Hydrologic impacts of climate change and urbanization in the Las Vegas Wash Watershed, Nevada

AE

actual evaporation

AR4

Assessment Reports

BASINS

Better Assessment Science Integrating Point and Nonpoint Sources

BMPs

best management practices

CA

cellular automata

CA-Markov

cellular automata-Markov chain model

CCRFCD

Clark County Regional Flood Control District

CN

Curve Number

DEM

Digital Elevation Model

GCMs

Global Climate Models

GHCND

Global Historic Climatology Network Daily

GIS

geographic information systems

HUC

hydrologic unit code

Ia

initial abstraction

IPCC

Intergovernmental Panel on Climate Change

LVW

Las Vegas Wash

MCE

Multi-Criteria Evaluation

MRLC

Multi-Resolution Land Characteristics Consortium

NLCD

National Land Cover Database

NOAA

National Oceanic and Atmospheric Administration

P

precipitation

PCMDI

Climate Model Diagnosis and Inter-comparison

PE

adjusted potential evapotranspiration

Q

runoff

S

retention

ST

soil moisture storage

SWB

Modified Thornthwaite-Mather Soil-Water-Balance Code for Estimating Groundwater Recharge

SWMM

Storm Water Management Model

USDA

US Department of Agriculture

USGS

US Geological Survey

WCRP

Working Group on Coupled Modelling Program

ΔST

change in soil moisture

In the United States, more than 75% of the population resides in urban areas. By 2030, more than 60% of the world population is expected to live in cities (Paul & Meyer 2001). Some researchers, for example, Paul & Meyer (2001) and Walsh et al. (2005), find that urbanization will affect not only the watershed ecosystem but also watershed hydrology. As urbanization increases the amount of area under impervious surfaces, a larger percentage of precipitation will contribute to surface runoff. The catchment will have a faster response to precipitation, and the time required to convert rainfall to runoff will be decreased. The magnitude of the peak flow and the frequency of small urban floods will also be increased (Shuster et al. 2005). Moreover, due to contaminated non-point source pollution from paved surfaces and industrial effluent, water quality will be degraded (Dunne & Leopold 1978; Klein 1979; LeBlanc et al. 1997).

Additionally, changes in climate will have significant impacts on watershed hydrology. Some studies have indicated that even modest variations in the amount of precipitation can have considerable effects on mean annual discharge (Whitfield & Cannon 2000; Muzik 2001). With global warming and climate change, an urban watershed may experience more extreme weather events, such as floods and droughts (Zhang & Chang 2013).

However, the hydrologic impacts of the changes in climate and land use, particularly urbanization, may work in tandem and it is often difficult to discern which factor will have a more dominant effect (Tomer & Schilling 2009). In an earlier study, Legesse et al. (2003) find that watershed hydrology is more sensitive to climatic variables (precipitation and air temperature), though land cover/land use also have considerable impacts. Bronstert et al. (2002) draw a similar conclusion that climate change has a significant relationship with peak discharge. Some other studies, such as those by Changnon & Demissie (1996) and Cognard-Plancq et al. (2001), however, conclude differently. They find that the changes in land use are responsible for the majority of the fluctuations in runoff. It seems, therefore, that the hydrologic effects of climate or land-use changes vary from place to place and from time to time. Such changes in the hydrologic cycle will undoubtedly affect water management (Xu & Singh 2004), and there will be challenges to sustainable urban development. To understand further the potential impacts of climate change and land-use change in watershed hydrology, it is important to be able to comprehensively assess the separate and combined hydrologic impacts of urbanization and climate change, and determine the predictor in watershed hydrology.

The objective of this study was to develop a cell-based hydrologic model to simulate the relationship between climate change and urbanization with surface flow. The model was developed using historical climate and land-use data. It was used to characterize the rainfall-runoff process and to predict the hydrologic changes in an urban area in the arid American Southwest, where future climate changes and urbanization can cause drastic consequences for water supply and demand. The watershed runoff was predicted under future changes in climate and urban development for the years 2030 and 2050, a time period generally sufficient for studying climate change effects (USGCRP 2009; NRC 2010). The results from the cell-based hydrologic model developed in this study may highlight watershed hydrologic response to the complex interplay of climate change and land-use change.

