Simulation of Watersheds Hydrology Under Different Hydro-Climatic Settings

Simulation of Watersheds Hydrology under Different Hydro-Climatic Settings

A dissertation submitted to the

Graduate School

of the University of Cincinnati

In partial fulfillment of the

requirement for the degree of

Doctor of Philosophy

In the Department of Geography

of the College of Arts and Sciences

Thushara D Ranatunga

B.Sc. Agricultural Engineering, University of Peradeniya, Sri Lanka, 2005

M.A. GIS and Remote Sensing, University of Cincinnati, 2008

Committee Chair: Dr. Susanna Tong

Committee members: Dr. Richard Beck

Dr. Hongxing Liu

Dr. Tomasz Stepinski

Dr. Jeffrey Yang

December 2014 ABSTRACT

Hydrological characteristics of a watershed are dependent on a variety of factors, including the local climate, land use, soil, and other anthropogenic influences. Changes in these factors unequivocally would affect the water resources. To ensure that we will have adequate water supply in the future, we need a methodology that would enable us to predict the hydrologic ramifications caused by potential changes of the above mentioned factors. Furthermore, there is a need for new integrative approaches that model not only the separate but also the combined impacts of these changes as they act in tandem with each other. It is also important to understand the relationships between different variables and the underlying watershed processes under different environmental settings in order to adapt, respond, and make efficient water resource management decisions.

The objectives of this dissertation are (i) to develop a tool/protocol to better understand the watershed systems and to help planners/resource managers to explore and predict the impacts of potential changes in climate, land use, and population, through basin scale watershed modeling; (ii) to introduce a total water management approach to help in managing the future potential changes in demand and supply of water; and (iii) to develop an approach to investigate the watershed characteristics and processes that control the hydrologic behaviors of the watersheds.

The watershed hydrologic model, the Hydrological Simulation Program-FORTRAN (HSPF) and

Normalized Root Mean Square Error (NRMSE) technique with Flow Duration Curve analysis were employed in this dissertation research project as the major assessment tools. These tools were applied to two watersheds, the Little Miami River (LMR) and the Las Vegas Wash (LVW) watersheds that are located under different hydro-climatic settings. From the results, it seemed that the tool developed could be helpful in facilitating the prediction of the plausible hydrologic consequences of climate, population, and land use changes. The simulation results also highlighted the extent to which different

ii global change factors could amplify the hydrologic effects at a watershed scale. Moreover, the results revealed that the hydrology in the LMR watershed is mostly sensitive to groundwater related parameters, whereas the LVW watershed is related to near surface soil parameters. Furthermore, high and medium flows are the most sensitive flow regimes to most of watershed processes.

iii

ACKNOWLEDGEMENT

First and foremost I would like to express my deepest gratitude to my supervisor, Dr. Susanna

Tong, whose constant guidance, patience and caring helped me in completing this study, from the dissertation proposal up to the publishing of journal manuscripts and submitting the final dissertation.

Without her guidance, endless advice and persistent help, this study would not have been possible.

There is no word or thought that can adequately capture my gratitude for all her support in shaping up my academic and professional career. Secondly, I would like to thank to Dr. Jeffery Yang, from the

USEPA, who provided me the opportunity to work at the USEPA and for his guidance and support throughout this study. I would also like to thank to the other members of my dissertation committee,

Dr. Richard Beck, Dr. Hogxing Liu, and Dr. Tomasz Stepinski for their valuable inputs and guidance.

To the faculty, staff, and most importantly to my colleagues at the Department of Geography, thank you for making each day of my stay at the department memorable. Without you all, my college experience would have been significantly less awesome.

I would like to thank my parents, parents-in-law, and my brother and sisters. They were always supporting me and encouraging me with their best wishes.

Finally to my wonderful wife, Sewwandi Rathnayake, thank you for all her support, love and understanding. She is always there cheering me up, and standing by me through the good and bad times.

TABLE OF CONTENT

Abstract …………………………………………………………………………………………………………………………………………………… ii

Acknowledgement …………………………………………………………………………………………………………………………………… v

Table of Content …………………………………………………………………………………………………………………………...... vi

List of Figures ………………………………………………………………………………………………………………………………………….. ix

List of Tables ……………………………………………………………………………………………………………………………………………. xi

Chapter 1: Introduction ……………………………………………………………………………………………………………………………. 1

Chapter 2: Predicting Plausible Impacts of Sets of Climate and Land Use Change Scenarios on Water

Resources ………………………………………………………………………………………………………………………………………………… 5

2.1. Introduction ………………………………………………………………………………………………………………………………. 5

2.2. Materials and Methods ……………………………………………………………………………………………………………… 6

2.2.1. Study Area ………………………………………………………………………………………………………………………… 6

2.2.2. Development of a hydrologic and water quality model …………………………………………………… 7

2.2.3. The hydrologic model for the LMR watershed …………………………………………………………………. 8

2.2.4. The water quality model for the LMR watershed …………………………………………………………….. 8

2.2.5. Generation of the future climate change scenarios …………………………………………………………. 8

2.2.6. Generation of the future land use change scenarios ………………………………………………………. 9

2.2.7. Development of a CA-Markov land use model for the LMR watershed ………………………….. 9

2.2.8. Prediction of future population growth ……………………………………………………………………………. 11

2.2.9. Incorporation of the population variable in the CA-Markov land use model …………………… 11

2.2.10. The 2050 land use change scenario for the LMR watershed ……….………………………………….. 11

2.2.11. Modeling of the hydrologic and water quality impact of climate, population and land use

changes………………………………………………………………………………………………………………………………………… 11

2.3. Results and discussion ………………………………………………………………………………………………………………… 14

2.4. Conclusion ………………………………………………………………………………………………………………………………….. 15

2.5. Reference ……………………………………………………………………………………………………………………………………. 16

Chapter 3: A Total Water Management Analysis of the Las Vegas Wash

Watershed, Nevada …………………………………………………………………………………………………………………………….. 18

3.1. Introduction ………………………………………………………………………………………………………………………………. 18

3.2. Study Area …………………………………………………………………………………………………………………………………. 20

3.3. Research Methodology ……………………………………………………………………………………………………………… 22

3.3.1. Developing the HSPF watershed model for the Las Vegas Wash watershed ……………………. 22

3.3.2. Formulating future climate change scenarios …………………………………………………………………… 25

3.3.3. Projecting population growth in the Las Vegas Valley ……………………………………………………… 27

3.3.4. Estimating future wastewater discharge ………………………………………………………………………….. 28

3.3.5. Generating future land use change scenario …………………………………………………………………….. 29

3.3.6. Simulation of Las Vegas Wash hydrology under the impacts of climate, land use, and

wastewater changes ………………………………………………………………………………………………………….. 31

3.3.7. Assessing the future water supply and demand ………………………………………………………………… 32

3.4. Results and Discussion ……………………………………………………………………………………………………………….. 32

3.4.1. Simulated stream discharge ……………………………………………………………………………………………… 32

3.4.2. Total water management …………………………………………………………………………………………………. 37

3.5. Conclusion ……………………………………………………………………………………………………………………………….... 39

3.6. References …………………………………………………………………………………………………………………………………. 40

Chapter 4: An Approach to Sensitivity Analysis in Watershed Hydrologic Modeling ………………………………. 43

4.1. Introduction ………………………………………………………………………………………………………………………………. 43

vii

4.2. Methods ……………………………………………………………………………………………………………………………………. 48

4.2.1. Study area ………………………………………………………………………………………………………………………….. 48

4.2.1.1. The LMR watershed ……………………………………………………………………………………………….. 48

4.2.1.2. The LVW watershed ……………………………………………………………………………………………….. 49

4.2.2. Model selection ………………………………………………………………………………………………………………… 51

4.2.3. Model Parameter selection ……………………………………………………………………………………………… 52

4.2.4. Defining flow regimes using flow duration curve analysis ……………………………………………….. 54

4.2.5. Sensitivity index ………………………………………………………………………………………………………………… 56

4.3. Results and discussion ……………………………………………………………………………………………………………….. 58

4.3.1. Flow duration curves and flow regimes of the two watersheds ………………………………………. 58

4.3.2. Parameter sensitivity ………………………………………………………………………………………………………… 60

4.3.3. Parameter sensitivity under different flow regimes …………………………………………………………. 67

4.4. Conclusion …………………………………………………………………………………………………………………………………. 72

4.5. References …………………………………………………………………………………………………………………………………. 76

Chapter 5: Conclusion ………………………………………………………………………………………………………………………………. 83

Bibliography …………………………………………………………………………………………………………………………………………….. 88

viii

LIST OF FIGURES

Figure 2.1: The Little Miami River watershed study area …………………………………………………………………………. 7

Figure 2.2: Land use maps of the Little Miami River Basin acquired from the 1980, 1992, and 2001 NLCD datasets …………………………………………………………………………………………………………………………………………………… 10

Figure 2.3: Projected 2001 land use map generated from CA-Markov …………………………………………………… 10

Figure 2.4: Projected 2001 population density map and 2001 MCE urban suitability map ……………………... 12

Figure 2.5: Projected 2001 land use map generated from the population coupled CA-Markov …………….. 13

Figure 2.6: Projected 2050 population density map, and 2050 MCE urban suitability map ……………………. 13

Figure 2.7: Projected 2050 land use map generated from the population coupled CA-Markov …………….. 14

Figure 3.1: The Las Vegas Wash watershed study area ………………………………………………………………………….. 20

Figure 3.2: HSPF model outputs ………………………………………………………………………………………………………………. 26

Figure 3.3: Population growth of the Las Vegas Valley ………………………………………………………………………….. 28

Figure 3.4: Observed wastewater discharge and population of the Las Vegas Valley …………………………… 29

Figure 3.5: Linear regression relationship between population and wastewater discharge ………………….. 29

Figure 3.6: Future population and wastewater projection of the Las Vegas Wash ………………………………… 30

Figure 3.7: Projected 2050 LULC map of the Las Vegas Wash watershed ……………………………………………… 31

Figure 3.8: HSPF simulated stream discharge with wastewater projection ……………………………………………. 33

Figure 3.9: Projected stream discharge in 2050 under base case, wet and dry climate with no-land use change scenarios ……………………………………………………………………………………………………………………………………… 34

Figure 3.10: Projected stream discharge in 2050 under wet and dry climate with land use change scenarios ………………………………………………………………………………………………………………………………………………….. 35

Figure 3.11: Total water demand and supply with return-flow credits …………………………………………………… 37

Figure 4.1: Maps of the study areas: the LMR and LVG watersheds ……………………………………………………….. 51

Figure 4.2: Flow duration curves for the LMR watershed and LVW watershed ………………………………………. 59

Figure 4.3: Comparison of parameter sensitivity for the entire flow ranges of LMR and LVW watersheds… 61

Figure 4.4: Average slope of the SI curves for the entire flow range ……………………………………………………… 62

Figure 4.5: LMR model parameter sensitivity in three flow regimes ………………………………………………………. 67

Figure 4.6: Average slope of the SI curves of three flow regimes in LMR watershed ……………………………. 68

Figure 4.7: LVW model parameter sensitivity in three flow regimes ……………………………………………………… 70

Figure 4.8: Average slope of the SI curves of three flow regimes in LVW watershed ……………………………. 71

LIST OF TABLES

Table 2.1: Land use categories in the LMR watershed in 1980, 1992, 2001 and 2050 by percentage ……. 7

Table 2.2: Calibration and validation results for the hydrological model ……………………………………………….. 8

Table 2.3: Calibration/validation targets for HSPF ………………………………………………………………………………….. 8

Table 2.4: Calibration and validation results for the mean daily TP concentrations in mg/L …………………. 9

Table 2.5: Calibration and validation results for the mean daily N concentrations in mg/L …………………… 9

Table 2.6: Hypothetical future climate scenarios ……………………………………………………………………………………. 9

Table 2.7: Validation results for 2001 land use projections …………………………………………………………………….. 11

Table 2.8: 2007 county population projections from the linear, geometric, and exponential models …… 12

Table 2.9: The p-value results from the t-tests between 2007 projected population values and the official estimates …………………………………………………………………………………………………………………………………………………. 12

Table 2.10: Modeling results of the effects of future climate and land use changes on stream flow ……. 14

Table 2.11: Modeling results of the effects of future climate and land use changes on TP ……………………. 15

Table 2.12: Modeling results of the effects of future climate and land use changes on N …………………….. 15

Table 3.1: Average monthly precipitation and temperature of the Las Vegas Valley, Nevada ………………. 21

Table 3.2: HSPF model parameters used in calibration …………………………………………………………………………… 24

Table 3.3: HSPF model calibration and validation results ……………………………………………………………………….. 25

Table 3.4: Land-use projection for 2050 …………………………………………………………………………………………………. 31

Table 3.5: Projected frequency of high flow occurrences and the highest flow values under each simulated scenario …………………………………………………………………………………………………………………………………… 36

Table 3.6: Average daily flow changes of each scenario from the base model ……………………………………… 36

Table 4.1: HSPF parameter descriptions ………………………………………………………………………………………………….. 53

Table 4.2: Flow rates for each regime in the LMR and LVW watersheds ……………………………………………….. 60

CHAPTER ONE

INTRODUCTION

Scientific evidence has suggested that the world is facing considerable uncertainties on future demand and availability of water resources over the next decades. Climate change can affect water supply by reducing or increasing the amount of water available, enhancing environmental degradation, and instigating water pollution problems (IPCC, 2008). However, climate change is just one of the factors affecting water resource management. In addition to climate change, population growth in an area can also lead to an increase in the demand for domestic, industrial, and agricultural water (Arnel

1999). Moreover, the increase in population can lead to an increase in wastewater discharge.

Furthermore, population growth and agglomeration in urban and sub-urban centers can accelerate the rate of changes of land use/land cover (LULC), which may lead to many consequences. The decrease in vegetative cover and an increase of impervious surfaces can alter the local energy budget and hydrologic cycle (Wu et al., 2007), affecting both the water quality as well as quantity. An increase of impervious surface can cause a decrease in local infiltration, percolation, and soil moisture storage, which in turn can lead to a reduction in natural interception and an increase in runoff and flood frequency (Brun and Band, 2000).

With an increasing population, a changing climate, and the expansion of human activities, it is imperative to manage our water resources properly (Sharma and Gosain, 2010). To address these challenges, extensive research has been conducted to assess the impacts of climate change, land use change, and urbanization on watershed hydrology. Simulation models have been employed to postulate their impacts on water quantity and quality. Unfortunately, most of these studies are unable to make a definite assessment due to the uncertainties of climate change as well as the projections of future population and land use patterns. There is a need to understand further the uncertainty and sensitivity

1 of these factors and how they affect watershed hydrology. Most importantly, we also need to develop an integrative approach that models not only the separate but also the combined impacts of these changes as they act in tandem with each other. It is also essential to understand the relationship between different variables and the water systems under different environmental settings in order to adapt, respond, and to make efficient water resource management decisions.

The hydrologic responses in a watershed to environmental change may vary spatially and temporally according to watershed characteristics and processes. Therefore, it is important for modelers to understand the underlying processes that drive the hydrologic behaviors in watersheds in order to conduct a more appropriate simulation and investigation of the potential impacts. To ascertain the impacts of global environmental changes and the hydrologic processes of watersheds under different hydro-climatic settings, different models are needed to simulate the hydrology of these watersheds.

Likewise, it is important to apply different models to simulate the different flow regimes within a watershed. Thus, more research is needed.

Watershed simulation models are considered as a useful tool in simulating the dominant watershed processes and examining the dynamic interactions between climate, land use, and water resources (Leavesley 1994). Such models describe mathematically spatial and temporal water and nutrient fluxes from the land surface and soil profile to the water bodies. They have been extensively used for flood forecasting, storm water management, water resources management, hydro-modification analysis, water supply availability analysis, and environmental impact assessment through continuous simulation of river basin response to land use and meteorological variables (Somura et al., 2009,

Mahamoud et al,. 2009, Middelkoop et al., 2001, Bekely and Knapp, 2010, Wu et al., 2007).

Furthermore, they can be used in assessing the impacts of climate change and other potential changes, such as land use and urbanization, on watershed hydrology.

In this dissertation, the primary goal was to develop a tool/protocol to better understand the watershed systems and to help planners/resource managers to predict and explore the impacts of such changes (e.g., land use, climate, and population changes), through basin scale watershed modeling. The secondary goals are (i) to introduce a total water management approach to help in managing the long and short term potential changes in demand and supply of water; and (ii) to develop a comparative sensitivity index that can be used to identify and compare the dominant watershed characteristics and processes that govern the hydrology of watersheds under different hydro-climatic conditions as well as different flow regimes within a watershed.

The Better Assessment and Science Integrating Point and Non-Point Sources (BASINS) program and the Hydrological Simulation Program-FORTRAN (HSPF) provided by the United States Environmental

Protection Agency (USEPA) were employed in this dissertation research project as the major watershed assessment tool. It was chosen not only because it is a comprehensive rainfall-runoff model linking surface dynamics to groundwater recharges through climate forcing data, but also because it provides a wide range of flexibility for model formulations and calibration processes. Further, it has the ability to produce the hydrology outputs at a specific point in the river network. For example, it can produce the total outflow at the very last outlet of the watershed. In order to understand the watershed processes and the impacts of different parameters to watershed hydrology, an approach was also developed using

Normalized Root Mean Square technique (NRMSE) and Flow Duration Curve analysis methodologies.

According to the requirements stipulated by the Department, this dissertation follows a three peer-reviewed articles format.

Chapter two contains the first journal article, which was published in Journal of Applied

Geography. It is entitled “Predicting plausible impacts of sets of climate and land use change scenarios on water resources”. Its goal was intended to derive an approach that can be used to examine the

3 hydrologic and water quality effects of population-driven land use change in concert with climate change. A watershed hydrologic and water quality model was developed to simulate the existing and future conditions of the Little Miami River (LMR) watershed, Ohio. The plausible impacts of varying climate and land use change scenarios on the river discharge and water quality conditions of the watershed were then assessed using the developed model.

Chapter three contains the second journal article, which was published in the Journal of Physical

Geography. It is entitled “A total water management analysis of the Las Vegas Wash Watershed,

Nevada”. In this study, an analytical tool/protocol was developed to simulate the stream hydrology of the Las Vegas Wash (LVW) watershed, Nevada. Using the developed simulation technique, the total water demand and supply conditions were simulated under the future potential climate, LULC, and population projections regimes.

Chapter four contains the third article that has submitted to the USEPA for clearance and will be submitted to the Journal of Environmental Modeling and Software. It is entitled “An Approach to

Sensitivity Analysis in Watershed Hydrologic Modeling”. The purpose of this article is to develop a comparative sensitivity index that can be used to identify and compare the dominant watershed characteristics and processes that govern the hydrology of two watersheds under two different hydro- climatic conditions as well as different flow regimes within a watershed. Results from LMR and LVW studies from the first and second articles revealed that the model parameters behave differently in different watersheds. In order to identify the predominant watershed characteristics and processes that drive the hydrology of each watershed, a tool was developed and applied to the two hydrologic models.

Applied Geography 32 (2011) 477e489

Contents lists available at ScienceDirect

Applied Geography

journal homepage: www.elsevier.com/locate/apgeog

Predicting plausible impacts of sets of climate and land use change scenarios on water resources

Susanna T.Y. Tong a,*, Yu Sun a, Thushara Ranatunga a, Jie He a, Y. Jeffrey Yang b a Department of Geography, University of Cincinnati, Cincinnati, OH 45221, USA b National Risk Management Research Laboratory, U.S. Environmental Protection Agency, Cincinnati, OH 45268, USA

abstract

Keywords: Our world is changing at an unprecedented rate in terms of climate and land use, but these changes can Climate change affect our water resources. Hence, we need a methodology that can predict both their individual and Land use change agglomerative ramifications. Using the Little Miami River (LMR) watershed as a case study, this paper Hydrologic regime describes a spatial analytical approach integrating mathematical modeling and geographical information Water quality CA-Markov sciences to quantitatively examine the relative importance of the separate and combined hydrologic and HSPF water quality impacts of climate and land use changes. The Hydrologic Simulation Program e Fortran (HSPF) model was chosen in this study to simulate stream flow and nutrient transport process. Five hypothetical climate change scenarios were used to cover the possible ranges of variability in the year 2050. An enhanced population-coupled Markov- Cellular Automata (CA-Markov) land use model was developed to predict the 2050 land use pattern. When these scenarios were incorporated into the HSPF model, the future conditions in the LMR basin were postulated. The findings demonstrated that: 1) the LMR watershed would experience an increase in flow and nutrients under the 2050 land use projection, 2) stream flow and water quality impacts would be amplified when both climate and land use changes were simultaneously considered, 3) land use change (and in the case of the LMR watershed, urbanization) could help to alleviate water shortage during the dry years, 4) total phosphorus and nitrogen would increase under all future climate and land use scenarios; the highest increase was found under the combined wettest and future land use scenarios, and 5) the described approach is effective in simulating the hydrologic and water quality effects of climate and land use changes in a basin scale. These results are relevant to planners; they can be useful in formulating realistic watershed management policies and mitigation measures. Ó 2011 Elsevier Ltd. All rights reserved.

Introduction precipitation by 10% in northern Ohio, but a decreasing trend by 10% in the south (NCSL & CIER, 2008). The rising temperature and Within the scientific community, it is generally agreed that our shifting precipitation patterns would have profound hydrologic climate is changing, and it will bring significant hydrologic conse- effects in the Midwest. quences. The Fourth Assessment of the Intergovernmental Panel on In addition to climate, population and land use are also Climate Change (IPCC, 2008) reports that with the increasing changing. Despite Ohio is not one of the fastest growth areas, it is concentration of greenhouse gases in the atmosphere, there will be the seventh most populous state in the nation (Ohio Department of a global average temperature increase of 2.3e6.2 C in this century. Development, 2010). Moreover, substantial changes in population In Ohio, the first signs of global warming are appearing. The average and land use have occurred in recent decades, especially in the temperature between 2000 and 2007 in Cleveland was 0.45 C Little Miami River (LMR) watershed in southwestern Ohio. Land use above the 30-year average prior to this period, and Cincinnati had change, such as urbanization, when amalgamated with climate experienced 25 days each year with a temperature of at least change, would undoubtedly affect both the quantity and quality of 32.2 C, which was six days more than the historical average before water resources. To maintain and sustain our water resources, 2000. Moreover, since 1900, there has been an increasing trend of appropriate watershed management policies and adaptation measures to future changes are necessary (Pielke, Prins, Rayner & Sarewitz, 2007; Yang, 2010). One of the objectives of this research * Corresponding author. Tel.: þ1 513 556 3435; fax: þ1 513 556 3370. is therefore to ascertain quantitatively the hydrologic and water E-mail address: [email protected] (S.T.Y. Tong). quality impacts of climate and land use changes.

