05-11-2016, 12:36 PM
1465276671-Hobsonthesis.pdf (Size: 2.6 MB / Downloads: 6)
ABSTRACT
Hobson, Adam N. (M.S. Civil, Environmental and Architectural Engineering)
Use of a Stochastic Weather Generator in a Watershed Model for Streamflow
Simulation
Thesis directed by Assistant Professor Balaji Rajagopalan
Current methods of streamflow forecasting rely on historic climate sequences
that are inadequate in length and have statistical relationships that are difficult to fit
and condition. Coupling a stochastic weather generator with a deterministic
watershed model can provide better streamflow forecasts. This study presents a
technique to couple a k-nearest neighbor stochastic weather generator and the
Precipitation-Runoff Modeling System (PRMS) watershed model to simulate historic
streamflow statistics and provide a framework for forecasting flows. The weather
generator uses weather data in the Upper Truckee River Basin on the California and
Nevada border (USA) to produce a simulated dataset. The simulated dataset
sufficiently preserves the statistics of the historic record of precipitation and
maximum and minimum temperature. Simulated weather variables were used as
input to PRMS, which adequately simulated modeled historic streamflows. A
conditioned forecast based on wet/dry years was used to demonstrate the utility of the forecasting framework.
INTRODUCTION
Realistic and accurate streamflow forecasts are an essential tool for water
resources planning and management. In many regions, agricultural, municipal, and
environmental water uses place increased demands on limited and variable freshwater
resources. In the Upper Truckee River Basin, like many basins in the Western United
States, planning and management of water resources is particularly challenging
because snowmelt from the Sierra Nevada Mountains is essentially the only source of
streamflow in the semi-arid desert of Western Nevada. As a result, inter-annual
variability of streamflow in such basins tends to be high. Streamflow forecasts can be
used as a tool to facilitate effective basin management by providing accurate forecasts
for water quality, volume, timing, and flow rates.
1.1. BACKGROUND
Simulated streamflows are used in conjunction with hydrologic and hydraulic
models to generate flow forecasts which can then be used to assist with water
resources management in a basin. The ensemble or extended streamflow prediction
(ESP) procedure is one method used to forecast streamflows in hydrologic models
used by the U.S. National Weather Service (NWS) and others [Day, 1985; Leavesley
et al, 2002; Mastin and Vaccaro, 2002]. The ESP method assumes that past
2
meteorological conditions are representative of what may occur in the future and
uses historic weather data as input variables to initialize a deterministic watershed
model. The model is then run in a simulation mode to forecast for the given time
period. For short term streamflow forecasts (i.e. one-week), the NWS uses seven-day
weather forecasts to drive the simulation model. For longer forecast periods (i.e. twoweeks,
monthly, seasonal), weather forecasts are unavailable. Therefore, historic time
series weather data from the representative time period are run through the watershed
model to generate an ensemble of flows that can be analyzed statistically to give a
stochastic forecast [Day, 1995].
The ESP approach has many advantages including simplicity in
implementation and assumption that historic weather conditions can be surrogates for
future weather. The method is easy to apply in any basin over any time scale where
historic weather data are available. By utilizing historic data records, the spatial and
temporal distributions of weather variables are inherently preserved. However, the
drawback to this method is that it relies on the length of the historic record to create
the members of the ensemble which, in many cases, does not provide enough series to
be statistically significant. For example, if only 19 years of basin-wide weather data
are available, only 19 flow hydrographs are possible, reducing the variability of the
simulated flows. Additionally, if a conditioned forecast is required, such as one based
upon large-scale climate indices or wet/dry relative forecast, the number of
possibilities is further reduced. Given the inherently short historic records in many
water-stressed basins and changing climate patterns a better streamflow forecasting
tool is required. Modifying the ESP approach by using a weather generator to create
3
statistically similar, synthetic weather data based on the historic record would
increase the number of records from which to generate ensembles and improve the
ability to make conditioned streamflow forecasts.
STUDY AREA
The Upper Truckee River Basin (Figure 1.1) is a snow-melt dominated
watershed that requires significant streamflow forecasting to meet competing
demands from various water users. The basin encompasses an area of approximately
3,060 square miles on the border of California and Nevada. The Truckee River
originates as the only drainage outflow from Lake Tahoe in California, runs
northeastward approximately 105 miles, and terminates in Pyramid Lake in Nevada.
The Truckee River has an average annual flow of 548,200 acre-feet (1973-1994
period of record) at the Farad gaging station on the California- Nevada border
[Horton, 1995].
The Upper Truckee Basin is a steep, high alpine or forested environment with
elevations reaching 9,000 to 10,000 feet [Horton, 1995]. The area receives 30 - 60
inches of precipitation annually, primarily in the form of snow [Taylor, 1998].
Consequently, most of the runoff results from snowmelt during the spring months.
The Upper Truckee River Basin has seven major storage reservoirs: Lake Tahoe,
Donner Lake, Independence Lake, Martis Creek Lake, Prosser Creek Reservoir,
Stampede Reservoir, and Boca Reservoir. These reservoirs are used for both flood
control and storage of water for downstream uses.
The Truckee River is used to produce hydropower at a plant upstream of the
Truckee Meadows in Nevada, an area that encompass the cities of Reno and Sparks.
4
Reno and Sparks use Truckee River water for municipal and industrial (M&I)
purposes. Downstream of the Truckee Meadows, the Truckee River flows to Derby
Dam where an annual average of almost 187,000 acre-feet of water is diverted from
the Truckee Basin through the Truckee Canal into Lahontan Reservoir for use in the
Newlands Project irrigation district. The Newlands Project diversion comprises the
most significant single withdrawal of the Truckee River’s waters.
The portion of the river that is not diverted continues through desert and into
Pyramid Lake within the Pyramid Lake Indian Reservation. Two fish, the endangered
cui-ui and the threatened Lahontan cutthroat trout live in Pyramid Lake and are
culturally and economically important to the Pyramid Lake Paiute Tribe. Low flows
and shallow depths in Truckee River below Derby Dam, however, have inhibited
spawning, egg incubation, and survival of these species [Taylor, 1998].
The Truckee River has been, and continues to be, crucial to the sustainment of
life in western Nevada. The river has played a major role in the settlement and
development of the area. Consequently, the policies and operations on the river
extend back to before the turn of the century and continue to be negotiated to this day.
Current negotiations seek to balance the demands of M&I for the cities of Reno and
Sparks, irrigation for Truckee Meadows, power production, as well as protection of
endangered and threatened species [Horton, 1995].