HUC is a watershed classification system developed by the US Geological Survey (USGS). Based on the characteristics of surface hydrology, the Agency delineated the watersheds of the United States and the Caribbean into hydrologic units (Seaber et al. 1987). These units are nested within each other, from the largest geographical regions to the smallest units. Each unit has a unique code, and each accepts surface water from a common upstream drainage area. The source area is either directly from a major river/stream or a series of streams with a combined drainage area. Moreover, all water from the unit drain to a single defined outlet point. In the USGS HUC system, there are six levels of classification, and the 8-digit HUC is the fourth-level cataloguing unit (Seaber et al. 1987).

The LVW Watershed is located in Clark County, Nevada. The watershed encompasses an area of approximately 4,854.7 km², extending about 65 km from the Spring Mountains in the west to Lake Mead in the southeast. The valley floor of this basin is broad and flat, sloping gently to the southeast.

The LVW is a primary drainage for the Las Vegas Valley. It has a perennial reach of about 19 km long, ending at the Las Vegas Bay in Lake Mead (Stave 2001). The tributaries to the LVW were historically ephemeral, but most of them have become perennial due to recent changes in landscape and local water management policies. As the tributaries pass through the built up areas in the Las Vegas Valley, they pick up treated wastewater, shallow subsurface ground water, overland flow from impervious surfaces, and storm water from the metropolitan area.

Compared to a vegetated watershed, an urbanized catchment has a different hydrologic condition and rainfall-runoff relationship because of its artificial impervious surface layer. This is especially the case under an arid environment. While there are many studies on the hydrologic conditions in urban watersheds, see for example the work of Burian & Shepherd (2005), Cheng & Wang (2002), and Rose & Peters (2001), not much work has been performed in a semi-arid watershed, such as the LVW Watershed. With a hot and dry environment, impending climate change due to global warming, expeditious population growth, and rapid urbanization, the LVW Watershed may face increasing problems and management challenges from increasing water demand and declining water availability in the future. Meanwhile, Lake Mead, the major water resource of Las Vegas, was found to have had a significant decrease in the amount of water during the same period of time (Barnett & Pierce 2008). Developing and maintaining a city in a desert area is challenging, especially in meeting its growing water demand. With the threats of water shortage, it is critical to have the ability to predict future hydrologic conditions in the LVW Watershed. Undoubtedly, this ability will facilitate a better understanding of the hydrologic impacts of urbanization and climate change in the area.

Many researchers have used various methods to simulate watershed hydrology. Among these methods, the Thornthwaite water balance model (Thornthwaite & Mather 1955) is widely used to estimate surface runoff and evapotranspiration (see, for example, Kolka & Wolf 1998; Keim 2010; Fish 2011). The model uses the monthly climate, land use/land cover, and soil type to estimate hydrologic inflows, storages, and outflows. As such, it offers a succinct report of the balance of rainfall and runoff and its seasonal variation (Ferguson 1996). Since the Thornthwaite model is based on monthly climate data, it is more flexible in terms of data requirements. Monthly climate data are generally available for many geographical locations. The monthly record is also in an appropriate temporal resolution for analyzing seasonal, yearly, and decadal trends. Moreover, the model is simple, efficient, and highly reliable. One drawback in the Thornthwaite model is the assumption that the direct runoff factor has a fixed linear relationship between precipitation and infiltration, an assumption that may not be true for all land-use and land-cover types and soil conditions (Ferguson et al. 1991).

In order to address this problem in this research, we employed the Curve Number (CN) method (US Soil Conservation Service 1986) in the calculation of direct runoff. The CN method takes land surface materials and hydrologic conditions into consideration. Schneiderman et al. (2007) have applied the CN method to analyze the hydrologic response to storm events in both an arid watershed and a humid watershed, and found that the CN method can help to overcome the drawbacks of the Thornthwaite model in calculating surface runoff. Ferguson et al. (1991) further combined these two methods into one model to calculate the urban water balance. The results seem to be much improved with a higher accuracy.

