University of Southern California Dissertations and Theses

A model for placement of modular pump storage hydroelectricity systems

(USC Thesis Other)

A model for placement of modular pump storage hydroelectricity systems

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

A Model for Placement of Modular Pump Storage Hydroelectricity Systems

by

Joseph Warren Rosenbery II

A Thesis Presented to the
Faculty of the USC Graduate School
University of Southern California
In Partial Fulfillment of the
Requirements for the Degree
Master of Science
(Geographic Information Science and Technology)

December 2019

Copyright © 2019 by Joseph Warren Rosenbery II

To Kelly and Olivia, I love you.

iv
Table of Contents
List of Figures ............................................................................................................................... vii
List of Tables .................................................................................................................................. x
Acknowledgments.......................................................................................................................... xi
List of Abbreviations .................................................................................................................... xii
Abstract ........................................................................................................................................ xiii
Chapter 1 Introduction .................................................................................................................... 1
1.1. Project Goal ........................................................................................................................3
1.2. Project Workflow ................................................................................................................5
1.3. Report Organization ............................................................................................................6
Chapter 2 Background .................................................................................................................... 8
2.1. The Electrical Grid and Renewable Energy ........................................................................8
2.2. Hydroelectric Reservoirs and Pump Storage ....................................................................10
2.3. Overlay Analysis for Suitability Modeling .......................................................................12
2.3.1. Boolean Overlay Analysis .......................................................................................13
2.3.2. Weighted Overlay Analysis .....................................................................................15
2.3.3. Fuzzy Overlay ..........................................................................................................17
2.4. Applications of Overlay Analysis in Renewable Energy Siting .......................................19
Chapter 3 Modeling Framework ................................................................................................... 24
3.1. Model Objective................................................................................................................24
3.2. Engineering Requirements ................................................................................................24
3.3. The Primary and Secondary Models .................................................................................28
3.4. Modeling Products ............................................................................................................30
3.5. Modeling Context .............................................................................................................31
3.5.1. Modeling Environment ............................................................................................31
v
3.5.2. Spatial and Temporal Scale .....................................................................................32
Chapter 4 Model Description ........................................................................................................ 34
4.1. Data Requirements ............................................................................................................34
4.1.1. Primary Model Data Requirements .........................................................................34
4.1.2. Secondary Model Data Requirement .......................................................................37
4.2. Primary Model Processes ..................................................................................................37
4.2.1. Construction Area Analysis .....................................................................................38
4.2.2. Refining the Search Area .........................................................................................40
4.2.3. Extracting a subset of the DEM ...............................................................................41
4.2.4. Relief Calculations ...................................................................................................42
4.2.5. Filtering Suitable Locations .....................................................................................44
4.2.6. Matching Points .......................................................................................................44
4.2.7. Filtering Connections ...............................................................................................45
4.2.8. Filtering Reservoir Locations ..................................................................................48
4.2.9. Model Results ..........................................................................................................49
4.3. Secondary Model Processes ..............................................................................................50
Chapter 5 Case Study: Model Processing, Outputs, and Evaluation of Results ........................... 52
5.1. Preliminary Steps ..............................................................................................................52
5.1.1. Processing the DEM ................................................................................................53
5.1.2. Binary Screen Creation ............................................................................................54
5.1.3. Restricted Lines .......................................................................................................55
5.2. Intermediate Results..........................................................................................................56
5.2.1. Construction area identification ...............................................................................57
5.2.2. Creating the Subset DEM ........................................................................................61
5.2.3. Searching for Relief .................................................................................................62
vi
5.2.4. Making Connections ................................................................................................64
5.3. Final Primary Model Results ............................................................................................65
5.4. Secondary Model ..............................................................................................................67
5.5. Model Performance ...........................................................................................................71
5.5.1. Primary Model .........................................................................................................72
5.5.2. Secondary Model .....................................................................................................82
5.5.3. Alternative Study Areas ...........................................................................................84
Chapter 6 Conclusions .................................................................................................................. 89
6.1. Model Assessment ............................................................................................................89
6.2. Uses ...................................................................................................................................94
6.2.1. Viability in Market ...................................................................................................94
6.3. Future Work ......................................................................................................................95
References ..................................................................................................................................... 97
Appendix A – Primary Model Process Table ............................................................................. 101

vii
List of Figures
Figure 1 – Basic Model Design ...................................................................................................... 3
Figure 2 – A graphic flowchart showing the basic model design workflow .................................. 6
Figure 3 – The traditional topographically constrained configuration for pump storage
hydroelectricity systems. Source: ClimateTechWiki.org 2006. ....................................... 11
Figure 4 – Esri Fuzzy Membership Function Plots (Esri 2016a).................................................. 18
Figure 5 – A simplified diagram showing the engineering parameters for relief and maximum
lateral runout used in the demonstration of the Primary Model. ...................................... 25
Figure 6 – The reservoir site leveling scale concept ..................................................................... 27
Figure 7 – The relationship between slope angle and cut/ fill volume needed to create a flat
construction surface. ......................................................................................................... 28
Figure 8 – A graphical depiction of the conceptual Primary Model ............................................. 29
Figure 9 – A graphical depiction of the conceptual Secondary Model ......................................... 30
Figure 10 – The portion of the model flow chart showing the steps to determine areas suitable
for construction. ................................................................................................................ 39
Figure 11 – A graphical depiction of the slope processes of construction area analysis .............. 40
Figure 12 – Modeling process for construction area identification. ............................................. 41
Figure 13 – The portion of the model flow chart showing the steps to create a new subset
DEM dataset...................................................................................................................... 42
Figure 14 – Model processes for upslope and downslope relief calculations. .............................. 43
Figure 15 – The portion of the model flow chart showing the steps to pair upper reservoir
location with lower reservoir locations ............................................................................. 45
Figure 16 – The portion of the model flow chart showing the steps to remove connections
between the upper reservoir and lower reservoir locations that cross named streams. .... 46
Figure 17 – A graphic depiction of conditions identified as false and real matched
connections where relief must be at least 300 m. ............................................................. 47
Figure 18 – The portion of the model flow chart showing the steps to identify and eliminate
false match connections. ................................................................................................... 48
viii
Figure 19 – The portion of the model flow chart showing the steps to eliminate upper and
lower reservoir locations that do not have a connection after the connection filtering
steps................................................................................................................................... 49
Figure 20 – The portion of the Secondary Model flow chart showing the steps that apply the
values of the fuzzy ranking dataset to the upper and lower reservoir locations. .............. 51
Figure 21 – The processed DEM for Los Angeles County used in the model example. .............. 54
Figure 22 – Areas not for placement consideration (left) and Binary Screening Layer (right) .... 55
Figure 23 – Restricted Lines dataset for Los Angeles County...................................................... 56
Figure 24 – Slope in degrees for Los Angeles County ................................................................. 57
Figure 25 – The raster dataset showing the maximum slope for each nine-cell neighborhood
in the study area. ............................................................................................................... 58
Figure 26 – The areas identified by the model as being suitable for the placement of a
terminal reservoir. ............................................................................................................. 59
Figure 27 – The areas identified as suitable for construction of a terminal reservoir after the
application of the binary screening layer. ......................................................................... 60
Figure 28 – The subset DEM created by using the areas deemed suitable for construction of a
terminal reservoir to extract elevation values from the study area DEM ......................... 62
Figure 29 – Intermediate datasets showing the maximum downhill (left) and upslope (right)
relief within 1,500m search radius. ................................................................................... 63
Figure 30 – Preliminary upper and lower reservoir location. ....................................................... 64
Figure 31 – A summary of final Primary Model outputs for Upper and Lower Reservoir.
Locations exaggerated for visual effect ............................................................................ 65
Figure 32 – A selection of final Primary Model results, La Canada Flintridge. ........................... 66
Figure 33 – The Fuzzy Layer for Los Angeles County. ................................................................ 69
Figure 34 – A selection of final model results near La Crescenta, California, after the
integration of the fuzzy layer. ........................................................................................... 70
Figure 35 – Histograms showing the distribution of fuzzy membership values for upper and
lower reservoir locations. The red vertical line on the histograms indicates the mean
membership value for each dataset. .................................................................................. 71
Figure 36 – Distribution of connection lengths ............................................................................ 73
Figure 37 – Distribution of point to point relief magnitude.......................................................... 74
ix
Figure 38 – A sample of Primary Model Results for upper and Lower reservoir locations ......... 75
Figure 39 – Model results near Palmdale, California. .................................................................. 76
Figure 40 – Example results from the Primary Model outputs near Universal City, Los
Angeles County, California. Note there is only one upper reservoir site indicated. ......... 77
Figure 41 – A graphical example of the one to many relationships between upper and lower
reservoir locations ............................................................................................................. 78
Figure 42 – Complex relationships in clustered reservoir locations ............................................. 79
Figure 43 – Results filtered to originate from a single upper reservoir location .......................... 81
Figure 44 – Distribution of alternative study areas in California ................................................. 85

x
List of Tables
Table 1 – Binary Criteria for Siting Wind Turbines (after Sparkes and Kidner 1996) ................ 13
Table 2 – Esri Raster Overlay Tool Summary (Esri 2016c) ......................................................... 16
Table 3 – Fuzzy Overlay Functions (Esri 2016b) ......................................................................... 18
Table 4 – Modified AHP results (Uyan 2013, 11-17) .................................................................. 20
Table 5 – Suitability criteria used to determine pump storage conversion potential for
existing hydroelectric facilities. ........................................................................................ 22
Table 6 – An example portion of the attribute table for one of the Secondary Model outputs
(n=40,868) ........................................................................................................................ 51
Table 7 – Variables used to generate the fuzzy dataset used in the Secondary Model for the
Los Angeles County case study ........................................................................................ 67
Table 8 – Summary of Primary Model Results for all Study Areas ............................................. 86

xi
Acknowledgments
First, I would like to thank my advisor Dr. Karen Kemp for all of her support and
flexibility while I worked through this project. I was glad that we were able not only to interact
as professor and student but through the thesis writing process as well.
I would also like to thank my thesis committee, Dr. John Wilson and Dr. Robert Vos, for
their input, encouragement, and support.
I would like to sincerely thank Dave Guis and Andy Wallace and the rest of the team at
Wallace Kuhl and Associates for supporting me during my tenure at USC and affording me the
opportunity to further my education. I would also like to thank Kurt Balasek for giving me the
idea for this research project and running through ideas with me.
Finally, I would like to thank my family: my Dad for always believing in whatever I
wanted to pursue; my wife Kelly, for putting up with all of the time I spent staring into the
computer and still giving me her unwavering support, and my daughter, Olivia for keeping me
smiling.

xii
List of Abbreviations
3DEP 3D Elevation Program
CGS California Geologic Survey
DEM Digital Elevation Model
GIS Geographic information system
GISci Geographic information science
MPSHS Modular pump storage hydroelectricity System
SSI Spatial Sciences Institute
USC University of Southern California
USGS United States Geologic Survey

xiii
Abstract
As the global energy market pushes toward the further development and integration of renewable
energy and reduced reliance on fossil fuels, the energy industry has looked to innovative
solutions to solve the shortcomings of green energy production. Diurnal fluctuation in electrical
production potential in solar and wind sources creates a need to develop ways to store surplus
energy resources for later deployment. Pump storage hydroelectricity, in which surplus energy is
used to pump water uphill to recharge a hydroelectric reservoir, holds a great deal of potential
when used in conjunction with other types of renewable energy. This report documents the
design and development of a two-phase analytical spatial model that identifies suitable locations
for the placement of paired top and bottom terminal reservoirs of a modular pump storage
hydroelectricity system (MPSHS). The first phase of the model applies user-defined search
criteria to identify locations for the construction of terminal reservoirs that meet the relief and
lateral run distance requirements. Further refinement of results from the first modeling phase
using secondary information can be used to rank suitable locations based on user-supplied
environmental, economic, and socio-demographic constraints and preferences. This thesis
presents details of model function as well as case study results for Los Angeles County.
1

Chapter 1 Introduction
With global concern over the deleterious effects of climate change, there has been a concerted
effort to generate a more significant percentage of electrical energy from renewable sources,
such as hydroelectricity, wind and solar, reducing the production of harmful greenhouse gas
emissions (Rosenberg 2008). To address this concern, the state of California has set agressive
goals concerning electricity and emissions. In 2018, the California State Senate passed Senate
Bill 100, which states that energy production in the State of California should achieve net-zero
greenhouse emissions by 2045 by focusing efforts on the development of renewable energy
sources (California Senate 2018).
Temporal fluctuations in the potential availability of conditions favorable to green energy
production such as wind and solar mean that green energy supply is often out of phase with
consumer demand. Without a way to store energy produced by renewable energy sources during
their peak productivity, fossil fuel generation must be available to meet the demand when it
exceeds the renewable energy production potential (Kaplan 2009).
Hydroelectric power is one suitable remedy. In 2017, it supplied nearly 17% of the
electrical power to the world by storing water in such a way that it can be used later for the
generation of electricity. However, financial, political, and practical constraints on the
construction of new water reservoirs for hydroelectric energy production largely hinder the
further development and exploitation of the resource (USGS 2018).
Traditional hydroelectric power generation facilities are a one-way system, meaning all
the water passing through the system moves downstream. In these systems, the electricity also
flows in only one direction, outward to the consumer (USGS 2018). Thus, the generating
potential is finite as the water used to turn the turbine-generator has to come from upstream in
2

the watershed and be stored in the reservoir (Madani, Guégan, and Uvo 2014, 153-163). The
Hoover Dam on the Colorado River is an example of this type of system.
One method that has been developed to store renewable energy for use on-demand is
pump storage hydroelectricity (Yang and Jackson 2011, 839-844). At its root, pump storage
hydroelectricity is a concept built on the more traditional hydroelectric model, where water
stored in a reservoir is released through a conduit (penstock) which feeds a turbine connected to
a turbine generator. However, pump storage systems have a unique attribute; they can move
water back uphill.
Pump storage hydroelectricity is a method by which the finite storage capacity is
augmented by the ability to pump water back to a higher potential energy state, ready for reuse in
the energy production cycle. While pumping requires energy, it can be supplied on demand, only
when there is unused and otherwise wasted energy available. With the implementation of pump
storage hydroelectricity production, the overall grid not only becomes less reliant on more
traditional fossil fuel-based forms of energy conversion but also provides alternatives to those
working to improve grid stability (Rehman, Al-Hadhrami, and Alam 2015, 586-598).
The research reported in this document supports efforts to identify potential solutions to
the problems currently facing the renewable energy market. How can these systems provide the
energy needed on demand without relying on traditional fossil fuel facilities for energy
production when environmental conditions limit the capabilities of renewable sources? Modular
Pump Storage Hydroelectricity Systems (MPSHS) can take advantage of the benefits afforded by
the traditional pump storage hydroelectricity model while avoiding the environmental and
political constraints of their larger counterpart. This type of technology has the potential to work
3

in tandem with other types of renewable energy products to reduce harmful greenhouse gas
emissions and move toward energy sustainability.
1.1. Project Goal
The goal of this research project was to develop a GIS model package within an Esri
ArcGIS ModelBuilder application that can be used by project developers and engineers (end-
user) in support of a decision-making process to determine the placement of MPSHS. This model
provides a preliminary assessment tool for identifying locations ideal for terminal reservoirs of
MPSHS. The model has two components. The Primary Model explores terrain within the
designated study area to identify locations suitable for the construction of reservoir tanks. The
Secondary Model builds on the Primary Model outputs by assigning aggregate suitability values
to each potential reservoir location using fuzzy logic datasets provided by the end-user
(Figure 1).