5 478 S.T.Y. Tong et al. / Applied Geography 32 (2011) 477e489

Although studies of the combined hydrologic and water quality Hamilton Counties, and portions of Montgomery, Clinton, Brown, effects of climate, land use, and population changes are desirable, Highland, and Madison Counties. most published investigations consider only one type of change Lying within the Till Plains section of the Central Lowlands (climate or land use change) or one type of impact (flow or water Physiographic Province, the underlying geology of the basin quality); while some examine the combined impacts, they only use consists of interbedded calcareous shale, limestone, and dolomite one or two scenarios. Few studies have investigated both changes from the Upper Ordovician or Silurian (Ohio DNR, 1964). Outwash and impacts under various scenarios at a basin scale. deposits composed of sand and gravel are usually found at the In the studies of climate change, most researchers employ the bedrock valleys and terraces. Due to their high permeability, these general circulation models (GCMs) or regional climate models glacial materials exert considerable impacts on surface water by (RCMs) to predict the future temperature and precipitation condi- absorbing large quantities of rainfall and releasing it throughout tions for a study area. By applying a water balance model and/or the year during the low-flow seasons. Further south where the till a water quality model, the projected climatic regimes are used to cover with impervious shale gradually replaces the glacial drift, the calculate runoff and simulate pollutant behavior. For example, dry-weather flow decreases, and the stream flow is often Gleick (1987) developed a monthly water balance model to augmented by wastewater discharges (Schneider, 1957). examine runoff and soil moisture changes under different climate The topography of the watershed is influenced by the last three change scenarios in the Sacramento River Basin. Mimikou, glaciations and the subsequent erosion by rivers. Most of the area is Kouvopolos, Cavadalis, and Vayianos (1991) examined the flat to gently rolling with steep-walled river valleys. In many places, regional effects of annual temperature increase on the spatial and outcropping and shale are exposed. The regional topographic temporal redistribution of water resources. Guo, Wang, Xiong, Ying, gradient is from north to south. The soils in the region belong to the and Li (2002) used a macro-scale and semi-distributed monthly GeneseeeWilliamsburg Association. Formed from silts, alluvial, water balance model to predict the impacts of climate change on and residual materials from the glacial deposits, the soils are deep the magnitude and timing of runoff and assess the vulnerability of and highly productive, but susceptible to erosion (Lerch, Hale, & the water resource in north China. Most of these studies evaluate Milliron, 1975). The northern portion of the watershed is charac- only the annual or seasonal stream flow, or the surface or ground terized by gently sloping land, low gradient streams, and areas of water quality. fertile soil. In the southernmost areas, the terrain is more dissected In land use studies, it is recognized that changes in land use will and hilly with a higher stream density and more drainage problems affect the rates of infiltration, evapotranspiration, and groundwater (Debrewer et al., 2000). Soils on the southeast older till plain are recharge, as well as the water quality in receiving water bodies, and less extensively cultivated than the younger till-derived soils in the climate change can further amplify the hydrologic effects. None- northwest. theless, the current land use modeling applications are often con- The LMR watershed has a cool temperate climate; summers are ducted with the assumption that the local climate would remain warm and humid with a high temperature of 30 C and a low constant for the simulation period. For example, Jones (1997) used temperature of 15 C, while winters are moderately cold with a few an export coefficient model to simulate the impacts of agricultural annual winter frosts and snowfall. Winter highs are around 0 C land change and catchment management strategies on nitrogen and lows about 10 C. Average annual air temperature ranges and phosphorus loadings. Zandbergen (1998) developed a concep- from 10 C in the north to 13 C in the south. The average annual tual model to demonstrate that the hydrologic impacts of urbani- precipitation for the area ranges from 90 to 110 cm; about one-third zation can be mitigated by changing the amount of impervious of the precipitation becomes surface runoff. Average snowfall in the areas and the riparian habitats. Bronstert, Niehoff, and Bürger watershed is 50e76 cm per year (Debrewer et al., 2000). (2002) used WaSiM-ETH model to assess the effects of urbanization and agricultural management on storm runoff. Rationale for choosing the LMR watershed as the study area In the face of impending climate and land use changes, there is The LMR watershed was selected to investigate the plausible a need for new integrative approaches that model not only the hydrologic and water quality impacts of climate and land use separate but also the combined impacts of these changes as they act changes because this predominately agricultural watershed has in tandem with each other. Hence, the second objective of this been undergoing a rapid urbanization process, and its water research was to derive an approach that could be used to examine resources are deteriorating. The LMR was once a pristine river the hydrologic and water quality effects of population-driven land system with minimal anthropogenic impacts. Containing some of use change in concert with climate change, highlight the extent to the most scenic and diverse riverine habitats in the Ohio River which the combination of climate and land use changes could Valley, it is home to more than 340 species of wildflowers, 250 bird amplify or ameliorate the hydrologic and water quality effects at species, and over dozens of fish species, some of which are rare, a watershed scale, and predict the plausible combined conse- threatened, or endangered (USDA, 1999). Because of its high water quences of climate, population, and land use changes under sets of quality, diversified aquatic communities of flora and fauna, pano- scenarios. ramic setting, and historic sites found along its banks, the Ohio Department of Natural Resources (Ohio DNR) designated the LMR Material and methods a state scenic river in 1969 and a national scenic river in 1973. It is the first state and national scenic river in Ohio. In addition, the Ohio Study area Environmental Protection Agency (Ohio EPA) listed the river as an Exceptional Warm Water Habitat Stream. It was also designated for The LMR watershed Primary Contact Recreation (East Fork LMR Collaborative, 2007). The LMR watershed was selected as a case study in this research Historically, most of the land area in the LMR watershed was (Fig. 1). Originating at the southeast of Springfield in southwestern used for agricultural activities. But in recent decades, the pop- Ohio, the LMR, a major tributary of the Ohio River, flows 169.78 km ulation in the watershed has grown substantially, and the water- from Clark County through several steep-sloped forested gorges to shed has undergone a rapid urbanization process. As calculated join the Ohio River at the confluence in Hamilton County, near the from the land use maps, from 1980 to 2001, the agricultural land eastern side of Cincinnati. Draining an area of 5840 sq km, the use has experienced a drastic decrease of 24.08%, whereas urban watershed encompasses Clark, Greene, Warren, Clermont, and land has increased by 6.78% (see Table 1). In the same period,

6 S.T.Y. Tong et al. / Applied Geography 32 (2011) 477e489 479

Fig. 1. The Little Miami River watershed study area. population from the ten counties in the basin has grown from organic compounds, and excessive amounts of nitrates were found 2,371,943 in 1980 to 2,530,076 in 2000, an increase of 6.67% (US in many locations of the LMR; some with concentrations at or Census Bureau, 2010). above drinking-water standards or guidelines for protecting With these recent developments, the water quality in LMR is aquatic life (USGS, 2004). As the population increases and declining. The northern half of the LMR watershed, still primarily urbanization continues, there will be an increase in water demand agricultural, is impacted by non-point source pollution. In the as well as wastewater discharge and combined sewer overflows, south, where there is extensive urbanization and suburban sprawl, which may further aggravate the conditions in the LMR, reducing the surface runoff is enriched with sediment discharge, rubber water availability and degrading water quality (Ohio EPA, 1996). fragments, motor oils, nutrients, heavy metals, hydrocarbons, The likelihood of climate change poses an additional threat to the chlorides, conservative solutes, and bacteria. Since the early 1990s, LMR watershed. In order to conserve the area and to protect our there have been indications that the LMR ecological system is limited and valuable water resources, it is crucial to be able to under stress. For instance, the Ohio EPA (2000) has discovered predict the hydrologic and water quality ramifications of future high levels of fish anomalies, as well as widespread presence of climate and land use changes so that effective water management Escherichia coli and Fecal coliform bacteria and pollution-tolerant actions can be implemented. The LMR watershed therefore is fish species. Janosy (2003) reported abnormally high levels of a good study area for this research. arsenic, cadmium, copper, mercury, selenium, and zinc in fish collected from the river. Atrazine, metolacholor, semivolatile Development of a hydrologic and water quality model

Table 1 Selection of a hydrologic and water quality model Land use categories in the LMR watershed in 1980, 1992, 2001 and 2050 by The Hydrologic Simulation Program e Fortran (HSPF) model percentages. (Bicknell, Imhoff, Kittle, Jobes, & Donigian, 2000) was selected in this Land use types 1980 1992 2001 Projected 2050 study to simulate the water quantity and quality. It is a robust, Water 0.68% 0.94% 0.97% 1.55% reliable, and comprehensive model commonly applied to ﬂood Urban 11.01% 17.72% 17.79% 39.22% forecasting, water quality modeling, as well as assessment of best Forest 7.60% 20.03% 23.65% 31.28% management practices and sensitivity of stream ﬂow to climate Agriculture 80.26% 70.93% 56.18% 26.46% change (Bicknell, Donigian, Jobes, & Chinnaswamy,1996; Chun et al., Others 0.45% 0.38% 1.41% 1.49% 2001; Cryer, Fouch, Peacock, & Havens, 2001; Donigian & Huber,

7 480 S.T.Y. Tong et al. / Applied Geography 32 (2011) 477e489

1991; Donigian & Crawford, 1976; Ng & Marsalek, 1992; Tsihrintzis, Table 3 Fuentes, & Gadipudi, 1997). Calibration/validation targets for HSPF. As a lumped conceptual model, HSPF uses defined meteoro- Very good Good Fair logical and watershed conditions to simulate stream flow compo- Hydrology/Flow <10a 10e15 15e25 nents and model water quality (Bicknell et al., 2000). There are Sediment <20 20e30 30e45 three basic modules in HSPF: PERLND is the module for pervious Nutrients <15 8e12 13e18 lands, IMPLND simulates the hydrodynamics on impervious land a The figure shown is the % difference between simulated and observed values segments, and RCHRES is the module for free-flow reaches in according to Bicknell et al. (2000). streams or lakes, each of which has different water balance equa- tions for the calculation of flow and water quality. both the calibration and validation results were within acceptable limits, it seemed that the developed hydrologic model was suffi- The hydrologic model for the LMR watershed cient for flow simulation in the LMR watershed. In order to construct a hydrologic model for the LMR watershed, a base model for the early 1980’s conditions was first compiled by The water quality model for the LMR watershed delineating the LMR watershed using the hydrologic unit code In this study, nitrates and nitrites as total nitrogen (N) and total (HUC) from the U.S. Geological Survey (USGS), stream characteris- phosphorus (TP) were used for water quality simulation because tics and Reach File coverages (RF3) from the U.S. Environmental these pollutants are commonly found in the LMR watershed and Protection Agency (USEPA), and digital elevation models at are considered important water quality indicators. Point source TP 30 30 m resolution (DEMs) from the USGS. Then the daily and N data from municipal and industrial wastewater treatment meteorological data from Jan 1, 1980 to Dec 31, 1984 from the plants that discharged to the channel reaches were retrieved from National Climatic Data Center (NCDC) for the climatic station at the Permit Compliance System (PCS) from the USEPA and Ohio EPA Dayton Airport, Ohio, the digitized soil layer from the national State and added to the model. Soil Geographic database (STATSGO) collected by the U.S. Depart- As in the development of the hydrologic model, the water quality ment of Agriculture (USDA), and the 1980 land use/land cover layer data from the USEPA’s STORET archive from 1980 to 1984 were used (LULC) from the USGS Geographic Information Retrieval and in model calibration, and those from 1985 to 1989 were used in Analysis System (GIRAS), as well as the default parameters model validation. Several parameters, including the storage of provided by HSPF were used to run the hydrologic model. The phosphorus on the pervious land segment (SQO), the rates of model was calibrated by comparing the simulated flow results with accumulation for phosphorus (ACQOP), and the maximum storage the observed daily discharge records from the USGS gauging station of phosphorus (SQOLIM), were adjusted. The final simulated mean near Milford (gauging number: 03245500) and iteratively adjusting daily TP concentration was 0.402 mg/L. The E statistics for the the values of the model parameters to match the local conditions of calibration period was 0.42 and that for the validation period was the river basin by trial-and-error until the error rate, which was 0.39. Although the E values were lower than desired, the correlation calculated as [(simulated value e observed value)/observed value], coefficient between the simulated and observed TP was 0.83 for the between simulated and observed stream flow was acceptable. After calibration period and 0.80 for the validation (Table 4), indicating numerous trials with adjustments of the lower zone nominal soil that the model could capture most of the variations in TP. moisture storage value (LZSN), the index to mean soil infiltration For the N modeling, the correlation coefficients between simu- rate (INFILT), the fraction of groundwater inflow entering deep lated and observed mean daily concentration were 0.87 for the groundwater (DEEPFR), the interflow parameter (INTFW), and the calibration period and 0.83 for the validation period. The E index interflow recession parameter (IRC), the error rate was approxi- was 0.54 and 0.46 for the calibration and validation periods, mately 11.74% (Table 2). According to Bicknell et al. (2000), an error respectively (Table 5). These results were in line with an earlier rate below 10% in flow simulation is considered as very good, and study (Liu & Tong, 2011), suggesting that the water quality model a range from 10 to 15% is regarded acceptable (Table 3). The cali- was acceptable. bration results also showed that the correlation coefficient between simulated and observed flows was 0.88 and the Nash-Sutcliffe Generation of the future climate change scenarios model efficiency coefficient, E (Nash & Sutcliffe, 1970), a statistic used to indicate how consistently observed values agree with Despite the continual development in GCMs and RCMs, uncer- predicted values, was 0.69 (Table 2). tainty exists on the magnitude of the changes in temperature and After calibration, the HSPF model was validated to ensure that precipitation (Wilby et al., 2006). Furthermore, these global and the model could fairly accurately simulate the real world conditions regional models are less reliable in simulating detailed spatial and even under different land use and climate regimes. In the valida- temporal features (Mailhot, Duchesne, Caya, & Talbot, 2007). Given tion, the 1992 land use, 1985e1989 weather data, and the param- the difficulties in deducing local climate patterns from regional eter values used in the calibration were utilized. The validation trends, some scientists, such as Hotchkiss, Jorgensen, Stone, and results showed a smaller difference (3.84%) between simulated and Fontaine (2000), Stonefelt, Fontaine, and Hotchkiss (2000), observed values. The correlation coefficient between simulated and Dagnachew, Vallet-Coulomb, and Gasse (2003), Zhu, Jenkins, and observed flows was 0.91, and the E statistic was 0.72 (Table 2). Since Lund (2005), Thomson, Brown, Rosenberg, Srinivasan, and

Table 2 Calibration and validation results for the hydrologic model.

Mean observed Mean simulated % Error between simulated and Average daily Nash-Sutcliffe daily flow (m3/s) daily flow (m3/s) observed flow valuesa correlation coefficient efficiency coefficient Calibration Period 1980e1984 37.81 33.37 11.74% 0.88 0.69 Validation Period 1985e1989 32.51 31.26 3.84% 0.91 0.72

a % error ¼ [(simulated - observed)/observed] 100. Source: Bicknell et al. (2000).

8 S.T.Y. Tong et al. / Applied Geography 32 (2011) 477e489 481

Table 4 Table 6 Calibration and validation results for the mean daily TP concentrations in mg/L. Hypothetical future climate scenarios.

Calibration Validation Climate scenario Changes in temperature/precipitation

Observed Simulated Observed Simulated Base case (BASE CASE) No change in temperature and precipitation data values data values Wettest (W2) þ2 C, þ20% precipitation Wet (W4) þ4 C, þ20% precipitation e e e e 1980 1984 1980 1984 1985 1989 1985 1989 Dry (D2) þ2 C, 20% precipitation Apr 0.263 0.372 DRIEST (D4) þ4 C, 20% precipitation May 0.392 0.348 0.484 0.449 Jun 0.404 0.383 0.510 0.489 Jul 0.512 0.488 1.845 0.942 Aug 0.442 0.452 1.082 0.916 coupling the Markov Cellular Automata (CA-Markov) model from Sep 0.508 0.529 0.715 0.721 IDRISI (2008) with a population variable was developed to generate Oct 0.333 0.303 0.838 0.801 Nov 0.370 0.351 a 2050 land use scenario. IDRISI is a geographic information system Mean 0.402 0.403 0.912 0.719 (GIS) and image processing software designed for spatial modeling, Correlation 0.83 0.80 decision support application, risk analysis, and spatial statistics fi coef cients computation. Through the Markov chain and/or the cellular Nash-Sutcliffe 0.42 0.39 efficiency automata (CA) modules, IDRISI is capable to simulate future land coefficient use conditions. Markov chain is a stochastic method for generating a series of random values. Their probabilities at a certain time interval depend Izaurralde (2005), and Muzik (2002), preferred to use a range of on the previous values. In land use prediction, Markov analysis uses hypothetical climate scenarios from the estimates of one or more the historical pattern of land use changes to calculate transition GCMs to simulate the watershed hydrologic responses. Since these matrices and determine the future land use pattern. Its major hypothetical scenarios cover a wide range of climatic regimes, this drawback is that it does not consider any geographical relation- approach is simpler and can save considerable time and effort. ships. But the use of CA in Markov can add the spatial dimension to Our purpose in this study was not to produce future climate the model, thereby capturing the temporal-spatial dynamics change predictions but to determine the plausible hydrologic (Tobler, 1970). In CA, the state of each cell at time tþ1 is determined impacts from a range of possible future climate change conditions; by its neighboring cells at time t according to the pre-defined for this reason, we adopted the latter approach to generate the transition rules (White & Engelen, 1993). By using the Multi- climate change scenarios. Based on the information provided by Criteria Evaluation (MCE) in IDRISI, a user-defined variable (such Karl, Knight, Easterling, and Quayle (1996) and USEPA (1998), four as a population variable) can also be coupled in the CA-Markov climate scenarios (W2, W4, D2, and D4) delineating the possible model to improve the performance of the original model. ranges in temperature (þ2 and þ4 C) and precipitation (þ20% and 20%) in Ohio by the horizon year 2050 were generated Development of a CA-Markov land use model for the LMR watershed (Table 6). These scenarios, together with the base case scenario To build the land use model for LMR watershed, two sets of (BASE CASE) where the climate regime was remained at the historical land use records were required to determine the pattern 1980e1989 level, were used at a later stage to model the impacts of of land use change, and an additional land use map was needed for different climatic conditions on water resources. validation. In this research, the USGS 1980 Land Use and Land Cover (LULC), and the 1992 and 2001 land use maps from the National Generation of the future land use change scenario Land Cover Data (NLCD) data were adopted (Fig. 2). The 1980 and 1992 maps were used as the base maps to develop the model, A common approach to derive future land use scenarios is to whereas the 2001 map was employed for validation. This was not adopt a land use model to simulate future land use conditions. This ideal as these two datasets had certain differences. The LULC data method enables one to predict the future conditions under certain were derived from aerial photographs from the GIRAS, and the known assumptions. In this study, an enhanced land use model NLCD data were derived from satellite imageries from the Multi- Resolution Land Characteristics Consortium (MRLC). They also

Table 5 differed in terms of mapping units. However, due to the lack of Calibration and validation results for the mean daily N concentrations in mg/L. comprehensive historical land use images, it was the only option available. Besides, they were collected by the same agency and Calibration Validation were classified using the same method from Anderson, Hardy, Observed Simulated Observed Simulated Roach, and Witmer (1976). To minimize inconsistence, we further data values data values re-sampled and re-classified the land use classes for all three 1980e1984 1980e1984 1985e1989 1985e1989 imageries into five categories using Anderson Level I classification Apr 2.70 2.42 method (see Table 1). These maps were also projected into NAD May 2.32 2.92 1.88 2.06 1983 State Plane Ohio South FIPS 3402 coordinate system and Jun 3.71 3.04 2.97 2.37 Jul 3.01 3.15 2.51 2.11 resized to ensure conformity in size and dimension. Aug 1.30 1.71 1.66 1.89 The 1980 and 1992 base maps were imported into IDRISI to Sep 1.95 1.76 2.39 2.2 project the land use pattern for 2001, the validation year. After Oct 1.37 1.77 2.39 2.38 running the Markov model, the transition probability matrix Nov 1.83 2.02 between each land use class, a transition area file, and a set of five Mean 2.28 2.35 2.30 2.18 Correlation 0.87 0.83 probability images for each land use class were created. When the coefficients transition area file and the probability images for each land use Nash-Sutcliffe 0.54 0.46 class were imported into the CA-Markov model, a projected land fi ef ciency use map for 2001 was generated (Fig. 3). Two validation statistics coefficient were applied to assess the accuracy of the model. The Relative

9 482 S.T.Y. Tong et al. / Applied Geography 32 (2011) 477e489

Fig. 3. Projected 2001 land use map generated from CA-Markov.

10 S.T.Y. Tong et al. / Applied Geography 32 (2011) 477e489 483

Operation Characteristic (ROC) examines the validity of a model by procedure for structuring and aiding complex decision-making comparing a suitability map showing the likelihood of a land use processes (Proctor, 2001). The Weighted Linear Combination class with a Boolean map showing where that land use class (WLC) function in MCE was employed to generate the population actually exists. A high ROC value indicates a better suitability map density mask by assigning a 50% weight for population growth and a better prediction result. In this study, the urban suitability based on the assumption that population growth had an equally map was compared with the actual urban area from the 2001 NLCD important contribution to urban expansion as the transition map. Another method to validate the accuracy of model is Kappa probabilities. After running the MCE, a new population coupled statistic, which compares the agreement of quantity and location of urban suitability map for 2001 was generated. ROC validation was cells in the categorical image. The ROC and Kappa statistic results used again to evaluate whether the new suitability map could (0.957 and 0.9495, respectively; Table 7) showed that the perfor- provide a better prediction than the original suitability map. The mance of the land use model was adequate. result indicated a higher ROC value with the population coupled suitability map than the one without the population variable. The Prediction of future population growth population coupled urban suitability map was then imported into The close relationship between population growth, urbaniza- the CA-Markov model to generate a new land use distribution map tion, and suburban sprawl is well known. It is likely that by intro- for 2001 (Fig. 5). To further ascertain the efficacy of the population ducing a population variant to the CA-Markov analysis, the coupled CA-Markov model in postulating the future land use projection capability of the land use model can be improved. As pattern, the new projected land use map was compared to the demonstrated by de Almeida et al. (2003), the incorporation of the actual land use distribution map in 2001 from NLCD using Kappa population density in the neural network analysis can improve the statistic (Fig. 2c). predictive power of the model. In order to ascertain this concept As shown from the results of ROC and Kappa index (Table 7), and to portray the importance of population growth in land use the projected 2001 land use pattern generated with the change, a population variable was developed and incorporated into population-coupled CA-Markov model displayed a better agree- the CA-Markov model by generating another set of a transition area ment with the actual 2001 land use than the one without the file and a transition probability map with a population variant. population variant. The ROC values increased from 0.957 to 0.977, In order to develop the population variable, the trend of pop- while the Kappa statistics changed from 0.9495 to 0.9507. These ulation growth has to be determined and the future population results suggested that the method of incorporating the pop- estimated. To this end, the 1990 and 2000 census block group ulation variable into the CA-Markov model was reasonable; it population data were abstracted from the U.S. Census Bureau could improve the model performance and help to better predict (2008) and projected into the year 2007 with the linear, the 2050 land use pattern. geometric, and exponential population models. The year 2007 was chosen because of the availability of population data from the Ohio The 2050 land use change scenario for the LMR watershed Department of Development (2005). The population projection Following similar procedures, the 2050 land use scenario was results from each method were then compared with the 2007 generated using the CA-Markov model in conjunction with the official data (Table 8). A two-tailed t-test was used to examine the 2050 population density variable generated from MCE. The base statistical differences between the predicted values and the official maps for 2050 projection were the 1992 and 2001 NLCD land use 2007 estimates (Table 9) and to determine the best model of maps (Fig. 2b and c). According to the linear growth model, the population growth in the area. This step was to ascertain the population for the horizon year of 2050 was projected and the accuracy of the population trend for postulating future population population density of the watershed was calculated. The WLC growth. Since the t-test results showed that the linear model (Stoto, method in MCE was used to generate the 2050 population density 1983) had provided the best estimation of the future population variable (Fig. 6a) and the population coupled suitability map of trend, it was selected to project the population for the validation urban area (Fig. 6b). The latter was imported into the CA-Markov year 2001. The projected population was converted into a pop- model to depict future population and urban growth and to ulation density map (Fig. 4a) and a ratio map with population postulate the land use pattern for the 2050 horizon year (Fig. 7). density standardized from low (0) to high (1), the latter was used as This scenario of land use change therefore was based not only on the urban suitability map imparting the influence of population the information of land cover change from 1992 to 2001, but also growth on land use change (Fig. 4b). on the projected population growth from 2001 to 2050. The projection results showed that there would be substantial urban Incorporation of the population variable in the CA-Markov land use development in the south-western and north-western portions of model the watershed. The urban suitability map was incorporated into CA-Markov through the Multi-Criteria Evaluation (MCE), which is an effective Modeling of the hydrologic and water quality impacts of climate, population, and land use changes

Table 7 To predict the 2050 hydrologic and water quality conditions in Validation results for 2001 land use projections. the LMR basin under future climate and land use change scenarios, Comparison ROC a new model had to be prepared by incorporating the projected Original suitability map without population variable versus 2001 0.957 2050 land use map and the four hypothetical climate scenarios for NLCD urban area 2050 and the BASE CASE scenario into the validated HSPF model. New suitability map with population variable versus 2001 NLCD 0.977 The model was then used to examine both the separate and the urban area combined impacts of climate change and population-driven land Comparison Kappa use change under the following scenarios: Projected 2001 land use without population variable versus 2001 0.9495 NLCD (1) The no change scenario (NO CHANGE) where climate was kept Projected 2001 land use with population variable versus 2001 0.9507 at the 1980e1989 BASE CASE level and land use remained as NLCD the 1992 pattern;

11 484 S.T.Y. Tong et al. / Applied Geography 32 (2011) 477e489

Table 8 2007 county population projections from the linear, geometric, and exponential models.

