate per erosion index for a unit plot on a specified soil (Arnold et al., 1995). A unit plot is 22.1 -m long with a uniform length-wise slope of 9 percentage, in a continuous fallow, tilled up and down the slope. The continuous fallows mean land that has been tilled and kept free of vegetation for more than 2 years. The unit of KUSLE in MUSLE equation is traditional English units of 0.01(ton acre hr)/(acre ftton inch). According to Wischmeier and Smith (1978) soil typically less erodible with decrease in silt fraction, regardless of whether the corresponding increasing is in the sand or clay fraction. Direct measurement of soil erodibility factor is more time consuming and costly so (Arnold et al., 1995) developed a general equation to calculate the fraction of soil particles.
K USLE =
0.00021.M1.14 .(12- OM) + 3.25.(csoilstr -2)+2.5.(c perm -3)
100 Where the particle-size parameter, M is calculated by
M=(msilt +mvfs ).(100-mc )
Eq. 4.23
Eq.4.24
Organic matter content (OM) = 1.72* orgC, (organic carbon content), Csoilstr = soil structure code in soil classification, Cperm = profile permeability class, Cover and management factor (USLE-C) is define as the ratio of soil loss from the land cropped under continuous to the corresponding loss from clean-tilled, continuous fallow (Wischmeire and smith, 1978). Erosion gets reduce due to canopy as it reduces rainfall energy of raindrops. Residue in the soil also disturbed the runoff which ultimately reduces velocity and carrying capacity. CUSLE is reflected by the cropping pattern and management practices on the land. It is communally used to differentiate the effect of management 43
practices on the soil conservation. In, SWAT updates CUSLE daily is calculated by using following equation.
CUSLE =exp( ln(0.8)-ln(C USLE,mn ) .exp -0.00115.rsd surf +ln C USLE,mn
Eq.4.25
Where, CUSLE,mn is the minimum value for the cover and management factor for the land cover, and rsdsurf is the amount of residue on soil surface (kg/ha). The minimum C factor can be estimate from a known average annual C factor using the following equation(Arnold et al., 1995), by considering average annual C factor for the land cover. CUSLEmn = 1.463ln[CUSLE,aa] + 0.1034
Eq.4.26
The support practice factor (PUSLE) has positive impact on soil loss and erosion rate. It is the ratio of soil loss with the specific practice factor to the corresponding loss with up and down slope practice. The support practice might be tillage, counter farming, strip-cropping on the contour, and terrace systems. Fixing the waterway and reducing the velocity of water is the important part of each practice system. Land slope has direct relationship with the soil erosion and sediment yield. The topographic factor (LSUSLE), is define as the ratio of soil loss per unit area from a field slope to that from a 22.1 m length of uniform 9% slope. The following equation is utilized in SWAT to calculate the topographic factor, LSUSLE =(
Lhill ).(65.41.sin(α hill )+4.56.sinα hill +0.065) 22.1
Eq.4.27
Where Lhill is the slope length(m), meter is in exponential term, and αhill is the angle of the slope length. The exponential term m is calculated by using following equation, m=0.6.(1-exp -35.835.slp )
Eq.4.28
And, the coarse fragmented factor used in MUSLE equation is calculated by using following equation, CFRG = exp (-0. 053*rock), where rock is the percentage of rock contain in soil layer. Sediment routing equation in SWAT In current version of SWAT, four additional physically based sediment routing method are incorporated. If user not define any one of these four methods, SWAT itself calculate the sediment by considering default Bagnold equation zero (“0”). If one of among these four physically based approach is selected, then sediment pool in six particles sizes are tracked by the model. In default Bagnold equation, all soil particles are considered as silt and the channel erosion is not partitioned between stream bank and stream bed and deposition is 44
assumed to occur only in the main channel, floodplain deposition is not considered, so in this study simplified Bagnold equation (ch_eq 1) was selected. Sediment transport in the channel network is the function of two processes, deposition, and degradation, operating simultaneously in the reach. SWAT calculate both deposition and degradation by considering the same with throughout the entire channel. So, the following equation is used to calculate amount of deposition and degradation of sediment in SWAT. Sed ch =sed ch,j -sed dep +sed deg
Eq.4.29
Where, Sedch= quantized of suspended sediment in reach (metric ton) Sedch,i = quantity of suspended sediment in the reach at the beginning of the period (metric ton) Seddep = quantity of sediment deposited in the reach segment (metric ton) Seddeg = quantity of sediment re-entrained in the reach (metric ton) The amount of sediment transported out of the reach is calculated by using following equation, Vout Vch Where, Sedout = quantity of sediment transported out of the reach (metric ton). Sedch = quantity of suspended sediment in reach (metric ton) Vout = volume of outflow during time steps (m3H2O) Vch = volume of water in the reach segment (m3H2O) Sed out =Sed ch .