Figure 1 – Basic Model Design
Using physical parameters provided by design engineers in the form of the constraints of
1) minimum head requirements, 2) maximum lateral run distances and 3) minimum tank
4

footprint size, topographically suitable terminal reservoir locations are identified by the Primary
Model. Accommodations for additional environmental variables that effectively eliminate areas
known to be unavailable for development have been incorporated into the model as two strategic
sets of variables acting as areal or linear prohibitions to construction.
The first set of environmental variables produce a binary screening layer compiled from
end-user provided datasets that identify areas known to be unsuitable for construction. For
example, the areas within the boundary of a National Park are likely not suitable for
construction. The binary screening data must be converted into a binary screening raster in which
areas that are suitable are coded 1 and areas that are not suitable are coded 0. All such areal
features are identified in advance by the model end-user and combined into one raster dataset for
use as a single model variable. This process is designed to fine-tune suitable location
identification and reduce the processing load.
The second set of environmental variables is used to assess the viability of the reservoir
connections. One of the fundamental components of the MPSHS is the connection, called a
penstock, that serves as the conduit from the upper reservoir to the lower reservoir. There are
many natural and man-made linear features that serve as continuous barriers to the construction
of penstocks, such as large streams and roadways. After the model has established the complete
set of possible connections between the upper and lower reservoirs, the model searches for
connections that cross linear features that cannot be crossed by a penstock and removes them
from inclusion in the Primary Model results.
Finally, an optional Secondary Model, further enhancing the products of the Primary
Model, provides the capability for the end-user to develop and apply their own additional
suitability layer using fuzzy logic, enabling further refinement in site selection capabilities.
5

As a core design requirement, this model was developed to use freely and widely
available spatial data. This approach allows the end-user of the model to gather all the required
data necessary to run the model to completion with minimal investment in time and capital with
respect to data procurement.
The datasets produced by the Primary Model include two 30m raster datasets, one for
each of the upper and lower reservoirs of the MPSHS, and a vector dataset of lines representing
viable connections between paired reservoirs. The Secondary Model applies the fuzzy logic
raster dataset to the two reservoir location datasets and produces a set of points at the center of
all raster cells that are suitable reservoir locations, each point attributed with its aggregate fuzzy
membership value.
1.2. Project Workflow
To accomplish the goals set forth in the section above, the model was developed using an
interactive approach common to projects of this type (Figure 2). After the development of the
research question, background research was conducted to understand better the environment
within which the problem was set. Following the research phase, a conceptual model was
developed for the study that incorporated the primary design elements required by project
engineers. With this basic understanding of the project goals, the models were developed and
evaluated through a versioning process that allowed for assessment through incremental
progress. The final product was then tested in multiple geographic locations to evaluate model
processes for issues caused by spatial and topographic variability, thus allowing for further
refinement of the final model.

6

Figure 2 - A graphic flowchart showing the basic model design workflow
1.3. Thesis Organization
This document is laid out as follows. First, Chapter 2 explores the basic components and
systems of the electrical grid in the United States including the incorporation of green energy, the
role that different types of renewable energy play in the energy production network, and some of
their significant disadvantages. Second, pump storage hydroelectricity is explored as a method
for recovery and storage of renewable energy as potential energy. Finally, the role of GIS and
suitability modeling in identifying locations for renewable resource projects is examined.
The modeling framework is discussed in Chapter 3. This includes a detailed examination
of the conceptual model on which the computational model was developed. The conceptual
model details the engineering requirements governing the MPSHS, the project constraints, a
description of model design, and the primary model outputs. Additionally, Chapter 3 discusses
the modeling context employed, including the modeling environment and relevant spatial and
temporal scales. This chapter also discusses the project design footprint and the advantages of
using county boundaries as the preferred study area limits.
7

Chapter 4 describes the structure of the model. Beginning with data gathering, this
section outlines the required data, processing steps, and a walkthrough of the processes
comprising the Primary Model developed for this research. The next section details the
Secondary Model components and processes.
While this model was developed over multiple study areas, Chapter 5 walks through the
model steps using Los Angeles County as an example modeling unit. First, the inputs to the
model are explored prior to model execution. The subsequent sections include a walkthrough of
the intermediate process output datasets and final model outputs for both the Primary and
Secondary models. A detailed evaluation of the final model outputs for both the Primary and
Secondary models completes the chapter.
Chapter 6 provides conclusions in the form of an overview of the model, its performance,
limitations, and ultimate usefulness. This chapter also provides details of opportunities for
continued research into this topic and examines the applicability of techniques developed to
other applications such as recreation, transportation, and engineering.
8

Chapter 2 Background
This chapter provides background information, setting the context for the research presented.
Beginning with an overview of the consumer electrical grid and the role of renewable energy, the
context of this chapter focuses on a discussion of the applications of pump storage
hydroelectricity. Additionally, this chapter provides a review of the application of GIS modeling
to the development of renewable energy infrastructure.
2.1. The Electrical Grid and Renewable Energy
The electrical grid in the United States is comprised of multiple components whose role
is to generate electricity from a variety of sources and distribute electricity to consumers. The
collection of systems that comprise the electrical grid is the byproduct of a multitude of small
systems that were built to meet the needs of local and regional customers. Over time, these small
systems have grown together to create the modern grid, which is comprised of three major
interconnected units. These cover the western states and western Canadian provinces (Western
Interconnection), the eastern states and eastern Canadian provinces (Eastern Interconnection),
and most of the state of Texas (ERCOT Interconnection) (Kaplan 2009).
With the rise of public and political awareness surrounding the need for increased
efficiency concerning energy generation and consumption, the U.S. is developing a technology-
driven grid management system currently being integrated into the existing system, called the
Smart Grid. Goals of the Smart Grid program allow for increased efficiencies in the demand/
supply curve, provide tools for end-user management, improve quality and reliability, and enable
the incorporation of renewable energy sources into the grid (Heirman 2012).
The U.S. electrical grid works on a demand-supply routine, which means energy
production must match demand and be available when demand for energy is high. The
9

fluctuation of demand on the grid is called load cycling. Thus, the supply of electrical energy
fluctuates continuously to meet the current demands of the electrical distribution grid. For some
types of renewable energy, diurnal fluctuations in generation potential may be temporally
displaced from that of the load cycle (Denholm et al. 2010).
Solar energy is a prime example of this effect as the peak generating potential is in the
middle of the day. The influence of solar energy on the grid produces a challenge for grid
managers due to its inconsistent contribution to the system. This is particularly difficult to
manage when solar energy is approaching the end of its diurnal cycle. At these times, demand is
typically increasing while conditions favorable to solar energy production decrease. The
implication is that alternative energy sources must compensate to stabilize the grid (California
Independent System Operator 2016).
One solution to the problem of load balancing is the application of energy storage on the
grid that can be tapped as needed (Suul, Uhlen, and Undeland 2008). The development of viable
long-term energy storage solutions could affect the broader implementation of renewable energy
by 20% (Benitez, Benitez, and van Kooten 2008). Current grid technology supports the
momentary deficit in electrical supply and captures oversupply with large capacitors that provide
temporary supply and storage, which allows for supply to match the demand curve (Chu and
Majumdar 2012). When the demand is displaced temporally from the renewable source peak
load cycle, the supply typically comes from traditional fossil fuel sources. To solve the electrical
grid storage problem, the widespread implementation of pump storage hydroelectricity facilities
has been a cost-effective and efficient tool to supply on-demand energy when needed (Denholm
et al. 2010; Chu and Majumdar 2012).
10

2.2. Hydroelectric Reservoirs and Pump Storage
Traditional hydroelectric energy facilities use water stored in a reservoir to generate
power by releasing water through a turbine which turns a generator. This technology is widely
used around the world and in 2017 accounted for approximately 17% of energy production
worldwide (USGS 2018). These systems not only produce clean, reliable energy, but they also
serve as storage reservoirs, allowing the rationing of water for agricultural and domestic
consumption, providing recreational space, and protecting large populations against devastating
floods. For all the benefits to be gained from building reservoirs and equipping them with
hydroelectric generation facilities, large topographically-constrained reservoirs can have a
deleterious effect on the environment (The National Geographic Society 2011).
There are also many reasons that building a new reservoir can be problematic. Large-
scale reservoirs have a high initial investment cost and can lead to extensive habitat loss and
truncation of riparian fisheries (USGS 2018). Downstream of reservoirs, the rivers are starved of
sediment crucial to natural habitats, and natural flow regimes are disrupted. Modification to a
naturally-regulated system in equilibrium can cause channel incision and irreparable damage to
the ecosystem (Pasternack, Wang, and Merz 2004).
The primary limitation for broader implementation of traditional pump storage systems is
the project scale. These traditional systems typically have a large, topographically-constrained
reservoir and a lower elevation impoundment for retaining water to be pumped up for reuse
(Rehman, Al-Hadhrami, and Alam 2015). Furthermore, many large reservoir projects are
confronted with fierce public opposition (Napier, Carter, and Bryant 1986). Environmentalist
organizations such as Friends of the River and the Sierra Club regularly form protests, work with
state and federal lobbyists, and file lawsuits in opposition of new reservoir construction making
11

the challenge even greater (Friends of the River 2018). While the proposed infrastructure
required for MPSHS would likely go through a rigorous public notification process prior to
construction, similar tank constructions exist in many urban areas without significant public
protest.
The concept of pump-storage hydroelectricity is based on the concepts of traditional
hydroelectricity but includes a design for reversing flows to recharge the system (Figure 3)
(Rehman, Al-Hadhrami, and Alam 2015, 586-598). Like traditional hydroelectric systems, when
demand for energy on the grid becomes high, stored water in the upper reservoir flows down
through a turbine generator and into a second, lower reservoir (USGS 2018). However, when
energy production from other sources of green energy is higher than off-peak demands, the load
can be balanced by activating the pump storage system, moving water back uphill, “recharging”
the system for the next deployment (Rehman, Al-Hadhrami, and Alam 2015, 586-598).

Figure 3 - The traditional topographically constrained configuration for pump storage
hydroelectricity systems. Source: ClimateTechWiki.org 2006.

12

To support the further development of green energy as a viable solution to the use of
fossil fuels, pump storage hydroelectricity works in tandem with renewable energy sources such
as wind and solar (Yang and Jackson 2011). The use of pump storage hydroelectricity generation
facilities complements other types of renewable energy sources by storing electrical energy as
potential energy. The ability to store energy on the grid produced by a renewable source would
dramatically increase the viability and eventually penetration of renewable energy technology
(Bueno and Carta 2006).
2.3. Overlay Analysis for Suitability Modeling
The application of Geographic Information Systems (GIS) in land-use suitability analysis has
been revolutionary to land use planning (Malczewski 2004). Furthermore, site suitability analysis
using GIS has been a powerful tool for the discovery of suitable sites on which to place
renewable energy infrastructure. At its most basic, site suitability models use environmental
variables, assign them values for suitability, and combine those values to determine suitable
locations (Mitchell 2012).
With a strong public, industrial, and political momentum behind the transition away from
fossil fuels, there is a significant push toward understanding what role spatial sciences can play
in renewable resource development and deployment. These proven renewable energy
technologies have precise and well-understood criteria regarding suitable locations with which to
design the most effective suitability models. Thus, site suitability is especially useful at siting
wind and solar generating potential (Henning Sten Hansen 2005).
Fundamental in the suitability modeling framework, overlay analysis is the process of
stacking spatial data and combining the layers to achieve a meaningful output value for all
locations using data found to be relevant to the objective (Bolstad 2005). In GIS, an overlay can
13

be accomplished with raster or vector datasets. In both cases, multiple datasets are combined
through arithmetic processes to produce a single output dataset containing a conclusion from a
combination of the inputs (O'Sullivan and Unwin 2010).
2.3.1. Boolean Overlay Analysis
As an example of the use of GIS-based suitability analysis for siting renewable energy
facilities, Sparks and Kinder combined 19 spatial datasets using simple Boolean overlay
processes to identify suitable locations for the constructions of wind turbine generators in
England. Their approach used simple proximity buffers surrounding spatial phenomena that
constrained the suitability of locations and assigned a simple binary qualifier for suitability,
meaning any location in the study area was either suitable, or it was not (Sparkes and Kidner
1996). Table 1 below shows the criteria used by Sparkes and Kidner to identify suitable
locations.
Table 1- Binary Criteria for Siting Wind Turbines (after Sparkes and Kidner 1996)

Feature Distance to Site must be greater than
Airports 3 km
National Parks 1 km
National trust property 1 km
Military danger zone 3 km
Scenic area 1 km
Forest park 1 km
Built-up area 2 km
City centroid 5 km
An urban centroid 2.5 km
Town centroid 1.5 km
Small town or village centroid 1 km
Small village, hamlet or isolated
settlement
750 meters
Lake, marsh or reservoir 250 meters
Motorway, A-road or B-road 300 meters
Railway 250 meters
River or canal 200 meters
Radio or TV mast 250 meters
14

Feature Distance to Site must be greater than
'Picturesque' or scenic feature 1 km
Elevation Site must be above 100 meters elevation

Creating the binary raster datasets shown above requires a multiple-step reclassification
process. In the case of suitability, where proximity to a known feature is the primary determining
factor in suitability, the distance must be determined for each criterion separately (Mitchell
2012). Whether data are provided in vector or raster format, a geoprocessing tool such as Esri’s
Euclidean Distance tool can be applied (Esri 2016e). This produces a raster dataset where each
cell is assigned a value equal to the distance of its center point from the closest input feature
(vector input) or non-null cell (raster input).
With this raster dataset indicating the distance values from the target phenomena, the
distance raster can then be reclassified based on the suitability criteria. Esri’s reclassification tool
interrogates the dataset and reassigns each cell with a value based on the reclassification
parameters defined in the tool (Esri 2016f). For example, if only the areas that are not within 200
meters of a river or canal are to be considered suitable, then the reclassification tool would assign
all those values within 200 meters of a river or canal with a zero, a negative suitability response.
Conversely, those areas outside of that buffer would be deemed suitable and assigned a value of
1, a suitable response (Mitchell 2012).
Combination of multiple binary datasets such as those outlined above by a multiplication
method similarly returns a binary response. This means that the resulting overlay dataset is
comprised of only those areas that returned a positive response for all of the considered criteria
(O'Sullivan and Unwin 2010).
Beyond binary overlay, the other two most common types of overlay are the weighted
overlay and the fuzzy overlay (Mitchell 2012). These methods move away from the Boolean
15

determination and provide results on a gradational basis in terms of suitability. By using these
more advanced methodologies, suitability can be graded, and suitable sites can be determined
relative suitability criteria (O'Sullivan and Unwin 2010).
2.3.2. Weighted Overlay Analysis
Weighted overlay has been used successfully to support decision making in suitability
modeling for the placement of renewable energy resources. The primary purposed of weighting
is to leverage the relative influence of individual criteria against each other. Applications for
these techniques have been applied in circumstances where many criteria are to be considered.
This becomes especially critical when considering a hierarchy of importance with respect to
criteria.
Aydin et al. (2013) used a weighted overlay to examine placement for hybrid renewable
energy systems. Their research produced a complex configuration of environmental variables
pertaining to both wind and solar. The overlay process accounted for the variables’ influence on
the generation potential for each energy type before combining the criteria to establish suitability
for the hybrid system.
Weighted overlay uses the same techniques as the Boolean method; however, this data-
driven approach assigns relative importance to each component considered in the overlay
process. This process is referred to as indexing (O'Sullivan and Unwin 2010). This provides a
better approach when looking at criteria that include conflicting attributes or objectives (Carver
1991).
The process by which multiple types of geospatial data are brought together in support of
decision making is called the multi-criteria evaluation method. While MCE is useful for
combining multiple datasets in suitability analysis, the process is complicated by not only the
16

selection of the component feature set but also by how the criteria are weighted in the overlay
process. Thus, the most critical component in weighted overlay analysis, also called weighted
linear combination, is the determination of layer weights prior to the overlay process.
Weights represent the relative perceived importance of the components of the overlay
(Carver 1991). The decisions regarding component weight in the overlay process are critical to
the proper use of the method and are often misapplied; this is called the multicriteria decision-
making problem. The relative importance of criteria should be founded in sound data-driven
reasoning, not ad hoc estimation (Malczewski 2000). The methods for determining component
weights should be determined through a process of data interrogation and evaluation. This
process typically employs a process to determine a hierarchy of importance within the criteria
considered.
Weighted overlay in GIS can be accomplished in multiple ways. Overlay tools provided
by Esri include Zonal statistics, combine, weighted overlay, and weighted sum. Table 2 provides
a brief description of each tool.
Table 2 – Esri Raster Overlay Tool Summary (Esri 2016c)
Tool Purpose
Tool Summarizes values in a raster layer by zones (categories) in another layer—
for example, calculate the mean elevation for each vegetation category.
Zonal Statistics Assigns a value to each cell in the output layer based on unique combinations
of values from several input layers.
Combine Automates the raster overlay process and lets you assign weights to each
layer before adding (you can also specify equal influence to create an
unweighted overlay).
Weighted Overlay Overlays several rasters, multiplying each by their given weight and
summing them together.