County 1990 Population 2000 Population 2007 estimate 2007 projection 2007 projection 2007 projection (ofﬁcial) (linear model) (geometric) (exponential) Brown 34966 42285 43956 47408 48302 42285 Butler 291479 332705 357888 361563 364986 332705 Clark 147548 144738 140477 142771 142803 144738 Clermont 150167 177450 193490 196548 1994467 177450 Clinton 35417 40543 43071 44131 44567 40543 Greene 136731 147886 154656 155695 156232 147886 Hamilton 866228 845303 842369 830656 830957 845303 Highland 35728 40875 42653 44478 44913 40875 Madison 37068 40216 41499 42420 42577 40216 Montgomery 573809 559062 538104 548739 548965 559062 Warren 113927 159013 204390 190573 200815 159013

Table 9 The p-value results from the t-tests between 2007 pro- ﬁ jected population values and the of cial estimates. - The wet climate change with 4 C increase in temperature and Model p-value 20% increase in precipitation only (W4); Linear model 0.995 - The dry climate change with 2 C increase in temperature and Geometric model 0.982 20% decrease in precipitation only (D2); Exponential model 0.954 - The driest climate change with 4 C increase in temperature and 20% decrease in precipitation only (D4); (2) The land use change only (LU) scenario where the projected 2050 land use scenario was used in conjunction with the BASE (7e10) the scenarios where climate change was considered CASE climate condition; together with land use change e

(3e6) the scenarios where only climate was changed e - The wettest climate and land use changes (W2 þ LU); - The wet climate and land use changes (W4 þ LU); - The wettest climate change with 2 C increase in temperature - The dry climate and land use changes (D2 þ LU); and and 20% increase in precipitation only (W2); - The driest climate and land use changes (D4 þ LU).

Fig. 4. (a) Projected 2001 population density map, (b) 2001 MCE urban suitability map.

12 S.T.Y. Tong et al. / Applied Geography 32 (2011) 477e489 485

Fig. 5. Projected 2001 land use map generated from the population coupled CA-Markov.

Fig. 6. (a) Projected 2050 population density map, (b) 2050 MCE urban suitability map.

13 486 S.T.Y. Tong et al. / Applied Geography 32 (2011) 477e489

Fig. 7. Projected 2050 land use map generated from the population coupled CA-Markov.

The simulation results from each scenario were compared to coupling the projected land use and climate changes, the future assess the individual and combined hydrologic and water quality stream flow could increase an additional 4.76 m3/s (11.41%) from influences of climate and land use changes. the LU scenario and an additional 4.27 m3/s (10.12%) increase from the W2 scenario. A similar case was found under the D4 þ LU scenario where land use change could ameliorate the dry condition Results and discussion by increasing the flow regime from 10.00 to 15.16 m3/s, a difference of 5.16 m3/s or an increase of 51.60%. Stream flow These results implied that the projected land use change could increase the flow, which would be a welcoming relief under dry When compared to the NO CHANGE scenario, where the climate conditions, mitigating the drought impacts of climate change. On and land use conditions were kept at the 1980s level, future climate the contrary, if our climate became wetter, than the impending change would alter the flow regime. It would induce an increase of urbanization in the river basin would exacerbate the flow regimes, 30.61% flow under the W2 scenario, a 12.84% increase under the perhaps causing more floods. W4 scenario, a 50.48% reduction under the D2 scenario, and When all the scenarios were compared, the greatest increase in a 69.05% reduction under the D4 scenario (Table 10). These results daily flow occurred under the W2 þ LU scenario (43.83%), while the showed that under the wet conditions, with a 2 C increase in greatest decrease took place under D4 scenario (69.05%). temperature, as in the case from W2 to W4, the daily flow would be reduced by 5.74 m3/s, which would be a 13.60% reduction. Similar situation was found under the dry conditions. A 2 C increase in Table 10 3 temperature would induce a reduction of 6 m /s, or 37.50% in daily Modeling results of the effects of future climate and land use changes on stream flow. But with a 40% precipitation change, for example, from W2 to flow. fl 3 D2, the amount of ow would be reduced by 26.2 m /s, or 62.09%. Scenario Mean daily % Difference Using the projected 2050 land use scenario, the simulated flows (m3/s) from NO CHANGEa amount of daily flow would increase 29.09% (Table 10). This might No change (NO CHANGE) 32.31 be attributed to the fact that, according to the projection, there Land use change only (LU) 41.71 29.09% would be more urban areas in the LMR basin in the year 2050. The Wettest climate change only (W2) 42.20 30.61% Wettest climate and land use changes 46.47 43.83% effect of the projected land use change on stream flow was higher (W2 þ LU) than the W4 climate scenario (12.84%), almost the same as the W2 Wet climate change only (W4) 36.46 12.84% scenario (30.61%), but much lower than under the D2 and D4 Wet climate and land use changes 37.55 16.22% conditions (50.48% and 69.05%, respectively). Hence, climate (W4 þ LU) change would have a more prominent hydrologic effect than land Dry climate change only (D2) 16.00 50.48% Dry climate and land use changes 18.12 43.92% use change if our environment became dry. (D2 þ LU) When climate change was considered together with land use Driest climate change only (D4) 10.00 69.05% change, the impact on discharge was more apparent. More daily Driest climate and land use changes 15.16 53.08% discharge (14.16 m3/s, a 43.83% increase) would be found under the (D4 þ LU) W2 þ LU scenario than the NO CHANGE scenario, indicating that by a % Difference ¼ (current scenario e NO CHANGE)/NO CHANGE.

14 S.T.Y. Tong et al. / Applied Geography 32 (2011) 477e489 487

Water quality change Table 12 Modeling results of the effects of future climate and land use changes on N.

Compared to the NO CHANGE scenario, the simulation results of Scenario Mean daily % Difference TP showed an increase in the mean daily concentration under all concentration with NO CHANGEa scenarios (Table 11). The greatest increase in the mean daily TP (mg/L) concentration was found under the W2 þ LU scenario (21.35%). No Change (NO CHANGE) 2.77 According to the USEPA Ambient Water Quality Criteria Recom- Land use change only (LU) 2.86 3.25% Wettest climate change only (W2) 2.99 7.94% mendations (USEPA, 2005), the level for phosphorus in freshwater Wettest climate and land use 3.09 11.55% systems is limited to 0.2e0.3 mg/L, beyond which there would be changes (W2 þ LU) a negative effect on aquatic ecosystems. The concentration of Wet climate change only (W4) 2.91 5.05% 0.415 mg/L found under this scenario would exceed this limit. The Wet climate and land use 2.97 7.22% changes (W4 þ LU) smallest increase of phosphorus occurred under the D4 scenario Dry climate change only (D2) 2.81 1.44% (2.63%); however, the concentration level under this scenario Dry climate and land use 2.85 2.89% (0.351 mg/L) would still be higher than the USEPA standard. changes (D2 þ LU) Land use change alone, the LU scenario, would produce a 4.09% Driest climate change only (D4) 2.83 2.17% increase of TP concentration from the NO CHANGE scenario, which Driest climate and land use 2.88 3.97% changes (D4 þ LU) would be less than the effects of climate change under W2, D2, and a ¼ W4 scenarios (14.33%, 7.60%, and 5.85%, respectively). These results % Difference (current scenario NO CHANGE)/NO CHANGE. indicated that climate change and, in a slightly lesser extent, land use change, would increase daily TP concentration. parameters are critical to the integrity of ecosystem and human The results from N modeling also revealed a moderate increase health. An increase in nutrient levels would likely cause more in the mean daily concentration under all scenarios (Table 12). incidences of algae blooms and eutrophication, degrading water Under the wet conditions, an increase of only 2 C in temperature quality. Hence, with the impending global warming and urbani- (in the case of W2) would induce an increase of the mean daily N zation, it seemed that the LMR basin would experience more concentration of 7.94% from the NO CHANGE scenario, whereas an nutrient enrichment problems. increase of 4 C (in the case of W4) could cause an increase of 5.05%. Under the dry conditions, the effects of temperature increase were Conclusion not as apparent, as the mean daily N concentration only increased 1.44% (under D2) to 2.17% (under D4) from NO CHANGE. This research attempted to derive an integrated spatial analyt- When climate change was combined with land use change, the ical methodology capable of postulating the possible hydrologic simulated mean daily N concentration increased somewhat. The and water quality ramifications of future changes in terms of largest effect of land use in such a combination was found under the climate, population, and land use. Using the approach outlined in W2 þ LU scenario, where the N concentration increased from 2.99 this paper, the separate as well as the combined impacts of climate under W2 scenario to 3.09 mg/L, a difference of 0.10 mg/L. This and land use changes on water resources in the year 2050 in the signified that when the agglomerate effects were considered, the LMR basin were examined. effects of land use change alone could only contribute a slight The hydrologic and water quality modeling results in the LMR increase in N concentration. basin revealed that stream flow would change in accordance with When compared with all the scenarios, the W2 þ LU had the climate change, although land use modification might be able to highest increase of nitrogen from NO CHANGE (11.55%), while the mitigate some of the effects. In general, simulations under the dry D2 scenario generated the least increase (1.44%). conditions paired with future land use change scenario (D2 þ LU Judging from these results, it was evident that both climate and D4 þ LU) produced an increase in daily flow in the LMR and change and land use change could have negative impacts on water a slight increase in the daily concentrations of TP and N than under quality. When the climate and land use changes occurred simul- the climate change only (D2 and D4) scenarios. The results from the taneously, the joint impacts could be intensified. In this study, both combined wettest climate scenario and land use change scenario TP and N simulations showed an increase in concentrations under (W2 þ LU) showed a much larger increase in daily flow as well as all climate and land use change scenarios. These two water quality a higher level of nutrients in the water. However, regardless of the future scenarios, it was predicted that the mean daily phosphorus (TP) and nitrates and nitrites (as N) concentrations in the receiving Table 11 water bodies would increase. Thus, nutrient enrichments would be Modeling results of the effects of future climate and land use changes on TP. a problem in the future, especially under the wet conditions. Scenario Mean daily % Difference By examining not only the separate but also the combined concentration (mg/L) with NO CHANGEa impacts of climate and land use changes on water resources, it No change (NO CHANGE) 0.342 would help us to further our understanding of the dynamics of the Land use change only (LU) 0.356 4.09% physical system in a watershed. The scenario results from this study Wettest climate change only (W2) 0.391 14.33% depicted a possible range of future flow and water quality condi- Wettest climate and land use 0.415 21.35% tions, which could be of values to the decision-makers in their changes (W2 þ LU) Wet climate change only (W4) 0.362 5.85% development of adaptation and mitigation strategies in preparation Wet climate and land use 0.370 8.19% for future climate and land use changes. Using this information, changes (W4 þ LU) better comprehensive and sustainable watershed protection Dry climate change only (D2) 0.368 7.60% programs, including erosion and sediment control, storm water Dry climate and land use 0.383 11.99% changes (D2 þ LU) management, and best management practices, could be devised to Driest climate change only (D4) 0.351 2.63% minimize the adverse impacts of flow and non-point source Driest climate and land use 0.365 6.73% pollution in the face of these impending changes. changes (D4 þ LU) The results also demonstrated the efficacy of HSPF in modeling a % Difference ¼ (current scenarioNO CHANGE)/NO CHANGE. water quantity and quality under a watershed scale. The application

15 488 S.T.Y. Tong et al. / Applied Geography 32 (2011) 477e489 of CA-Markov model coupling with a population variable also IPCC. (2008). Contribution of working group I to the fourth assessment report of the proved to be more effective than the one without the population intergovernmental panel on climate change, 2007. Cambridge, United Kingdom and New York, NY, USA: Cambridge University Press. variant in simulating future land use changes, providing a more Janosy, S. D. (2003). Trace elements and synthetic organic compounds in streambed realistic land use pattern for the year 2050. This comprehensive sediment and fish tissue in the Great and Little Miami River Basins, Ohio and Indiana, e approach seemed to be reliable and might provide a reasonable tool 1990 98. National Water-Quality Assessment Program, Water Resources Inves- tigations Report 02-4305. Columbus, Ohio: U.S. Geological Survey. for predicting the long-term impacts of land use and climate Jones, J. B. (1997). Benthic organic matter storage in streams: influence of detrital changes on water resources, useful to environmental scientists, import and export, retention mechanisms, and climate. Journal of the North state and local agencies, watershed managers, and regional American Benthological Society, 16(1), 109e119. Karl, T. R., Knight, R. W., Easterling, D. R., & Quayle, R. Q. (1996). Indices of climate planners. change for the United States. Bulletin American Meteorological Society, 33, 279e292. Lerch, N. K., Hale, W. F., & Milliron, E. L. (1975). Soil survey of Clermont County. Ohio: Acknowledgments Soil Conservation Services, U.S. Department of Agriculture. Liu, Z., & Tong, S. T. Y. (2011). Using HSPF to model the hydrologic and water quality impacts of riparian land-use change in a small watershed. Journal of Environ- This research was partially funded by the U.S. Environmental mental Informatics, 17(1), 1e14. Protection Agency, through its Office of Research and Development. Mailhot, A., Duchesne, S., Caya, D., & Talbot, G. (2007). Assessment of future change The authors are grateful to the agency for the financial support. The in intensity-duration-frequency (IDF) curves for Southern Quebec using the e manuscript has been subjected to the Agency’s administrative Canadian Regional Climate Model (CRCM). Journal of Hydrology, 347,197 210. Mimikou, M., Kouvopolos, Y., Cavadalis, G., & Vayianos, N. (1991). Regional hydro- review and has been approved for external publication. Any opin- logical effects of climate change. Journal of Hydrology, 123,119e146. ions expressed in this paper are those of the authors and do not Muzik, I. (2002). A first-order analysis of the climate change effect on flood necessarily reflect the views of the Agency; therefore, no official frequencies in a subalpine watershed by means of a hydrological rainfall-runoff model. Journal of Hydrology, 267,65e73. endorsement should be inferred. Any mention of trade names or Nash, J. E., & Sutcliffe, J. V. (1970). River flow forecasting through conceptual models. commercial products does not constitute endorsement or recom- Part 1: a discussion of principles. Journal of Hydrology, 10(3), 282e290. mendation for use. NCSL, CIER (2008). Ohio: assessing the costs of climate change, economic and environmental costs of climate change reports. Paper presented at the National Conference of State Legislatures and the Center for Integrative Environmental Research, University of Maryland. http://www.ncsl.org/IssuesResearch/ References EnvironmentandNaturalResources/ClimateChangePublications/tabid/13242/ Default.aspx Accessed 01.05.11. de Almeida, C. M., Batty, M., Monteiro, A. M. V., Camara, G., Soares-Filho, B. S., Ng, H. Y. F., & Marsalek, J. (1992). Sensitivity of streamflow simulation to changes in Cerqueira, G. G., et al. (2003). Stochastic cellular automata modeling of urban climatic inputs. Nordic Hydrology, 23,257e272. land use dynamics: empirical development and estimation. Computers, Envi- Ohio Department of Development. (2005). Reports in population and housing e ronment and Urban Systems, 27, 481e509. population estimates. Annual population estimates for Ohio counties: Anderson, J. R., Hardy, E. E., Roach, J. T., & Witmer, R. E. (1976). A land use and land 1960e2009. Census and preliminary estimates of the intercensal population of cover classification system for use with remote sensor data. Geological Survey Ohio and Counties, 2000e2009. http://development.ohio.gov/research/ Professional Paper 964. Washington, D.C: U.S. Government Printing Office. Reports_In_Population_and_Housing-Population_estimates.htm Accessed Bicknell, B. R., Donigian, A. S., Jobes, T. H., & Chinnaswamy, R. (1996). Modeling 18.08.08. nitrogen cycling and export in forested watershed using HSPF. Athens, GA: Office Ohio Department of Development. (2010). Population and housing: Ohio’s pop- of Research and Development, U.S. Environmental Protection Agency. ulation. http://development.ohio.gov/research/files/p0006.pdf Accessed 6 Bicknell, B. R., Imhoff, J., Kittle, J., Jobes, T., & Donigian, A. S. (2000). Hydrological October 2010. simulation program e Fortran user’s manual. Release 12. Washington, D.C.: Office Ohio DNR. (1964). Water inventory of the Little Miami and Mill Creek Basins and of Water, U.S. Environmental Protection Agency. adjacent Ohio River tributaries. Ohio Water Plan Inventory Report 18. Ohio Bronstert, A., Niehoff, D., & Bürger, G. (2002). Effects of climate and land-use change Department of Natural Resources. on storm runoff generation: present knowledge and modeling capabilities. Ohio EPA. (1996). Ohio water resource inventory, executive summary: Summary, Hydrological Processes, 16(2), 509e529. conclusions, and recommendations. Columbus, Ohio: Division of Surface Water Chun, K. C., Chang, R. W., Williams, G. P., Chang, Y. S., Tomasko, D., LaGory, K., et al. and Monitoring Assessment Section. Ohio Environmental Protection Agency. (2001). Water quality issues in the Nakdong River Basin in the Republic of Ohio EPA. (2000). Biological and water quality study of the Little Miami River Basin, Korea. Environmental Engineering and Policy, 2,131e143. 1998. OEPA Technical Report Number MAS/1999-12-3. Columbus, Ohio: Ohio Cryer, S. A., Fouch, M. A., Peacock, A. L., & Havens, P. L. (2001). Characterizing Environmental Protection Agency. agrochemical patterns and effective BMPs for surface waters using mechanistic Pielke, R., Jr., Prins, G., Rayner, S., & Sarewitz, D. (2007). Lifting the taboo on modeling and GIS. Environmental Modeling and Assessment, 6,195e208. adaptation. Nature, 445,597e598. Dagnachew, L., Vallet-Coulomb, C., & Gasse, F. (2003). Hydrological response of Proctor, W. (2001). Multi-criteria analysis and environmental decision-making: a case a catchment to climate and land use changes in tropical Africa: case study of study of Australia’s forests. PhD Thesis, Canberra: Australian National University. south central Ethiopia. Journal of Hydrology, 275(1e2), 67e85. Schneider, W. J. (1957). Relation of geology of streamflow in the Upper Little Miami Debrewer, L. M., Rowe, G. I., Reutter, D. C., Moore, R. C., Hambrook, J. A., & Baker, N. T. Basin. The Ohio Journal of Science, 57(1), 11. (2000). Environmental setting and effects on water quality in the Great and Little Stonefelt, M. D., Fontaine, T. A., & Hotchkiss, R. H. (2000). Impacts of climate change Miami River Basins, Ohio and Indiana. Water Resources Investigations Report 99- on water yield in the Upper Wind River Basin. Journal of American Water 4201. National Water-Quality Assessment Program. Columbus, Ohio: U.S. Resources Association, 36(2), 321e336. Geological Survey. Stoto, M. A. (1983). The accuracy of population projections. Journal of the American Donigian, A. S., & Crawford, N. H. (1976). Modeling nonpoint pollution from land Statistical Association, 78(381), 13e20. surfaces. EPA-600/3-76-083. Athens, GA: U.S. Environmental Protection Agency. Thomson, A. M., Brown, R. A., Rosenberg, N. J., Srinivasan, R., & Izaurralde, C. (2005). Donigian, A. S., & Huber, W. C. (1991). Modeling of nonpoint source water quality in Climate change impacts for the conterminous USA: an integrated assessment, urban and non-urban areas. EPA-600/3e9/039. Athens, GA: U.S. Environmental Part 4: water resources. Climate Change, 69,67e88. Protection Agency. Tobler, W. R. (1970). A computer movie simulating urban growth in the Detroit East Fork LMR Collaborative. (2007). An innovative approach to identifying key region. Economic Geography, 46, 234e240. priorities for improving water quality in the East Fork Little Miami River: A national Tsihrintzis, V. A., Fuentes, H. R., & Gadipudi, R. K. (1997). GIS-aided modeling of demonstration project for watershed management. Final Grant Report. Ohio: nonpoint source pollution impacts on surface and ground water. Water Clermont Office of Environmental Quality. Resources Management, 11, 207e218. Gleick, P. H. (1987). The development and testing of a water balance model for US Census Bureau. (2008). Census 2000 datasets. Population Division. U.S. Bureau of climate impact assessment: modeling the Sacramento Basin. Water Resources the Census. http://www2.census.gov/census_2000/datasets Accessed 08.08.08. Research, 23(6), 1049e1061. US Census Bureau. (2010). United States census 2000. Bureau of the Census. http:// Guo, S., Wang, J., Xiong, L., Ying, A., & Li, D. (2002). A macro-scale and semi- www.census.gov/ Accessed 01.08.09. distributed monthly water balance model to predict climate change impacts USDA. (1999). The wild and scenic river study process. Technical Report Prepared for in China. Journal of Hydrology, 268(1-4), 1e15. the Interagency Wild and Scenic Rivers Coordinating Council. Oregon: U.S. Hotchkiss, R. H., Jorgensen, S. F., Stone, M. C., & Fontaine, T. A. (2000). Regulated Forest Service and National Park Service, U.S. Department of Agriculture. river modeling for climate change impact assessment: the Missouri River. USEPA. (1998). Climate change in Ohio. EPA-236-F-98e007s. Washington, D.C.: Journal of American Water Resources Association, 36(2), 375e386. Office of Policy, U.S. Environmental Protection Agency. IDRISI. (2008). IDRISI. Clark Lab, Clark University. http://www.clarklabs.org USEPA. (2005). National recommended water quality criteria. Washington, D.C.: Accessed 02.11.09. Office of Water. U.S. Environmental Protection Agency.

16 S.T.Y. Tong et al. / Applied Geography 32 (2011) 477e489 489

USGS. (2004). Heavily used pesticides found in more than 90% of stream samples in Yang,Y.J.(2010).Redeﬁne water infrastructure adaptation to a nonstationary southwestern Ohio. News Release, May 2004. U.S. Geological Survey. http://oh. climate. Journal of Water Resources Planning and Management, 136, water.usgs.gov/miam/MIAM_NAWQ_SUM_RPT Accessed 27.12.07. 297e301. White, R., & Engelen, G. (1993). Cellular automata and fractal urban form: a cellular Zandbergen, P. A. (1998). Urban watershed risk assessment using GIS: a case study modeling approach to the evolution of urban land-use patterns. Environment of the Brunette River watershed in British Columbia, Canada. Journal of and Planning A, 25(8), 1175e1199. Hazardous Materials, 61,163e173. Wilby, R. L., Whitehead, P. G., Wade, A. J., Butterﬁeld, D., Davis, R. J., & Watts, G. (2006). Zhu, T., Jenkins, M. W., & Lund, R. (2005). Estimated impacts of climate warming on Integrated modelling of climate change impacts on water resources and quality California water availability under twelve future climate scenarios. Journal of in a lowland catchment: River Kennet, UK. Journal of Hydrology, 330, 204e220. American Water Resources Association, 41(5), 1027e1038.