Although the Thornthwaite model and the CN method are proved to be reliable, they can only calculate the water balance at certain points in a watershed. Since both the input data and the output results are in a point format, it will be difficult to capture and portray the spatial heterogeneity of hydrologic conditions within a watershed. As the analyses of watershed hydrology are becoming more complex with more spatial requirements (Beven & Feyen 2002), many researchers turn to the use of geographic information systems (GIS), which can provide the computational capabilities and the abilities to manage and process spatial hydrologic and physiographic information (Singh & Woolhiser 2002; Olivera et al. 2006). Besides, GIS allows a comprehensive consideration of environmental factors, such as land use, soil, and elevation, in a flexible spatial resolution setting (Cuo et al. 2008).

One GIS based hydrologic model is the Storm Water Management Model (SWMM) (USEPA 2009), which has been used in different parts of the world under various climatic conditions to analyze urban runoff and to design urban drainage systems, such as urban stormwater systems and the combined and sanitary sewers. For example, Tsihrintzis & Hamid (1998) have applied SWMM to model storm events in several small (about 0.04–0.2 km²) urban catchments. Other researchers, such as Wang et al. (2012) and Krebs et al. (2013), have also used SWMM, and they find that the model is reliable. But the results of their study cannot be easily generalized to other areas due to the different environmental conditions, such as land-use patterns. Moreover, SWMM uses the average precipitation to depict the amount of rainfall in each subcatchment. The model assumes not only the amount of rainfall received but also that the land-use pattern and the type of surface materials and soils in each subcatchment are homogeneous. Since SWMM is mostly applied in studies of small urban areas in subcatchments, this assumption can be valid. Nevertheless, when the study area is as large as thousands of square kilometers, such as the LVW Watershed, SWMM may not work as well. To consider the spatial variation within a watershed, a large number of subcatchments along with hydrologic parameters and water transportation networks will need to be set up. As such, it will require substantial data for input and computing time for analyses. Despite the fact that SWMM can perform well in small urban catchments under various climatic conditions (USEPA 2015), it can be challenging to use it to simulate the hydrologic conditions in watersheds encompassing a large area with a mix of land-use types and different hydro-climatic conditions.

The GIS-based SWB model (Modified Thornthwaite-Mather Soil-Water-Balance Code for Estimating Groundwater Recharge) overcomes the drawback of SWMM by using raster layers as input data (Westenbroek et al. 2010). This method requires climate, land use/land cover, and soil data to perform the Thornthwaite water balance calculation. The model has been successfully applied to water balance studies (see, for example, Dripps 2003; Dripps & Bradbury 2007; Hart et al. 2012). Nevertheless, SWB was originally designed for estimating groundwater recharge, and it has not been widely used in surface water research.

As an alternative method, the cell-based hydrologic model not only combines hydrologic modeling with GIS but also uses cells (or pixels) as the basic unit. Since each cell contains its environmental information, for example its hydrologic characteristics, this method can simulate the physical processes within each cell and the interactions between the neighboring cells. Hence, the model has the potential to predict accurately the temporal and spatial rainfall-runoff responses of the watershed. Moreover, the use of a cell-based model may help to improve flexibility in hydrologic modeling, as it can be used with different spatial resolutions. According to Krysanova et al. (1998), the cell-based model that they have developed for the Elbe drainage basin is reasonably accurate in simulating water quantity. Ragettli & Pellicciotti (2012) have also used a cell-based rainfall-runoff model to study the Juncal River Basin, where streams are fed by ice and snow. They find that their cell-based model can provide accurate simulations. For these reasons, we developed a cell-based model to simulate the hydrologic conditions in the LVW Watershed.

The map of the 8-digit hydrologic units for the LVW Watershed was derived from the Better Assessment Science Integrating Point and Nonpoint Sources (BASINS) software (USEPA 2004). In addition, this investigation used several other types of data: (1) climate data, including the monthly precipitation and temperature data, to develop and validate the model; (2) Digital Elevation Model (DEM) information for determining the flow direction of runoff within the LVW Watershed from its upper reaches to the outlet; (3) discharge data at the outlet of the LVW Watershed for calibration and validation of the hydrologic model; (4) land-use and soil type data used to calculate the water balance for each cell; and (5) future climate and land-use data based on some realistic future scenarios.