17

2.3.3. Fuzzy Overlay
Many objects have hard physical boundaries. For example, a building site has definite
boundaries, and an electrical transmission line has a precise linear pathway. This defined
boundaries approach is apparent in the Boolean overlay example presented above by Sparkes and
Kidner (1996) where they could not consider areas within the buffer distance of the criteria
selected. On one side of the buffer, the area is suitable, while on the other, it is not. The boundary
is, therefore, defined as sharp.
However, other phenomena have attributes and extents that vary with respect to location
and definition. The term for these conditions is Fuzzy. Fuzziness is a way of representing a
gradational property of spatial phenomena, such as soil, which varies across its boundaries,
transitioning from one type or category to another over a distance. In other words, the boundaries
between values are not defined by a definite border (O'Sullivan and Unwin 2010).
In overlay analysis, fuzzy datasets provide criteria that have suitability values that change
gradually with a change in location. For example, the suitability of a soil class for the
construction of a building may vary as its attribute values move towards the center of the class’
defined attribute range. Thus, as a location of interest moves away from the boundary between
soil classes, fuzziness can be stated as a value related to the strength of membership in a
suitability set related to distance from the boundary.
The fuzzy membership values can be determined by a membership function that
examines the range of values in the fuzzy dataset and assigns a fuzzy value ranging from zero to
one, where one has the highest membership. Common types of membership functions can assign
high membership to low values, high values, or values centered around an ideal value. Esri’s
fuzzy membership tool contains a variety of membership functions for assigning membership
18

(Esri 2016a). Resulting raster datasets containing the fuzzy membership data are referred to as
fuzzy set datasets. Figure 4 shows membership plots for these functions.

Figure 4 - Esri Fuzzy Membership Function Plots (Esri 2016a)
The fuzzy overlay is the process of combining the individual fuzzy sets based on a
predefined method. As with the creation of the fuzzy sets, there are multiple ways to combine
data using the fuzzy overlay tool. Table 3 identifies the methods and their applicability in the
Esri fuzzy overlay tool (Esri 2016b).
Table 3 – Fuzzy Overlay Functions (Esri 2016b)
Type Function Uses
Fuzzy And Returns the minimum
value of the sets
considered
Useful for determining the least common
denominator for suitability criteria
Fuzzy Or Returns the maximum
value of the sets
considered
Useful for identifying the highest membership
value of the input criteria
19

Type Function Uses
Fuzzy Product Returns the product of
the values considered
Not often used, this value will be lower than all of
the individual input criteria, as each variable will be
a fraction of the full membership (1)
Fuzzy Sum Returns the sum of the
values considered
Not often used, this value will be a linear
combination of all of the input criteria. A sum of
the inputs of a fuzzy overlay will not necessarily
produce a dataset where the highest membership
values are the most universally suitable.
Fuzzy Gamma Returns an algebraic
product of the fuzzy
sum and the fuzzy
product, both raised to
the power of gamma
This can be used to produce a value typically
intermediate to that of the product and sum
overlays. However, this is a compromise function
between those two approaches.
2.4. Applications of Overlay Analysis in Renewable Energy Siting
According to the United States Energy Information Administration (EIA), renewable
energy production accounted for approximately 11.5 % of the total energy consumed in the U.S.
Of that 11.5 %, solar (0.95 %), wind (2.5 %), and hydroelectricity (2.7%) make of more than
half (6.5%) (United States Energy Information Administration 2019). The remainder of the
renewable market is divided between biomass energy (5.1%) and geothermal (0.2%).
Of the primary renewable energy sources currently utilized for production, solar, wind,
and hydroelectricity have the most direct application scenarios for Site suitability analyses. Each
of these types of energy infrastructure has their own unique circumstances to which suitability
modeling can be approached and, in every case, multiple criteria need to be evaluated in order to
develop a functional operation successfully.
Solar projects on a commercial scale have enormous structures that often cover huge
tracts of land. MCE has been shown to be an increasingly crucial component of site suitability
analyses for siting solar projects. However, as previously mentioned, the relative importance of
each criterion considered in the analysis is crucial to successful implementation. Uyan applied
20

the analytical hierarchy process (AHP) to help determines the relative importance of each
component used in the study (Uyan 2013).
The study used five criteria in the suitability analysis. These criteria included the distance
from residential areas, land use, distance from roads, slope, and distance to transmission lines.
Additionally, binary constrains as buffers were placed around residential areas, roads, hydrologic
features, and environmentally protected areas. An essential consideration in this study is that
because of the uniform distribution of solar energy potential in the study area; it was not
considered as an important criterion in the suitability analysis.
The criteria were processed using AHP, a mathematical method used in multicriteria
decision making. In the AHP analysis, each criterion is compared pairwise with all other criteria,
eventually determining the influence on the outcome, to which each criterion contributes by
assigning an unbiased weight to each criterion. Table 4 presents a modified list of criteria and
weights based on the AHP analysis from this study.
Table 4 – Modified AHP results (Uyan 2013, 11-17)
Criteria Weight
Distance to Residential 0.14
Land Use 0.41
Distance to Roads 0.03
Slope 0.08
Distance to Transmission Lines 0.34

These results indicate that of the five criteria used in the study, weights were distributed
asymmetrically. Weights show that the most influential component of the analysis was land use,
then the distance to transmission lines. The remainder of the criteria accounted for roughly 25%
of the remaining influence, with the distance to roads being weighted at just 3% (Uyan 2013).
21

Aydin applies a similar approach to suitability modeling for wind generating facilities but
employs the use of fuzzy logic criteria. First and foremost, wind production must be an
evaluation of wind energy potential, which was developed as a fuzzy set as wind energy varies
across space. The second set of criteria were developed using clearly defined environmental
objectives. These objectives included distance limitation s from nature reserves, residential areas,
and airports, in addition to habitat concerns (Aydin, Kentel, and Duzgun 2010).
Wind energy was quantified into a fuzzy dataset which needed only modification into a
fuzzy set for the purposes of integration into the model, where the higher potential for wind
resulted in higher membership values. The remainder of the objective environmental datasets had
to be transformed into fuzzy sets based on membership functions. In each case, a linear
membership function was used to assign membership, such that criteria showed zero
membership until the value was at least half of the distance limit for that particular criteria.
Thereafter, a linear membership was assigned until reaching the distance limit, where all values
exceeding that limit were granted full membership (Aydin, Kentel, and Duzgun 2010).
Through a process of multicriteria decision making for analysis of the environmental
variables, two different outcomes resulted. One operation produced a worst-case scenario, and
the other produced the best-case. This was accomplished by using different fuzzy overlay
operators. Fuzzy “and” resulted in a scenario where locations were ranked by the lowest valued
objective. Fuzzy “or” results in locations ranked by their highest value objective.
Each of the two environmental impact scenarios was then combined with the wind energy
potential membership set to produce two, scenario-based final evaluations of suitability. The
final step in the study was to validate the model determined suitability values by assessing the
values at the locations of existing wind farms in the region. Aydin et al. concluded their model
22

successfully uses the presented criteria to estimate the locations with high potential for use as
locations for generating wind energy (Aydin, Kentel, and Duzgun 2010).
Suitability modeling can be used for identification of areas suitable for placement of new
energy infrastructure, but it can also be used to evaluate existing infrastructure. In the research
presented by Fitzgerald, GIS is used to explore the potential for the conversion of traditional
unidirectional hydroelectric facilities into pump storage hydroelectric generating facilities
(Fitzgerald et al. 2012).
The criteria considered for this study focused on topographic analysis surrounding
existing dams and focused on the potential for placement of a lower reservoir within 5 kilometers
of the dam in an area meeting the design criteria. Lower reservoir location was selected based on
average slope and elevation over the area. Additionally, there needed to be at least 150 m of head
between the dam and the selected lower reservoir location. The parameters and constraints used
in this model are described in Table 5.
Table 5 – Suitability criteria used to determine pump storage conversion potential for existing
hydroelectric facilities.
Transformation, Topography & Physical Characteristics
Minimum volume of existing reservoir 1,000,000 m³
Maximum distance between existing reservoir and potential
lower reservoir site
5 km
Minimum head 150 m
Maximum slope of second reservoir area 5°
Assumed minimum of new, second reservoir surface area 70,000 m²
Minimum distance from new reservoir to inhabited sites 500 m
Minimum distance from new reservoir to existing
transportation infrastructure
200m
Minimum distance from new reservoir to an UNESCO site
5 km
5 km

23

The model outputs identify all of the reservoirs where an area exists that meets the
criteria set forth in Table 4 and identifies an areal footprint for the lower reservoir. The research
shows potential for the use of GIS in suitability analysis surrounding preexisting hydroelectric
generation facilities and the conversion to pump storage capabilities. However, the authors
acknowledge that the limitations to the construction of large reservoirs in the environment
remain a challenge.
24

Chapter 3 Modeling Framework
This chapter describes the workflow of the model developed in this study. This chapter also
explores the conceptual model used to develop the model and discusses the modeling context.
Both of these elements are integral to model development and understanding.
3.1. Model Objective
The model developed for this study was designed to assist in identifying suitable
locations for MPSHS. At its core, the model was expected to find a potential location for the
placement of paired upper and lower reservoir components of the MPSHS given parameters
provided by project development engineers in conjunction with other spatial components and
optional pathways for further refinement of location selection. This model is intended to be used
as a core component of a broader site selection process to be expanded upon and field verified by
project engineers. It is designed to use freely and easily obtainable spatial data with coverage
spanning the continental United States.
3.2. Engineering Requirements
Based on the fundamental principles of pump storage hydroelectricity systems previously
described, the MPSHS has basic physical requirements for the placement of its primary
components, the upper and lower reservoirs. The foundational spatial relationship between these
components controls the potential energy stored in the system and is defined by the change in
head over distance. While conceptual designs are available, for the purposes of this study, the
system parameters used in the model are hypothetical and do not reflect specific construction
requirements, acknowledging only that these parameters exist as variables. However, the values
25

used throughout, are based loosely on the conceptual design parameters. Thus, the two core
requirements that the primary model uses for the demonstration implementation are (Figure 5):
1. Relief or the change in elevation between the upper and lower reservoirs must be
greater than 300 meters; and,
2. The lateral distance between the upper and lower reservoirs cannot exceed 1,500
meters.

Figure 5 - A simplified diagram showing the engineering parameters for relief and maximum
lateral runout used in the demonstration of the Primary Model.

For the system to function, the reservoirs must be large enough to contain sufficient water
to support the sustained operation of the hydroelectric generator. This implies that a specific
areal footprint is necessary to construct tanks of the required size. For the purposes of the
demonstration of this model, the area available for construction of each tank pad must be at least
8,100 square meters (90 meters by 90 meters). Additionally, for construction to be viable based
26

on slope stability and earthwork considerations, the area to be selected for construction of the
tank pad cannot exceed a slope angle of 15°.
The restriction on slope is governed by a feasibility problem with respect to both
geotechnical limitations and cost of construction. If it is assumed that the reservoir tank must be
placed on level ground, then any candidate site must be leveled prior to construction. This
process requires soil to be moved from one part of the site to the other. Leveling a site could be
accomplished by cutting a portion of the upslope soil material and placing it on the downslope
side (Figure 6). Thus, the slope angle has a direct relationship to cut and fill volumes and
downslope stability concerns (Connolly, MacLaughlin, and Leahy 2009).
27

Figure 6 – The reservoir site leveling scale concept
Figure 7 demonstrates the relationship between slope angle and the volume of soil needed
to be moved to level a 900m² area (30m by 30m), which is approximately 1/9
th
of the modeled
construction area, or one cell in the suggested resolution for model analysis. When these volumes
are applied to an entire 8,100 area, volumes of soil that may require excavation, transportation,
28

and re-compaction on steeper slopes will become prohibitively expensive. Thus, the
demonstration model’s slope limitation of 15° is deemed an appropriate starting point.

Figure 7 - The relationship between slope angle and cut/ fill volume needed to create a flat
construction surface.
3.3. The Primary and Secondary Models
This model was designed to determine potential areas suitable for the construction of
terminal reservoirs in the MPSHS. In the Primary Model, suitable locations for upper reservoirs
that, within the model search area, have an associated area suitable for the construction of a
lower reservoir are identified. Figure 8 depicts the major functional components of the Primary
Model.
0
500
1000
1500
2000
2500
3000
3500
4000
0° 5° 10° 15° 20° 25° 30° 35° 40° 45° 50°
Cut/ Fill Soil Volume (m³)
Pre-construction Slope Angle
Cut/ Fill V olume to Level 900 m² Area
29

Figure 8 – A graphical depiction of the conceptual Primary Model

The core dataset required for the analysis is a Digital Elevation Model (DEM) dataset for
the study area. The Primary Model also accommodates two optional screening components to be
provided by the model end-user. These include 1) a binary screening raster layer identifying
parts of the landscape that are unsuitable for construction of reservoirs (e.g., built-up areas or
30

national parks) and 2) a vector line dataset representing crossover restricted lines (e.g., major
highways or rivers) that connections cannot cross (called “Restricted Lines” in Figure 8).
The Secondary Model uses a dataset derived using fuzzy logic that is applied to the
Primary Model outputs to determine the best candidate sites for deployment of MPSHS. The
Fuzzy Layer is composed of end-user-provided data, which is combined by application of a fuzzy
overlay then converted into points with attributes noting their suitability relative to their fuzzy
membership (Figure 9).

Figure 9 - A graphical depiction of the conceptual Secondary Model
3.4. Modeling Products
The purpose of the Primary Model is to find locations that meet the siting requirements
identified by project engineers for the placement of the upper and lower reservoirs of the system.
Each cell in the Primary Model output raster datasets represents the center of an area suitable for
either upper or lower reservoir construction. Additionally, each cell in the output raster identified
as a suitable location for the construction of either an upper or lower reservoir is within the
required proximal distance of its companion reservoir location. The model also identifies
31

connections such that each upper reservoir location is paired with all of the possible lower
reservoir locations within the designated search area, a one-to-many scenario.
Finally, application of the Secondary Model assesses further the suitability of both the
upper and lower reservoir locations based on end-user-defined criteria. This process is
accomplished by application of a fuzzy overlay and returns vector points for both upper and
lower reservoir locations that are assigned attribute data regarding their aggregate suitability
relative to the individual components of the fuzzy dataset. The conversion to vector points was
found to facilitate better data interrogation when examining final results at the single system
level.
3.5. Modeling Context
This model is intended to be used as a core component of a broader site selection process.
This section discusses the context of the model.
3.5.1. Modeling Environment
This model was built using Esri GIS software. Working with the Esri ArcGIS Pro
software suite, analysis tools offered within the spatial analysis and spatial statistics extensions
were utilized for the terrain analysis and data processing portions of this research. While these
methods are not unique to Esri GIS software, the model developed for this thesis was created
using Esri’s GIS environment and their ModelBuilder application. ModelBuilder is an
application developed to create and manage models as workflow routines. These link a series of
geoprocessing tools together using a visual programing language (Esri 2016g). The choice to
work exclusively within the ArcGIS Pro environment was due mainly to its ubiquity in industry
and ease of use with respect to both model construction and distribution.
32

3.5.2. Spatial and Temporal Scale
The spatial resolution for this analysis was determined based on commonly available
DEM data, which has a spatial resolution of 30 meters. The Primary Model uses DEM data
easily obtained from a source such as the USGS 3DEP which has a vertical accuracy of 3.04
meters at a 95% confidence interval (Gesch et al. 2014). At this scale, regional terrain
characteristics relevant to this model are accurately captured. Small scale geomorphic processes
that may operate over short time-scales are largely masked at this coarser scale, but the temporal
currency of the DEM is not critical.
Additional data selected by the end-user for inclusion in the model for either the binary
screening layer or the restricted lines dataset could potentially have a significant effect on the
end result. Therefore, careful consideration of their temporal currency and spatial accuracy
should be made when selecting these datasets. For example, if the lands contained within the
boundary of a National Park were to be considered unsuitable for deployment, it is vital to
confirm that the input datasets have spatial accuracy appropriate for use with a 30m DEM and
that the data are current. Using Death Valley National Park as an example, in 2019, the park
expanded by 90,000 acres. Working today with an older boundary layer would have a significant
impact on the size of the area to be evaluated for suitable MPSHS sites.
This model was developed to operate at the scale of an average county in California,
approximately . During the development process, it was found that at this scale, the model
performed nominally with respect both to the quality of the model outputs and the processing
resources required to run the model. Designing the model for use at a county extent also
determines the extent of the model outputs. By adopting this extent as the unit of analysis, the
spatial datasets and spatial extent can be standardized, allowing the model to be broadly
applicable. Furthermore, the county is often the extent of existing political and administrative
33

constraints on the implementation of MPSHS technology. These political and administrative
constraints could range from land use permitting to construction regulations. Keeping the
MPSHS within a single county jurisdiction may also cut down on costs associated with long term
operation of the facility.
However, while his model is designed to be implemented at the county scale, there is no
built-in limitation for the size or shape of the modeled area. Average county sizes vary
nationally. In California, for instance, counties range in area from 121 square kilometers to
51,948 square kilometers. In areas where counties are smaller and denser, such as those in the
eastern United States, multiple counties could be considered for analysis by merely merging
county-level datasets.