17 Physical Geography, 2014 Vol. 35, No. 3, 220–244, http://dx.doi.org/10.1080/02723646.2014.908763

A total water management analysis of the Las Vegas Wash watershed, Nevada Thushara Ranatungaa, Susanna T.Y. Tonga*,YuSuna and Y. Jeffrey Yangb

aDepartment of Geography, University of Cincinnati, 401 Braunstein Hall, ML 131, Cincinnati, OH, USA; bNational Risk Management Research Laboratory, U.S. Environmental Protection Agency, Cincinnati, OH, USA (Received 10 October 2013; accepted 20 March 2014)

Climate change, land-use change, and population growth are fundamental factors affecting future hydrologic conditions in streams, especially in arid regions with scarce water resources. Located in the arid southwest within the Las Vegas Wash watershed, Las Vegas is one of the fastest growing metropolitan areas of the coun- try. In the past 30 years, because of climate and land-use changes, it has experienced a decrease in clean water supply but an increase in water demand. To alleviate some of these problems, large amounts of water have been pumped into the city from different sources, such as Lake Mead, and the urban wastewater is treated and returned back to the reservoir for water augmentation. However, in the face of continual global climate change and urbanization in the watershed, long-term planning for sustainable water management is critical. This research was designed to provide a comprehensive analysis incorporating hydrologic modeling, population projection, land-use change modeling, and water management policies to examine the total water balance and management options in this arid and rapidly urbanizing watershed under various scenarios of climate regime, population growth, land-use change, and total water management programs for the year 2050. Keywords: watershed hydrologic modeling; Las Vegas; total water management; climate change; population change; land-use/land-cover change

1. Introduction Watershed hydrologic and water quality models are useful tools in conceptualizing the

Downloaded by [thushara ranatunga] at 11:31 15 October 2014 relationships between climate, land use, and water resources. Further, they can be used to assess the impacts of climate change and other potential anthropogenic changes on watershed hydrology. For more than two decades, these models have been used to assess the impacts of climate change, land-use change, and urbanization on watershed hydrology (e.g., Leavesley, 1994). Moreover, they have been applied to simulate the responses of river basins to climatic variables, ﬂood forecasting, storm water management, environmental impact assessments, and for the design and operation of water- control facilities. Unfortunately, most of these models cannot accurately ascertain the impacts on water quantity and quality due to uncertainty of climate change and population, and land-use projections. There is a paucity of research evaluating the future hydrologic conditions and total water balance, especially for some semi-arid watersheds. There is also a need for additional work to improve the assessment method of

*Corresponding author. Email: [email protected]

18 Physical Geography 221

watershed hydrology. To address these research needs, a comprehensive analytical tool coupling a suite of methodologies, including hydrologic modeling, population projection, and land-use change modeling, was developed in this research to assess and predict the future hydrologic conditions and to analyze the total water management. An arid urbanized watershed, the Las Vegas Wash watershed, was selected as a case study to assess the potential hydrologic effects and the total water balance under various climate, population growth, land-use change, and urban wastewater discharge scenarios for the year 2050. By revealing the plausible hydrologic impacts of the changes in climate, land use, and population on an arid and highly urbanized watershed in the United States, this research can contribute to a more nuanced understanding of watershed hydrology under such an environment. Climate change, land-use change, and population growth are fundamentally important factors affecting future hydrologic changes, especially in semi-arid regions where water is a limiting factor to the growth and survival of ﬂora, fauna, and human settlements (Woodhouse, 2004). Any future changes in stream discharge and watershed hydrology will undoubtedly affect water availability (Somura et al., 2009). To efﬁ- ciently plan for and manage water resources, one has to consider both the supply and demand of water over the next few decades. Climate change can reduce or increase the amount of water supply and induce certain environmental consequences. For example, the decrease in precipitation, coupled with an increase in temperature, may reduce the availability of fresh water supply and the ability of a water body to dilute and transport pollutants. This may further degrade water quality and reduce the amount of fresh water for human consumption. Conversely, with a larger population and further agglomeration in cities, the demand for domestic, industrial, and agricultural water is expected to increase, thereby affecting the ways that water is managed (Arnell, 1999). In some special circumstances, the amount of urban wastewater, which will increase as the urban population grows, if recycled properly, can be used as a major source of water for human uses. In this research, the watershed of the Las Vegas Wash in Clark County of southern Nevada was chosen as the study area because of its rapid population growth. For the past 30 years, the Las Vegas metropolitan area has been one of the fastest growing areas in the United States, and it is the largest population center in Nevada. Its current population is approximately 2 million (Southern Nevada Water Authority [SNWA], 2009). Moreover, the area has been experiencing much drier and hotter weather. Sev-

Downloaded by [thushara ranatunga] at 11:31 15 October 2014 eral studies have estimated the impacts of climate change on water resources of the region. Christensen, Wood, Voisin, Lettenmaier, and Palmer (2004) showed that stream flow in the region is highly sensitive to precipitation and temperature changes. They estimated that a 10–15% precipitation decrease by mid-century could cause a considerable decrease in stream discharge. Furthermore, much of the watershed is undergoing rapid urbanization and land-use change. With more urban lands and impervious surfaces, the hydrology of the watershed will be different. Additionally, the municipalities in this watershed use the treated urban wastewater discharge to augment clean water supply. Based on the amount of treated wastewater that is returned to Lake Mead via Las Vegas Wash, a certain amount of credit is appropriated. These return-flow credits can then be used toward withdrawing clean water from the Colorado River. In the return-flow credit strategy, treated wastewater therefore becomes a resource, as it can substantially increase the total amount of water apportioned to the watershed. This strategy is mainly used to encourage municipalities to release treated wastewater back into streams, which also helps to balance the water levels in the supply source. Since

19 222 T. Ranatunga et al.

water conservation or reuse can lead to a decrease in water returned to the water body (Qaiser, Ahmad, Johnson, & Batista, 2011), this practice will lead to a reduction in return-flow credits. A municipality will have to consider all possible options so as to derive the best solution to increase water supply, while reducing water usage. The Las Vegas Wash watershed therefore provides an ideal study area, not only for examining the hydrologic changes caused by climate change and urban development, but also for analyzing total water management. In this study, the hydrologic modeling results under the future scenarios of climate, population growth, and land-use changes were used to (1) estimate the total discharge of the Las Vegas Wash to Lake Mead, (2) calculate the return-flow credit, (3) estimate the future water demand in mid-twenty-first century, and (4) calculate the total water balance while considering the other sources of water available to Las Vegas.

2. Study area The Las Vegas Valley is located in southern Nevada (Figure 1). It lies within both the Great Basin and Mojave Desert sections of the Basin and Range physiographic province. The arid valley is bounded by the West Spring Mountains to the west and Ground Gunnery Range to the north. The watershed is approximately 4850 km2, and the elevation of its valley floor is approximately 610 m. Most of the soils in the Las Vegas Valley are composed of gravel, windblown sands, and fine-grained silts and clays. The soils in the valley floor typically have a low field capacity and a high permeability. On steep slopes, especially along the Wash, disturbed soils are susceptible to erosion (Bureau of Land Management, 2004). Downloaded by [thushara ranatunga] at 11:31 15 October 2014

Figure 1. The Las Vegas Wash watershed study area (ﬁgures are in color in the online version of this article).

20 Physical Geography 223

The climate of the Las Vegas Valley is hot and arid. The average annual precipitation is about 106 mm, and it occurs mostly as high-intensity, short-duration storms in July and August and low-intensity rainfall in winter months. The average monthly temperatures range from 9 to 34 °C, and the normal average annual temperature is about 21 °C (Table 1). Average daily relative humidity ranges from 32 to 56% in mid-winter and from 11 to 28% in mid-summer. Evapotranspiration is high because of the high summer temperatures, high solar radiation, cloudless skies, low humidity, and the frequent high summer wind (Morris, Devitt, Crites, Borden, & Allen, 1997; Stave, 2001). Because of the arid Mediterranean climate, most surface water in the basin is from summer flash floods, winter rains, and flow from artesian springs (Morris et al., 1997). The urban run-off drains to the Las Vegas Wash, which is a barren, gently sloping, sandy channel from the Las Vegas Valley to the Las Vegas Bay, an arm of Lake Mead. When the city of Las Vegas was first established in 1905 as a way station for the San Pedro, Los Angeles, and Salt Lake Railroads, the Las Vegas Wash was dry with low flows. It contained discharge only during brief periods of major storm run-off. In the last century, a series of social and economic developments, including legalized gaming, the construction of Hoover Dam, industrial production for the Second World War, atomic testing, tourism, and the advent of the modern mega resort, gradually attracted residents and businesses into southern Nevada. As the communities in the Las Vegas Valley began to grow, the demand for water increased (SNWA, 2009). Conversely, the rising population has increased the volume of wastewater discharged into the Las Vegas Wash. Furthermore, the rapid urbanization and the increase in impervious surfaces have allowed more storm water run-off to flow directly into the Wash rather than be absorbed by the soil (Piechota & Bastista, 2003). By the 1950s, the growing urban area discharged enough urban run-off and wastewater into the Las Vegas Wash to create a small but perennial stream flow (Piechota & Bastista, 2003). By the end of the 20th Century, the ephemeral stream had been transformed into an active river channel with a size of about 40 m wide flowing into Lake Mead (Buckingham & Whitney, 2007; Uni- ted States Bureau of Reclamation, 2009). Today, the flow of the Las Vegas Wash is composed of treated domestic and industrial wastewater effluent, wet weather run-off, and groundwater seepage. There are three major municipal wastewater treatment plants located along the Las Vegas Wash, which collect and treat all the municipal wastewater generated in the Las Vegas Valley. In addition, there are nine other permitted discharges along the Wash that contribute to the flow of the Wash (Piechota & Bastista, 2003). fl

Downloaded by [thushara ranatunga] at 11:31 15 October 2014 Currently, the Las Vegas Wash is facing many challenges. The increased water ow in the Wash not only has accelerated soil erosion and destabilized the stream channel, but also has degraded wetland areas and contributed excessive amounts of sediment to the Las Vegas Bay (United States Environmental Protection Agency [USEPA], 2012b).

Table 1. Average monthly precipitation and temperature of the Las Vegas Valley, Nevada.

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Annual Normal average 9.2 11.6 15.5 19.6 25.2 30.4 33.6 32.5 28.1 20.8 13.6 8.7 20.7 temperature (°C) Normal average 13.7 19.3 11.2 3.8 3.0 1.8 10.2 8.4 6.4 6.9 9.1 12.7 106.4 precipitation (mm)

Source: National Oceanic and Atmospheric Administration (2013).

21 224 T. Ranatunga et al.

Other than wastewater discharge, climate change, in the form of overall drying of the region and the increased frequency of extreme high precipitation events, has also created considerable impacts to the region (Christensen et al., 2004; United States Geolog- ical Survey [USGS], 2000). Since Las Vegas is a major ﬁnancial center and one of the fastest growing cities of the region, it is important to have a better picture of the conditions of its future water resources. With a rapid increase in water demand but a decrease in the supply of natural clean water, the need to further understand its watershed hydrology and have the ability to forecast its water balance for the near- and long-term future is becoming ever more essential. Due to these reasons, the watershed of Las Vegas Wash provides a unique setting for examining the hydrologic impacts of urban growth and climate change as well as analyzing the total water management.

3. Research methodology Hydrologic forecasting is necessary for planning, especially for the water-deﬁcient Las Vegas; however, the long-term forecasting accuracy is often limited due to the uncertainties of climate and socioeconomic changes. Thus, there is an urgent need to develop an enhanced method to simulate the future hydrologic conditions and to assess different water management options. Although the state and local agencies, such as the SNWA, Las Vegas Wash Coordination Committee (LVWCC), and the water authorities of Clark County, have used various methods to assess the future conditions of water resources, no one method can be considered as a complete method in estimating the future hydrology of the watershed. By incorporating several methods in predicting the hydrologic conditions under different future climate, population growth, and land-use change scenarios, and assessing the total water management, this research provides a more comprehensive picture of the future hydrologic conditions in the watershed. To develop an analytical tool capable to postulate the future hydrologic conditions, this research entailed seven major steps: (1) Building a basin scale semi-distributed watershed model for the Las Vegas Wash watershed using the Hydrological Simulation Program-FORTRAN (HSPF) model; (2) Development of potential climate change scenarios; (3) Projection of the population in Las Vegas for the year 2050; (4) Projection of wastewater discharge; (5) Incorporation of population growth into land-use change modeling to predict the future land-use pattern in the watershed; (6) Simulation of the

Downloaded by [thushara ranatunga] at 11:31 15 October 2014 total stream discharge for the 2050 horizon year in the Las Vegas Wash under different scenarios of climate change, population growth, land-use change, and wastewater discharge; and (7) Assessing the total water management options in the Las Vegas Valley for 2050. The horizon year of 2050 was selected because most water infrastructure systems have a life-span of about 30–40 years, and planning for water resource management is often for a 30–40 years range.

3.1. Developing the HSPF watershed model for the Las Vegas Wash watershed This study used the HSPF model in the Better Assessment Science Integration Point and Nonpoint Sources (BASINS) to simulate the watershed hydrology of the study area. BASINS is a software package compiled by the USEPA, which includes not only data input and output interfaces and enhancements, but also a few hydrologic models and GIS metadata sets (USEPA, 2012a; Whittemore & Beebe, 2000). HSPF can

22 Physical Geography 225

simulate run-off from non-point sources and pollutant loadings from point sources. Besides, it models the hydrologic behavior and fate of pollutants in a watershed (Bicknell, Imhoff, Kittle, Jobes, & Donigian, 2000). HSPF was chosen in this study because it is a comprehensive watershed hydrologic model. It has been extensively used in various watershed analyses, such as in the assessments of the impacts of climate change (Albek, BakırÖğütveren, & Albek, 2004; Göncü & Albek, 2010), land-use change, and urbanization (Brun & Band, 2000; Chung, Park, & Lee, 2010; Endreny, Somerlot, & Hassett, 2003), as well as in water budget estimation (Said, Stevens, & Sehike, 2005). Besides, it provides a wide range of flexibility for model formulations and calibration processes. In HSPF, there are three modules. PERLND simulates hydrologic run-off and water quality processes on pervious land segments over which an appreciable amount of water infiltrates into the ground. IMPLND simulates impervious land segments, such as paved urban surfaces, over which infiltration is negligible. The processes occurring in the free-flowing reaches and reservoirs are simulated with the RCHRES module (Albek et al., 2004; Endreny et al., 2003; Mishra, Kar, & Singh, 2007). The meteorological and hydrologic data necessary for the simulations are in the form of time series, which are fed into the model by utilizing a standalone data management program, the watershed data management program (WDM). The parameters are supplied to HSPF via the input file, the Use’s Control Input file. The first task of this research was to develop a watershed model for the study area. A base model was constructed for water years (starting from October) 1991–1995. This period was selected because it included both wet and dry conditions. The watershed of Las Vegas Wash was delineated using the eight-digit hydrologic unit code watershed boundaries (HUC8) from the USGS, stream characteristics and Reach File coverage (RF3) from the USEPA, and digital elevation models (DEMS) at a 30 × 30 m resolution from the USGS. The 1991–1995 meteorological time series data from the National Cli- mate Data Center for the climatic station at Las Vegas Airport, Nevada, the digitized soil layer from the national State Soil Geographic database (STATSGO) collected by the US Department of Agriculture, and the 1992 land-use/land-cover layer (LULC) from the USGS National Land Cover Data set (NLCD) were also used to run the model. Point-source data for wastewater discharges were obtained from USEPA Permit Compliance System database. The basin area was partitioned into five different land-cover segments based on the

Downloaded by [thushara ranatunga] at 11:31 15 October 2014 following land uses: urban areas, forests, shrubs/grass, barren, and water. The urban areas were simulated as both pervious and impervious land segments, whereas the other four land-use types were simulated as pervious land segments. The percentage of the urban impervious build-up area was set as 50% as estimated from the land-use map. The rest of the default parameters provided by HSPF were kept unchanged. After the hydrologic model for the Las Vegas Wash was developed, it was calibrated by comparing the simulated flow results with the observed daily discharge records from the USGS gauging station at the Las Vegas Wash Boulevard, Lake Las Vegas, Henderson, NV (Gauge number: 09419790). Model calibration is an iterative process that requires manual adjustments of the model parameters to match the local conditions of the watershed. The adjustments are done by trial and error until the error rate between the simulated and observed stream flow is acceptable. In this study, the major parameters that were adjusted included the lengths and the slopes of the overland flow plains (LSUR and SLSUR), upper and lower zone nominal soil moisture storage (UZSN and LZSN), the index to mean soil infiltration rate (INFILT), the fraction of

23 226 T. Ranatunga et al.

groundwater inflow entering deep groundwater (DEEPFR), the base groundwater recession (AGWRC), Manning’s n for overland flow (NSUR), and the interflow recession parameter (IRC) (Table 2). After several adjustments, the error rate of the model is 3.08%, the correlation coefficient between the simulated and observed flows is 0.97, and the Nash–Sutcliffe model efficiency coefficient (Nash & Sutcliffe, 1970) is 0.94 (Table 3). As shown in the duration curve (Figure 2), the hydrograph of the modeled flow and observed flow follow similar patterns with only a slight exception under the very high flow situations. After calibration, the HSPF model was validated to ensure that the model could accurately simulate actual conditions under different land-use and climate conditions. The validation period was chosen from water years of 1998 to 2002, and the land-use pattern of 2001 and meteorological data from 1998 to 2002 were used. The outputs were compared with the observed data, and model performances were evaluated (Figure 2). The validation results show an error estimate of −5.73%, correlation coefficient of 0.90, and Nash–Sutcliffe efficiency coefficient of 0.89 (Table 3). According to Bicknell et al. (2000), if the model has an error rate below 10%, the model is considered as a very good model, capable of simulating the hydrologic conditions in the watershed. Since both the

Table 2. HSPF model parameters used in calibration.

Possible valuesa Values used Module Parameter Units Function of Min Max in the model PWATER LZSN – Lower zone nominal inches Soils, climate 2 15 10 soil moisture storage INFILT – Index to infiltration in/hr Soils, land use 0.001 0.5 0.5 capacity LSUR – Length of overland feet Topography 100 700 300 flow SLSUR – Slope of overland ft/ft Topography 0.001 0.3 0.0573 flow plane AGWRC – Base groundwater none Baseflow 0.85 0.999 0.92 recession recession DEEPFR – Fraction of ground none Geology, 0 0.5 0.7 water inflow to deep recharge Groundwater recharge UZSN – Upper zone nominal inches Surface soil 0.05 2 0.5

Downloaded by [thushara ranatunga] at 11:31 15 October 2014 soil moisture storage conditions, land use NSUR – Manning’sn none Surface 0.05 0.5 0.45 (roughness) for overland ﬂow conditions, residue, etc. IRC – Interﬂow recession none Soils, 0.3 0.85 0.45 parameter topography, land use

IWATER LSUR – Length of overland feet Topography 50 250 300 flow SLSUR – Slope of overland ft/ft Topography 0.001 0.15 0.0573 flow plane NSUR – Manning’sn none Surface 0.01 0.3 0.1 (roughness) for overland flow conditions, residue, etc.

aUSEPA (2000).

24 Physical Geography 227

Table 3. HSPF model calibration and validation results.

Observed Simulated % Error between Nash– average daily average daily Correlation observed and Sutcliffe flow (m3/s) flow (m3/s) coefficients simulated coefficient Calibration 5.33 5.30 0.973 3.08 0.94 (10/1991– 09/1995) Validation (10/ 6.73 6.53 0.902 −5.73 0.89 1998–09/ 2002)

calibration and validation results have error rates less than 10%, the Las Vegas Wash hydrologic model developed in this study was considered as an appropriate model for simulating the ﬂow conditions in the Las Vegas Wash watershed.

3.2. Formulating future climate change scenarios Numerous studies have suggested that global warming may cause extreme weather events as well as a shift in winter precipitation patterns and, hence, the winter and/or spring discharge (see for example, Barnett & Pierce, 2008; Intergovernmental Panel on Climate Change [IPCC], 2007, 2008; Karl, Melillo, & Peterson, 2009). Alterations in precipitation and evapotranspiration rates can affect the amount of annual run-off (Albek et al., 2004; Lee & Chung, 2007), groundwater, and soil moisture (Somura et al., 2009). Since 1901, precipitation has increased at an average rate of more than 6% per century in the lower 48 states and nearly 2% per century worldwide. However, shifting weather patterns have caused certain areas, such as Las Vegas and other parts in the American southwest, to experience less precipitation than before (Congressional Budget Ofﬁce, 2009; Miller, Bashford, & Strem, 2003; National Science and Technol- ogy Council, 2008; USEPA, 2010). Although there are many future climate projections, most of them are based on different downscaling and inferential methods. The projected results are varied, and the degree of uncertainties is high. Since our aim was not to produce an accurate future climate change prediction, but rather to analyze the plausible hydrologic impacts from a range of possible future climate change conditions, we chose to use a range of possible Downloaded by [thushara ranatunga] at 11:31 15 October 2014 climate change scenarios based on the 2000 and 2009 annual reports published by the United States Global Change Research Program (USGCRP). Results from the state-of- the-science climate models from the Hadley Centre in the United Kingdom and the Canadian Centre for Climate Modeling and Analysis, climate change information from IPCC (2007, 2008), and data from historical observations were used to generate these scenarios (Karl et al., 2009; USGCRP, 2000). These climate models predict that human-induced emissions of heat-trapping gases will cause warming effects in the future. With no explicit climate policies to reduce greenhouse gas emissions, average global air temperature is projected to rise 2.4–6.4 °C by the end of this century, and the temperature of the study area is projected to increase 4 °C by the horizon year of 2050 (IPCC, 2007; Karl et al., 2009). In terms of precipitation in 2050, the projection results differ from the different global climate models. Among the climatic community, the consensus is that the study area will be drier in the future. Most of the published reports (see for example, Barnett & Pierce, 2008; Karl

25 228 T. Ranatunga et al. Downloaded by [thushara ranatunga] at 11:31 15 October 2014

Figure 2. HSPF model outputs: (a) calibration hydrographs depicting the modeled results and observed values, (b) calibration duration curve, (c) validation hydrograph, and (d) validation duration curve. 26 Physical Geography 229

et al., 2009), including those from the IPCC (2007, 2008), predict that there will be a decrease in the availability of fresh water resources in the study area. Nevertheless, some studies suggest that the area may have an increasing trend of precipitation (Miller et al., 2003; USGCRP, 2000; Wolock & McCabe, 1999). For example, the Hadley and Canadian models project that precipitation in the region will increase. Because of the incongruent results, two future climate scenarios (Wet and Dry) were generated for this study. The wet scenario (Wet) considered a 20% increase in precipitation and 4 °C increase in temperature, whereas the dry scenario (Dry) considered a 20% decrease in precipitation and 4 °C increase in temperature. According to these scenarios, historical temperature and precipitation time series data were adjusted and imported to the HSPF hydrologic model through the WDM utility program.

3.3. Projecting population growth in the Las Vegas Valley The close relationship between population growth of an area and its watershed hydrology is well known (Buckingham & Whitney, 2007). In the Las Vegas Wash, the average daily ﬂow in 1970 was 0.736 m3/s. But in 2000, with the rapid increase in population and urban development, and the ensuing increase in wastewater and urban run-off, it has increased to 6.23 m3/s (Piechota & Bastista, 2003). To better postulate the future hydrologic conditions in the Wash, it is essential to estimate the population growth of the study area. This estimate can be used as a ﬁrst approximation to predict the amount of urban wastewater discharge to the Wash. Besides, it can be used to facilitate a more accurate projection of land-use changes. According to Brida and Accinelli (2007), Harris (2005), and Miranda and Lima (2010), the growth of population can often be portrayed by a logistic function. Logistic population growth model implies that when the environment has adequate resources, the population will grow exponentially, but when the population is in proportion to the amount of natural resources and in alignment with the environmental carrying capacity, it will grow at a much slower rate. When the resources become limited and population is larger than the environmental carrying capacity, the growth rate will be reduced and will ultimately become zero (Brida & Accinelli, 2007). Before the mid-20th Century, the population growth in southern Nevada was slow. But, from 1970 through 2000, the average annual population growth was 7% per year. This was the time period with an exponential population growth. By 2000, the southern Nevada’s population had

Downloaded by [thushara ranatunga] at 11:31 15 October 2014 increased to nearly 1.5 million people (SNWA, 2009). After the end of the 20th Cen- tury, the population growth became slower, growing at a rate of 3.4% in 2009. Accord- ing to the model predictions from the Center for Business and Economic Research (CBER) at the University of Nevada, by 2015, the growth rate will be at 2.6%, and by 2030, it will be leveled off at around 1%. In the year 2050, the growth rate will be 0.8% (CBER, 2012). These growth trend and estimates seem to indicate that the population growth of the Las Vegas Valley resembles that of a logistic growth curve (Figure 3). In this study, the logistic function available in the SPSS statistics software package, release version 16.0.2 (SPSS Inc., 2008), was employed. A curve ﬁtting equation was generated, and the statistics were estimated based on the historical population data of the Valley. The SPSS output for the population logistic function shows an R-squared value of 0.995, indicating a high level of signiﬁcance. Using the logistic equation generated, the future population was estimated up to the horizon year of 2050 (Figure 3). From the analysis, it seems that the population of the Valley would be about

27 230 T. Ranatunga et al.

Figure 3. Population growth of the Las Vegas Valley.

3.77 million by 2050. This estimate is similar to the predictions of the CBER, which estimated a population of 3.85 million.

3.4. Estimating future wastewater discharge According to the 2009 SNWA water resources plan, more than 90% of the total water use of the Las Vegas Valley is population dependent (SNWA, 2009). With the increase in urban population, more water is needed. Conversely, as Morris et al. (1997) and Buckingham and Whitney (2007) have suggested, with the increase in population and tourism, more wastewater and treated sewage efﬂuent will be discharged into the Las Vegas Wash. Besides, the increase in the area covered by impervious surfaces would cause a higher volume of urban run-off (Buckingham & Whitney, 2007). Apart from

Downloaded by [thushara ranatunga] at 11:31 15 October 2014 that, the water authorities of the Valley also prefer to have most of the treated wastewater discharged into Lake Mead with minimal reuse as it will allow them to earn return- flow credits and obtain additional apportionment of water from the Colorado River. Since most of the urban wastewater discharge is generated by the population, there may be a strong relationship between these two variables. To ascertain this relationship, a visual assessment of the historical trends of the amount of wastewater and the population was conducted. The visual comparison shows an unequivocal relationship between these two variables (Figure 4). Different curve fitting functions, including polynomial, exponential, and linear regression, were used to describe the relationship between population and wastewater discharge. Among all these models, it was found that the linear regression model gave the best fit between population and wastewater, and the Pearson correlation coefficient between these two variables was highly significant with a value of 0.97 and a standard error of 0.002 (Figure 5). It therefore suggests that the future wastewater discharge could be estimated by the population data. Based on the linear

28 Physical Geography 231

Figure 4. Observed wastewater discharge and population of the Las Vegas Valley.

regression equation derived, the amount of the future wastewater discharge was projected for the year 2050 (Figure 6). The result was then imported into the HSPF model as a point-source input to estimate the future total ﬂow of the Wash.