The discharge data were abstracted from the USGS. The data are from the gauge station located at 36 °06′01.35″N latitude and 114 °56′35.95″W longitude, about 800 m upstream from the outlet of the LVW (USGS 2013). The available discharge data from this gauge station are confined to two periods: 1989–1997 and 2006–2011. To be consistent with the climate data available from NOAA, two time-periods were selected to develop the model: 1992–1996 and 2008–2010.

Since the primary goal of this study was to develop a cell-based hydrologic model capable of characterizing watershed hydrology and postulating the plausible future hydrologic conditions, hypothetical future climate and land-use scenarios were used for model input.

To obtain a future land-use scenario, many scholars use land-use modeling. A Markov chain is a stochastic method, which uses two sets of historical land-cover images to calculate the probabilities of change in each land-use class and to model the potential for future changes (Muller & Middleton 1994). Although it is capable of simulating land use change over time and the randomness in the land use change process, one major problem is that the model does not consider the geographical spatial relationship (Ye & Bai 2008). To address this problem, the cellular automata (CA) can be coupled into the Markov chain model.

CA is a discrete dynamic system in which the state of each cell is determined by its neighboring cells according to the pre-defined transition rules (White & Engelen 1993). When the cellular automata-Markov chain model (CA-Markov) is used in land-use simulation, the state of each cell in each time step is determined not only by the probabilities of land-use change from one category to another category, but also by its neighboring cells (Sang et al. 2011). Consequently, it can simulate the change process over time as well as space (Eastman 2009). Moreover, by using the Multi-Criteria Evaluation (MCE) method and a set of weighting factors (such as the rate of population growth), the CA-Markov model can assess the impacts of various spatial or temporal variables on future land-use change (Zhang et al. 2008). Because of its superior capabilities in producing land-use maps that take into account the historical trend of land-use development, the randomness of land-use distribution, the proximity effects of existing land-use classes, and other causal factors of land-use change, this research used the CA-Markov module in IDRISI (Eastman 2009).

To simulate the rainfall-runoff process in the LVW Watershed, two traditional hydrologic models were used to estimate the total surface runoff from each cell: the Thornthwaite water balance model (Thornthwaite & Mather 1955), which was employed to simulate the conversion from precipitation to infiltrated water or water surplus, and the CN method (US Soil Conservation Service 1986) to predict direct runoff. Their runoff equations were imbedded in the cell-based model. Model parameterization includes the spatial relationship, the water balance within each cell and among neighboring cells, and hydrologic properties of both the natural and man-made systems, such as vegetation types, urban land use/land cover, water supplies, and water surplus.

In this research, a different CN was used to approximate a different moisture condition in different seasons and to adjust the amount of direct runoff. Moreover, the CN method was combined with the Thornthwaite water balance method, and the rainfall-runoff simulation was performed at each cell. In theory, it could produce a better approximation of the hydrologic conditions than if the Thornthwaite water balance method was used alone, particularly in the LVW Watershed where the drainage area is large and there is a mix of urbanized land use and vegetated land cover.

The cell-based hydrologic model was developed using the historical 1992–1996 climate data and the 1992 NLCD land-use data. The 1992 NLCD data were chosen because it is the only available meta-data set from the Consortium for that period, and a search using the Google Earth Engine (2012) shows that the land-use pattern of the LVW Watershed remained almost the same during that time period. Monthly surface runoff and river discharge were simulated and the model results for the period of 1992–1996 were calibrated using the observed discharge records from the USGS gauge station at 36 °06′01.35″N and 114 °56′35.95″W, which was the only gauging station with continuous historical monitoring data in the study area. As it is the closest station to the outlet of the watershed, collecting runoff from the entire watershed, it is therefore appropriate to compare its monitored records with the simulated values from the cell-based model for the purposes of calibration and validation.