34

Chapter 4 Model Description
This section describes the model developed in this study. Principal model components are
separated into groups representing the main functions of the model. While the model was
constructed within the Esri environment, the terms used below describe the GIS processing steps
underlying the packaged tools provided by Esri. These sections describe the major steps in the
model to generate the final product.
This model uses Esri ArcGIS Pro software and the Esri ModelBuilder application to
create two packaged models. The Primary Model outputs the set of locations suitable for
reservoir construction and the set of lines between paired upper and lower reservoirs. The
Secondary Model combines the Primary Model outputs with a fuzzy joint membership function
layer to further refine the output by identifying the most suitable candidate sites. Thirty-four
individual processes are combined to form the Primary Model, and four are used in the
Secondary Model.
4.1. Data Requirements
This model was developed to utilize data that are widely and freely available throughout
the continental United States in order to facilitate its widespread deployment.
4.1.1. Primary Model Data Requirements
The Primary Model examines the terrain for the placement of the system. The following
sections describe the necessary data and the related variables used as model inputs required to
generate model parameters.
35

4.1.1.1. Study Boundary
The study area boundary defines the lateral limits of the area to be examined. As
discussed in the previous chapter, the model was developed to work at a county scale. Study
boundary data used in the model will typically be vector files of county boundaries obtained
from a government source. For use in the model, this data should be projected with as little areal
distortion as possible. This requires the use of an equal area type projection such as an Albers
projection. Once a suitable projected Coordinate System is selected, all spatial data should be re-
projected into that system and registered against the root DEM dataset.
4.1.1.2. DEM
The base dataset of the Primary Model is a DEM. To ensure that the model has
applicability over as wide an area as possible, it was designed to utilize the United States
Geological Survey’s 3D Elevation Program (3DEP) 1/3 arc-second (10 meters) to 1 arc second
(30 meters) elevation products. The 3DEP data provided at this scale is a seamless DEM dataset
with full coverage of the continental US. The 3DEP products are distributed by the USGS in 1-
degree panels and are seamlessly combinable to cover large areas comprised of multiple panels
(USGS 2019).
For integration into the model, the DEM panels must be combined to create a single
DEM dataset with coverage over the entire study area. This is accomplished by generating a
mosaic dataset from multiple panels. Once combined, the dataset must be projected into the
designated projected coordinate system for the model and converted into the designated project
resolution of 30 meters by 30 meters using a bilinear interpolation for resampling. The elevations
in the new raster DEM dataset should be presented in meters relative to mean sea level. Finally,
the DEM is cropped to the lateral limits of the study area to eliminate unnecessary processing.
36

It is important to note that the DEM layer selected for the Study Area DEM resolution and
position sets the “Model spatial framework” that everything else must be registered to.
4.1.1.3. Binary Screen Components
The Binary Screening Raster is an optional component of the Primary Model. It is
included to provide the end-user with an opportunity to eliminate areas predetermined to be
unavailable for construction of the MPSHS. This data can include administrative and other areas
that will not be considered in the analysis. By default, the model incorporates a blank screening
layer.
The Binary Screening Raster must be a raster dataset in the same extent and spatial
resolution as the input DEM. For areas that are not to be considered in the suitability analysis,
cell values should equal zero; for areas considered, cell values should equal one. If the end-user
does not want to incorporate a binary screening dataset into the workflow, a constant value raster
(all values = 1) can be substituted.
4.1.1.4. Restricted Crossings
The Primary Model provides a method for recognizing linear features that exist in the
study area that connections (i.e., penstocks) cannot cross. This type of data would be presented
by a vector line dataset and could represent streams, utility corridors, political boundaries, or
other barriers that cannot be crossed. In the Primary Model, this feature dataset is referred to as
Restricted Lines.
The Restricted Lines dataset is to be provided by the end-user and is to be composed of a
single vector line dataset. In this dataset, each line represents a feature that a connection cannot
cross when connecting the two terminal reservoirs. The dataset can be comprised of any number
of different linear components or phenomena.
37

4.1.2. Secondary Model Data Requirement
The Fuzzy Layer is the input dataset that is applied to the Primary Model outputs by the
Secondary Model. While the data used in the binary screening layer in the Primary Model
eliminate certain areas from consideration as reservoir sites, the fuzzy layer is comprised of
spatial variables that have varying impact on the suitability of locations for the placement of the
MPSHS components. For example, if the proximity to a roadway improves the suitability of a
potential reservoir location, a fuzzy membership layer could be created such that cells closer to
roadways have a higher fuzzy membership than those further away. This dataset could be one
dataset or a combination of several spatial datasets in which phenomena have been given a fuzzy
membership value and combined into a single fuzzy set for incorporation into the Secondary
Model through a joint membership function such as a fuzzy overlay. The cell values of the fuzzy
dataset will be closer to one for those areas that have high suitability for reservoir development,
and less suitable areas will have cell values approaching zero.
As with the binary dataset, the fuzzy dataset must be in the same spatial extent and spatial
resolution as the input DEM.
4.2. Primary Model Processes
The Primary Model performs the terrain analysis that examines the selected Study Area
DEM for locations suitable for the construction of MPSHS. As noted in the chapter on model
design, the components of the model answer four basic questions 1) where is the terrain suitable
for construction of a reservoir? 2) where are suitable locations for the upper reservoirs? 3) where
are suitable locations for lower reservoirs? and 4) which upper locations are paired with which
lower locations?
38

Both the Primary and Secondary Models were constructed in Esri ArcGIS Pro using the
ModelBuilder application. The Primary Model is comprised of a series of component processes
and produces three final outputs. The only required input for the model is a 30-meter DEM in an
appropriate coordinate system with elevations provided in meters. Optional inclusions into the
model are a Binary Screening Layer and a Restricted Lines dataset.
The following sections provide a detailed look at the major steps in the Primary Model. A
complete presentation of Primary Model processes along with related inputs, outputs, tools, and
parameters is presented as Appendix A.
4.2.1. Construction Area Analysis
The construction area requirements are provided by the design engineer in the form of an
areal footprint. For example, a reservoir of appropriate size may be a cylindrical reservoir 70
meters in diameter. Application of a 10m radial buffer for necessary equipment dictates that the
minimum size requirement for the area of construction must contain a circle of at least 90m. At
the spatial resolution of 30m, the area required to support the construction of the design reservoir
is a nine-cell Moore neighborhood, with a square footprint measuring approximately 90 meters
by 90 meters, or 3x3 30m cells.
Additional constraints on construction dictate the maximum slope angle that can be
considered suitable for construction as provided by the design engineer. As a model default, 15°
is used as the slope angle limit. This means that for a location to be considered suitable, the slope
between the center focal cell and the center of the eight surrounding cells cannot exceed the
slope angle limit. To allow for this model to be used in varying scenarios, the search area (size
and shape of the neighborhood) and the maximum slope criteria are model parameters that can
be adjusted to suit the user's needs.
39

This analysis is completed in a three-step process, shown in Figure 10, and is based on
the slope in the study area. First, the slope is calculated from the Study Area DEM using the
slope tool. This tool identifies the maximum difference in elevation value between each cell and
each of its eight neighbors (Esri 2016d). The tool produces a raster dataset (Study Area Slope
Raster) where each cell value represents the maximum slope in degrees at each cell in the
dataset.

Figure 10 - The portion of the model flow chart showing the steps to determine areas suitable for
construction.
Next, using the Study Area Slope Raster derived from the input Study Area DEM, a
moving window analysis is performed over the entire study area using the focal statistics tool to
identify the maximum slope in the neighborhood of each cell. This is accomplished by use of the
focal statistics tool. Parameters for this stage in the model include the radius of the search area
for the moving window analysis and the shape of the search area. The search area is defined as
the radius in cells around the focal cell. For example, given the 8,100 square meter footprint
demonstration requirement (3x3 30m cells), for each cell in the grid, by default, this tool looks at
values of the focal cell and the eight neighbor cells (a two-cell radius) and assigns the maximum
observed slope angle to the focal cell in the output raster dataset. Values in the newly created
Max Slope Raster represent the maximum slope within the search neighborhood (equivalent to
the construction area footprint) for each focal cell in the study area.
40

The maximum neighborhood slope raster is then reclassified into a binary raster dataset
where focal cells with all neighbors not exceeding the Slope Limit are assigned a value of 1, and
those with exceedances are assigned a value of 0 (Figure 11). The Slope Limit is a key parameter
in the model, and the End-User is given the ability to modify this parameter. The resultant
Suitable Area Raster identifies those cells within the study area that are the center of a
neighborhood where the slope and areal extent is suitable for the construction of reservoir
components. This estimation is based solely on the local topography expressed by the DEM.

Figure 11 – A graphical depiction of the slope processes of construction area analysis
4.2.2. Refining the Search Area
Upon completion of the construction area analysis, the optional Binary Screening Raster
is applied. The Binary Screening Raster is provided by the user and identifies areas that are
excluded from the area to be considered for development. These areas are represented in the
user-provided binary raster as zero values. Areas that are to be considered should have a value of
one.
Application of the Binary Screening Raster to the Suitable Area Raster is accomplished
by multiplying the construction area raster by the binary screen. The net effect eliminates from
consideration areas that were previously identified as suitable by the Construction Area Focal
Analysis, but which should not be included in the final set of suitable sites.
41

The product of this stage of the model is the Construction Area Raster, where areas
comprised of grid cells fit for the construction of the technology are characterized by a value of 1
and those that do not have a value of 0. Using the raster to vector conversion tool, the model
then converts the Construction Area Raster to a vector polygon layer, which is then reduced to
only those areas deemed suitable through the slope analysis by using the select tool to create a
new vector dataset, Construction Area Polygons, which are comprised of only areas deemed
suitable for construction of the terminal reservoirs (Figure 12).

Figure 12 – Modeling process for construction area identification.
4.2.3. Extracting a subset of the DEM
To reduce processing time, the Construction Area Polygons dataset is used as a mask
feature to extract from the Study Area DEM a raster layer whose data includes elevation for only
those areas suitable for construction. The result is the Study Area DEM Subset, as shown in
Figure 13. Additionally, the polygons representing areas suitable for construction are exported.
While not a critical component of the model outputs, the Final Suitable Areas Polygons provide
the end-user with a graphical representation of the areas considered that could be useful in
presentation of the model outputs or independent validation.
42

Figure 13 - The portion of the model flow chart showing the steps to create a new subset DEM
dataset.
4.2.4. Relief Calculations
Relief calculations are the critical engineering component of the suitability model. As
shown in Figure 14, parallel modeling processes identify locations where the downslope and
upslope relief within the given search distance meet the system design requirement. The net
result is two vector point datasets, Upper Locations as Points and Lower Locations as Points,
where each point represents the center of a 90m by 90m square that is either a potential upper or
lower reservoir location as determined by the maximum change in elevation within the
maximum connection distance identified by the Relief Search Radius Parameter.
43

Figure 14 – Model processes for upslope and downslope relief calculations.
For upper reservoir citing, calculations are applied to only the Study Area DEM Subset
extracted by the previous step. A moving window analysis is applied to each focal cell in the
Study Area DEM Subset. Each cell within the 1,500m relief search radius around the focal cell is
examined to determine the minimum elevation therein. Finally, the value of the elevation of the
lowest cell within each cell’s search area is assigned to the focal cell, creating the Maximum
Elevation in Focal Search Raster. The model steps for identifying potential lower reservoir
locations from amongst the suitable building sites is accomplished the same way by modifying
the search criteria to identify the maximum elevation within the search area.

To identify the maximum potential relief for each cell in an area suitable area for
construction of the system, maximum upslope relief is calculated by application of raster math,
where, for each cell, the Maximum Elevation In Focal Search Raster is subtracted from the
elevation of the same cell in the Study Area DEM Subset, producing a new raster dataset where
44

each cell contains a value indicating the maximum downslope relief within the search area. This
process is mirrored and repeated to generate the raster dataset for maximum Downhill relief.
4.2.5. Filtering Suitable Locations
Following the relief calculations, the model now has two raster datasets that demonstrate
potential relief within the study area for locations suitable for construction. This dataset must
then be filtered to eliminate those cells in which the maximum (upslope or downslope) potential
relief does not meet the design requirements for the system. Filtering is accomplished by
reclassifying both the upslope and downslope maximum relief raster such that cells in which the
value meets or exceeds the system design specification for relief are given a value of 1 and cells
that do not meet the requirement are reclassified as NODATA. Using the NODATA eliminates
irrelevant results from each dataset as they will not be needed in later calculations.
Each raster is then converted into a vector points dataset such that each cell that meets the
relief requirements and represents the centroid of an area suitable for construction is represented
by a single point. Again, two datasets are created identifying potential reservoir locations as
points, and these are referred to in the model as the Upper Location Points and Lower Location
Points (Figure 14). The purpose of this step is to provide two vector point datasets representing
potential reservoir locations to serve as inputs into the near analysis.
4.2.6. Matching Points
The next stage in the model works to identify and pair potential upper reservoir sites to
their respective lower reservoir sites (Figure 15). The previous model components have produced
two sets of points, the Upper Location Points and Lower Location Points, that have undefined
spatial relationships to each other. The relief calculation and associated processes have created
45

these datasets independently, and it is necessary that each potential upper reservoir point is
paired with at least one potential lower reservoir point.
To identify these one-to-many relationships, a table is generated, using the Near tool, that
identifies all of the lower reservoir points within the previously established Relief Search Radius
for each upper reservoir point. This table has two fields for upper reservoir location coordinate
pairs (X and Y) and two fields for lower reservoir location coordinate pairs. Each row in the
table represents a unique connection between an upper reservoir point and a lower reservoir point
that fall within the lateral search radius of each other.

Figure 15 - The portion of the model flow chart showing the steps to pair upper reservoir
location with lower reservoir locations
Using the relationship table generated by the near analysis, lines are generated using a
tool that creates a straight line from each upper reservoir location to all of the paired lower
reservoir locations within the search area. The resulting output is a vector line dataset
(Connection Lines) representing all of these connections.
4.2.7. Filtering Connections
At this stage in the workflow, the one-to-many connections created by pairing all of the
upper reservoir locations to their respective lower locations has two significant complications.
The first problem is that there are likely linear features such as roads or streams that a system
46

connection cannot cross. The second problem is that not all lower reservoir points located within
designated search areas are from the upper reservoir locations that genuinely meet the
requirement because the near analysis discussed above indiscriminately identifies all target
points within the search radius as a match. In other words, just because a potential lower
reservoir is within the search radius of a potential upper reservoir, that pair may not meet the
relief requirement.
4.2.7.1. Filtering Restricted Line Crossings
Using the Restricted Lines dataset, the model identifies reservoir connections that
intersect barriers to connection. The selected connections are then removed from the connection
dataset by selecting the inverse of the connections that cross Restricted Lines, then copies those
features to a new dataset. The result is a dataset is comprised of only Connections That Do Not
cross Restricted Lines (Figure 16).

Figure 16 - The portion of the model flow chart showing the steps to remove connections
between the upper reservoir and lower reservoir locations that cross named streams.
4.2.7.2. Filtering False Match Connections
The processes creating the connection and removing those that cross Restricted Lines still
leaves the second problem to solve. Application of the Near analysis has identified lines that
47

connect each of the upper reservoirs to all lower reservoir locations within the Relief Search
Area distance; false match connections are made. Figure 17 demonstrates this concept.