3.5. Generating the future land-use change scenario For the last 50 years, with the expansion of the city and the increase in population, the study area has undergone remarkable increases in urban area and decreases in agricul- Downloaded by [thushara ranatunga] at 11:31 15 October 2014

Figure 5. Linear regression relationship between population and wastewater discharge.

29 232 T. Ranatunga et al.

Figure 6. Future population and wastewater projection of the Las Vegas Wash.

tural and pasture lands (Adhikari, Acharya, Shanahan, & Zhou, 2011). An increase in impervious surface will cause local decreases in natural interceptions, infiltration, percolation, and soil moisture storage. Consequently, the amount of run-off will increase. Less infiltration in times of storms and the increase in the overflow of effluents will create higher peak flows (SNWA, 2009). Besides, the surface run-off will have larger volumes, greater velocities, and shorter lag times between peak rainfall and the highest flow concentration (Brun & Band, 2000). It is therefore common that watersheds with large amounts of impervious cover have a much reduced groundwater recharge and an increase in storm flow and flood frequency. As the main drainage in a highly urbanized watershed, the Las Vegas Wash shows similar characteristics in its hydrograph, with sudden peak discharges that lasted only for a short time period. To further understand the plausible hydrologic impacts of urbanization, the future land-use pattern has to be predicted. A common approach to derive future land-use scenarios is to use a land-use model to simulate the future land-use conditions. In this study, the Markov cellular automata

Downloaded by [thushara ranatunga] at 11:31 15 October 2014 (CA-Markov) land-use model in IDRISI (2008) developed for the Las Vegas Wash watershed by Sun, Tong, Fang, and Yang (2013) was used to simulate the land-use conditions in the Las Vegas Wash watershed for the year 2050. This is an enhanced model coupling the CA-Markov model with a population variant to depict the effects of population growth in land-use change (Tong, Sun, Ranatunga, He, & Yang, 2012). Two sets of historical land-use records were used to determine the pattern and the trend of land-use change, and an additional map was used for validation. The 1992 and 2001 LULC maps from the USGS NLCD were adopted to develop the model, whereas the 2006 NLCD was used to validate the model. Additionally, the population estimate of 2050 from the logistic equation was incorporated into the CA-Markov land-use model to postulate the land-use pattern in 2050. The projection results suggest that there would be a great increase in urban area and city expansion throughout the Valley (Table 4). Figure 7 depicts the simulated 2050 land-use map of the Las Vegas Wash watershed.

30 Physical Geography 233

Table 4. Land-use projection for 2050.

Land-use/cover type 1992 (km2) 2050 (km2) % change from 1992 Urban 411.47 1875.77 355.87 Forest 287.13 367.31 27.93 Water 2.54 43.41 1610.20 Range/grass land 3730.44 2258.34 −39.46 Barren land 396.35 309.79 −21.84

Downloaded by [thushara ranatunga] at 11:31 15 October 2014 Figure 7. Projected 2050 LULC map of the Las Vegas Wash watershed.

3.6. Simulation of Las Vegas Wash hydrology under the impacts of climate, land-use, and wastewater changes To predict the 2050 hydrologic conditions in the Las Vegas Wash watershed under future climate, wastewater, and land-use change scenarios, a new model was prepared by incorporating the two climate change scenarios (Wet: an increase in 4 °C temperature and 20% precipitation; and Dry: an increase in 4 °C temperature and a decrease in 20% precipitation), the wastewater discharges generated from population growth (WW), and the projected 2050 land-use patterns (LU) into the validated HSPF model. The model was then used to simulate the impacts of wastewater discharge, climate change, and land-use change under the following scenarios:

31 234 T. Ranatunga et al.

(1) Wastewater discharge with no change in climate and land use (Base + WW) (2) Wastewater discharge with a wet climate and no land-use change (Wet + WW) (3) Wastewater discharge with a dry climate and no land-use change (Dry + WW) (4) Wastewater discharge with a wet climate and land-use change (Wet + WW + LU) (5) Wastewater discharge with a dry climate and land-use change (Dry + WW + LU)

3.7. Assessing the future water supply and demand Currently, water in the Colorado River is apportioned among the seven Colorado River Basin states for their total consumption. Nevada receives 3.7 × 108 m3 of water per year from the Colorado River for consumptive use (SNWA, 2009). Since Las Vegas is the main population center of the state, almost all of the apportioned water to Nevada is drawn by the city. If the Las Vegas Valley Water Authorities return some of the treated used water to Lake Mead via the Las Vegas Wash, they can claim a return-flow credit. Each year, through this return-flow credit, southern Nevada is able to divert more water than its consumptive use (Cooley, Hutchins-Cabibi, Cohen, Gleick, & Herger, 2007; SNWA, 2009). The return-flow credit is allowed only for the water returned that was originally from the Colorado River. For most years, about 90% of the total discharge is considered as the water originally diverted from the Colorado River system (Cooley et al., 2007; Peterson, 2007; SNWA, 2009). Using this information and the projected wastewater discharge derived from population growth, the amount of return-flow credit was estimated. Assuming that the same amounts of water would be withdrawn in the future, the total amount of water that would be drawn from the river was estimated as (3.7 × 108 m3 + the “return-flow credit”). Other sources of water supply, such as from the Las Vegas Valley ground water, in-state groundwater resources, and Tributary Conservation and Imported Intentionally Created Surplus (ICS), are considered as the major non-Colorado River supplies. By further assuming that similar per capita demand would continue in the future, the demand of domestic and industrial water uses of the Las Vegas Valley was estimated based on the logistic population estimates, the SNWA conservation goals, and the 2010 per capita demand, which is 945 LPD (liters per day) (SNWA, 2009).

4. Results and discussion Downloaded by [thushara ranatunga] at 11:31 15 October 2014 4.1. Simulated stream discharge As about 90% of the total stream discharge of the Wash is composed of treated wastewater (Cooley et al., 2007; Piechota & Bastista, 2003; SNWA, 2009), the amount of urban wastewater generated from the population in the watershed is one of the most important determinant factors for future flow estimates. Although climate and land-use changes affect only 10% of the total stream discharge, considerable changes in the peak flow can be observed under different climate and land-use change scenarios. Figure 8 depicts the results from the HSPF simulation under the base case (i.e., no climate or land-use changes) and projected future urban wastewater discharge. According to the average daily stream discharge hydrographs generated from the model, with the increase in population and the amount of wastewater discharge, there would be major increases in base flow as well as peak flows. The projected flow of the Wash shows a 2.5-fold increase in the daily discharge by 2050 when compared to the observed flow

32 Physical Geography 235

in 1992. The average annual flow rate would increase from 6.23 m3/s in 1992 to 17.69 m3/s by 2050 (Figure 9(a)). Undoubtedly, this increase in flow would further widen the Wash, making it more vulnerable to head-cutting and bank erosion. It would also reduce the amount of wetlands in the Wash, deteriorate water quality, as well as increase sedimentation into Lake Mead (SNWA, 2009). Buckingham and Whitney (2007) have estimated that for the last century, about 6.6 million m3 of sediments have been deposited to Lake Mead due to the continuous increase in wastewater discharge. When the 2050 Wet and Dry climate change scenarios were added to the model, the discharge output of the watershed shows substantial changes. Under the hypothetical Wet climate scenario, the peak flows would be higher than normal, whereas the base flow would only change slightly (Figure 9(b)). Since it was the wastewater discharge that constituted most of the base flow in the basin, the Wet climate regime would only affect the peak flows during storms. In this simulation exercise, the model projects more frequent large flow events with a magnitude of over 70 m3/s. On an average year, there would be four occurrences of high peak flow events with a magnitude of greater than 83.25 m3/s (Table 5). In addition, the total average flow would slightly increase by about 1.95% relative to the projected average flow rate in base condition (Base + WW) (Table 6). Water authorities in the area may need to implement some flood and erosion controlling measures to minimize the possible flood damage to human settlements as well as wetlands in the Wash. Conversely, in the Dry climate change scenario, the peak flows that could happen in 2050 would not be as high (Figure 9(c)). The highest peak flow according to the model output would be 57.77 m3/s (Table 5). The total average annual flow rate also would decrease by 1.78% (Table 6). Further, under the Dry climate scenario, the decrease in the low flow conditions especially during the summer months would not be as noticeable. This may be attributed to the fact that the urban Downloaded by [thushara ranatunga] at 11:31 15 October 2014

Figure 8. HSPF simulated stream discharge with wastewater projection.

33 236 T. Ranatunga et al. Downloaded by [thushara ranatunga] at 11:31 15 October 2014

Figure 9. Projected stream discharge in 2050 under: (a) the base case (no climate and land-use change scenario) and projected wastewater trends, (b) wet climate change and no land-use change scenario, (c) dry climate change and no land-use change scenario.

34 Physical Geography 237

wastewater discharge is the predominant factor controlling the base flow even under dry climate conditions. In this study, the hydrologic impacts of land-use changes were also examined. The modeling results show that land-use changes could have a substantial impact on stream discharge. Using the CA-Markov land-use model developed for this study area (Sun et al., 2013), it was predicted that from 1992 to 2050, the urban area would increase by 355% (Table 4). Because of the increase in impervious surfaces, there would be major impacts on stream discharge, especially for the peak flow during storm events. The HSPF model predicted that under the future land-use patterns, there would be flow events with a higher magnitude (Figure 10). Downloaded by [thushara ranatunga] at 11:31 15 October 2014

Figure 10. Projected stream discharge in 2050 under: (a) wet climate change and land-use change scenario, (b) dry climate change and land-use change scenario.

35 238 T. Ranatunga et al.

Under the Wet climate change and land-use change scenario, larger flow events would occur with higher frequencies (33 times a year) (Table 5). Most importantly, the peak flow could reach 373.78 m3/s (Figure 10(a)). According to the USGS (2000), there had been several destructive flood events in the past. In 1984, there was a flood with a flow of 219.74 m3/s; in 1998, there was one with 271.83 m3/s; and in 1999, there was another with 302.9 m3/s. The flood event that occurred in 1999 was the most devastating one in history. But in this study, the model results show that the floods that would occur in 2050 could be much larger than those recorded. The projected flow (373.78 m3/s) could be very destructive. This scenario gives the highest percentage increase in total discharge, which is 16.59% with an average flow rate of 20.6 m3/s (Table 6). In the face of future climate and land-use changes, better flood controlling measures and infrastructures are needed. Under the Dry climate change and land-use change scenario, the peak flow would also increase, but with a lesser magnitude in comparison with the Wet climate and land-use change scenario (Figure 10(b)). On the other hand, the frequency of the peak flow events would be higher than under the base scenario (Base + WW) with no climate and land-use changes (Table 5). Among all the scenarios studied, the projections of this scenario are critical. Most of the other recent modeling efforts predict hot and dry trends in the future and also an increase in urbanization in the study region (Barnett & Pierce, 2008; IPCC, 2007, 2008; Karl et al., 2009; USEPA, 2010). Therefore, the projected results from this scenario complement those studies. In our analysis, as the temperature increases and precipitation decreases in the future, there would be a higher number of large flow events. Table 5 shows that there would be 16 flow events with an average flow larger than 70 m3/s. The largest flow event would be 106.19 m3/s, which could cause considerable flood damages. Overall, under the Dry and future land-use scenario, the total average flow would increase by 4.28% compared to the base scenario (Table 6). This could be due to the larger run-off volumes during storm events as a result of an increase in impervious surfaces.

Table 5. Projected frequency of high ﬂow occurrences and the highest ﬂow values under each simulated scenario.

Highest Flow Occurrences Frequency of ﬂoods over 70 m3/s Highest value

Downloaded by [thushara ranatunga] at 11:31 15 October 2014 Base + WW 1 70.51 Wet + WW 4 83.25 Dry + WW 0 57.77 Wet + WW + LU 33 373.78 Dry + WW + LU 16 106.19

Table 6. Average daily ﬂow changes of each scenario from the base model.

Average ﬂow rate (m3/s) % change from Base + WW Base + WW 17.66 Wet + WW 18.01 1.95 Dry + WW 17.35 −1.78 Wet + WW + LU 20.60 16.59 Dry + WW + LU 18.42 4.28

36 Physical Geography 239

Generally, the results show a 2.5-fold increase in average daily discharge in the Wash by the mid-twenty-first century if the projected population growth continues. Apart from the increase in the average flows, there would be more extreme flood events caused by climate and land-use changes. These kinds of destructive events can be con- trolled by applying best management practices, such as by constructing detention basins or dry ponds (Welty, 2009). But in this watershed, excessive use of detention basins is not encouraged as it may reduce the amount of run-off that will channel to Lake Mead, and the Las Vegas Valley may not get as much return-flow credits. However, in order to control floods, it is necessary to have some detention basins. Currently, there are 39 detention basins throughout the watershed and another 30 basins have been planned (LVWCC, 1999). These detention basins are designed to reduce peak storm water flows by detaining water and releasing it over a period of less than seven days.

4.2. Total water management In light of these results, this study evaluated the total water balance between water supply and demand up to the year 2050. The analysis was based on our projected stream discharge. Figure 11 depicts major ﬁxed fresh water supplies and the total water demand estimated from the population projections in the Las Vegas Valley up to 2050. According to the total water supply and demand comparison, the Las Vegas Valley system would have adequate water supply from existing sources until the early 2020s. The plot analysis shows the cut-off year as 2024. After this time, the demand would exceed the total amount of water available, and the Valley would need to ﬁnd other alternative sources of fresh water. The estimated demand of water by 2024 would be 9.55 × 108 m3 per year, and it would increase up to 1.33 × 109 m3 per year by 2050. Based on the 2010 SNWA annual report, the main sources of water supply for the area Downloaded by [thushara ranatunga] at 11:31 15 October 2014

Figure 11. Total water demand and supply with return-ﬂow credits.

37 240 T. Ranatunga et al.

are from the Colorado River allocation, Las Vegas Valley ground water, Virgin/Muddy Rivers Tributary and Coyote Spring Valley Groundwater ICS conservations, and Drop 2 Reservoir System Efficiency ICS. These supplies can provide adequate water until 2024, and after that time, the model predicts that there may not be enough water to meet the demand. According to the projected discharge analysis of Las Vegas Wash, the return-flow credit can be a good source of water. With an increase in population, the total indoor and outdoor water uses will be increased. Studies suggest that a higher outdoor water use may lead to a higher loss of water through evaporation (Qaiser et al., 2011; Stave, 2003). Therefore, it may be advantageous not to reuse outdoor wastewater. This will increase the portion of flows that is discharged back to Lake Mead and increase the return-flow credit. Furthermore, it will help to maintain water level of the lake. However, it does not mean that water conservation indoors would not be beneficial; any amount that is conserved will reduce the amounts required from other means of supply. In this assessment, we assumed that the Colorado River water allocation to Nevada would remain unchanged, but in the future, due to the drying of Lake Mead, the amount of water allocation may change. Such analysis is beyond the scope of this paper as it depends on water rights consideration and will require the assessment of the Colorado River system as a whole. There have been some attempts to assess the future water availability of the Colorado River (see for example the work of Barnett & Pierce, 2008; Christensen & Lettenmaier, 2007; McCabe & Wolock, 2007; Woodhouse, 2004). Most of them reported that the amount of water in the Colorado River will drop drastically by the 2020s. The potential deficit after 2024 will be a challenge to the water authorities in the Las Vegas Valley. The total deficit of water requirement would increase up to nearly 2.46 × 108 m3 per year by 2050 (Figure 11). According to the SNWA (2009), there are potential sources that the authorities have targeted to bring more water to the Valley. There are plans to draw water from other groundwater resources in Clark, Lincoln, and White Pine Counties. SNWA is also planning to build a massive pipe- line system that would take underground water from the Great Basin aquifer system, located about 482.8 km north of Las Vegas, and pump it to Las Vegas. The plan calls for transferring up to 2.22 × 108 m3 of water per year from rural Nevada to the Las Vegas Valley (Progressive Leadership Alliance of Nevada, 2006). But there are

Downloaded by [thushara ranatunga] at 11:31 15 October 2014 major environmental concerns that extensive environmental degradations could occur in the Great Basin area if the ground water levels are lowered. The other potential water supplies are sea water desalinization, brackish water desalinization, and with- drawal from water banking, such as the Arizona, California, and Southern Nevada water banks. The best way to bridge the gap between future water supply and demand is by reducing water consumption. This is particularly the case considering that the amount of water in the Colorado River may be reduced in the future. Through water conservation practices, such as education, water pricing, regulations and incentives, and water smart landscaping, SNWA is planning to reduce the per capita water use by 753 LPD (liters per day) per capita (20%) by 2035. With conservation practices, SNWA has already achieved a reduction of consumptive use by roughly 7.9 × 107 m3 annually between 2002 and 2008 (SNWA, 2009). It is anticipated that the conservation plan would save 3.4 × 108 m3 (20%) of water annually by 2035.

38 Physical Geography 241

5. Conclusion The main goal of this study was to assess the future discharge and total water balance in the Las Vegas Wash watershed. The results of this study show that the total water flow of the stream is highly affected by the wastewater discharges from the urban areas in the Las Vegas Valley. The amount of wastewater discharged to the stream and city population is linearly related with the correlation coefficient of 0.97. Therefore, in this study, the estimation of the urban wastewater discharge was based on the trends of population growth of the area. According to the historical population data, population growth of the area followed the shape of a logistic function. Using the fact that 90% of the flow in the Las Vegas Wash is attributed to urban wastewater, it was estimated that the average flow of the Las Vegas Wash would be 17.67 m3/s by 2050. Many studies suggest that there will be severe hydrologic impacts of climate change in the American southwest. But the results from this study show that in the Las Vegas Wash, the impacts of wastewater discharge are more important than that of climate change, except under the wet weather conditions. Under such conditions, there would be more frequent large peak flow events. Land use is the other factor that has a notable impact on the flow in the Wash. Population increase is the major driving force for the changes in land use over the last several decades. With the increase in population, there would be a 355% increase in urban surfaces by the 2050, which would change the shape of the hydrograph as the peak flows would have higher magnitudes. These higher peak flows would cause sudden floods and damages to the area. Compared with the most devastating historical peak flow of 302.9 m3/s in 1999, the projected flows of 373.78 m3/s under the scenario of combined climate, population, and land-use changes would be even more devastating. Hence, the local water authorities and city planners need to make necessary management adjustments for adaptation. This study further finds that the current existing water supplies and future planned supplies would not meet the increasing demand by 2024. After 2024, the gap between water supply and demand would grow further, and by 2050, an additional 2.46 × 108 m3 of water would be needed every year to meet the additional demand. Therefore, it is essential for the water authorities to find alternatives or other sources of water after 2024. In the future, the return-flow credit may play a crucial role as it will allow the Las Vegas system to have a higher water apportionment. With less outdoor water reuse, the portion of water that would be discharged as wastewater would increase, and conse-

Downloaded by [thushara ranatunga] at 11:31 15 October 2014 quently the amount of water that could be claimed back as return-flow credit would increase. But, there are other environmental considerations with an increased amount of wastewater discharge, such as the water quality in Lake Mead, soil erosion through widening the Wash, loss of wetland habitats, and an increase in the chance of flooding. Besides, this assessment was based on the assumption that the Colorado River water allocation would remain the same in the future. According to other assessments, the Colorado River is predicted to be much drier, and there would be less water available for domestic and industrial use. Hence, further assessment of the water supply from the Colorado River and the amount of wastewater discharge is necessary in order to have a better picture of the future water balance in the Las Vegas Valley. Though Las Vegas is used as the study area in this research, the findings may be useful to other arid and urbanized watersheds. This research introduces an approach to analyzing the combined effects of climate, population, and land-use changes. The total water management analysis also suggests that the use of treated wastewater as a

39 242 T. Ranatunga et al.

resource can be an adaptation method for attaining the amount of required water for Las Vegas and similar regions.

Acknowledgments This work was partially supported by the US Environmental Protection Agency [grant number EP- C-11-006]. The authors are grateful to the agency for the financial support. The manuscript has been subjected to the Agency’s administrative review and has been approved for external publication. Any opinions expressed in this paper are those of the authors and do not necessarily reflect the views of the Agency; therefore, no official endorsement should be inferred. Any mention of trade names or commercial products does not constitute endorsement or recommendation for use.

References Adhikari, A. R., Acharya, K., Shanahan, S. A., & Zhou, X. (2011). Removal of nutrients and metals by constructed and naturally created wetlands in the Las Vegas Valley, Nevada. Envi- ronmental Monitoring and Assessment, 180,97–113. doi:10.1007/s10661-010-1775-y Albek, M., BakırÖğütveren, Ü., & Albek, E. (2004). Hydrological modeling of Seydi Suyu watershed (Turkey) with HSPF. Journal of Hydrology, 285, 260–271. doi:10.1016/ j.jhydrol.2003.09.002 Arnell, N. W. (1999). Climate change and global water resources. Global Environmental Change, 9, S31–S49. doi:10.1016/S0959-3780(99)00017-5 Barnett, T. P., & Pierce, D. W. (2008). When will Lake Mead go dry? Water Resources Research, 44,1–10. doi:10.1029/2007WR006704 Bicknell, B. R., Imhoff, J., Kittle, J., Jobes, T., & Donigian, A. S. (2000). Hydrological simulation program Fortran user’s manual. Release 12. Washington, DC: Office of Water, U.S. Environmental Protection Agency. Brida, J. B., & Accinelli, E. (2007). The Ramsey model with logistic population growth. Eco- nomics Bulletin, 3,1–8. Brun, S. E., & Band, L. E. (2000). Simulating runoff behavior in an urbanizing watershed. Com- puters, Environment and Urban Systems, 24,5–22. doi:10.1016/S0198-9715(99)00040-X Buckingham, S. E., & Whitney, J. W. (2007). GIS methodology for quantifying channel change in Las Vegas, Nevada. Journal of the American Water Resources Association, 43, 888–898. doi:10.1111/j.1752-1688.2007.00073.x Bureau of Land Management. (2004). Las Vegas valley disposal boundary: Environmental impact statement. Retrieved from the US Department of Interior. Web site at http://www.blm.gov/nv/ st/en/fo/lvfo/blm_programs/planning/las_vegas_valley_disposal.html Center for Business and Economic Research. (2012). Population forecasts: Long-term projections for Clark County, Nevada 2012–2050. Las Vegas, NV: University of Nevada-Las Vegas. Christensen, N. S., & Lettenmaier, D. P. (2007). A multimodel ensemble approach to assessment Downloaded by [thushara ranatunga] at 11:31 15 October 2014 of climate change impacts on the hydrology and water resources of the Colorado River basin. Hydrology and Earth System Sciences, 11, 1417–1434. doi:10.5194/hess-11-1417-2007 Christensen, N. S., Wood, A. W., Voisin, N., Lettenmaier, D. P., & Palmer, R. N. (2004). The effects of climate change on the hydrology and water resources of the Colorado River basin. Climatic Change, 62, 337–363. doi:10.1023/B:CLIM.0000013684.13621.1f Chung, E., Park, K., & Lee, K. S. (2010). The relative impacts of climate change and urbanization on the hydrological response of a Korean urban watershed. Hydrological Processes, 25, 544–560. doi:10.1002/hyp.7781 Congressional Budget Office. (2009). Potential impacts of climate change in the United States. Retrieved from the Congress of the United States. Web site at http://www.cbo.gov/publication/ 41180 Cooley, H., Hutchins-Cabibi, T., Cohen, M., Gleick, P. H., & Herger, M. (2007). Hidden oasis: Water conservation and efficiency in Las Vegas. Retrieved from the Pacific Institute and Western Resources Advocates. Web site at http://www.pacinst.org/reports/las_vegas/hidden_oasis.pdf Endreny, T. A., Somerlot, C., & Hassett, J. M. (2003). Hydrograph sensitivity to estimates of map impervious cover: A WinHSPF BASINS case study. Hydrological Processes, 17, 1019– 1034. doi:10.1002/hyp.1178