The calibration was assessed by (1) percentage bias, calculated by ∑ (estimated value – observed value)/∑ observed value and (2) the correlation coefficient between the observed values and the simulated values. After the model was calibrated, model validation was required to ascertain the model validity and reliability in hydrologic simulation and predictions. For the LVW Watershed, model validation was performed by using the historic climate data from 2008 to 2010. Since the 2011 version of the NLCD data were not released at the time when this research was conducted, the 2006 NLCD data, the latest available land-use data, were used. From the Google Earth Engine search, it was found that only a few minor changes to the urban area were observed. The validation process was similar to that of the calibration, and the estimated and observed values of surface runoff for each month were compared by percentage bias and correlation coefficient.

Based on the calibrated and validated model, the hydrologic conditions in the LVW Watershed in 2030 and 2050 were simulated under future climate and land-use projections. Three sets of model simulations were conducted to analyze the individual or combined impacts of climate and land-use changes on future hydrologic conditions. The first simulation used the IPCC's future climate data and historical land-use map of 2006 to examine the hydrologic effects of climate change. The second simulation employed the predicted future land-use data but the historic climate data of 2010 to investigate the influence of land-use change on hydrology. The last simulation utilized both the future climate and future land-use data to study the combined impacts of climate and land-use changes in the future hydrologic conditions.

Another possible cause for the overestimation is the presence of detention basins in the Clark County Regional Flood Control District (CCRFCD). The occasional burst of rainfall in summer often causes flash floods in the area as storm water flows onto the valley floor. The detention basins are built by the CCRFCD as best management practices (BMPs) to control flash floods by temporarily storing the water and releasing it to the LVW later at a controlled rate. Currently, the total capacity of the detention basin is 20.96 million m³ (CCRFCD 2013). Hence, the effect of detention basins on surface runoff simulation can be substantial. But water detention and evaporation in the detention basins are not represented in the cell-based modeling because the actual amount of surface flow stored in each detention basin varies from storm to storm and place to place. Besides, each detention basin differs in its maximum capacity and the rate of water released. Most importantly, the impacts of detention basins or other BMPs on surface runoff are usually more pronounced under a finer spatial resolution. Thus, most analyses on BMPs are conducted in a neighborhood scale (see, for example, Lee et al. 2012). The LVW Watershed is a large basin consisting of many subcatchments and numerous neighborhoods. To incorporate detention basins in model development and analysis for the LVW Watershed is therefore beyond the scope of this paper. The presentation of the analyses and results of the hydrologic impacts of BMPs on a much smaller subwatershed is deferred to another paper.

These findings are consistent with the predictions by the IPCC model ensemble. Since the land-use pattern was assumed to remain the same, the projected decrease in winter precipitation and increase in summer precipitation would certainly have direct impacts on the surface runoff. In this study, the simulation results show that future winters will have less runoff, and summers will have more runoff. Hence, it is likely that there will be a higher risk of drought in winters and flood in summers.

Figure 18(b) also shows the hydrologic impacts of land-use change. If the future climate were the same, the amount of monthly surface runoff would increase slightly for both winter and summer. Under the land-use change only scenario, the rates of increase in 2050 will be larger than that of 2030 by 5.14% in July to 21.22% in February. This finding indicates that with urbanization, the expansion of impervious surface will cause a persistent and substantial increase in surface runoff in both winter and summer seasons. This result agrees with the common notion that by increasing the impervious surface and decreasing vegetation cover, urbanization will reduce the amount of infiltration, decrease the travel time of surface runoff, and increase the surface runoff (LeBlanc et al. 1997).

When the hydrologic impacts of climate and land-use changes are considered in tandem, the results are similar to those when climate change is considered alone (Figure 18(c)). There will be less runoff in winter and more runoff in summer by the horizon years of 2030 and 2050. The decrease in surface runoff in January and February is the largest (−51.07% and −52.86% in 2030 and −43.84% and −38.92% in 2050, respectively). The increase in July and August will likely be 48.11% and 40.03% in 2030 and 52.42% and 65.29% in 2050, respectively.