Eq.4.30
4.6.3 SWAT Nutrient process simulation The assessment of the nitrogen and nitrogen cycle is very important because it is very important for plant growth and is being hot topic for researcher. The nitrogen cycle is very complex system with dynamic nature containing the water, atmosphere, and soil. As compare to other elements like carbon, protein, oxygen, hydrogen, Nitrogen is very essential for plants. So, SWAT is a powerful tool to calculate the nitrogen in the soil profile and shallow aquifer. Two basic system involves in the nitrogen cycle one is nitrogen processes in the surface and subsurface soil and another is amount of nitrogen receiving by water bodies from soil. The organic nitrogen which is associated with humus, mineral forms of nitrogen held by soil collides, and mineral forms of nitrogen in solution are the major three forms of nitrogen in soils. The amount of nitrogen in soils can be increased by various means like use of fertilizer, manure application, fixation by symbiotic or non- symbiotic bacteria and rain. Nitrogen taken by plants, underground leaching, denitrification, volatilization, and soil erosion are the major causes of removing the nitrogen from soils. The following Figure 4.4 (a) shows the detail components of nitrogen cycle (Neitsch et al., 2011). 45
Figure 4. 4 (a) Nitrogen Cycle in soil and (b) Partitioning of Nitrogen in SWAT Soil and water assessment tool (SWAT) monitor the five different pools of nitrogen in soil. Two pools are inorganic forms of nitrogen like Ammonia-nitrogen and nitrate-nitrogen , others three are organic nitrogen(Neitsch et al., 2011). Those five pools are shown in given Figure 4.4 (b). In SWAT user, can define the initial nitrate-nitrogen level in the soil, if the user has less knowledge and able to fix the initial nitrate- nitrogen concentration, the SWAT itself fix by using following equation. The initial nitrate is varied with the soil depth and has positive relationship, -Z NO3,conc =7.exp 1000
Eq.4.31
Where, NO3conc is the initial concretion of nitrate at depth Z (ppm), Z is the depth from the soil surface (mm). The solution of above equation in dramatic form is shown below;
Organic nitrogen levels are assigned that the C: N ratio for humic materials is 14:1. The following equation is used in SWAT to calculate the concentration of humic nitrogen in soil layer,
orgCly Orgn hum,ly =104 14
Eq.4.32
46
Where orgnhum,ly is the concentration of humic organic nitrogen in the soil layer (ppm), orgCly is the amount of organic carbon in the layer (%). To separate the humic organic nitrogen following two equation are used in SWAT(Neitsch et al., 2011), ogrNact,ly =orgN hum,ly .fractN
Eq.4.33
orgNsta,ly =orgN hum,ly .(1-fractN )
Eq.4.34
Where, oegNact,ly is the concentration of nitrogen in active organic pool(mg/kg), orgNhum,ly is the concentration of humic organic nitrogen in the layer (mg/kg), fractN is the fraction of humic nitrogen in the active pool, and orgNsta,ly is the concentration of nitrogen in the stable organic pool (mg/kg), the fraction of humic nitrogen in the active pool, fractN 0,02. Nitrogen in fresh organic pool is set to zero in all layers except the top 10mm of soil. In the top 10mm, the fresh organic nitrogen pool is set to 0.15% of the initial amount of residue on the soil surface. orgNfresh,surf = 0.0015.rsdsurf
Eq.4.35
Where orgNfresh,surf , is the nitrogen in the fresh organic pool in the top 10mm (kg N/ha), and rsdsurf is the material in the residue pool for top 10mm of soil (Kg/ha). The ammonia pool for soil nitrogen, is initialized to 0 ppm. Nitrate routing equation in SWAT Nitrate from soil surface may be transported by percolation, runoff, and sub-surface lateral flow. To calculate the amount of nitrate load moved with water, calculation of concentration of nitrate of mobile water is necessary. After calculating concentration, total amount of nitrate lost can be easily calculate by multiplying concentration with the volume of moving water in each pathway. The following equation is used to calculate the concentration of nitrate in moving water. NO3ly (1-exp Conc NO3,mobile =
-w mobile ) (1-θ e ).SATly
Eq.4.36
w mobile
Where concNO3,mobile is the concentration of nitrate in mobile water for a given layer (KgN/mm H2O), NO3ly is the amount of nitrate in the layer (KgN/ha), wmobile is the amount of mobile water in the layer(mm H2O), θe is the fraction of porosity from which anions are excluded, and SATly is the saturated water content of the soil layer (mm H2O).
47
4.6.4 Performance evaluation of SWAT model There are so many statistical parameters to evaluate the model performance such as Efficiency Index (EI), Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Percentage Bias (PBIAS), and Coefficient of Determination (R2), Nash Sutcliffe Efficiency (NSE), Volume Error (Dv) etc. In this research performance of SWAT is checked by Coefficient of Determination, Percent Bias, and Nash Sutcliffe Efficiency, RMSEobservation standard deviation ratio and weighed R2. The coefficient of determination provides the measure of how well the model can simulate the output. This value lies in between 0 and 1. Zero indicates poorest simulation and 1 indicates very good simulation. NS shows the goodness of fit of observed and simulated data. PBIAS analyze the trend of observed and simulated data, it observed whether the simulated value is smaller or larger than observed value. According to many research NSE should be greater than 0.6 for perfect match between the observed and simulated value, whereas the PBIAS should be less than 15% for perfect match (Shrestha et al., 2016). 𝑅2 =
𝑁∑𝑋𝑌−∑𝑋∑𝑌
Eq. 4.37
(√𝑁(∑𝑋 2 )−(𝛴𝑋)2 )−(√𝑁(𝛴𝑌 2 )−(𝛴𝑌)2 ) N
PBIAS
(Qi Qi ') i 1
*100
N
Qi
Eq. 4.38
i 1
N
NSE 1
(Qi Qi ')
2
i 1 N
Eq. 4.39
(Qi Q
Mean
i 1
)
2
n
PSR=
RMSE = STDEVobs
(Q -Q ) i
' 2 i
i=1
Eq. 4.40
n
(Q -Q i
mean
)
i=1
Where, X = Observed Value of Daily Discharge, Y = Simulated Value of Daily Discharge, Qi = Measure Daily Discharge, Qi’ = Simulated Daily discharge, Qmean = Average daily Discharge for Simulated Period, N = Number of Daily Discharge Value. (Moriasi et al., 2007) Table 4. 2 Model performance rating criteria of modeling Performance rating Very good Good Satisfactory Unsatisfactory
PSR 0≤PR≤0.5 0.5