40 Physical Geography 243

Göncü, S., & Albek, E. (2010). Modeling climate change effects on streams and reservoirs with HSPF. Water Resources Management, 24, 707–726. doi:10.1007/s11269-009-9466-6 Harris, S. (2005). Diffusive logistic population growth with immigration. Applied Mathematics Letters, 18, 261–265. doi:10.1016/j.aml.2003.03.009 IDRISI. (2008). IDRISI. Clark Lab. Retrieved from Clark University. Web site at http:// www.clarklabs.org Intergovernmental Panel on Climate Change. (2007). Contribution of working group II to the fourth assessment report of the Intergovernmental Panel on Climate Change. Cambridge: Cambridge University Press. Intergovernmental Panel on Climate Change. (2008). Climate change and water. Technical paper of the Intergovernmental Panel on Climate Change. Geneva: IPCC Secretariat. Karl, T. R., Melillo, J. M., & Peterson, T. C. (2009). Global climate change impacts in the Uni- ted States. New York, NY: Cambridge University Press. Las Vegas Wash Coordination Committee. (1999). Las Vegas Wash comprehensive adaptive management plan. Retrieved from Las Vegas Wash Project Coordination Team. Web site at http:// www.lvwash.org/html/resources_library_lvwcamp.html Leavesley, G. H. (1994). Modeling the effects of climate change on water resources – A review. Climatic Change, 28, 159–177. doi:10.1007/BF01094105 Lee, K. S., & Chung, E. S. (2007). Hydrological effects of climate change, groundwater with- drawal, and land use in a small Korean watershed. Hydrological Processes, 21, 3046–3056. doi:10.1002/hyp.6513 McCabe, G. J., & Wolock, D. M. (2007). Warming may create substantial water supply shortages in the Colorado River basin. Geophysical Research Letters, 34, L22708. doi:10.1029/2007GL031764 Miller, N. L., Bashford, K. E., & Strem, E. (2003). Potential impacts of climate change on Cali- fornia hydrology. Journal of the American Water Resources Association, 39, 771–784. doi:10.1111/j.1752-1688.2003.tb04404.x Miranda, L. C. M., & Lima, C. A. S. (2010). On the logistic modeling and forecasting of evolu- tionary processes: Application to human population dynamics. Technological Forecasting and Social Change, 77, 699–711. doi:10.1016/j.techfore.2010.01.006 Mishra, A., Kar, S., & Singh, V. P. (2007). Determination of runoff and sediment yield from a small watershed in sub-humid subtropics using the HSPF model. Hydrological Processes, 21, 3035–3045. doi:10.1002/hyp.6514 Morris, R. L., Devitt, D. A., Crites, A. M., Borden, G., & Allen, L. N. (1997). Urbanization and water conservation in Las Vegas valley, Nevada. Journal of Water Resources Planning and Management, 123, 189–195. doi:10.1061/(ASCE)0733-9496(1997)123:3(189) Nash, J. E., & Sutcliffe, J. V. (1970). River flow forecasting through conceptual models. Part 1: A discussion of principles. Journal of Hydrology, 10,282–290. doi:10.1016/0022-1694(70)90255-6 National Oceanic and Atmospheric Administration. (2013). The climate of Las Vegas, Nevada. Retrieved from National Weather Service Office, Las Vegas, Nevada. Web site at http:// www.wrh.noaa.gov/vef/climate/LasVegasClimateBook/ClimateofLasVegas.pdf National Science and Technology Council. (2008). Scientific assessment of the effects of global

Downloaded by [thushara ranatunga] at 11:31 15 October 2014 change on the United States. A report of the Committee on Environment and Natural Resources. Retrieved from National Science and Technology Council. Web site at http:// www.climatescience.gov/Library/scientiﬁc-assessment/Scientiﬁc-AssessmentFINAL.pdf Peterson, M. (2007). Accounting for Nevada’s use of Colorado River water. Retrieved from Colorado River Commission of Nevada. Web site at http://crc.nv.gov/docs/ env-confrence-2007-04_CRCN_Conference_RFC.pdf Piechota, T., & Bastista, J. (2003). Source water assessment for the Las Vegas valley surface waters. Carson City, NV: State of Nevada, Bureau of Health Protection Services. Progressive Leadership Alliance of Nevada. (2006). Las Vegas valley groundwater development project: Where does it start? Where will it end? Retrieved from Progressive Leadership Alli- ance of Nevada. Web site at www.planevada.org Qaiser, K., Ahmad, S., Johnson, W., & Batista, J. (2011). Evaluating the impact of water conservation on fate of outdoor water use: A study in an arid region. Journal of Environmental Management, 92, 2061–2068. doi:10.1016/j.jenvman.2011.03.031 Said, A., Stevens, D. K., & Sehlke, G. (2005). Estimating water budget in a regional aquifer using HSPF-MODFLOW integrated model. Journal of the American Water Resources Associ- ation, 41,55–66. doi:10.1111/j.1752-1688.2005.tb03717.x

41 244 T. Ranatunga et al.

Somura, H., Arnold, J., Hoffman, D., Takeda, I., Mori, Y., & Di Luzio, M. (2009). Impact of climate change on the Hii River basin and salinity in Lake Shinji: A case study using SWAT model and a regression curve. Hydrological Processes, 23, 1887–1900. doi:10.1002/hyp.7321 Southern Nevada Water Authority. (2009) Water resources plan 09. Southern Nevada Water Authority, Las Vegas, Nevada. Retrieved from Southern Nevada Water Authority. Web site at http://www.snwa.com/ws/resource_plan.html SPSS Inc. (2008). SPSS base 16.0.2 for Windows user’s guide. Chicago, IL: SPSS. Stave, K. A. (2001). Dynamics of wetland development and resource management in Las Vegas Wash, Nevada. Journal of the American Water Resources Association, 37, 1369–1379. doi:10.1111/j.1752-1688.2001.tb03645.x Stave, K. A. (2003). A system dynamics model to facilitate public understanding of water management options in Las Vegas, Nevada. Journal of Environmental Management, 67, 303–313. doi:10.1016/S0301-4797(02)00205-0 Sun, Y., Tong, S. T. Y., Fang, M., & Yang, Y. J. (2013). Exploring the effects of population growth on future land use change in the Las Vegas Wash watershed: An integrated approach of geospatial modeling and analytics. Environment, Development and Sustainability, 15, 1495–1515. doi:10.1007/s10668-013-9447-z Tong, S. T. Y., Sun, Y., Ranatunga, T., He, J., & Yang, Y. J. (2012). Predicting plausible impacts of sets of climate and land use change scenarios on water resources. Applied Geography, 32, 477–489. doi:10.1016/j.apgeog.2011.06.014 United States Bureau of Reclamation. (2009). Colorado River basin salinity control project-Las Vegas Wash Unit-Title II. Retrieved from U.S. Department of Interior. Web site at http://www.usbr.gov/projects/Project.jsp?proj_Name=CRBSCP+-+Las+Vegas+Wash+Unit +-+Title+II#Group608790 United States Environmental Protection Agency. (2000). BASINS technical note 6: Estimating hydrology and hydraulic parameters for HSPF.Ofﬁce of Water 4305: EPA-823-R00-012. Washington, DC: U.S. Environmental Protection Agency. United States Environmental Protection Agency. (2010). Climate change indicators in the United States. Washington, DC: National Service Center for Environmental Publications, United States Environmental Protection Agency. United States Environmental Protection Agency. (2012a). Better assessment science integrating point and nonpoint sources. User’s manual.Ofﬁce of Water, EPA-823-C-07-001, Washington, DC: U.S. Environmental Protection Agency. United States Environmental Protection Agency. (2012b). Nonpoint source program success story: Best management practices drastically reduce sediment and restore water quality in Las Vegas Wash. Retrieved from US Environmental Protection Agency. Web site at http://water. epa.gov/polwaste/nps/success319/state_nv.cfm United States Geological Survey. (2000). Flood of July 8, 1999, in Las Vegas Valley, Southern Nevada. USGS Factsheet 80, US Department of the Interior. United States Global Change Research Program. (2000). Climate change impacts on the United States: The potential consequences of climate variability and change. Retrieved from the

Downloaded by [thushara ranatunga] at 11:31 15 October 2014 National Assessment Synthesis Team. Web site at http://downloads.globalchange.gov/nca/ nca-2000-report-overview.pdf Welty, C. (2009). The urban water budget. In L. A. Baker (Ed.), The water environment of cities (pp. 17–28). New York, NY: Springer Science Business Media LLC. Whittemore, R. C., & Beebe, J. (2000). EPA’s BASINS model: Good science or serendipitous modeling? Journal of the American Water Resources Association, 36, 493–499. doi:10.1111/ j.1752-1688.2000.tb04281.x Wolock, D. M., & McCabe, G. J. (1999). Estimates of runoff using water balance and Atmo- spheric General Circulation Models. Journal of the American Water Resources Association, 35, 1341–1350. doi:10.1111/j.1752-1688.1999.tb04219.x Woodhouse, C. A. (2004). A paleo perspective on hydroclimatic variability in the Western United States. Aquatic Sciences, 66, 346–356. doi:10.1007/s00027-004-0723-8

42 CHAPTER FOUR

An Approach to Sensitivity Analysis in Watershed Hydrologic Modeling

ABSTRACT:

The hydrologic responses in a watershed may vary spatially and temporally according to watershed characteristics and inputs. In this study, we developed an index to evaluate the sensitivities of the model parameters to the hydrologic processes in a watershed. The index was then applied to two watersheds with different hydro-climatic settings, the Little Miami River (LMR) watershed and the Las

Vegas Wash (LVW) watershed. To ensure that we can make comparisons between the two watersheds, we used NRMSE as the index to estimate the parameter sensitivity of each of the hydrologic models. We also combined the NRMSE index approach with the flow duration curve analysis technique to measure the parameter sensitivity under different flow regimes. The results show that the parameters related to groundwater are highly sensitive in the LMR watershed, whereas the LVW watershed is primarily sensitive to near surface and impervious parameters. Generally, the high and medium flows are more impacted by most of the parameters. Low flow regime was highly sensitive to groundwater related parameters.

Keywords: Sensitivity analysis, HSPF, NRMSE, Flow Duration Curve

1. Introduction

Watershed hydrologic models are a simplification of reality, and they are mostly based on conceptual representations of the physical processes that govern the flow of water through and over the soil. Parameters for such models are generally related to soil properties, vegetation characteristics, topographic properties, stream characteristics, crop characteristics, and climate and atmospheric

43 conditions (Bergström and Grahm, 1998). These models typically have two types of parameters:

“physical” parameters and “process” parameters. Physical parameters represent physically measurable properties of the watershed, such as, area of the watershed, stream length, surface slope, impervious surface percentage, etc. The watershed properties that are not directly measurable are considered as process parameters. Examples of such properties include near surface soil moisture storage, groundwater recession rate, deep percolation to the groundwater storage, and lateral interflow rate

(Sorooshian and Gupta, 1995).

The hydrologic responses in a watershed may vary spatially and temporally according to watershed characteristics and inputs. Therefore, it is important for modelers to identify the dominant parameters controlling model behavior. One approach to gain this understanding is through sensitivity analysis. Sensitivity analysis evaluates the influences of parameters on the model output (Freer et al.,

1996; Hall et al., 2005; Hornberger and Spear 1981; Liang and Guo, 2003; Pappenberger et al., 2006;

Wagener et al., 2001; Sieber and Uhlenbrook, 2005; Tang, 2007). As such, they are valuable tools for simplifying and improving model structure (Cariboni et al., 2007). Besides, they can facilitate the identification of the most influential model parameters and the least important parameters that can be eliminated from the final model (Hamby, 1994) as well as any unrecognized problems that affect the changes in the hydrologic process of a watershed. They can also be used to test model adequacy, relevance, and conceptualization, and if the model behaves according to underlying assumptions (Ma et al., 2000; Tang, 2007; Wagener et al., 2003). They also provide valuable information on model calibration by allowing the modeler to identify the parameters that are responsible for the hydrologic behavior of the watershed. By helping the users to focus on only the relevant or sensitive parameters for the model and the critical regions in input space (Christiaens and Feyen, 2002), sensitivity analyses can enhance the efficiency in model application and facilitate the determination of priorities for future research and field measurements (Bahremand and De Smedt, 2008; Sieber and Uhlenbrook, 2005). As

Doherty (2009) noted, another possible use of the sensitivity analysis is in evaluating the potential use of a complex, detail model in addressing one or more issues facing a particular study area.

Generally, sensitivity analyses are categorized in two groups: local sensitivity and global sensitivity (Muleta and Nicklow, 2005; Saltelli et al., 2000; van Griensven et al., 2006). Local sensitivity analysis approaches identify the output responses by varying the model parameter sequentially one at a time while keeping all the other parameters fixed at their nominal (base model) values (Holvoet et al.,

2005; Spruill et al., 2000; Turanyi and Rabitz, 2000). On the other hand, global sensitivity approaches explore the entire range of parameters, examining parameter interaction and the impacts of the parameters on model outputs (Muleta and Nicklow, 2005; Saltelli et al., 1999; Tang, et al., 2007).

Typically, global sensitivity methods are more expensive and time consuming to conduct. They require powerful computational techniques, such as supercomputing, in order to get fast and accurate results.

There are two well-known local parameter sensitivity analysis methods; the nominal range and the differential analysis (Frey and Patil, 2002; Helton and Davis, 2003). Nominal range sensitivity analysis estimates the percentage change of outputs with respect to the change of model inputs in relation to their baseline (nominal) values. The percentage change is considered as the sensitivity of the corresponding input. Differential analysis uses the partial derivatives of the model outputs in response to the perturbations of the model input. The derivative values are considered as the measurement of sensitivity (Tang et al., 2007). The main advantage of the nominal range and differential analysis methods is that they are straightforward to implement while they only have modest computational demands. The major disadvantage of these methods is their incapability to estimate the parameter interactions; therefore, it may lead to an underestimation of true model sensitivities.

Evaluation of local sensitivity of model parameters in hydrologic modeling utilizes a number of statistics and techniques. Root Mean Square Error (RMSE) is a commonly used error index (Chu and

Shirmohammadi, 2004; Singh et al., 2005; Vazquez-Amábile and Engel, 2005). It is normally accepted that the lower the RMSE, the better the model performance, but the value of RMSE depends on the magnitudes of each occurrence. In our case, it is the daily flow values of the watershed hydrologic model.

In many studies, sensitivity analyses are conducted in one watershed, and the model parameter sensitivity is analyzed over the entire flow regime for the simulation period or for a specific season. But in this study, our goal was to compare the sensitivities of the parameters in two watersheds and in different flow regimes. Since stream discharges in different watersheds are not the same, the sensitivity of each parameter in the model may vary between different watersheds and under different flow regimes. Besides, there are different forms of time series data. Hence, it is not feasible to compare and capture the level of sensitivity of model parameters between multiple watershed models. The same problem can also arise when comparing sensitivity between different flow regimes in one model. For this limitation, it is necessary to apply a standardization method. Moriasi et al. (2007) has applied baseline standard deviation and used the RMSE-observations standard deviation ratio (RSR) as a standardization method. Another similar approach is Normalized RMSE (NRMSE). It is a standardization method for estimating the ratio between RMSE and a range of baseline data by normalizing all the

RMSEs to a comparable scale. NRMSE has been widely applied in various fields, such as artificial neural network analysis (ANN) (Plutowski, 1994), climatology (Kim and Valdés, 2003), hydrology (Su et al.,

2008; Kourgialas et al., 2008), market predictability analysis (Qian and Rasheed, 2004), and image analysis (Nayak et al., 2004; Guizar-Sicairos et al., 2008). NRMSE is a powerful index, which generates a standard value for time series with different levels of magnitudes. In this study, we used NRMSE as an index to estimate the sensitivity of the parameters in two watershed hydrologic models with different hydro-climatic settings as well as among different flow regimes.

There have been some studies on analyzing parameter sensitivity in the Hydrological Simulation

Program – FORTRAN (HSPF) model (see for example, Donigian and Love, 2007; Fonseca et al., 2014).

There are also studies that estimate parameter sensitivity under different flow regimes using other watershed models, such as the Soil and Water Assessment Tool (SWAT). For example, in their study of

SWAT, Cibin et al. (2010) have shown that the model is more sensitive to soil evaporation coefﬁcient under low stream flow conditions than under high stream flow conditions. However, sensitivities between different flow regimes in the HSPF model are yet to be elucidated. In order to attain a more accurate simulation of watershed hydrology using the HSPF model, it is therefore imperative to understand the influence of parameters under different flow conditions.

In this study, we performed sensitivity analyses on the calibrated and validated HSPF models developed for two different watersheds with different watershed conditions. Moreover, we evaluated the sensitivity of each parameter on the model under different flow regimes. When compared with other sensitivity analysis methods for measuring both local and global sensitivity, the approach used in this study is innovative in the way that sensitivity of different flow regimes are considered in the local sensitivity analysis. The specific objectives of this study were to evaluate the sensitivity of each model parameter for the watershed processes: 1) under two hydro-climatic conditions, and 2) over different flow regimes.

With this analysis, the most appropriate range of parameter in simulating the watershed processes and hydrologic behavior in each watershed would be identified. These results can help in future modeling activities in the study areas or other watersheds with similar climatic and watershed characteristics and settings. The results can also be used to obtain information on the behavior of HSPF parameters as well as on related parameters in other watershed models under different flow regimes.

2. Methods

2.1. Study Area

For our study, we selected two watersheds, the Little Miami River (LMR) watershed and the Las

Vegas Wash (LVW) watershed, which are located under different environmental and climate settings.

Figure 1 shows the spatial locations and land characteristics of these two study areas. In our earlier studies, we had developed HSPF hydrologic models for these two watersheds and used them to study the hydrologic impacts of climate change and land use change (Tong et al., 2012; Ranatunga et al.,

2014). Our results revealed that for different watersheds, the hydrologic effects of different watershed characteristics and processes are different. In this study, in order to improve our original models and to attain better information for water resource management, we investigated the sensitivity of various parameters in our models. Since the distribution of flow regimes in these two watersheds are not the same, we also determined the sensitivities of the model parameters under different flow conditions. To further our understanding of the relationships between watershed characteristics and the hydrology of the two watersheds, we compared the differences in the sensitivities of the model parameters in each watershed by deriving an index, which will not be affected by the magnitudes of stream flows.

2.1.1 The LMR watershed

Originating at the southeast of Springfield in southwestern Ohio, the LMR is a major tributary of the Ohio River. It flows 169.78 km from Clark County through several steep-sloped forested gorges to join the Ohio River at the confluence near the eastern side of Cincinnati in Hamilton County. Its watershed is predominately agricultural with a draining area of about 5840 sq. km. It has a cool temperate climate; summers are warm and humid with a high temperature of 300C and a low

48 temperature of 150C, while winters are moderately cold with a few winter frosts and an average annual snowfall of 50-76 cm. Average annual air temperature ranges from 100C in the north to 130C in the south. The average annual precipitation for the area ranges from 90 to 110 cm; about one-third of the precipitation becomes surface runoff (Debrewer et al., 2000; Tong et al., 2012). Most of the area in the watershed is flat to gently rolling with steep-walled river valleys. The northern portion of the watershed is characterized by gently sloping land, low gradient steams, and areas of fertile soil. In the southernmost areas, the terrain is mostly dissected and hilly with a higher steam density and more drainage problems (Debrewer et al., 2000). The soils in the watershed belong to Genesee-Williamsburg

Association. Formed from silt, alluvial, and residual materials from the glacial deposits, the soils are deep and highly productive but susceptible to erosion (Lerch et al., 1975). Soils in the southeast older till plain are less extensively cultivated than the younger till-derived soils in the northwest.

2.1.2. The LVW watershed

Located in southern Nevada, the LVW is the main channel conveying water from the Las Vegas

Valley to the Las Vegas Bay, an arm of the Lake Mead. The flow of LVW is mostly composed of treated domestic and industrial wastewater effluents. Dry and wet weather runoff and groundwater seepages are the other types of flows contributed to the flow of LVW (Piechota and Bastista, 2003). Rising population in the Las Vegas Valley has increased the volume of wastewater discharge into the LVW.

Furthermore, the rapid urbanization and the increase of impervious surfaces have allowed more storm water runoff to flow directly into the Wash rather than be absorbed by the soil (Piechota and Bastista,

2003).

The LVW watershed is bounded by the West Spring Mountain to the west and Ground Gunnery

Range to the north. The areal extent is approximately 4850 km2, and the elevation of its valley floor is approximately 610 m. It is a predominantly urbanized watershed. The climate of the Valley is hot and arid. The average annual precipitation is about 106 mm, and it occurs mostly as high-intensity, short- duration storms in July and August and low intensity rainfall in winter months. The average monthly temperature ranges from 90C to 340C, and the normal average annual temperature is about 210C. The average daily relative humidity ranges from 32 to 56 percent in mid-winter and from 11 to 28 percent in mid-summer. Evapotranspiration is high because of the high summer temperatures, high solar radiation, cloudless skies, low humidity, and the frequent high summer wind (Morris et al., 1997; Stave, 2001). The soils of the Las Vegas Valley are composed of gravel, windblown sands, and fine grained silts and clays.

The soils on the valley floor typically have a low field capacity and a high permeability. On the steep slopes, especially along the Wash, the disturbed soils are susceptible to erosion (BLM, 2004).

Figure 1: Maps of the study areas: the LMR and LVG watersheds

2.2. Model selection

The most important step in watershed hydrologic simulation is to select an appropriate model based on the scale of application, data demand, and computing requirements. In this study, we used the

HSPF model (Bicknell et al., 2001). The HSPF is a continuous, watershed-scale hydrologic model developed to simulate hydrologic and water quality processes in natural and man-made water systems

(Bicknell et al., 2001). The model combines a set of parameters relating to different factors, such as the time series of rainfall, temperature, and evaporation, and other parameters related to land use patterns, soil characteristics, and agricultural practices, to simulate the processes that occur in a watershed. By representing a watershed as a collection of land segments and channels (reaches) and subdividing the

watershed into sub-watersheds, HSPF uses a mass balance approach, where water and water quality

constituents are routed through appropriate pathways. There are several modules and sub-modules to

simulate various processes in each sub-watershed. The module PERLND is used to simulate the

processes that occur on pervious land areas, IMPLND is used for impervious land areas, and RCHRES for

reaches and reservoirs. The sub-modules PWATER and IWATER simulate runoff from PERLND and

IMPLND, respectively; and the sub-module HYDR routes water through the RCHRES.

2.3. Model parameters selection

In this study, we used the information provided by the HSPF model technical notes (USEPA,

2000) and literature (Abdulla et al., 2009; Donigian and Love, 2007; Fonseca et al., 2014; Mishra et al.,

2007) to select the model parameters for our sensitivity analyses. We selected nine key model

parameters to capture the major processes that occur in each of the watersheds. Six module parameters

were from PERLND and three module parameters were from IMPLND. These are the parameters we

used in calibrating the models when we first developed the LMR and LVW models. Table 1 summarizes

the selected parameters, their definitions, and the range of the values (minimum to maximum) for each

parameter. Additionally, the category of the watershed processes (for example, groundwater, soil, land

use, climate, geology, topography, and surface conditions) is given. The calibrated parameter values

used in the model for each watershed are also listed in the table. These values were decided by an

iterative procedure of parameter evaluation comparing the simulated flow values against observed

values. The models with the values of these calibrated parameters were considered as the base models,

and their simulation results were defined as the ‘base case’. The parameter sensitivities were estimated

in reference to these base cases.

All the model parameters considered in this research are process-based parameters. They are not directly measurable. Instead, they are related to a combination of different watershed characteristics and climate inputs (Sorooshian and Gupta, 1995). In other words, the sensitivity of a specific watershed is a function of the specific combination of the parameter values that reflect the climate and watershed characteristics that control the hydrologic response (Donigian and Love, 2007).

For example, LZSN is related to both precipitation (climate) and soil characteristics of the region. The

INFILT parameter is related with LZSN as well as DEEPFR.