Figure 17 - A graphic depiction of conditions identified as false and real matched connections
where relief must be at least 300 m.
False match connections occur when a potential location for a lower reservoir location is
within the near search for an upper reservoir location, but that connection lacks the relief
required by the model design parameters. To correct this problem, a model component was
developed to identify and eliminate these false match connections (Figure 18).
48

Figure 18 - The portion of the model flow chart showing the steps to identify and eliminate false
match connections.
While not a complicated computation, there are many steps in the process. The elevation
data from suitable locations must be joined with connections. After each connection has
elevations for both terminal ends, the relief can be calculated by subtracting the two values. The
final step in this process is selecting those connection lines that meet or exceed the design relief
requirement to generate a Final Suitable Connections dataset.
4.2.8. Filtering Reservoir Locations
With the filtering of the connection lines completed, some locations previously identified
as being suitable for the placement of the upper or lower reservoir locations may no longer have
a connection to a reservoir pair. Using the reservoir locations represented as points (Lower/
Upper Reservoir Points with Elevations), the connections are used to select all of the points in
each dataset that do not intersect a remaining connection endpoint. The selected points are then
deleted from each of the point datasets. The newly created Final Upper Reservoir Locations
Points and Final Lower Reservoir Locations Points datasets are then converted into new raster
49

datasets, Final Upper Reservoir Locations Raster, and Final Lower Reservoir Locations Raster,
respectively (Figure 19).

Figure 19 - The portion of the model flow chart showing the steps to eliminate upper and lower
reservoir locations that do not have a connection after the connection filtering steps.
4.2.9. Model Results
The final Primary Model outputs are the Final Lower Reservoir Location Raster, Final
Upper Reservoir Location Raster, Construction Area Polygons, and Final Reservoir
Connections. Reservoir locations are provided as raster datasets at the same resolution as the
input DEM dataset. The centroid of each cell represents the center of a 90m by 90m area with
potential for use as a modular reservoir location.
The Construction Area Polygons are simply areas where the slope limit criteria have been
met, and the construction of the modular reservoir is possible.
Final Reservoir Connections are exported separately in a parallel process as vector lines
with attributes indicating relief and distance between the paired reservoirs. Since these
connections are an integral component of the filtering of reservoir locations, they are considered
intermediate data and not preserved in the ModelBuilder workflow.
50

4.3. Secondary Model Processes
The Secondary Model builds on the outputs from the Primary Model and employs a user-
provided fuzzy dataset to analyze the suitability of the paired upper and lower reservoir dataset.
The user-provided data selected for this analysis can be comprised of many types of spatial
phenomena. Each fuzzy set represents a spatial factor that has been assigned a fuzzy membership
value. All of the separate fuzzy sets are then combined using the fuzzy overlay tool to produce
the Fuzzy Layer, used in the Secondary Model.
The Secondary Model uses the Final Model Lower Reservoir Location Raster and the
Final Model Lower Reservoir Location Raster from the Primary Model as the two primary input
raster datasets. It should be recalled that the Primary Model assigns the value of 1 to all suitable
grid cells within both the upper and lower reservoir location raster datasets, while all other cells
have a “NODATA” value. The Fuzzy Raster created by the end-user, with values ranging from 0
to 1, is the third input dataset. In parallel processes, the upper and lower reservoir raster datasets
are multiplied by the fuzzy overlay raster. The resulting intermediate datasets are comprised of
raster layers where the grid value in areas previously identified as being suitable for a reservoir
location is equal to its corresponding fuzzy membership value (Figure 20).

51

Figure 20 - The portion of the Secondary Model flow chart showing the steps that apply the
values of the fuzzy ranking dataset to the upper and lower reservoir locations.

The final step in each of the parallel processes transforms the Final Model Lower
Reservoir Location Raster and the Final Model Lower Reservoir Location Raster to points
containing the membership value corresponding with the co-located cell in the Fuzzy Layer. The
vector points now represent the centroid of an area suitable for construction of a reservoir that
meets the relief requirement when matched with a paired reservoir and contains attribute values
indicating that location’s fitness for use when considering the variables present in the fuzzy
overlay. Table 6 presents a typical attribute field layout for Secondary Model products.
Table 6 – An example portion of the attribute table for one of the Secondary Model outputs
(n=40,868)
OBJECT_ID point_id Fuzzy Membership Value
(Grid_Value)
1 1 .99554
2… 2 .75456
…40,868 40,868 .01213
52

Chapter 5 Case Study: Model Processing, Outputs, and Evaluation of Results
This chapter examines the intermediate and final outputs of the Primary and Secondary Models
applied to the county of Los Angeles. This area was chosen because of its proximity to USC and
the potential for terrain suitable for deployment of the technology.
During the model development process, several areas in California were used as test
areas. These areas included Los Angeles County, Mono County, Santa Clara County, Butte
County, and Yolo County. These areas provided a diverse cross-section of topography and
landform geomorphology. Los Angeles County, California, served as the primary study area and
the subject of the case study provided in this chapter. The remaining study areas are briefly
discussed at the end of this chapter.
This chapter takes a step by step approach to examine model processes, intermediate
data, and final products.
5.1. Preliminary Steps
Beyond data procurement, raw data must be converted into data types and formats
suitable for use in the model. This includes processing the Study Area DEM, assembling the
Binary Screening Layer and the Restricted Lines dataset.
Additionally, because the model was designed and tested using study areas located within
the state of California, the projected coordinate system used herein is the North American Datum
1983 (2011) California Teale Albers Coordinate system in meters. This coordinate system was
chosen because it is an equal-area projection with minimal areal distortion, covering the entire
state.
53

5.1.1. Processing the DEM
To serve as the principal input into the Primary Model, the DEM must be in a single
raster dataset at the correct spatial resolution and in the correct coordinate system. The DEM
created for the Los Angeles County example is comprised of four 3DEP 1-degree panels. The
3DEP panels were combined to create a single mosaic DEM dataset and converted into the North
American Datum 1983 (2011) California Teale Albers Coordinate system with horizontal units
in meters. The elevation is provided by the USGS 3DEP program in meters by default; thus, no
modification is required, as the engineering requirements and projected coordinate system use
meters as the unit of measure. The final input DEM was then extracted such that the extent of
coverage is coincident with the county boundary (Figure 21).
54

Figure 21 - The processed DEM for Los Angeles County used in the model example.
5.1.2. Binary Screen Creation
The binary screen is an optional element in the model that serves to help the end-user
eliminate areas from consideration in the early stages of the modeling process. This served two
major functions. First, eliminating areas from consideration reduces the area which needs to be
processed, in turn, reducing model run times. Second, it reduced the potential for false match
results. For example, without eliminating waterbodies from consideration, the model can identify
lakes as large flat areas suitable for placement of the lower reservoir.
55

For the Los Angeles County example, NHD waterbodies, National Parks, and State Parks
were used to create a single binary screening layer using a simple presence or absence test
(Figure 22; left). From these vector areas, a raster dataset was created in the same coordinate
system and spatial resolution (30m) as the DEM. Where an exclusion feature was present, the
raster was given a value of 0, if no exclusion feature was present, that cell was assigned a value
of 1 (Figure 22; right).

Figure 22 – Areas not for placement consideration (left) and Binary Screening Layer (right)
5.1.3. Restricted Lines
The final input into the Primary Model is the Restricted Lines dataset consisting of linear
features such as roads, streams, or utility corridors that the reservoir connections cannot cross.
For this example, a filtered version of the NHD streamline dataset was used.
The logic used to filter the stream line dataset for this case study assumed that all features
where the name field is not null are protected by the Water Quality Act based on their
prominence in the dataset. Thus, these streams cannot be crossed without submitting to the
Environmental Impact Assessment process. Smaller tributaries or drainages, while still having
56

the potential to cause problems during project development, are less likely to have a significant
impact. Figure 23 shows the named stream features for Los Angeles County.

Figure 23 – Restricted Lines dataset for Los Angeles County
5.2. Intermediate Results
The Primary Model produces many intermediate datasets that are superfluous and not
preserved in the final model outputs. However, understanding these data in the context of their
application is essential to understanding how the model functions as a whole.
57

5.2.1. Construction area identification
The first stage of the model uses the DEM covering the entire study area (Study Area
DEM) and identifies areas suitable for construction of the terminal reservoirs. The first step
produces a slope raster from the Study Area DEM for Los Angeles County, where slope angles
are represented in degrees (Figure 24).

Figure 24 - Slope in degrees for Los Angeles County
The second step performs the moving window analysis producing a new raster dataset
where each cell is given the value of the maximum slope angle in its 9 cell Moore neighborhood.
58

Each cell in the Maximum Slope Raster identifies the maximum slope in an area approximately
90m by 90m required for reservoir construction(Figure 25).

Figure 25 - The raster dataset showing the maximum slope for each nine-cell neighborhood in
the study area.
Reclassifying the Maximum Slope Raster performs a binary pass/fail test on the dataset
where cell values are modified to represent the suitability of an area for construction. In this
example, if a cell value is found to exceed the design specification of 15° for maximum slope in
59

the neighborhood, the cell is given a value of zero, and if it does not exceed that value, the cell is
reassigned a value of one (Figure 26).

Figure 26 - The areas identified by the model as being suitable for the placement of a terminal
reservoir.
The Binary Screening Raster is multiplied by the Reclassified Maximum Slope Raster
(binary), where both datasets are comprised of values of either zero or one, thereby eliminating
the areas that would be suitable for construction but fall within an area deemed not suitable by a
component of the binary screening dataset. Areas identified as suitable in both datasets are given
60

a value of 1. Figure 27 shows the effects of this overlay on the areas considered suitable for
construction of a terminal reservoir.

Figure 27 - The areas identified as suitable for construction of a terminal reservoir after the
application of the binary screening layer.
The final step in this stage is comprised of two processes. The first process converts the
areas deemed suitable for construction to vector polygons. The second process removes from the
polygon dataset those polygons representing areas not suitable for construction, resulting in a set
61

of polygons covering all of the areas suitable for constructing either the upper or lower reservoir
components. This dataset is used to extracting the Subset DEM (described below).
5.2.2. Creating the Subset DEM
The Subset DEM for Los Angeles County was used to perform the remainder of the
terrain analysis. Because only the areas identified as being suitable for construction area relevant
to the remainder of the model processes, eliminating the extraneous information from the Study
Area DEM reduced the processing load. Figure 28 shows the new Subset DEM extracted using
the areas identified as suitable for construction. The subset DEM demonstrated the drastic
reduction in areas to be considered, which eliminated most areas located in mountainous areas.
62

Figure 28 - The subset DEM created by using the areas deemed suitable for construction of a
terminal reservoir to extract elevation values from the study area DEM
5.2.3. Searching for Relief
Working with the subset DEM, relief is calculated by a moving window analysis which
looks in a 50 cell (1500m) radius from the focal cell to identify the lowest point within the search
radius. A new raster dataset is then created where the focal cell assumes the value of the lowest
point within the search radius. The same function is repeated in mirror, where the focal cell
searches the same radius for the highest point, creating another raster dataset. Each raster is then
63

compared to the subset DEM using a raster math function yielding two raster datasets that
describe the maximum relief in both the upslope and downslope directions (Figure 29).

Figure 29 - Intermediate datasets showing the maximum downhill (left) and upslope (right) relief
within 1,500m search radius.
The relief datasets were then reclassified using the design parameters for the system. In
this example, the minimum relief from the upper reservoir to the lower reservoir is 300 meters.
Thus, each of the datasets was processed to remove all cells where relief did not meet the
minimum requirement. Figure 30 shows the reclassified maximum relief layers for both
downslope (left) and upslope (right). These areas identify the preliminary reservoir locations for
both the upper and lower reservoir sites, independently.
64

Figure 30 – Preliminary upper and lower reservoir location.
5.2.4. Making Connections
The process of making connections utilizes a near analysis which applies a many to one
search, eventually producing a table matching each potential upper reservoir location to all of the
potential lower reservoir locations within its 1,500-meter search radius. For the Los Angeles
County example, the table included 3,156,665 lower reservoir location matches for 18,338 upper
reservoir locations.
First, this step eliminates those connections that cross restricted line features. In this case
study, as described above, a subset of the NHD streamlines dataset was used as the only
component of the Restricted Lines dataset. Second, the model assigns the elevations at each of
the connection’s terminal ends as attributes to each connection, then calculates relief for each
connection. Connections that do not meet the relief requirement are identified as false match
connections and eliminated from the dataset. This filtering process reduced the potential
connections in Los Angeles County from 3,156,665 to 2,207,844.
65

5.3. Final Primary Model Results
The final steps in the model use an intersect function to examine the potential tank
locations to ensure they are coincident with a previously identified as suitable is found not to be
coincident with a connection, it is eliminated from the population. The filtering of the upper and
lower reservoir locations (Figure 30) further reduces the locations identified as suitable for
reservoir placement by 5% and 8% for the upper and lower reservoir locations, respectively
(Figure 31).

Figure 31 – A summary of final Primary Model outputs for Upper and Lower Reservoir.
Locations exaggerated for visual effect
66

The Primary Model produces three final datasets covering the entire County of Los
Angeles. Figure 32 shows a close up of these results for a small part of LA County. This
perspective shows an array from a cluster of upper reservoir locations to a broader area identified
for lower reservoirs. Additionally, Figure 32 demonstrates the nature of reservoir placement
relative to the terrain.

Figure 32 – A selection of final Primary Model results, La Canada Flintridge.
67

5.4. Secondary Model
For the example used herein, a fuzzy dataset was constructed using four spatial
components. These include epicenter locations for historical earthquakes, green energy-
producing facilities, roadways, and landslide susceptibility.
Each of these datasets was converted into fuzzy data using the conditions outlined in
Table 6. Landslide susceptibility was obtained as a raster ranked from 1 to 10 where 1 is low
susceptibility, and 10 is high. It was converted into a fuzzy set by reclassifying the value of areas
less susceptible to landslides having a higher membership. Each of the vector datasets (Historical
Earthquakes, Green Energy Facilities, and Roads) was converted into a distance raster where cell
values denote the Euclidian distance from each feature.
Table 7 - Variables used to generate the fuzzy dataset used in the Secondary Model for the Los
Angeles County case study
Input Dataset Primary
Data
Type
Spatial
Scale
Fuzzy Membership Membership
Type
Midpoint
Distance to Historic
Earthquakes
Points California Higher membership with increased
distance from Earthquakes
Large 30 km
Distance to Green
Energy Production
Facilities
Points California Lower membership with increased
distance from Facilities
Small 20 km
Distance to Roads Lines California Lower membership with increased
distance from Roads
Small 1,000 m
Landslide
Susceptibility
Raster California Dataset ranked 1 to 10; fuzzy
membership ranked .1 to 1, where
higher values have a lower
membership
Small 5

The resulting Fuzzy Raster incorporates the four fuzzy sets into a single raster dataset
using a with values ranging from 0 to 1, infinitely.
Review of both the fuzzy sum and fuzzy product outputs indicated that these functions
provide the extreme ends of the membership spectrum. The fuzzy product output provides a
68

dataset where low membership dominates, while the sum function provides the opposite where
most of the study area has higher membership.
Ultimately, the overlay used in the case study was produced using a fuzzy gamma
function, where gamma was equal to 0.9. Documentation for the Fuzzy Overlay Tool provided
by Esri reports that the gamma function performs an algebraic combination of both the product
and sum fuzzy overlay functions serves as a compromise function, and therefore, may provide
adequate results for the demonstration of this model feature (Esri 2016b).
Figure 33 shows the Fuzzy Layer for Los Angeles County before integrating it with the
Primary Model results. Using the fuzzy sets considered, the overlay shows that large portions of
low-lying areas have high membership, while areas in the mountainous areas show lower
membership. Membership values for the county range from near zero to 0.99. Isolated points of
low membership in the Los Angeles basin correspond to epicenters of historical earthquakes.
69

Figure 33 - The Fuzzy Layer for Los Angeles County.

Figure 34 shows results of the Secondary Model outputs for both the upper and lower
reservoir locations in a selected area of Los Angeles County. Membership value for model
outputs indicates the fitness of that location to support the placement of the designated terminal
reservoir. In the area shown, all areas identified as suitable for upper reservoir locations have
70

high membership values for suitability. For lower reservoir location, most areas have been
identified as having low membership.