A comparison of the changes in the surface runoff in the LVW Watershed with those under a climate change only or land-use change only condition shows the amalgamated hydrologic effects of urbanization with future climate. Land-use change, largely through urbanization, will likely ameliorate the reduction in surface runoff caused by climate change in the winter season and further increase the surface runoff in summer. Urbanization may help to increase the surface runoff by 0.43 to 8.18% in winter and 0.24 to 9.54% in summer. These results imply that in the arid LVW Watershed, climate change has a relatively more dominant hydrologic effect. With the continued urban development, a larger impervious layer will result in more surface runoff (Kang et al. 1998; Weng 2001; Olivera & DeFee 2007). But in the LVW Watershed, the impacts of urbanization on the rainfall-runoff relationship are at a lesser extent.

As the modeling results have revealed, in the future, the surface runoff in summer will increase in the LVW Watershed, which may lead to a higher risk of flooding. The current detention basin projects by the CCRFCD can be instrumental in reducing flash floods. According to the CCRFCD (2013), the total capacity of detention basins will be increased approximately five times in the next decades. This mitigation measure will greatly help to control floods. On the other hand, the results from this study show that in winter, the surface runoff in the study area will decrease drastically. The reduction in winter precipitation and snow melt may pose challenges to sustainable water management policies in the watershed. The information on the predicted changes and seasonality of LVW discharge to Lake Mead may be useful to government agencies in devising urban development plans and water management policies.

This cell-based hydrologic simulation aims to explore the hydrologic relationship of climate change and urbanization. Monthly rainfall-runoff simulation yields quantitative assessment of the plausible individual and combined impacts of climate change and land-use change in the years 2030 and 2050.

Through the modeling exercise, climate change is found to be the primary factor in determining the surface runoff in the watershed. Urbanization exacerbates runoff generation in summer and reduces the degree of decline in climate-induced runoff in winter. While there are many studies of the effects of climate and land-use changes on watershed hydrology (see, for example, Changnon & Demissie 1996; Cognard-Plancq et al. 2001; Bronstert et al. 2002; Legesse et al. 2003), the results of this investigation of an arid and rapidly urbanizing watershed are noteworthy. CA-Markov land-use projections indicate that significant urban sprawl will occur by the horizon years of 2030 and 2050, creating a larger amount of impervious surface, which inevitably will affect the hydrologic cycle, and particularly the infiltration process. But the future change in precipitation will very likely determine surface runoff and watershed hydrology in the study area.

The calibration and validation results of the cell-based model also show that by combining the traditional hydrologic modeling methods (Thornthwaite water balance model and CN method) with the cell-based and GIS simulation, it can produce a reasonable model prediction of surface runoff from the LVW Watershed. Because a wide range of factors affecting watershed hydrology are taken into consideration, this model is capable of expressing the spatial heterogeneity of the hydrologic variables and depicting how runoff would respond to various combinations of climate, land-use cover, imperviousness layer, and soil types. Nonetheless, the model is rather crude. Future improvement can be made to account for the temporal and spatial heterogeneity of precipitation. This may be particularly useful in estimating the risks of floods and droughts in a semi-arid environment.

Another area of improvement is to include the effects of detention basins on runoff in a smaller watershed. To this end, we have conducted another case study on a subwatershed of the LVW Watershed, the Duck Creek Subwatershed. The results show that the existing detention basins in this small subwatershed will be inadequate under the future climate and land-use change scenarios, and additional infiltration and detention BMPs will be needed to curb the projected increase in storm runoff in summer.

In this paper, we only intended to examine the hydrologic impacts of urbanization and climate change. For this reason, we focused our discussion on the application of the cell-based model in simulating urbanized areas, and have not specifically reported the sensitivity of our model to non-built up areas. However, we believe that the model can be applicable in non-urbanized land use. This is because, in our cell-based model, different land-use types are represented by different CN numbers. In fact, as illustrated, less than 20% of the LVW Watershed encompasses built-up areas. Moreover, we have conducted another study on the Upper Mill Creek Watershed in Ohio, which is a less developed area, and the results have indicated that the model is capable of simulating the hydrologic conditions in that watershed as well.

This research was partially funded by the US Environmental Protection Agency. The authors are grateful to the agency for the financial support.

The US 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.