Model Possible Parameter Module Parameter Units Function of Values1 Values Min Max LMR LVG LZSN - Lower Zone Nominal Soil mm Soils, climate 50.8 381 162.56 127 Moisture Storage INFILT - Index to Infiltration Capacity mm/hr Soils, land use 0.028 12.7 12.7 31.75 AGWRC - Base groundwater None Base flow recession 0.85 0.999 0.85 0.92 recession PWATER DEEPFR - Fraction of GW inflow to Geology, Groundwater None 0 0.5 0.45 0.7 deep recharge recharge UZSN - Upper zone nominal soil Surface soil conditions, mm 1.27 50.8 20.32 12.7 moisture storage land use IRC - Interflow recession parameter None Soils, topography, land use 0.3 0.85 0.75 0.45

LSUR - Length of overland flow meters Topography 15.24 76.2 76.2 91.44 NSUR - Manning’s n (roughness) for Surface conditions, None 0.01 0.3 0.05 0.08 IWATER overland flow residue, etc. RETSC - Retention (Interception) Retention potential of mm 2.54 12.7 2.54 2.54 Storage capacity impervious surfaces

1 Range of possible values for each model parameter (USEPA, 2000)

Table 1: HSPF parameter descriptions

Both the LMR and LVW models were run by changing the values of each chosen parameter

separately while holding the base values of the rest of the parameters. The values of each chosen

parameter were changed by 25% increments starting from its minimum and up to its maximum. In

other words, it is the 25% perturbation from the mean values, increases and decreases, for each model

parameter. Thus, each of the model parameter was tested six times, including the base model value.

Since there were nine parameters to be tested, we performed 54 model scenario runs for each

watershed model. The daily flow outputs were recorded for each run for these 54 runs.

2.4. Defining flow regimes using flow duration curve analysis

River flow regimes can be defined using qualitative descriptions or a wide range of quantitative

values. The latter is referred to as flow duration curve analysis. It is a method involving the analysis of

the historical flow data. A flow duration curve generally illustrates the percentage of time, or

probability, that flow in a stream will equal to or exceed a particular value. As such, the duration curve

approach allows for characterizing stream flow into different flow regimes based on their probability of

occurrence. A basic flow duration curve measures high flows to low flows. The X-axis represents the

percentage of time (known as duration or frequency of occurrence) that a particular flow value is

equaled to or exceeded. The Y-axis represents the quantity of flow at a given time step, e.g., cubic

meters per second (m3/s), associated with the duration. Flow duration intervals are expressed as

percentage of exceedance, with zero corresponding to the highest stream discharge in the record (i.e.,

flood conditions) and 100 to the lowest (i.e., drought conditions). For instance, a flow duration interval

of 60% associated with a stream discharge of 10 m3/s implies that 60% of all observed daily average

stream discharge values equal to or exceed 10 m3/s. Typically, low flows (flow during prolonged dry

54 spells) are exceeded a majority of the time, while high flows, such as those resulting in floods, are exceeded infrequently (USEPA, 2007)

In this study, observed flow data during the modeled time period in each watershed were used in generating flow duration curve. For the LMR watershed, the time period for the analysis was from

1990 to 1991, while for the LVW watershed, it was from 1991 to 1995. Using the USEPA guideline

(USEPA 2007), flow duration curves were developed for each watershed. The observed flow data were first arranged in descending order and ranked from 1 to N. Then the frequency of occurrence

(exceedance probability) was estimated using the following formula.

푅 퐹 = 100 ∗ Equation 1 푁+1

where F is the frequency of occurrence (express as % of time a particular flow value is equaled or exceeded), R is the rank, N is the number of observations.

In every model setup, there can be outliers, especially from the extremely high peak flow values, that can have an impact on the general trends of the watershed. In order to reduce bias, we ranked the flow values and removed the highest and the lowest 1% of flow values. After eliminating the outliers, the sorted flow rate was plotted against the exceedance probability in a semi-log curve to generate the flow duration curve. A common way to look at the duration curve is by dividing it into five zones based on percent exceedance: high flows (10 % exceedance), moist condition (10-40 % exceedance), mid-range flows (40-60 % exceedance), dry condition (60-90 % exceedance), and low flows (90-100 percent exceedance) (USEPA, 2007). In this study, three different hydrologic conditions (flow regimes) were identified using intervals of the flow duration curves. They were high flow, medium flow, and low flow, representing percent exceedance 10%, 10-60%, and 60%-100%, respectively. Based on this criterion, the flow outputs for each model sensitivity runs were separated according to these three flow regimes.

2.5. Sensitivity Index

In this study, the development of the sensitivity index was based on the RMSE estimation technique. It is a measure of difference, or residuals, between the values predicted by each sensitivity scenario run and the values actually observed from the base case model. RMSE is one of the widely used methods to estimate parameter sensitivity and model performances among the watershed hydrologic modeling community (Tang et al., 2007, Moriasi et al., 2007). The total RMSE for the entire flow range as well as for each flow regime was calculated by summing the individual RMSEs. By aggregating the individual RMSEs, the total RMSE therefore provides a single measure of predictive power. For our analysis, we calculated the RMSEs for the output daily flow values of all the sensitivity runs with respect to the base model outputs. Equation 2 depicts the formula for RMSE estimation.

∑푛 (푋 −푋 )2 푅푀푆퐸 = √ 푖=1 푏푎푠푒,푖 푠푐푒푛푎푟푖표,푖 Equation 2 푛

where RMSE is Root Mean Square Error, X is the flow value, and n is the number of observations.

In order to standardize the RMSE values into a common and comparable scale between each flow regime, it is essential to normalize the calculated RMSEs. In this study, the flow range for each flow regimes were used as the denominator in normalizing. The total RMSE of each flow regime was divided by the range of the flow values in the base model under each regime. The equation for Normalized

RMSE (NRMSE) is provided below (Equation 3).

푅푀푆퐸 푁푅푀푆퐸 = Equation 3 (푋푏푎푠푒,푚푎푥−푋푏푎푠푒,푚푖푛)

where NRMSE is Normalized Root Mean Square Error and X is the flow value

The developed NRMSE value was used as the sensitivity index (SI) to identify the sensitivity of model parameters to watershed hydrology in each HSPF model. According to the index, higher SI values correspond with higher sensitivity to watershed hydrology, and lower index values are related to lower sensitivity. Using this index, we first identified the most sensitive parameters in each watershed to its hydrology at the entire flow range. In order to determine their responses in the two watersheds, the sensitivity indices were plotted against each parameter perturbation for the entire range of flow values.

To measure the magnitude of the response change, the average slope was estimated for each parameter in the two watersheds. The average slope of a sensitivity curve depicts the degree of variation in sensitivity across the entire parameter range. Furthermore, the trend of hydrologic response to the increase of parameter values from its lowest to maximum was derived by observing the shapes of the sensitivity curves.

Then, the SI of each flow regime, which was obtained based on the above flow duration curve analysis, was calculated for each sensitivity run. Thus, three SI values were obtained for one perturbation in each model parameter. When these SI values were plotted against each perturbation for each parameter separately, the sensitivity of the changing parameter values to watershed hydrology in these three flow regimes could then be identified by observing the shapes of the curves and the magnitudes of the slopes as the parameter values increased from minimum to maximum.

3. Results and discussion

3.1. Flow duration curves and flow regimes of the two watersheds

The flow duration curves developed for the two watersheds are shown in Figure 2. Based on the

USEPA guidelines of flow duration curve development (USEPA, 2007), we defined the ranges of the flow discharge rates for high, medium, and low flow regimes for the two watersheds. For the LMR watershed, high, medium, and low flow regimes correspond to the flow discharge rates of greater than 129 (10% exceedance), 129 to 19 (10-60% exceedance), and less than 19 m3/s (60-100% exceedance), respectively. For the LVW watershed, they correspond to greater than 6 (10% exceedance), 6 to 5 (10-

60% exceedance), and less than 5 m3/s (60-100% exceedance), respectively (Table 2). The flow rates recorded during the study periods indicated that the LMR watershed is characterized by higher flow rates than the LVW watershed (Figure 2). The flow rates of the LMR watershed span a wider range from

3 m3/s to 673 m3/s, while in the LVW watershed, the range of flow rates is relatively narrow, from 3 m3/s to 23 m3/s. Thus, in the LVW watershed, the flow rates do not show much variability across the three flow regimes. But in the LMR watershed, the flow values are highly varied when compared to LVW flow regimes. These variations between each flow regimes also cause different levels of sensitivity to model parameters. For example, the parameter sensitivity between medium and low flow regimes in the LVW watershed may not be as visible as in the LMR watershed.

Figure 2: Flow duration curves for the (a) LMR watershed; (b) LVW watershed

Although the areal extents of the LMR and LVW watersheds are very similar (5840 and 4850 sq. km, respectively), the differences between flow rates in these two watersheds are quite large. These may be attributed to their watershed characteristics. The LMR watershed is predominantly agricultural,

59 and it has a cool temperate climate regime. Since a large portion of the watershed is covered by vegetation land types, the inflows are predominantly from naturally occurring overland flow, subsurface runoff, and ground water seepages (Wang, 2001). The frequent wet weather events in the LMR watershed may lead to a higher discharge rate (Tong et al., 2012). On the other hand, the LVW watershed is under a hot and dry climatic regime, with less wet weather events throughout the year

(NOAA, 2013). As the watershed is mostly urbanized, the discharge in LVW is mainly from treated wastewater from the urban centers in Las Vegas. Only a very small portion of the discharge (about 10%) is from the rainfall events (SNWA, 2009). As discussed from a previous hydrologic modeling study by

Ranatunga et al. (2014), for the LVW, there are only 14 to 20 average annual flash rain events, which may have contributed to the high flow portion of the duration curve.

Flow rates (m3/s)

Low (>60%) Medium (10%-60%) High (<10%)

LMR <19 19-129 >129

LVW <5 5-6 >6

Table 2: Flow rates for each regime in the LMR and LVW watersheds

3.2. Parameter sensitivity

The estimated sensitivity indices described in this study provide qualitative and quantitative assessments of the responses of the HSPF parameters to the model output in different watersheds. They can assist in identifying the most and the least sensitive parameters to the hydrology of the watershed.

The two dimensional curves between sensitivity index and perturbations of each parameter (Figure 3) and average slope of each curve (Figure 4) are used to explain the relationships of each parameter to

60 the model output. In the LVW watershed, eight out of nine parameters are sensitive to its watershed hydrology, whereas in the LMR watershed, only six parameters are sensitive. The results also show that almost all the parameters have a non-linear relationship to the model output in both the watersheds.

According to Muleta and Nicklow (2005), this nonlinear relationship between parameter values and output values is typical in hydrologic models.

Figure 3: Comparison of parameter sensitivity for the entire flow ranges of LMR and LVW watersheds

0.045 0.04 0.035 LMR LVG

0.03 0.025 0.02

0.015 Curve Slope Curve 0.01 0.005 0

Parameter

Figure 4: Average slope of the SI curves for the entire flow range

In the LMR watershed, DEEPFR, INFILT, and AGWRC are the three most sensitive parameters among the nine recommended hydrologic parameters. Among these three parameters, DEEPFR and

INFILT have a decreasing sensitivity and AGWRC has an increasing sensitivity to watershed hydrology as their parameter values increase. According to the average slope of the sensitivity curve, DEEPFR

(average slope 0.034) is the most sensitive parameter, followed by INFILT and AGWRC. DEEPFR is the fraction of infiltrating water that is lost to deep aquifers. It also represents any other losses that may not be measured at the flow gage used for calibration (SJRWMD, 2012). Deep permeable soil layers and limestone and shale bedrock (Wang, 2001) associated with the LMR watershed may influence the relationship of this parameter to watershed hydrology. The shape of the DEEPFR curve (Figure 3) shows a linear decrease of sensitivity as the parameter value is increased. It also indicates that the values towards the higher end of the parameter are more appropriate in simulating hydrology in watersheds similar to the LMR. This high SI value also implies that if the watershed properties associated with the

DEEPFR parameter are disturbed, it can alter the watershed hydrology substantially.

INFILT is the second highest sensitive parameter with a slope value 0.0267 (Figure 4) in the LMR watershed. INFILT is the parameter that effectively controls the overall division of the available moisture from precipitation (after interception) into surface and subsurface flow and storage components

(USEPA, 2000). It has a comparatively higher sensitivity at the lower values of the parameter than at its higher values, as the sensitivity reduces drastically with the increase of parameter value. Hence, careful adjustment of the parameter values between its minimum to first perturbation is important during the calibration process. This result also indicates that high values of the infiltration parameter are more appropriate for this watershed, which is in accordance to the fact that soils of the watershed have a high infiltration capacity (Debrewer et al., 2000). This relationship also implies that decreasing the infiltration in the watershed by adding more impervious surfaces from development activities could lead to considerable alterations in the watershed hydrology of this basin.

The sensitivity of AGWRC is increased with increasing parameter value. AGWRC is the groundwater recession rate, which controls the shape of the hydrograph after a storm event (USEPA,

2000). The overall AGWRC is a complex function of watershed conditions, which include climate, topography, soil condition, and land use/cover. A significant increase can be observed towards the high end of the parameter values. This relationship indicates that changing the parameter values at its high end is not appropriate for this watershed, as the higher values of AGWRC have a higher impact on the simulated flow. The higher sensitivity in AGWRC parameter in the LWR watershed may be attributed to the fact that the majority of the area in the LMR watershed is either under forested or agricultural land use. Besides, it has a moist climate. Hence, a considerable fraction of the water from precipitation is percolated, some become ground water, and others become lateral flows. This fraction of water will be slowly released to the streams. In this watershed, low values of AGWRC are more suitable for accurate hydrologic simulation.

In other HSPF modeling studies, LZSN is often found to be one of the most sensitive parameters to hydrology (Donigian and Love, 2007; Mishra et al., 2007). But, surprisingly in our analysis, LZSN shows a relatively small impact to the hydrology of the watershed. That may be because LZSN is mainly a climate (precipitation) dependent parameter (Marsalek and Ng, 1989, USEPA, 2000), and previous studies have suggested that the hydrology in the LMR watershed is affected more by land use variations rather than climate changes (see for example, Tong et al., 2012). Furthermore, the presence of thick layer of fertile soil in the watershed (Lerch et al., 1975) may have a higher level of moisture storage capacity. Therefore, the parameter value may vary substantially without showing a significat impact to the model.

UZSN and IRC are the two other parameters recommended by the HSPF modeling community to be considered in the development of the model. Our analysis for LMR watershed indicates that these two parameters show a small impact to the model output as the parameter values vary from minimum to maximum values (Figure 3 and 4). For UZSN, the convention is to use 10% of the value as the value for

LZSN (See Table 1) (SJRWMD, 2012). Since the sensitivity of LZSN is low in the LMR model, UZSN also shows a low sensitivity. Depending on the agricultural conditions, tillage, and other practices, UZSN may change over the course of the growing season (SJRWMD, 2012). The mixed land use nature of the watershed (although we considered LMR as an agricultural watershed) and higher moisture holding capacity of the soil layers could be the reasons for the lower sensitivity for this parameter.

From our results, it is obvious that all impervious parameters, including LSUR, NSUR, and RETSC, are insensitive in the LMR watershed. This may be related to the fact that the LMR watershed does not have a large urban area. Hence, the impacts from changing the impervious parameter values are negligible.

Regarding the LVW watershed, INFILT, DEEPFR, and LZSN are the three most sensitive parameters. AGWRC and RETSC parameters also show considerable degrees of sensitivity to the hydrology of the watershed, but in comparison with INFILT, DEEPER, and LZSN, their sensitivities are lower. According to the average slope of the sensitivity curve (Figure 4), INFILT is the most sensitive parameter to watershed hydrology with a value of 0.0396 within its parameter range, whereas the average slopes of the curves are 0.0361 and 0.0177 for DEEPFR and LZSN, respectively. All these three parameters show a decreasing sensitivity as their parameter values increase. This indicates that the higher values of these parameters are more appropriate for modeling the hydrology in this dry and urbanized watershed. Out of the nine parameters, DEEPFR is the only parameter with a linear relationship between model output and parameter values.

The sensitivity of INFILT drastically changes within the first two perturbations and stays more or less constant with further increase of parameter value, implying that lower levels of infiltration have a higher impact on watershed hydrology than high infiltration values. But, when compared to the LMR watershed, the magnitude of the impact of INFILT is higher in the LVW watershed. This is true in most of the watersheds with more impervious surfaces, as under this situation, the infiltration will be reduced, and the amount of runoff will be increased. An increase of INFILT reduces the immediate surface runoff; therefore, lower INFILT values are more sensitive in this watershed. Overall, infiltration related watershed processes play an important role in hydrologic simulation in this watershed.

According to the HSPF technical notes (USEPA, 2000), LZSN is one of the main factors that highly contributes to evapotranspiration. Several studies also suggest that INFILT, LZSN and DEEPFR are highly sensitive parameters for annual stream flow (Singh, et al. 2005; Mishra et al., 2007). In the LVW watershed, evapotranspiration is usually high, and it is an important factor affecting watershed hydrology. Furthermore, because LVW watershed has sandy soils with low water storage capacity (MLB,

2004), it is more sensitive to the LZSN parameter. A small change to the factors related to LZSN would lead to a considerable impact to the hydrologic behavior of this watershed.

Different to the LMR watershed, all three impervious parameters (I-LSUR, I-NSUR, and RETSC) in the LVW watershed show some degree of sensitivity to the hydrologic outcome of the watershed.

Moreover, in the LVW watershed, the overall sensitivities of almost all the parameters are comparatively higher than in the LMR watershed. This indicates that when the parameter values are altered, the impacts to the model output will be higher in the LVW watershed than LMR watershed.

Since the average annual stream discharge in LVW is smaller than LMR (5.3 and 52.2 m3/s, respectively), the results may imply that the sensitivity in a watershed with a smaller stream may be higher than the watershed with a larger river. It also shows that the sensitivity index developed in this research can be used to compare the sensitivity of parameters between different watersheds. The parameters with the largest difference in sensitivity between the two watersheds are LZSN and the three impervious parameters mentioned above.

Generally, this analysis indicates that the LMR watershed is sensitive mostly to the ground water related parameters (DEEPFR and AGWRC) than near surface parameters, whereas LVW is primarily sensitive to near surface parameters (INFILT and LZSN) and to the impervious parameters. Thus, in the beginning of the modeling development process, modelers should start calibrating these highly sensitive parameters, since a little adjustment of these parameters has a significant impact on the results in watershed hydrology. Figure 3 is a good guide to identify the most appropriate parameter ranges for the two watersheds.

3.3 Parameter sensitivity under different flow regimes

As described in section 3.2, changing the values of certain parameters can result in different levels of response in the overall watershed hydrology. In this research, in addition to the entire flow regime, we further explored the sensitivity of the parameters under different flow regimes (high, medium, and low). This is because we wanted to understand whether the parameter sensitivity in the entire flow range is similar to individual segments of the flow regimes. Figures 5 to 8 illustrate the sensitivity index of each parameter for the three flow regimes in the two watersheds.

Figure 5: LMR model parameter sensitivity in three flow regimes

0.16 0.14

0.12 High

0.1 Medium 0.08 Low

0.06 Curve Slope Curve 0.04 0.02 0

Parameter

Figure 6: Average slope of the SI curves of three flow regimes in LMR watershed

In the LMR watershed, according to the slopes of the each curve (Figure 6), high flow is the regime most affected by all parameters. Similar to the behavior for the entire flow regime explained earlier, DEEPFR shows the highest sensitivity than other parameters to the high flow condition, followed by INFILT, AGWRC, UZSN, IRC, and LZSN. Based on the shape of the SI curve (Figure 5), altering DEEPFR near its low values will largely change the runoff of the watershed for all three flow regimes. It also shows that high values of DEEPFR are appropriate in simulating all three flow regimes. According to the

HSPF technical notes (USEPA, 2000), decreasing DEEPFR value will increase the overland flow, especially during the wet weather events, which usually create high flows. Figure 6 further shows that the impact of this parameter to the medium and low flows conditions is moderate when compared to the sensitivity of other parameters under the same flow conditions. However, it is likely that the overall high sensitivity for the entire flow range described above is due to the impacts from the high flow sensitivity.

As the second most sensitive parameter, INFILT shows similar sensitivity at both high and medium flow regimes and a low sensitivity to the low flow conditions. Both high and medium flow conditions indicate the same pattern (the graph in Figure 5 shows a sudden drop at the beginning and then the curve gradually decreases) while the low flow condition indicates a gradual decrease throughout. Similar to the sensitivity for the entire flow regime, the hydrologic response in all three flow regimes is low when the parameter value decreases up to the second perturbation. Further decrease in the parameter values shows a substantial impact on the high and medium flows. This behavior further indicates that higher values of INFILT parameter are more suitable for the LMR watershed in simulating the hydrology of all flow conditions. Based on the average sensitivity to the INFILT parameter, the model provides a better simulation for high and medium flows rather than low flows.

For the AGWRC in LMR, the sensitivity increases gradually in all three flow regimes (Figure 5).

The changing of AGWRC values has more impacts to the high flow conditions than the other two flow conditions. Medium flow has the lowest impact from the change of parameter value, whereas low flow shows comparatively a high response. However, lower values of the parameter are appropriate for all three flow conditions. In general, if the purpose of modeling is to simulate high flow conditions in a watershed, DEEPFR, INFILT, and AGWRC parameters will be the appropriate parameters for investigation.

Although LZSN, UZSN and IRC did not show a huge impact to model output, their behavior is important in simulating the detailed patterns of watershed hydrology. According to Figure 5, LZSN shows a similar pattern of sensitivity to all three flow regimes. But interestingly, it shows a higher sensitivity to the low flow condition compared to the other two flow regimes. Therefore, LZSN can be used as a parameter to simulate low flow conditions in watersheds similar to that of LMR. In a similar manner, IRC is another parameter suitable in modeling medium flows. However, as shown in Figure 5,

69 moderate values of LZSN, UZSN, and IRC are appropriate for accurate simulation of watershed hydrology in all three flow regimes. Similar to sensitivity to the entire flow range, all three impervious parameters show almost no sensitivity to all three flow regimes.

Figure 7: LVW model parameter sensitivity in three flow regimes

0.16 0.14

0.12 High

0.1 Medium 0.08 Low

0.06 Curve Slope Curve 0.04 0.02 0

Parameter

Figure 8: Average slope of the SI curves of three flow regimes in LVW watershed

As illustrated from Figures 7 and 8, in most cases, the sensitivities of each parameter in the three flow regimes of LVW have different behaviors than the LMR watershed. There is a distinct variability of response between each flow regime. This infers that using a common value for a parameter is not appropriate in simulating all three flow regimes in the LVW watershed for most of the parameters.

Therefore, due to the variations between flow regimes, simulation of all hydrologic regimes in the LVW watershed (and in other watersheds with similar hydro-climatic conditions) is a complex process. It is noticeable that the impacts from most of the watershed parameters, such as INFILT, DEEPFR, and LZSN, to the model output in low flow regime are less apparent compared to the other two flow conditions.

This could be related to the fact that a large portion of LVW base flow (low flow) is contributed from the point source discharges (from wastewater treatment plants) which are directly connected to the main stream (SNWA, 2009). Therefore, changing the parameter values shows a low variation in model output in low flow conditions. For this reason, use of these parameters to simulate low flows is not recommended. On the other hand, since AGWRC shows a very high sensitivity to low flows, it can

71 certainly be used to simulate low flows. According to the SI curve for AGWRC (Figure 7), high values of the parameter show high sensitivity, whereas low to medium range shows more or less no sensitivity.

This relationship also indicates that an increase in the ground water recession rate in this watershed can cause a significant impact in base flows of the stream. However, it is recommended to use a value from a higher range for INFILT and DEEPFR parameters and a lower range value for AGWRC parameter in simulating all three flow regimes. Out of the three impervious parameters, RETSC shows a comparatively higher sensitivity to all three flow regimes. Although the LVW watershed is considered as an urban watershed, the relative impacts from these impervious module parameters are less than pervious parameters.

4. Conclusion

The goal of this study was to use a simple and effective analytical approach to develop an index that can be used to evaluate and compare the parameter sensitivity to different watershed models under different hydro-climatic settings and to different flow regimes in one watershed. For this purpose, we used NRMSE as the index to estimate the parameter sensitivity of the HSPF hydrologic model. The index provides a standardized value that can be used to compare the parameter sensitivity in model outputs with different magnitudes. We combined the NRMSE index approach with flow duration curve analysis technique to measure the parameter sensitivity under different flow regimes. This approach may contribute to the literature in HSPF modeling as well as to other hydrologic modeling using different programs due to its ability to generate a sensitivity index that can be applied to compare the impacts of the watershed processes to the hydrology of watersheds under different hydro-climatic conditions and flow regimes. This tool is an improvement to other local sensitivity analysis methodologies, which usually measure the parameter sensitivity of the entire flow regime.

Using this tool, we derived the following conclusions that may contribute to the literature of hydrologic modeling of Little Miami River (LMR) watershed and Las Vegas Wash (LVW) watershed (also to other watersheds that are under similar hydro-climatic conditions). The outcomes of this study are useful in understanding the most and least important watershed characteristics and processes in the

LMR and LVW watersheds that affect the hydrologic behavior of the watersheds.