Figure 34 – A selection of final model results near La Crescenta, California, after the integration
of the fuzzy layer.
In this case study the distribution of membership for the reservoirs, both upper and lower,
occurred in a trimodal pattern, where membership for reservoir locations was clustered near zero,
near one, or near the center. The mean membership for upper reservoirs was slightly higher than
center while the mean for lower reservoir membership was slightly less than center (Figure 35).
71

Figure 35 – Histograms showing the distribution of fuzzy membership values for upper and
lower reservoir locations. The red vertical line on the histograms indicates the mean membership
value for each dataset.
5.5. Model Performance
Review of the final Los Angeles County data outputs for both the Primary and Secondary
Models shows that there are many possibilities for the deployment of MPSHS. The model
identified 12,716 locations suitable for the placement of the upper reservoir and 40,868 potential
lower reservoir locations. The disparity of these populations due primarily to the fact that in this
geomorphic setting the toe of most slopes and valley floors are geomorphically more stable and
therefore have a higher likelihood of supporting land suitable for construction of the required
infrastructure.
Geomorphic controls on terrain vary by study area location and have a significant impact
on the expression of terrain features. In California, there are many different geomorphic
72

provinces that each contain unique topographic features that this model can exploit. Using
examples from Los Angeles County, it is apparent that the model placement of lower reservoir
location tends to prefer the wide valley floors, where sediment has accumulated as alluvium. Due
to its arid nature, low precipitation and corresponding low amounts of runoff create the extensive
alluvial plains abutting steep, rugged slopes with minimal soil development (Norris and Webb
1990). This corresponds to an abundance of lower reservoir location relative to the population of
corresponding upper reservoirs.
5.5.1. Primary Model
In Los Angeles County, the Primary Model identified 2,207,884 possible connections
linking upper reservoir locations to lower reservoir locations. Relief calculated for these
connections ranged from the minimum 300m to a maximum of 768m. The mean relief for the
population of connections was 343m, indicating that the bulk of the connections identified by the
model occurred near the design limit.
Another useful indicator of viability is the relief gradient, or the lateral distance needed to
achieve the desired change in head. For the Los Angeles County example, the mean lateral
distance required was found to be 1,352m with a standard deviation of 127m. These statistics
show that placement of upper and lower reservoirs in Los Angeles County approached the
engineering limitations set forth by the system design, with 414,589 of the 2,207,884
connections existing at the lateral limit of 1,500m. In other words, the number of viable
connections increased as the lateral connection distance increased. Figure 36 demonstrates the
relationship between the number of connections meeting the relief requirement and the lateral
connection distance.
73

Figure 36 – Distribution of connection lengths

Intuitively, this relationship is logical. As the design parameters allow for increased
distances between the upper and lower reservoirs, the number of potential reservoir locations and
in turn connections will increase. In other words, as allowable lateral distances increase, so does
the number of potential reservoir locations; thus, more placement opportunities. Conversely, a
similar trend can be found for the minimum relief parameter. As the minimum required head
(relief) decreases, the number of potential reservoir locations increases (Figure 37).
74

Figure 37 – Distribution of point to point relief magnitude

This observation shows that the lower the minimum relief requirement demanded by the
MPSHS design, the higher the potential for suitable locations. The relationship between the
minimum relief required for the MPSHS to function and the number of potential locations for
placement is an essential component when considering the spatial extent for the analysis. For
example, in an area where relief gradients are less, connections may need to be longer than 1,500
meters. Thus, engineering requirements would need to be modified, and system components
reevaluated.
The model results for the upper and lower reservoirs were compared to topographic data
and features to determine the validity of selections. Figure 38 demonstrates a collection of these
observations in an area near Palmdale, California. It should be noted that each node, represents
the center of a 90m by 90m area, suitable for construction of a terminal reservoir.
75

Figure 38 – A sample of Primary Model Results for upper and Lower reservoir locations

The placement of upper reservoir locations occurs along a ridgeline where the topography
is relatively flat. Some upper reservoir locations occur on the top of the slope while others occur
on both flanks. While all of these meet the relief requirement of 300m to the toe of the slope
where the cluster of lower reservoir locations, it would likely be impractical to utilize suitable
locations on the backslope of the ridge (the side opposite the paired lower reservoir). Figure 39
demonstrates the application of the restricted line constraints as there are no lower locations
76

present on the side opposite the stream feature; in this case, the California Aqueduct.
Additionally, proximity to renewable energy is demonstrated as the photovoltaic solar panels are
shown adjacent to a cluster of suitable sites in the lower right of Figure 39.

Figure 39 – Model results near Palmdale, California.

Lower reservoir locations are determined first based on the identification of paired upper
reservoir locations through the use of the near analysis, as most study areas are dominated by flat
terrain suitable for construction of a lower reservoir. Conversely, upper reservoir locations are
77

often limited to geographically isolated areas such as ridge tops (Figure 39) or, more confining
features such as hilltops (Figure 40). Based on this relationship, clusters of lower reservoir
locations often occur 1) at the toes of the slopes on which their paired upper reservoir is located
and 2) form a fan shape with the leading edge representing the maximum lateral distance for a
viable connection.

Figure 40 – Example results from the Primary Model outputs near Universal City, Los Angeles
County, California. Note there is only one upper reservoir site indicated.
78

The modeled locations for lower reservoir connections shown in Figure 40 also
demonstrate the effectiveness of the Construction Area Focal Analysis, whereby any area with
topography not suitable for construction is eliminated from placement consideration.
The number of connections in the Los Angeles outputs is enormous, with over 2 million
connections. This large quantity of connections is due to the relationships between the upper and
lower reservoir locations, as it is only in rare circumstances where a single reservoir node exists
in isolation. Typically, reservoir locations occur in clusters and have one-to-many relationships
with other clustered reservoir pairs that meet the criteria. In this case, one upper reservoir may
serve as a match for many lower reservoir locations. Such is also true for other upper locations in
the cluster. The result is an array of connections from each upper location to each of the paired
lower reservoir locations (Figure 41).

Figure 41 – A graphical example of the one to many relationships between upper and lower
reservoir locations

The usability of the model results is complicated by the enormous dataset produced for
the connections. As described above, there are over 2 million connections for the Los Angeles
79

County study area. A closer look at the results shows how clustered results overlay one another
such that any one of the connections of the upper reservoir and paired lower connections are
difficult to discern making the practicality of this dataset questionable without an interpretation
workflow focused on refinement and filtering of the results (Figure 42). The large areas of black
shown in the figure are the overlapping connections linking all of the upper reservoir locations to
all the possible matched lower reservoir locations.

Figure 42 – Complex relationships in clustered reservoir locations
80

Selecting a viable location for the MPSHS could use a top-down process similar to the
model itself, which may be ideal in study areas where the limiting factor in suitability is upper
reservoir placement. In other words, first, find a suitable upper location from the final dataset and
then filter out all of the connections that do not originate from that point. The final selection can
then proceed from the paired lower reservoirs. As shown in Figure 43, this filtering process
would eliminate a significant portion of the possible lower reservoir locations. Of course, the
process could also be done in reverse where all of the possible upper pairs are found using a
lower reservoir location.
81

Figure 43 – Results filtered to originate from a single upper reservoir location
The filtering process can be easily implemented using the attributes of the connection
lines themselves. Each line has the feature ID for both its origin (upper locations) and its
endpoint (Lower Locations). By running a query to select all connections that are associated with
the target feature, either an upper or lower reservoir location, a simplified set of connections can
be identified.
82

Examining the results for connections further indicate that there is an unknown
component with regard to the character of the terrain between the reservoir locations. While this
is obviously a variable in the construction feasibility for construction, it is outside of the scope of
this study.
5.5.2. Secondary Model
Secondary Model results for the Los Angeles study area show strong clustering of
membership values in the fuzzy dataset when applied to the locations identified for reservoir
placement. Using the Fuzzy Layer dataset described above, the Secondary Model for Los
Angeles County resulted in reasonable results with respect to suitability. As expected, in the
greater Los Angeles area, distance to roads was not a major limiting factor within the fuzzy
dataset. Historical seismic activity expressed as isolated incidences of low membership.
Additionally, green energy production criteria graded over a long distance in its fuzzy set, so this
variable would be unlikely to affect small scale suitability (i.e., reservoir location within a
cluster). Therefore, the dominant influencer on differences in the membership of clustered
reservoir locations in the Fuzzy Layer is most likely landslide potential dataset.
Additionally, the scale of the fuzzy sets applied the membership function over an
expansive area with highly variable distributions in phenomena. For example, historic
earthquakes were used as a base dataset. Conversion into a fuzzy set was accomplished by
applying a higher membership based on distance from the epicenter.
While sound logically, the resulting fuzzy set was biased toward those parts of the state,
not in seismically active areas. Furthermore, the midpoint function was based on a static 30km
radius, decreasing from that point. This ensured that the closer to an epicenter (zero to 30km)
83

membership would be below 0.5. Beyond 30km the membership would be higher at a rate
normalized to the furthest distance in the Euclidian input distance to epicenter raster.
If the Fuzzy Layer and its parent fuzzy sets were created at the study area scale, the
outcome might have been not only different but more applicable to the area of interest. In these
cases, more study area-specific choices in variables could be considered.
Another critical element of the Secondary Model is the selection of the Fuzzy Sets
themselves. The four variables selected for use in this case study were selected for their
simplicity of presentation and processing. The parameters by which they were transformed into
fuzzy sets were equally as simple. Primarily, the functions looked at a membership based on
simple distance parameters (i.e., higher membership values for proximity to roads). This likely
contributed to the uniformity in outcome with respect to the statistical distribution of
membership values, especially in the chosen study area, where the density of the selected
variables is significantly different from the remainder of the state.
Another consideration for potential future use of the Secondary Model is the joint
membership function used in the overlay process. For this example, a gamma function (gamma =
0.9) was used to provide a result that acted as a compromise between the product and sum
functions. This used an algebraic combination of both the sum and product functions raised to
the power of gamma. As with other aspects of overlay analysis, careful consideration should be
applied to all components when selecting the parameters for the Joint Membership Function.
Although the development of the Fuzzy Overlay used in the model case study was
intended for demonstration only, it does expose some of the difficulties with respect to scale,
fuzzy criteria, and fuzzy membership function selections. The Fuzzy Layer created from the
overlay process was constructed to apply a uniform fuzzy overlay to the Los Angeles County
84

study area and each of the alternate study areas to which the model was applied during
development. This means that the fuzzy membership functions were applied uniformly to the
entire statewide datasets, rather than on a study area-specific scale. Parameters of the fuzzy
membership functions were chosen based on limited data and many assumptions. In a practical
use scenario, there should much more consideration placed on the creation of fuzzy sets and the
decisions regarding the fuzzy membership functions for the overlay analysis.
Thus, the Fuzzy Layer used in the Secondary Model case study for Los Angeles County is
a simple version of an end-user provided Fuzzy Layer and was developed to show feasibility,
rather than present defensible real-world application.
5.5.3. Alternative Study Areas
During the development process, the model was tested in multiple locations within the
State of California to ensure that it would work under various geographic conditions. These
locations were Mono County, Butte County, Santa Clara County, and Yolo County. These
alternative study areas are shown in Figure 44.
85

Figure 44 – Distribution of alternative study areas in California
The same process described above for Los Angeles County was applied to the other study
areas. To maintain consistency in results for each study area, the Binary Screening Layer,
Restricted Lines dataset, and the Fuzzy Layer were created to cover the entire state. For each
study area, the county boundary was used to extract the data to the study area limits.
Collectively, the test areas presented a diverse range of terrain types to test the Primary
Model. They span the geomorphic provinces of the Central Valley, Coast Range, Basin and
Range, Sierra Nevada, and Transverse ranges, which provide a diverse collection of landforms
86

and topography. Additionally, they cover regions from densely populated, highly developed
areas to mostly rural undeveloped counties that also presented variation in the fuzzy sets
included in the Fuzzy Layer used in the Secondary Model. For each study area, the county
boundary was used to extract the data and to define the study area limits. Table 7 presents a
summary of the results in different case study areas.
Table 8 – Summary of Primary Model Results for all Study Areas

Study Area
Elevation
Range
(Meters)
County Area
(Square
Kilometers)
Potential
Upper
Reservoir
Locations
Potential
Lower
Reservoir
Locations
Number of
Connections
Upper/
Lower
Reservoir
Ratio
Los Angeles
County
0 – 3,064 12,310 17,216 40,868 2,207,844 0.42
Santa Clara
County
0 - 1,336 3,377 0 0 0 -
Yolo County 0 - 955 2,652 1,388 1,060 75,196 1.31
Mono County
1,273 –
4,342
8,110
County datasets too large for computational resources;
over 7 million connections
Butte County 14 – 2,175 4,340 187,537 52,564 4,809,203 3.56

The results demonstrate that the model produces a large number of locations for both
upper and lower reservoir locations in each study area. A closer examination of the results
relative to the geomorphological terrain of each study area yields an insight into the suitability of
specific settings to MPSHS deployment. Additionally, these results show that the total change in
elevation within a study area is not alone, an indicator of MPSHS suitability potential.
In Santa Clara County, there were no locations that were found to be suitable for the
construction of the MPSHS. This was likely due to the low countywide relief and the
components of the Binary Screening Layer, which occluded a significant portion of the higher
elevations from considerations.
Results for Yolo County found areas suitable for MPSHS on the eastern margin of the
Coast Ranges. Yolo County occupies an area with moderate relief in the hills and canyons on its
87

western flank and flat alluvial valleys and delta lands of the Sacramento Valley. Topography in
the western portion is characterized by moderately steep slopes with peaks and ridgelines lining
leaner fault-controlled valleys. Model results indicate that the terrain is suitable for MPSHS
placement and there were around 75,000 suitable connections.
Mono County is situated on the east side of California at the margin between the Sierra
Nevada and the Basin and Range Province. The Sierra Nevada mountains are a westward
dipping fault block characterizes by high relief on its eastern boundary. Due to the massive
potential for relief meeting the criteria of the model, the intermediate datasets proved too large
for the computational resources at the scale chosen, with the near analysis producing over 7 ½
million matches. For an area of this type, it would be beneficial to refine further the Binary
Screening Layer to minimize the areas considered prior to using this model. Due to the limitation
in time, further efforts to complete the analysis in this study area were abandoned.
Butte County offers unique terrain when compared to the other test areas selected.
Located at the northern limits of the Sierra Nevada, Butte County is often referred to as the
tablelands. This is due to its large flat westward dipping buttes comprised of volcanic mudflows.
Rivers and streams have incised through the mudflow to create a topography of plateaus and
river canyons, which flow westward toward the central valley. This unique topographic profile
provided a difference in outcome for the model results. The results showed a high number of
upper reservoir locations could be placed on the tops of the laterally extensive plateaus while
limiting the number of lower reservoir locations due to the steeply sloped canyons cross-cutting
the terrain.
Overall the Primary Model provides results that are satisfactory to the goals set forth. The
model outputs identify areas likely suitable for placement of both upper and lower reservoir
88

locations that meet the engineering design specifications. Binary raster screening and restricted
line screening processes effectively eliminate from contention MPSHS configurations that
intersect features identified as not being suitable for MPSHS placement. The Secondary Model
applies the attributes of the user-provided Fuzzy Layer to the areas identified as suitable for
placement of a terminal reservoir, adding further value to the model results.
89

Chapter 6 Conclusions
This chapter provides an assessment of the success of the model developed during this study,
including the overall model outcomes, potential uses, implications of the final product, and
viability in the marketplace. This section also discusses areas for future development and other
applications where the concepts developed herein could be applied.
Overall the Primary Model was found to succeed in identifying areas likely suitable for
the construction of the MPSHS. Areas selected as suitable were found to meet the design
specifications in both having an areal footprint which meets the slope limitations supporting
construction feasibility of terminal reservoirs and meets or exceeds the relief requirements from
the upper reservoir to the lower reservoir.
6.1. Model Assessment
Examination of intermediate datasets provides an explicit acknowledgment that within
each study area, there are many locations which are suitable for placement of a reservoir, without
meeting the additional engineering criteria. Using the design footprint method of screening
proved to be the most significant topographic indicator for placement with respect to
construction feasibility. While other topographic variables such as topographic position index
(TPI) and terrain roughness were initially considered, they were found to be too limiting and
difficult to apply over a diverse range of geomorphic conditions without calibration.
The moving window analysis is used for the construction footprint exploration (see
Section 4.2.1) and both relief analyses (see Section 4.2.4). For the construction footprint
analysis, this technique is able to identify all locations within the study area that are “flat”
enough to be considered suitable over the entire design footprint. Review of intermediate
datasets that identify areas meeting the construction footprint criteria indicates, as expected that
90

most low-lying areas within the study area are identified as suitable. This is contrasted by results
from upland areas, where suitability is typically far less common. Additionally, by making the
search radius for the moving window analysis a model parameter, the end-user can modify the
required footprint for construction as well as the slope angle limits, adding situational flexibility
to the application. Applying basic concepts of geomorphology to modeled results, this method
was shown to be capable of producing acceptable results under limited review. With little effort,
the model could be modified to support the search for upper and lower reservoir locations where
the area required for construction was not identical.
Application of the moving window analysis for the relief calculations worked to identify
all grid cells in the subset DEM that met or exceeded the minimum design relief. By using the
method to identify both upslope relief and downslope relief, the model was able to identify both
the upper reservoir locations and the lower reservoir locations, independently. After
identification of the upper and lower reservoir sites, connections could be made linking upper
reservoir locations to lower reservoir locations.
Early in the model development process, it was apparent that the connections had to be
filtered to provide more viable results. The connection culling process began by identifying the
apparent flaws of connection identification by examining model results with respect to mapped
features such as connections that crossed streams due to meandering channels and topographic
features caused by fluvial morphology. Another glaring problem, as discussed in Section 4.2.7.2,
was false match connections which were found to be a byproduct of the connection process
linking upper reservoir locations to lower reservoir locations, without regard for point to point
relief, to which a solution was created and integrated into the Primary Model.
91

A fundamental component of the hydroelectric generating potential is the motive force
produced by the difference in head pressures between the upper and lower reservoirs. As this is
an essential consideration for the suitability of one reservoir pair over another, connections
linking reservoir pairs contain attribute data indicating the relief by connection allowing the user
to filter and rank the connections by relief. In conjunction with connection length, this
information can be further summarized by gradient or the ratio of the change in head over the
connection distance. While these are important considerations for the selection of MPSHS
placement, they are not incorporated into the model developed in this study as there are currently
no engineering design criteria for these variables.
While auditing model outputs for observable breaches of the conceptual model, it became
apparent water bodies such as lakes and oceans were identified as suitable for reservoir
placement. This selection was due to the nature of water bodies to appear as large flat areas. It
should be noted that in the case study discussed in Chapter 5, water bodies were included in the
binary screening dataset, not as a standard model parameter. While this condition can be seen as
a deviation from the conceptual model, case studies in the literature have shown that some viable
projects can consider natural water bodies as lower reservoir options. Recognizing this reality,
the inclusion of water bodies in the binary screening layer remains a recommendation rather than
a standard component (Bueno and Carta 2006, 312-340).
Incorporation of the binary screening option into the model workflow was added as a
method of reducing the area to be considered for placement of the system before the more
resource-heavy computations were performed. This was found to be an essential step when the
model moved from smaller study areas where terrain capable of supporting the technology was
limited to a small fraction of the overall study area like Los Angeles County and Yolo County to
92

larger, more mountainous areas like those in Mono County, located on the western boundary of
the Basin and Range geomorphic province.
Furthermore, when considering data to be included in the binary screening dataset,
potential areas to be excluded from the overall analysis were found to be highly variable and
regionally specific. For example, in Los Angeles County, the user may wish to exclude
developed areas due to the high population density while in Inyo County, lands belonging to
Death Valley National Park may be essential to include in the binary layer. The same is true for
the restricted line crossings, which will change drastically based on location and experience.
In Santa Clara County, one of the study areas considered as an example; there were no
potentially viable solutions to the analysis. This was likely due to a large amount of terrain likely
suitable being screened by the Binary Screening component and the low countywide relief.
Results for other study areas varied with respect to the ratio of upper reservoir location to
lower reservoir locations. Some underlying correlations can be made between geomorphic
expression, erosional patterns in geologically unique terrains, and total study area relief.
A clear example of terrain effects on the Primary Model outcome is in the comparison
between the results from Los Angeles County and those of Butte County. Los Angeles County is
an arid climate where the topography is dominated by the San Gabriel Mountains, a faulted and
steeply sloped terrain with little soil development creating jagged peaks and ridgelines abutting
nearly flat alluvial valleys. In contrast, the Sierra Nevada foothills of Butte County are capped by
volcanic mudflows called the tablelands, expressed as flat-topped, shallowly westward dipping
features cut by deeply incised east-west trending canyons. The topography of Butte County
provides ample room for upper reservoirs while limiting potential lower reservoir site. Los
Angeles County’s Upper Reservoir to Lower Reservoir ratio is 0.42, while Butte County's is
93

3.56. This implies that the terrain in Butte County is more suitable for upper reservoirs, while the
opposite is true for Los Angeles County.
Due to the nature of the MPSHS, which relies on gravity and hydraulic head pressures to
function, it is intuitive that the topographic character of a study area would have a dramatic
influence on the model outcomes. While a surface-level examination of the data implies that an
empirically quantifiable correlation likely exists between landform development and suitability
for MPSHS, a thorough examination was not included in this study.
The Secondary Model provided integration of a fuzzy dataset to determine further
suitability for the final Primary Model results. This provides the end-user an opportunity to
further refine the suitability analysis by applying additional variables to the locations identified
by the Primary Model. Similar to the binary screening dataset, the fuzzy dataset would likely
incorporate a diverse and regionally-specific collection of variables. In the example provided
above, four datasets with spatially variable degrees of influence are considered and are shown to
have a diverse impact on the suitability of output features. It is important to note that the data
used to develop the fuzzy dataset used in the case study was not weighted with respect to the
relative importance of each input dataset. As a result, although accounted for in the creation of
each fussy set, each fuzzy input had an equal degree of influence over the final ranking value.
The membership functions and the joint membership function of the fuzzy analysis were
found to have an enormous degree of influence on the suitability analysis when incorporated into
the project workflow. However, the application of defensible criteria selection in the fuzzy
datasets was outside of the scope of this study.
94

6.2. Uses
The model developed for this study identifies potential locations for the development of
MPSHS. Based on model parameters, reservoirs must be centered on an area 90 meters by 90
meters where the maximum slope cannot exceed the design specification. The upper reservoir
must be located in such a position that the relief to a location suitable for the placement of a
lower reservoir cannot be less than the design parameter of 300 meters. Additionally, the lateral
distance between the upper and lower reservoirs cannot exceed 1,500 meters. In order to make
the model as useful as possible with respect to the project-specific engineering constraints, the
search area parameters for the construction footprint, the lateral relief search radius and the
minimum relief requirements are modifiable by the model user.
This model produces output products to assist in solving the design problem of MPSHS
placement. Components of this model may have application in other areas of engineering and
suitability analysis. The moving window analyses used herein demonstrate the ability to
characterize large datasets into easily digestible outputs designed to answer a specific problem
based on terrain features. For example, relief, used herein as an average gradient between
upslope and downslope points, may have applications in other material transport problems such
as large conveyor design or fluid transport problems.
6.2.1. Viability in Market
The concept of MPSHS is an emerging market, and the scenario explored herein has yet
to be put into action. While the current market for this model may be limited, the potential for
this suitability modeling to be utilized as a component of a broader push toward energy
sustainability is promising. Coupled with broader political and environmental motivation, this
95

model could be used to support further consideration MPSHS as an alternative or complement to
other resource types.
6.3. Future Work
This model focused on the initial steps of suitability analysis for MPSHS. The outputs are
limited in scope to only locations where there is a potential for deployment of the technology.
Future development in this area can work toward further refinement of the output products and
refinement in suitability analysis.
While the model develops the connections between upper and lower reservoir locations,
with the exception of the consideration of restricted lines, the character of the terrain between
those connections is not considered. Project viability may be significantly impacted by the terrain
roughness or topographic barriers between paired reservoirs. For example, if there is a ridgeline
between terminal reservoirs, it may be impractical to construct the system at that location due to
construction costs required to overcome that type of topographic features. The incorporation of
further terrain analysis techniques could be applied to overcome these problems and further
refine the model products to account for topographic barriers to placement not accounted for in
the model developed to date.
Additional work in terrain analysis could expand on the suitability criteria, specifically,
with respect to connections and suitability for placement. One significant variable in the viability
for construction of MPSHS is the ability to install all of the system components as cost-
effectively as possible. An addition to the model could work to characterize the terrain
underlying the connection for suitability.
Additionally, the model could be further enhanced by a process that evaluates the
suitability of the connections with respect to their path. One option may be to perform a buffer
96

analysis for connections and examine the effects of external variables spanning the distance
between terminal reservoirs. Careful consideration will be needed, as the factors affecting the
connections may be drastically different from those considered for reservoir placement.
Reservoir locations presented as outputs of this model are provided as raster values in the
Primary Model and then as points once the fuzzy dataset is applied. While these both give an
idea of the location identified as suitable, further work could be undertaken to aggregate these
outputs into areas that can be summarized and graded based on a larger spatial footprint to
provide more tangible results to model end-users.
The model developed in this study has presented a reasonable solution to the design
problem presented. At the county scale, freely available elevation data products with coverage
across the contiguous United States allow this model to identify suitable locations for the
placement MPSHS adequately and with flexibility for variations in engineering and design
criteria. Parameters built into the model allow adjustments for system design modifications and
incorporation of third-party screening and suitability layers for further refinement of the model
results.
97

References
Aydin, Nazli Yonca, Elcin Kentel, and Sebnem Duzgun. 2010. "GIS-based environmental
assessment of wind energy systems for spatial planning: A case study from Western
Turkey."Renewable and Sustainable Energy Reviews 14(1): 364-373.
Aydin, Nazli Yonca, Elcin Kentel, and H. Sebnem Duzgun. 2013. "GIS-based site selection
methodology for hybrid renewable energy systems: A case study from western
Turkey." Energy Conversion and Management 70: 90-106.
Benitez, Pablo C., Liliana E. Benitez, and G. Cornelis van Kooten. 2008. "The Economics of
Wind Power with Energy Storage." Energy Economics 30(4): 1973-1989.
Bolstad, Paul. 2005. GIS Fundamentals: A First Text on Geographic Information Systems. 2
nd

ed. White Bear Lake, Minnesota: Eider Press.
Bueno, C., and Jose A. Carta. 2006. "Wind powered pumped hydro storage systems, a means of
increasing the penetration of renewable energy in the Canary Islands." Renewable and
Sustainable Energy Reviews 10(4): 312-340.
California Independent System Operator. 2018. "What the Duck Curve Tells Us About
Managing a Green Grid." Accessed 11/22/18,
https://www.caiso.com/Documents/FlexibleResourcesHelpRenewables_FastFacts.pdf.
Carver, Stephen J. 1991. "Integrating multi-criteria evaluation with geographical information
systems." International Journal of Geographical Information System 5(3): 321-339.
Chu, Steven, and Arun Majumdar. 2012. "Opportunities and challenges for a sustainable energy
future." Nature 488(7411): 294.
ClimateTechWiki.org. 2019. "Energy Storage: Pumped Storage." Accessed 4/4/19,
https://www.climatetechwiki.org/technology/jiqweb-ph.
Connolly, David, Shane MacLaughlin, and Martin Leahy. 2009. "Locating Potential Sites for
Pumped Hydroelectric Energy Storage." In Proc. of the Electrical Energy Storage
Applications and Technologies (EESAT) Conference, Seattle, USA, pp. 4-7.
California Senate. 2018. SB-100 California Renewables Portfolio Standard Program: emissions
of greenhouse gases (September 10, 2018). Sacramento, CA.
Denholm, Paul, Erik Ela, Brendan Kirby, and Michael Milligan. 2010. Role of Energy Storage
with Renewable Electricity Generation (Report Summary)(Presentation). Report
NREL/PR-6A2-49396. National Renewable Energy Lab, Golden, CO.
Esri. 2019a. "How Fuzzy Membership Works." Accessed 7/14/2019,
http://desktop.arcgis.com/en/arcmap/10.3/tools/spatial-analyst-toolbox/how-fuzzy-
membership-works.htm.
98

Esri. 2019b. "How Fuzzy Overlay Works." Accessed 6/16/19,
http://desktop.arcgis.com/en/arcmap/10.3/tools/spatial-analyst-toolbox/how-fuzzy-
overlay-works.htm.
Esri. 2019c. "Overlay Analysis.” Accessed 7/14/2019,
http://desktop.arcgis.com/en/arcmap/10.3/analyze/commonly-used-tools/overlay-
analysis.htm.
Esri. 2019d. "Tool Help: Slope.” Accessed 6/9/2019,
http://desktop.arcgis.com/en/arcmap/10.3/tools/spatial-analyst-toolbox/slope.htm.
Esri. 2019e. "Understanding Euclidean Distance Analysis.” Accessed 7/14/2019,
http://desktop.arcgis.com/en/arcmap/10.3/tools/spatial-analyst-toolbox/understanding-
euclidean-distance-analysis.htm.
Esri. 2019f. "Understanding Reclassification.” Accessed 7/14/2019,
https://pro.arcgis.com/en/pro-app/tool-reference/spatial-analyst/understanding-
reclassification.htm.
Esri. 2019g. "What is ModelBuilder?” Accessed 6/15/19,
http://desktop.arcgis.com/en/arcmap/10.3/analyze/modelbuilder/what-is-
modelbuilder.htm.
Fitzgerald, Niall, Roberto Lacal Arántegui, Eamon McKeogh, and Paul Leahy. 2012. "A GIS-
based model to calculate the potential for transforming conventional hydropower
schemes and non-hydro reservoirs to pumped hydropower schemes." Energy 41(1): 483-
490.
Friends of the River. 2018. "Rivers Under Threat.” Accessed 4/1/2018,
https://www.friendsoftheriver.org/our-work/rivers-under-threat/.
Gesch, D.B., Oimoen, M.J., and Evans, G.A., 2014, Accuracy assessment of the U.S. Geological
Survey National Elevation Dataset, and comparison with other large-area elevation
datasets —SRTM and ASTER. U.S. Geological Survey Open-File Report 2014–1008.
Heirman, D., 2012. “What makes Smart Grid—Smart—And who is in the “game”?” IEEE
Electromagnetic Compatibility Magazine 1(2): 95-99.
Hansen, Henning Sten. 2005. "GIS-based multi-criteria analysis of wind farm development."
In Proceedings of the 10th Scandinavian Research Conference on Geographical
Information Sciences, 13th-15th June 2005, pp. 75-87.
Kaplan, Stan M. 2009. "Electric power transmission: background and policy issues." Library of
Congress, Congressional Research Service.
Madani, Kaveh, Marion Guégan, and Cintia B. Uvo. 2014. "Climate change impacts on high-
elevation hydroelectricity in California." Journal of Hydrology 510: 153-163.
99

Malczewski, Jacek. 2000. "On the use of weighted linear combination method in GIS: common
and best practice approaches." Transactions in GIS 4(1): 5-22.
Malczewski, Jacek. 2004. "GIS-based land-use suitability analysis: a critical
overview." Progress in planning 62(1) : 3-65.
Mitchell, Andy. 2012. The ESRI Guide to GIS analysis: modeling suitability, movement, and
interaction. Vol. 3. Redlands, CA: Esri Press.
Montello, Daniel, and Paul Sutton. 2006. An introduction to scientific research methods in
geography. SAGE Publications, Inc.
Napier, Ted L., Michael V. Carter, and Elizabeth G. Bryant. 1986. "Local perceptions of
reservoir impacts: A test of vested interests." American Journal of Community
Psychology 14(1): 17-37.
Norris, Robert Matheson, and Robert W. Webb. 1990. Geology of California. 2nd ed. Hoboken,
NY: John Wiley & Sons.
O'Sullivan, David, and George LW Perry. 2013. Spatial simulation: exploring pattern and
process. Hoboken, NY: John Wiley & Sons.
O’Sullivan, David, and David J. Unwin. 2010. Geographic Information Analysis, 2nd ed.
Hoboken, NY: John Wiley & Sons.
Pasternack, Gregory B., C. Lau Wang, and Joseph E. Merz. 2004. "Application of a 2D
hydrodynamic model to design of reach ‐scale spawning gravel replenishment on the
Mokelumne River, California." River Research and Applications 20(2): 205-225.
Rehman, Shafiqur, Luai M. Al-Hadhrami, and Md Mahbub Alam. 2015. "Pumped hydro energy
storage system: A technological review." Renewable and Sustainable Energy Reviews 44:
586-598.
Rosenberg, Ronald H. 2008. "Diversifying America's energy future: The future of renewable
wind power." Virginia Environmental Law Journal 26: 505.
Sparkes, Andrew, and David Kidner. 1996. "A GIS for the environmental impact assessment of
wind farms." In ESRI European User Conference, London, 1996, pp. 2-4.
Suul, Jon Are, Kjetil Uhlen, and Tore Undeland. 2008. "Variable speed pumped storage
hydropower for integration of wind energy in isolated grids: case description and control
strategies." In Nordic Workshop on Power and Industrial Electronics (NORPIE/2008),
June 9-11, 2008, Espoo, Finland. Helsinki University of Technology.
The National Geographic Society. 2011. "Reservoir." Accessed 3/24/2019,
https://www.nationalgeographic.org/encyclopedia/reservoir/.
100