The flow duration curves indicate that the river flow patterns and magnitudes in the two watersheds are different, although they have similar drainage area. In the LVW watershed, the flow rates do not show much variability across the high, medium, and low flow regimes, whereas in the LMR watershed, the flow values are highly varied across the three flow regimes.

Based on the sensitivity index analysis for the entire flow range, eight out of the nine parameters used in the model development in the LVW watershed are sensitive to watershed hydrology, whereas only six parameters are sensitive in the LMR watershed. The sensitivity index curves developed in this study is a useful indicator to identify the pattern of changes in sensitivity across its possible parameter range. This study shows that almost all of the parameters show a non-linear relationship to the model output in both watersheds. Furthermore, this information can be used to decide the most appropriate value for each parameter in the two watershed models.

Using the average slope of each sensitivity curve, we identified the most to least sensitive parameters for each watershed. The highest sensitive parameters have the greatest impacts to hydrologic simulation in each watershed. For the LMR watershed, the order of parameter sensitivity from the highest to lowest are DEEPFR, INFILT, AGWRC, UZSN, IRC, and LZSN, respectively. For the LVW watershed, INFILT, DEEPFR, LZSN, AGWRC, RETSC, I-NSUR, I-LSUR, and UZSN are the parameters that have the highest to lowest sensitivity, respectively. Overall, this study shows that the groundwater related parameter (DEEPFR and AGWRC) are the most sensitive parameters to the LMR watershed than

73 near surface parameters, whereas LVW is primarily sensitive to near surface parameters (INFILT and

LZSN) and the impervious parameters. This information will be useful in selecting the most appropriate watershed processes not only for HSPF modeling but also for other models, such as SWAT and SWMM, in simulating hydrology. For example, parameters related to groundwater processes should be used in modeling the LMR watershed or other similar watersheds.

In the LVW watershed, the overall sensitivity of almost all the parameters is comparatively higher than in the LMR watershed. This shows that if the parameter values are changed, the impacts to the model output are higher in the LVW watershed than the LMR watershed. This study further reveals that the parameters for the impervious module (LSUR, NSUR and RETSC) are not sensitive in watersheds that consist of less urban surfaces, as these impervious parameters are insensitive at the LMR watershed, but they are sensitive at the LVW watershed.

Using the sensitivity index, we were able to identify the model parameters that have the highest to the lowest impact to each of the three flow regimes in the LMR and LVW watersheds. Regardless of the differences between land-use classes, soil, and hydro-climatic conditions, high and medium flows are the flow regimes that usually have a greater impact from all the parameters, except AGWRC, in both the watersheds. AGWRC usually shows a large impact to the low flow conditions than other two flow regimes in both the watersheds, which indicates that AGWRC is an important parameter to use when simulating low flow conditions in these two watersheds. The impact to the low flow conditions from all the parameters in the LVW watershed is lower when compared to the LMR watershed. This is attributable to the fact that the base flow from the LVW watershed is mostly from the discharges from wastewater treatment facilities.

To improve the parameter sensitivity information obtained from this study, one can increase the frequency of the parameter value perturbation. In this study, we used 25% increment from its minimum

74 possible value. But, according to Kourgialas et al. (2008), 10% perturbation can be used in local parameter sensitivity analysis.

5. Acknowledgement

This research was partially funded by the U.S. Environmental Protection Agency. The authors are grateful to the agency for the financial support. The primary author also acknowledges the supports and guides provided by Dr. Susanna Tong and Dr. Jeffery Yang.

6. Disclaimer

The U.S. Environmental Protection Agency, through its Office of Research and Development, funded and managed, or partially funded and collaborated in, the research described herein. It has been subjected to the Agency’s administrative review and has been approved for external publication. Any opinions expressed in this paper are those of the author (s) and do not necessarily reflect the views of the Agency, therefore, no official endorsement should be inferred. Any mention of trade names or commercial products does not constitute endorsement or recommendation for use.

References

Abdulla, F., Eshtawi, T., Assaf, H. 2009. Assessment of the impact of potential climate change on the water balance of a semi-arid watershed. Water Resources Mange. 23: 2051-2068.

Bahremand, A. De Smedt, F. 2008. Distributed Hydrological Modeling and Sensitivity Analysis in Torysa

Watershed, Slovakia. Water resources management. 22 (3), 393-408.

Bergström, S., Graham, L. 1998. On the scale problem in hydrological modelling. Journal of

Hydrology. 211 (1), 253-265.

Bicknell, B. R., Imhoff, J., Kittle, J., Jobes, T., Donigian, A. S. 2001. Hydrological Simulation Program -

Fortran (HSPF). User's Manual for Release 12. U.S. EPA National Exposure Research Laboratory, Athens,

GA.

BLM (Bureau of Land Management). 2004. Las Vegas valley disposal boundary: Environmental impact statement. U.S. Department of Interior, Las Vegas Field Office, Las Vegas, Nevada.

Cariboni, J., Gatelli, D., Liska, R. and Saltelli, A. 2007. The role of sensitivity analysis in ecological modelling. Ecological Modelling. 203 (1–2), 167-182.

Christiaens, K., Feyen, J. 2002. Use of sensitivity and uncertainty measures in distributed hydrological modeling with an application to the MIKE-SHE model. Water Resources Research. 38 (9), 8-1.

Chu, T. W., Shirmohammadi, A. 2004. Evaluation of the SWAT model's hydrology component in the

Piedmont physiographic region of Maryland. Transactions of the ASAE. 47 (4), 1057-1073.

Cibin, R., Sudheer, K. P., Chaubey, I. 2010. Sensitivity and identifiability of stream flow generation parameters of the SWAT model. Hydrological Processes. 24 (9), 1133-1148.

Debrewer, L.M., Rowe, G.I., Reutter, D.C., Moore, R.C., Hambrook, J.A., Baker, N.T. 2000. Environmental setting and effects on water quality in the Great and Little Miami River Basins, Ohio and Indiana.

National Water-Quality Assessment Program, Water Resources Investigations Report, 99-4201, U.S.

Geological Survey, Columbus, Ohio.

Doherty, J., Hunt, R. J. 2009. Two statistics for evaluating parameter identifiability and error reduction. Journal of Hydrology. 366 (1), 119-127.

Donigian Jr, A. S., Love, J. T. 2007. The Housatonic River watershed model: model application and sensitivity/uncertainty analysis. In: 7th International IWA Symposium on System Analysis and Integrated

Assessment in Water Management. Website at: www.aquaterra.com/resources/pubs/pdf/donigian-

2007.pdf

Fonseca, A., Ames, D. P., Yang, P., Botelho, C., Boaventura, R. 2014. Watershed Model Parameter

Estimation and Uncertainity in Data-limited Enviromnets. Environmental Modeling and Software. 51: 84-

93.

Freer, J., Beven, K., Ambroise, B. 1996. Bayesian estimation of uncertainty in runoff prediction and the value of data: An application of the GLUE approach. Water Resources Research. 32 (7), 2161-2173.

Frey, C.H., Patil, S. R. 2002. Identification and review of sensitivity analysis methods. Risk analysis, 22 (3),

553-578.

Guizar-Sicairos, M., Thurman, S.T. Fienup, J.R. 2008. Efficient subpixel image registration algorithms.

Optics letters. 33 (2), 156-158.

Hall, J., Tarantola, S., Bates, P., Horritt, M. 2005. Distributed sensitivity analysis of flood inundation model calibration. Journal of Hydrological Engineering. 131 (2), 117-126.

Hamby, D.M. 1994. A Review of Techniques for Parameter Sensitivity Analysis of Environmental Models.

Environmental Monitoring and Assessment. 32, 135-154.

Helton, J.C. Davis, F.J. 2003. Latin hypercube sampling and the propagation of uncertainty in analyses of complex systems. Reliability Engineering & System Safety, 81 (1), 23-69.

Holvoet, K., van Griensven, A., Seuntjens, P., Vanrolleghem, P. A. 2005. Sensitivity analysis for hydrology and pesticide supply towards the river in SWAT. Physics and Chemistry of the Earth. Parts A/B/C, 30(8),

518-526.

Hornberger, G. and Spear, R. 1981. An approach to the preliminary analysis of environmental systems. J.

Environ. Manage. 12, 7-18.

Kim, T. W., Valdés, J. B. 2003. Nonlinear model for drought forecasting based on a conjunction of wavelet transforms and neural networks. Journal of Hydrologic Engineering. 8 (6), 319-328.

Kourgialas, N., Karatzas, G. and Nikolaidis, P. 2008. Simulation of the flow in the Koi-liaris River basin

(Greece) using a combination of GIS, the HSPF model and a karstic-snow melt model. In: iEMS 2008:

International Congress on Environmental Modeling and Software, pp. 512-520.

Lerch, N.K., Hale, W.F., Milliron, E.L. 1975. Soil survey of Clermont County. Soil Conservation Services,

U.S. Department of Agriculture, Ohio.

Liang, X., Guo, J. 2003. Intercomparison of land-surface parameterization schemes: sensitivity of surface energy and water fluxes to model parameters. Journal of Hydrology. 279 (1), 182-209.

Ma, L., Ascough, J. C., Ahuja, L. R., Shaffer, M. J., Hanson, J. D., Rojas, K. W. 2000. Root zone water quality model sensitivity analysis using Monte Carlo simulation. Transactions of the ASAE. 43 (4), 883-

895.

Marsalek, J., Ng, H. Y. F. 1989. Evaluation of pollution loadings from urban nonpoint sources: methodology and applications. Journal of Great Lakes Research. 15 (3), 444-451.

Mishra, A., Kar, S., Singh, V. P. 2007. Determination of runoff and sediment yield from a small watershed in sub-humid subtopics using the HSPF model. Hydrological Processes. 21, 3035 – 3045.

Moriasi, D., Arnold, J., Van Liew, M., Bingner, R., Harmel, R. Veith, T. 2007. Model evaluation guidelines for systematic quantification of accuracy in watershed simulations. Trans. ASABE. 50 (3), 885-900.

Morris, R. L., Devitt, D. A., Crites, A. M., Borden, G., Allen, L. N. 1997. Urbanization and water conservation in Las Vegas valley, Nevada. Journal of Water Resources Planning and Management. 123

(3), 189-195.

Muleta, M. K., Nicklow, J. W. 2005. Sensitivity and uncertainty analysis coupled with automatic calibration for a distributed watershed model. Journal of Hydrology. 306 (1), 127-145.

Nayak, J., Bhat, P. S., Acharya, R., Niranjan, U. C. 2004 Simultaneous storage of medical images in the spatial and frequency domain: A comprehensive study. BioMedical Engineering OnLine. 3:17.

NOAA (National Oceanic and Atmospheric Administration). 2013. The climate of Las Vegas, Nevada.

Retrieved from National Weather Service Office, Las Vegas, Nevada. Website at: www.wrh.noaa.gov/vef/climate/LasVegasClimateBook/ClimateofLasVegas.pdf. Last Accessed

September, 2014

Pappenberger, F., Iorgulescu, I., Beven, K. J. 2006. Sensitivity analysis based on regional splits and regression trees (SARS-RT). Environmental Modelling & Software. 21 (7), 976-990.

Piechota, T., Bastista, J. 2003. Source water assessment for the Las Vegas valley surface waters. Carson

City, NV: State of Nevada, Bureau of Health Protection Services.

Plutowski, M. 1994. Selecting training exemplars for neural network learning. Doctoral dissertation,

University of California, San Diego.

Qian, B., Rasheed, K. 2004. Hyrts Exponent and financial marker predictability. Website at: http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.137.207. Last Acessed: August, 2014.

Ranatunga, T., Tong, S. T., Sun, Y., Yang, Y. J. 2014. A total water management analysis of the Las Vegas

Wash watershed, Nevada. Physical Geography. 35 (3), 1-25.

Saltelli, A., Tarantola, S., Campolongo, F. 2000. Sensitivity analysis as an ingredient of modeling. Statistical Science. 15 (4), 377-395.

Saltelli, A., Tarantola, S., Chan, K. P. S. 1999. A quantitative model independent method for global sensitivity analysis of model output, Technometrics. 41, 39–56.

Sieber, A., Uhlenbrook, S. 2005. Sensitivity analyses of a distributed catchment model to verify the model structure. Journal of Hydrology. 310 (1), 216-235.

Singh, J., Knapp, H. V., Arnold, J. G., Demissie, M. 2005. Hydrological modeling of the Iroquois river watershed using HSPF and SWAT. Journal of American Water Resources Association. 41 (2): 343 – 360.

SJRWMD (St. Jhons River Water Management District). 2012. St. Jhons River Water Supply Impact

Study. Scientific Referance Center, Palatka, Florida.

SNWA (Southern Nevada Water Authority). 2009. Water Resources Plan 09. Southern Nevada Water

Authority, Las Vegas, Nevada. Retrieved from Southern Nevada Water Authority. Web site at http://www.snwa.com/ws/resource_plan.html. Last Accessed: September, 2014

Sorooshian, S., Gupta, V. K. 1995. Model Calibration, in: V. P. Singh (Eds.), Computer Models Of

Watershed Hydrology. Water Resources Publications, Colorado, pp. 23-68.

Spruill, C. A., Workman, S. R., Taraba, J. L. 2000. Simulation of daily and monthly stream discharge from small watersheds using the SWAT model. Transactions of the ASAE. 43 (6), 1431-1439.

Stave, K. A. 2001. Dynamics of wetland development and resource management in Las Vegas Wash,

Nevada. Journal of the American Water Resources Association. 37, 1369–1379.

Su, F., Hong, Y., Lettenmaier, D.P. 2008. Evaluation of TRMM multisatellite precipitation analysis (TMPA) and its utility in hydrologic prediction in the La Plata Basin. Journal of Hydrometeorology. 9(4), 622 –

640.

Tang, Y., Reed, P., Wagener, T. and Van Werkhoven, K. (2007.) Comparing sensitivity analysis methods to advance lumped watershed model identification and evaluation. Hydrology & Earth System Sciences. 11

(2), 797-817.

Tong, S. T., Sun, Y., Ranatunga, T., He, J., Yang, Y. J. 2012. Predicting plausible impacts of sets of climate and land use change scenarios on water resources. Applied Geography. 32 (2), 477-489.

Turanyi, T., Rabitz, H. 2000. Local methods, in: Saltelli, A., Chen, K., Scott, E. M. (Eds.), Sensitivity

Analysis. John Wiley & Sons Ltd, Chichester, pp. 81-99.

USEPA. 2000. Estimating hydrology and hydraulic parameters for HSPF. BASINS Technical Note – 6.

Office of Water, United States Environmental Protection Agency, Washington DC.

USEPA. 2007. An approach for using Load Duration Curves in the development of TMDLs: EPA 841-B-07-

006. Office of Wetlands, Oceans and Watersheds, United States Environmental Protection Agency,

Washington DC. van Griensven, A., Meixner, T., Grunwald, S., Bishop, T., Diluzio, M., Srinivasan, R. 2006. A global sensitivity analysis tool for the parameters of multi-variable catchment models. Journal of hydrology. 324 (1), 10-23.

Vazquez-Amabile, G. G., Engel, B. A. 2005. Use of SWAT to compute groundwater table depth and stream flow in the Muscatatuck River watershed. Transactions of the ASAE. 48 (3), 991-1003.

Wagener, T., Boyle, D. P., Lees, M. J., Wheater, H. S., Gupta, H. V., Sorooshian, S. 2001. A framework for development and application of hydrological models. Hydrology and Earth System Sciences. 5 (1), 13-26.

Wagener, T., McIntyre, N., Lees, M. J., Wheater, H. S., Gupta, H. V. 2003. Towards reduced uncertainty in conceptual rainfall‐runoff modelling: Dynamic identifiability analysis. Hydrological Processes. 17 (2), 455-

476.

Wang, X. 2001. Intigrated water-quality management nad land-use planning in a watershed context.

Journal of Environmental Management. 61, 25-36.

CHAPTER FIVE

CONCLUSION

This dissertation research provides various approaches/tools that can be used to investigate the impacts of changes in climate, population, and land use on watershed hydrology and water quality constituents through basin scale watershed modeling. It compares the hydrologic impacts of these variables in two different environmental and hydro-climatic settings, the cool-temperate LMR watershed in southern Ohio and the hot and arid LVW watershed in southern Nevada. A total water management approach is also introduced to help in managing the future potential changes in demand and supply of water in the Las Vegas urban center. Additionally, it investigates the watershed characteristics and processes that control the hydrologic behaviors of the two watersheds.

The findings of this dissertation research can add to the existing literature of the fields of watershed hydrology and hydrologic modeling. In contrast to previous other studies, this dissertation evaluates the combined impacts of a variety of potential changes that can affect the hydrology and water quality conditions of a watershed system. One of the outcomes from these studies indicates that the combined scenarios of climate, land use, and population projections provide more detail information on potential hydrologic impacts than the individual scenarios. When climate change was considered together with land use change, the impact on stream discharge and water quality was more apparent. Furthermore, combined scenarios can minimize the degrees of uncertainty of climatic and land use projections (Hejazi and Moglen, 2008). On the question of whether climate or land use would play a greater role in influencing the hydrologic behaviors, this dissertation suggests that climate is more important. Climate changes show a greater impact on the high flow conditions, which generate more frequent large peak flow events during the wet weather conditions compared to the other flow situations. Moreover, the watersheds that have a high population growth rate, such as the Las Vegas

Wash watershed, population driven wastewater discharge plays a greater role compared to climate and land use changes.

Although high amounts of wastewater discharges is not recommended in most of the watershed systems, the results from this study indicate that the LVW watershed is different because of its special conditions. In this watershed, the return flow credit concept plays a crucial role as it allows the Las

Vegas system to have a higher water apportionment from its major water supply source, the Lake Mead.

With less indoor-outdoor water reuse, the portion of water that would be discharged as wastewater would increase, and consequently, the amount of water that could be claimed back as return flow credit would increase.

This dissertation also reveals that the basin scale HSPF watershed models can be successfully used in assessing the short- and long-term impacts of future climate and land use changes on water resources. It is a useful tool, which can be used to further our understanding of the dynamics of the physical system in a watershed.

Furthermore, NRMSE based analytical approach together with flow duration curve analysis is an effective method that can be used to compare the sensitivities of the model parameters between different watersheds as well as between different flow regimes in a watershed. The NRMSE index provides a scale to compare the sensitivity of the hydrology of a watershed to the hydrologic parameters. The approach requires modest computational needs; it is cost effective and can be applied with less special skills.

The other two findings from this dissertation are that the cool temperate, agricultural watersheds with mixed land use, such as the LMR watershed, are sensitive to the ground water related characteristics and processes, whereas watersheds that are located in hot and dry climatic settings with

84 predominantly urban landscapes, such as the LVW watersheds, are sensitive to the near surface soil parameters. Generally, high and medium flow regimes are highly affected by the changes in most of ground water related watershed processes in the LMR watershed. Whereas in the LVW watershed, low flows show relatively higher sensitivity to some of the ground water related parameters (for example, groundwater recession rate), while high and medium flows have high sensitivity to the near surface watershed parameters.

Many potential applications can be obtained from this research. The results can be used to improve planning for future water resources in these two watersheds. The modeling approach can provide an alternative method in analyzing, evaluating, and predicting the possible impacts of other management scenarios on water resources. Furthermore, the results acquired from this dissertation can be used to facilitate the resource managers, local government agencies, or planners in their decision making processes, including planning for water conservation practices, adjusting land use and farming practices, constructing flood control facilities, and installations of BMPs. It can assist in the development of realistic mitigation strategies for better use of water resources to meet future water demands. An example of a potential application of the results from the LMR modeling exercise is the

Water Resources Adaptation Program (WRAP) at the USEPA.

In summary, findings from this dissertation study provide mainly three aspects of new and improved methodologies and information on hydrologic and watershed modeling that can contribute to the existing literature of the subject. They help to answer the following questions: (1). How to incorporate combined scenarios of climate, land use, and population change into hydrologic and watershed models, and what influences they will have on future hydrological behaviors of a stream? (2).

How to simulate the behavior of the wastewater discharges in a watershed hydrologic system and what potential influences would occur from such inputs on total water management in a urbanized

85 watershed? and (3). How to estimate the sensitivity of a watershed processes/parameter on each separate flow regimes of a stream.

Future Research:

When simulating the future stream hydrology of a watershed with a rapid population growth, it is important to include the socio-economic changes in the modeling exercise so as to achieve a more accurate assessment of the system. Socio-economic changes have a significant impact on hydrologic behaviors of the watershed than other environmental changes (Alcamo, et al., 2007). For example, large scale industries and businesses demand comparatively higher amounts of water than domestic uses.

Therefore, it is important to consider the socio-economic developments in the area along with climate, land use, and population projections. Results from this study can be further improved if one can incorporate such variables into the models.

According to Mailhot (2007), the climate projections are often considered as the most uncertain parameter in simulating the features at a finer spatial and temporal resolution. In this study, we used the hypothetical climate change scenarios based on several global and regional climate modeling results.

This is because of the uncertainty in deducing local climate patterns from a global/regional trend. If we can employ a local climate change scenario using the local boundary conditions, temperature, precipitation, and other local climate variables, the hydrologic projections of the study area can be further improved.

In estimating the sensitivity of model parameters, this dissertation employed 25% increments from a parameter’s minimum possible value. To improve the parameter sensitivity, one can increase the frequency of the parameter value perturbation. For example, one can use 10% perturbation as

86 suggested by Kourgialas et al (2008). Moreover, information obtained on the sensitivity of watershed processes to the hydrologic regimes can be further improved by increasing the number of classes in the flow duration curve.

BIBLIOGRAPHY

Alcamo, J., Florke, M., Marker, M. 2007. Future long-term changes in global water resources driven by socio-economic cha climatic changes. Hydrological Science, 22 (2): 247 – 275

Arnell, N. W. (1999). Climate change and global water resources. Global Environmental Change, 9: 31-49

Bekele, E. G., and Knapp, H. V. 2010. Watershed Modeling to Assessing Impacts of Potential Climate Change on Water Supply Availability. Water Resources Management, 24 (13): 3299-3320

Brun, S.E., Band, L.E. 2000. Simulating runoff behavior in an urbanizing watershed. Computers, Environment and Urban Systems, 24, 5-22

Hejazi, M. I., Moglen, G. E. 2008. The effect of climate and land use change on flow duration in the Maryland Piedmont region. Hydrological Processes, 22: 4710n – 4722

IPCC. (2008). Contribution of working group I to the fourth assessment report of the intergovernmental panel on climate change, 2007. Cambridge, United Kingdom and New York, NY, USA: Cambridge University Press

Kourgialas, N., Karatzas, G. and Nikolaidis, P. 2008. Simulation of the flow in the Koi-liaris River basin (Greece) using a combination of GIS, the HSPF model and a karstic-snow melt model. In: iEMS 2008: International Congress on Environmental Modeling and Software, pp. 512-520.

Leavesley, G. H. 1994. Modeling the effects of climate change on water resources – a review. Climatic Change, 28, 159-177

Mailhot, A., Duchesne, S., Caya, D., Talbot, G. (2007). Assessment of future change in intensity-duration- frequency (IDF) curves for Southern Quebec using the Canadian Regional Climate Model (CRCM). Journal of Hydrology 347 (1-2): 197-210

Middelkoop, H., Daamen, K., Gellens, D., Grabs, W., Kwadijk, J. C. J., Lang, H., Parmet, B.W.A.H., Schädler, B., Schulla J., and Wilke, K. 2001. Impact of Climate Change on Hydrological Regimes and Water Resources Management in the Rhine Basin. Climate Change, 49 (2): 105-128

Mohamoud, Y. M., Sigleo, A. C., AndParmar, R. S. 2009. Modeling the Impacts of Hydromodification on Water Quantity and Quality. U.S. Environmental Protection Agency, Washington, DC, EPA/600/R-09/116 (NTIS PB2010-104718)

Sharma, K. D., Gosain, A. K. 2010. Application of climate information and predictions in water sector: Capabilities. Procedia Environmental Sciences, 1: 120 -129

Somura, H., Arnold, J., Hoffman, D., Takeda, I., Mori, Y., & Di Luzio, M. (2009). Impact of climate change on the Hii River basin and salinity in Lake Shinji: A case study using SWAT model and a regression curve. Hydrological Processes, 23, 1887-1900

Wu, W., Hall, C. A. S., Scatena, F. N. 2007. Modelling the impact of recent land-cover changes on the stream flows in northeastern Puerto Rico. Hydrological Processes, 21 (21) : 2944 – 2596