United States Energy Information Administration. 2019. "Total Energy Consumption of
Renewable and Nonrenewable Energy Sources." Accessed 7/14/2019,
https://www.eia.gov/renewable/data.php.
USGS. 2018. "Hydroelectric Power Water Use." Accessed 3/24/2019,
https://water.usgs.gov/edu/wuhy.html.
USGS. 2019. "About 3DEP Products and Services.” Accessed 3/5/2019,
https://www.usgs.gov/core-science-systems/ngp/3dep/about-3dep-products-services.
Uyan, Mevlut. 2013. "GIS-based solar farms site selection using analytic hierarchy process
(AHP) in Karapinar region, Konya/Turkey." Renewable and Sustainable Energy Reviews
28: 11-17.
Yang, Chi-Jen, and Robert B. Jackson. 2011. "Opportunities and barriers to pumped-hydro
energy storage in the United States." Renewable and Sustainable Energy Reviews 15(1):
839-844.
101

Appendix A – Primary Model Process Table
Process
ID
Process
Name
Input Data
Name
Input
Data
Type
Tool
Major
Parameter(s)
Output Data
Name
Output
Data
Type
Purpose
1 Slope
Study Area
DEM
30m
Elevation
Raster
Slope
Slope in
degrees
Study Area
Slope Raster
30m
Slope
Raster
Creates a Slope raster for use in
identifying suitable locations
for construction of terminal
reservoirs
2
Construction
Area Focal
Analysis
Study Area
Slope Raster
30m
Slope
Raster
Focal
Statistics
Search
Distance: 1
Cell; Search
Pattern:
Rectangle
Max.
Neighborhood
Slope Raster
30m
Raster
Examines each of the eight
neighbor cells for each cell and
assigns the parent cell value
equal to the highest cell in the
neighborhood
3
Binary
Reclassify
Construction
Areas
Max.
Neighborhood
Slope Raster
& Reclass
Expression
30m
Raster
Reclassify
Reclass
Expression:
If the slope is
greater than
15 then 0;
else 1
Max Slope
Binary Raster
30m
Raster
Reclassifies each cell into
binary code to eliminate areas
that are not suitable for
construction based on slope.
4
Apply
Binary
Screen
Max Slope
Binary Raster
& Binary
Screening
Raster
30m
Rasters
Times Multiply cells
Construction
Area Raster
30m
Raster
Creates a raster dataset that
additionally eliminates the
areas identified by the binary
screening raster from
consideration in terminal
reservoir placement
5
Convert
Raster to
Polygon
Construction
Area Raster
30m
Raster
Raster to
Polygon
Based on the
value field
Construction
Area
Polygons
Vector
Polygons
creates a vector polygon
dataset that identifies all of the
areas deemed suitable for the
construction of a terminal
reservoir independent of relief
(1) and areas that are not
suitable (0).
102

Process
ID
Process
Name
Input Data
Name
Input
Data
Type
Tool
Major
Parameter(s)
Output Data
Name
Output
Data
Type
Purpose
6
Select
Suitable
Areas for
Construction
Construction
Area
Polygons
Vector
Polygons
Select
Based on the
value field
Suitable
Areas for
Construction
Vector
Polygons
Creates a vector dataset of
polygons representing ONLY
those areas suitable for
placement of a terminal
reservoir
7
Extract
Suitable
Areas from
Study Area
DEM
Suitable
Areas for
Construction
(Mask) &
Study Area
DEM (Target
Layer)
30m
Raster &
Vector
Polygons
Extract by
Mask
Extract from
Study Area
DEM by
Suitable
Areas for
Construction
Study Areas
DEM Subset
30m
Raster
Extracts the portions of the
DEM that are coincident with
8
Export
Suitable
Areas
Polygons
Suitable
Areas for
Construction
Vector
Polygons
Export None
Final Suitable
Areas
Vector
Polygons
Exports intermediate dataset
9
Search for
Maximum
Elevation
Study Areas
DEM Subset
30m
Raster
Focal
Statistics
Search
Distance: 50
Cells
(1500m);
Search
Pattern:
circular
Maximum
Elevation in
Focal Search
Raster
30m
Raster
Searches a specified radius for
the maximum elevation and
assigns the target cell that
value
10
Search for
Minimum
Elevation
Study Areas
DEM Subset
30m
Raster
Focal
Statistics
Search
Distance: 50
Cells
(1500m);
Search
Pattern:
circular
Minimum
Elevation in
Focal Search
Raster
30m
Raster
Searches a specified radius for
the minimum elevation and
assigns the target cell that
value
103

Process
ID
Process
Name
Input Data
Name
Input
Data
Type
Tool
Major
Parameter(s)
Output Data
Name
Output
Data
Type
Purpose
11
Calculate
Maximum
Uphill
Relief
Maximum
Elevation in
Focal Search
Raster &
Study Areas
DEM Subset
30m
Raster
Minus
(Maximum
Elevation in
Focal Search
Raster) -
(Study Areas
DEM)
Maximum
Uphill Relief
30m
Raster
Subtracts the maximum
elevation within the search area
by the true elevation at each
location.
12
Calculate
Maximum
Downhill
Relief
Minimum
Elevation in
Focal Search
Raster &
Study Areas
DEM Subset
30m
Raster
Minus
(Study Areas
DEM Subset)
- (Minimum
Elevation in
Focal Search
Raster)
Maximum
Downhill
Relief
30m
Raster
subtracts the true elevation at
each location by the minimum
elevation in the focal search
area.
13
Reclassify
Uphill
Relief
Maximum
Uphill Relief
30m
Raster
Reclassify
Reclass
Expression:
If relief is
greater than
300 then 1;
else
NODATA
Lower
Locations
Raster
30m
Raster
Creates a rater dataset
comprised only of cells that
meet the design relief
requirement in the uphill
direction.
14
Reclassify
Downhill
Relief
Maximum
Downhill
Relief
30m
Raster
Reclassify
Reclass
Expression:
If relief is
greater than
300 then 1;
else
NODATA
Upper
Locations
Raster
30m
Raster
Creates a rater dataset
comprised only of cells that
meet the design relief
requirement in the downhill
direction.
15
Convert
Lower
Locations to
Points
Lower
Locations
Raster
30m
Raster
Raster to
point
None
Lower
Location
Points
Vector
Points
Creates a point vector dataset
from the raster dataset
104

Process
ID
Process
Name
Input Data
Name
Input
Data
Type
Tool
Major
Parameter(s)
Output Data
Name
Output
Data
Type
Purpose
16
Convert
Upper
Locations to
Points
Upper
Locations
Raster
30m
Raster
Raster to
point
None
Upper
Location
Points
Vector
Points
Creates a point vector dataset
from the raster dataset
17
Search for
Lower
Reservoir
points near
upper
reservoir
points
Upper
Location
Points &
Lower
Location
Points
Vector
Points
Near
Search
distance:
1500m
Start and End
Points Table
Table
Generates a one to many tables
where each upper reservoir
location is paired with every
lower reservoir point within the
prescribed search radius
18
Connect
Upper
Reservoir
Points to
Lower
Reservoir
Points
Start and End
Points Table
Table
XY to
Line
Start Point
(X,Y) and
end Point
(X,Y)
Connection
Lines
Vector
Lines
Creates a vector line dataset of
lines connecting the upper
reservoir to lower reservoir
locations
19
Select
Connections
that Do Not
Cross
Named
Streams
Connection
Lines &
Restricted
Lines
Vector
Lines
Select
Select the
invert of lines
that intersect
Restricted
Lines
Selection of
Connection
Lines that Do
not Cross
Vector
Lines
Selects the inverse of features
that cross the restricted lines
20
Copy
Selected
Features
Selection of
Connection
Lines that Do
not Cross
Vector
Lines
Copy
Features
Copy
Selected
Connections
that Do not
cross
Restricted
Lines
Vector
Lines
Creates a new vector dataset
comprised of features that DO
NOT cross the Restricted Lines
Dataset
105

Process
ID
Process
Name
Input Data
Name
Input
Data
Type
Tool
Major
Parameter(s)
Output Data
Name
Output
Data
Type
Purpose
21
Assign
Elevations
to Lower
Points
Study Areas
DEM Subset
& Lower
Location
Points
Vector
Points
Extract
Values to
Points
None
Lower
Reservoir
Points with
elevations
Vector
Points
adds the attribute of elevation
from the DEM subset to each
point selected as suitable for
construction of a lower
reservoir
22
Assign
Elevations
to Upper
Points
Study Areas
DEM Subset
& Upper
Location
Points
Vector
Points
Extract
Values to
Points
None
Upper
Reservoir
Points with
elevations
Vector
Points
adds the attribute of elevation
from the DEM subset to each
point selected as suitable for
construction of an upper
reservoir
23
Spatial Join
Upper
Locations to
Connections
Connections
that Do not
Cross
Streams &
Upper
Location
Points
Vector
Lines &
points
Spatial
Join
Join one to
many
Connections
with upper
elevations
Vector
Lines
Adds the attribute of upper
reservoir elevation to each
connection line
24
Spatial Join
Lower
Locations to
Connections
Connections
with upper
elevations
Vector
Lines &
points
Spatial
Join
Join one to
many
Connections
with upper
and lower
elevations
Vector
Lines
Adds the attribute of lower
reservoir elevation to each
connection line
25
Add a field
for Relief
Connections
with upper
and lower
elevations
Vector
Lines
Add Field
Field Type :
Float
Connections
with Relief
Field
Vector
Lines
Adds a field for relied on the
connection dataset
26
Calculate
Relief
Connections
with Relief
Field
Vector
Lines
Calculate
Field
(Upper
Location
Elevation) -
(Lower
Location
Elevation)
Connections
with Relief
Vector
Lines
Calculates the relief value for
each connection
106

Process
ID
Process
Name
Input Data
Name
Input
Data
Type
Tool
Major
Parameter(s)
Output Data
Name
Output
Data
Type
Purpose
27
Select
Connections
that meet the
Relief
Requirement
Connections
with Relief
Vector
Lines
Select by
Attribute
Select
Connections
where Relief
exceeds
300m
Final Suitable
Connections
Vector
Lines
creates a new dataset with only
the connections that meet the
relief requirement
28
Export
Connections
Final Suitable
Connections
Vector
Lines
Export None
Final
Reservoir
Connections
Vector
Lines
Exports Final Modeled
Connections
29
Select upper
points that
DO NOT
have a
connection
Final Suitable
Connections
& Lower
Reservoir
Points with
elevations
Vector
Points
Select
Invert Select
points that
intersect
connections
Selected
lower points
without
connections
Vector
Points
Selects all points in the lower
points dataset that do not
intersect with a connection
30
Select lower
points that
DO NOT
have a
connection
Final Suitable
Connections
& Upper
Reservoir
Points with
elevations
Vector
Points
Select
Invert Select
points that
intersect
connections
Selected
Upper Points
Without
Connections
Vector
Points
Selects all points in the upper
points dataset that do not
intersect with a connection
31
Delete lower
points
without
connections
Selected
lower points
without
connections
Vector
Points
Delete
Delete
Selected
Final Lower
reservoir
points
Vector
Points
Deletes points from the dataset
that do not interest a
connection line
32
Delete
Upper
points
without
connections
Selected
Upper Points
Without
Connections
Vector
Points
Delete
Delete
Selected
Final Upper
reservoir
points
Vector
Points
Deletes points from the dataset
that do not interest a
connection line
107

Process
ID
Process
Name
Input Data
Name
Input
Data
Type
Tool
Major
Parameter(s)
Output Data
Name
Output
Data
Type
Purpose
33
Convert
Lower
Reservoir
Points to
Raster
Final Lower
reservoir
points
Vector
Points
Point to
Raster
None
Final Lower
Reservoir
Location
Raster
30m
Raster
Converts the final modeled
lower reservoir locations to a
raster dataset at the Study Area
DEM Resolution
34
Convert
Upper
Reservoir
Points to
Raster
Final Upper
reservoir
points
Vector
Points
Point to
Raster
None
Final Upper
Reservoir
Location
Raster
30m
Raster
Converts the final modeled
Upper reservoir locations to a
raster dataset at the Study Area
DEM Resolution

Abstract (if available)

Abstract As the global energy market pushes toward the further development and integration of renewable energy and reduced reliance on fossil fuels, the energy industry has looked to innovative solutions to solve the shortcomings of green energy production. Diurnal fluctuation in electrical production potential in solar and wind sources creates a need to develop ways to store surplus energy resources for later deployment. Pump storage hydroelectricity, in which surplus energy is used to pump water uphill to recharge a hydroelectric reservoir, holds a great deal of potential when used in conjunction with other types of renewable energy. This report documents the design and development of a two-phase analytical spatial model that identifies suitable locations for the placement of paired top and bottom terminal reservoirs of a modular pump storage hydroelectricity system (MPSHS). The first phase of the model applies user-defined search criteria to identify locations for the construction of terminal reservoirs that meet the relief and lateral run distance requirements. Further refinement of results from the first modeling phase using secondary information can be used to rank suitable locations based on user-supplied environmental, economic, and socio-demographic constraints and preferences. This thesis presents details of model function as well as case study results for Los Angeles County.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Building a geodatabase design for American Pika presence and absence data

PDF

Modeling historic structure preservation candidacy on Fort Ord

PDF

Building a spatial database of biochar research and practice with Web-GIS

PDF

Site suitability analysis for implementing tidal energy technology in southern California

PDF

Onshore wind power systems (ONSWPS): a GIS-based tool for preliminary site-suitability analysis

PDF

Eye.Earth Pro (Beta v1.0): application development and spatial financial analysis utilizing the PESTELM framework

PDF

Preparing for immigration reform: a spatial analysis of unauthorized immigrants

PDF

Congestion effects on arterials as a result of incidents on nearby freeway: When should you get off the highway?

PDF

Roadway hazard analysis: a safe ride for motorcycles

PDF

Creating a water quality geodatabase for the West Hawai‘i Island region

PDF

Closed landfills to solar energy power plants: estimating the solar potential of closed landfills in California

PDF

Coastal vulnerability assessment for archaeological sites on San Clemente Island and San Nicolas Island, California

PDF

Geographic object based image analysis for utility scale photovoltaic site suitability studies

PDF

Suitability analysis for wave energy farms off the coast of Southern California: an integrated site selection methodology

PDF

Assessing the transferability of a species distribution model for predicting the distribution of invasive cogongrass in Alabama

PDF

An application of aerial drones in high definition mapping for autonomous vehicles

PDF

Reducing maternal mortality by improving medical facility accessibility: a methodology demonstrated for the Democratic Republic of the Congo

PDF

Developing art-based cultural experiences in North Kohala: A community engagement project with OneIsland

PDF

Risk assessment to wildlife from Ohio on-shore wind farm development: a landscape model approach

PDF

Creating a Web GIS to support field operations and enhance data collection for the Animal and Plant Health Inspection Service (APHIS)

Asset Metadata

Creator Rosenbery, Joseph Warren, II (author)

Core Title A model for placement of modular pump storage hydroelectricity systems

School College of Letters, Arts and Sciences

Degree Master of Science

Degree Program Geographic Information Science and Technology

Publication Date 09/17/2019

Defense Date 07/31/2019

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag analysis,ArcGIS Pro,GIS,green energy,hydroelectricity,Model,OAI-PMH Harvest,pump storage,renewable,spatial,topography, terrain

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Kemp, Karen (committee chair), Vos, Robert (committee member), Wilson, John (committee member)

Creator Email jwrosenb@usc.edu,jwrosenbery2@yahoo.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-217301

Unique identifier UC11673883

Identifier etd-RosenberyJ-7815.pdf (filename),usctheses-c89-217301 (legacy record id)

Legacy Identifier etd-RosenberyJ-7815.pdf

Dmrecord 217301

Document Type Thesis

Rights Rosenbery, Joseph Warren, II

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA