dynamic modeling to inform environmental management
TRANSCRIPT
University of Vermont University of Vermont
UVM ScholarWorks UVM ScholarWorks
Graduate College Dissertations and Theses Dissertations and Theses
2008
Dynamic Modeling to Inform Environmental Management: Dynamic Modeling to Inform Environmental Management:
Applications in Energy Resources and Ecosystem Services Applications in Energy Resources and Ecosystem Services
Mark Gately University of Vermont
Follow this and additional works at: https://scholarworks.uvm.edu/graddis
Recommended Citation Recommended Citation Gately, Mark, "Dynamic Modeling to Inform Environmental Management: Applications in Energy Resources and Ecosystem Services" (2008). Graduate College Dissertations and Theses. 89. https://scholarworks.uvm.edu/graddis/89
This Thesis is brought to you for free and open access by the Dissertations and Theses at UVM ScholarWorks. It has been accepted for inclusion in Graduate College Dissertations and Theses by an authorized administrator of UVM ScholarWorks. For more information, please contact [email protected].
DYNAMIC MODELING TO INFORM ENVIRONMENTAL MANAGEMENT:
APPLICATIONS IN ENERGY RESOURCES AND ECOSYSTEM SERVICES
A Thesis Presented
by
Mark Gately
to
The Faculty of the Graduate College
of
The University of Vermont
In Partial Fulfillment of the Requirements
for the Degree of Master of Science
Specializing in Community Development and Applied Economics
October, 2008
Accepted by the Faculty of the Graduate College, The University of Vermont, in partial fulfillment of the requirements for the degree of Master of Science, specializing in Community Development and Applied Economics.
Thesis Examination Committee:
r
olodinsky, Ph.D.
Chairperson Robert Costanza, Ph.D.
Vice President for Research and Dean of Graduate Studies
Date: April 3, 2008
Abstract
Two original computer simulation models are presented in this thesis. Although these
models differ in their temporal, spatial, and structural dimensions, they are unified by a
common purpose: to build quantitative understanding of environmental resources and
better inform their future management.
According to the U.S. Department of the Interior’s Minerals Management Service, there
are significant undiscovered reserves of oil and natural gas located in the Gulf of Mexico
Outer Continental Shelf region. While the existence of these energy resources is critical
to the nation’s future economic well-being, of equal importance is the amount of already
extracted energy that will be required to deliver the new fuel to society in a useful form;
the difference between the two quantities is the net energy supply. “Energy return on
investment” (EROI) is an indicator of the net productivity of an energy supply process;
specifically, it is the ratio of gross energy production to total, direct plus indirect, energy
cost. Chapter 1 describes a dynamic model designed to calculate the EROI of offshore
energy extraction in the U.S. Gulf of Mexico from 1985 to 2004 under differing
assumptions regarding energy cost and technology. In 2004, the EROI of the process is
estimated to range from 10 to 25 depending on how comprehensively energy costs are
defined. In comparison, the EROI of U.S. onshore petroleum extraction in the 1930s was
at least 100.
Ecosystem services are those functions of ecosystems that support, directly or indirectly,
human welfare. Although interest in ecosystem services has surged in recent decades and
is currently still on the rise, these phenomena have yet to be universally quantified. The
current Multi-scale, Integrated Models of Ecosystem Services (MIMES) project is an
ambitious attempt to do so through dynamic, spatially explicit modeling. As a part of this
broad initiative, Chapter 2 details the development and testing of a model designed to
measure and map the ecosystem service “water regulation” at multiple scales. The model
is an extension of the well known and widely used “runoff curve number” method
originally developed by the U.S. Department of Agriculture; it is applied to the Winooski
watershed (Vermont, U.S.A.) and to the entire globe.
ii
Citation
Material from this thesis has been published in the following form:
Gately, M.. (2007). The EROI of U.S. offshore energy extraction: A net energy analysis
of the Gulf of Mexico. Ecological Economics, 63, 355-364.
Material from this thesis has been submitted for publication to Ecosystems on May 5,
2008 in the following form:
Gately, M.. On semantics and scale: Developing a dynamic multi-scale model of the
ecosystem service water regulation. Ecosystems.
iii
Table of Contents
Page
Citation ................................................................................................................................ ii
List of Tables ...................................................................................................................... v
List of Figures .................................................................................................................... vi
Introduction ......................................................................................................................... 1
Chapter 1. The EROI of U.S. offshore energy extraction: A net energy analysis of the
Gulf of Mexico .................................................................................................................... 6
1. Introduction ................................................................................................................ 7
2. EROI theory ............................................................................................................... 9
2.1. The stationary state ............................................................................................. 9
2.2. EROI dynamics ................................................................................................. 13
3. The offshore process ................................................................................................ 14
4. Model specification .................................................................................................. 17
4.1. Temporal and spatial articulation ...................................................................... 17
4.2. EROI calculation ............................................................................................... 18
4.3. Technology ....................................................................................................... 21
5. Results and conclusions ........................................................................................... 23
References ..................................................................................................................... 29
Appendix A ................................................................................................................... 31
Appendix B ................................................................................................................... 33
iv
Chapter 2. On semantics and scale: Developing a dynamic multi-scale model of the
ecosystem service water regulation .................................................................................. 35
1. Introduction .............................................................................................................. 35
2. Semantics ................................................................................................................. 39
3. Scale ......................................................................................................................... 41
4. A formal model of water regulation ........................................................................ 43
5. Results and discussion ............................................................................................. 50
5.1. Calibration and validation ................................................................................. 50
5.2. Scenario analysis ............................................................................................... 56
5.3. Scale effects ...................................................................................................... 59
5.4. The global application ....................................................................................... 61
6. Conclusion ............................................................................................................... 63
References ..................................................................................................................... 65
Literature Cited ................................................................................................................. 67
v
List of Tables
Page
Table 1.1 – Activity phases of the offshore process and their annual per unit costs by
water depth (Dismukes et al., 2003). ........................................................................ 19
Table 1.2 – Expenditure profile for platform abandonment and removal (Dismukes et al.,
2003). ........................................................................................................................ 21
Table 2.1 – A list of ecosystem services and their corresponding ecosystem functions
(Costanza et al., 1997). ............................................................................................. 36
Table 2.2 – Default CN parameterization for the 11-biome MIMES land cover scheme
and HSG. ................................................................................................................... 46
Table 2.3 – Results of model calibration (1992) and validation (2001). .......................... 54
Table 2.4 – Calibrated parameters. ................................................................................... 55
Table 2.5 – Scenario analysis............................................................................................ 57
Table 2.6 – Scale effects. .................................................................................................. 60
vi
List of Figures
Page
Fig. 1.1 – A generic energy supply process defined over the duration [0,T]...................... 9
Fig. 1.2 – The energy cost tier structure. .......................................................................... 12
Fig. 1.3 – A typical energy supply system, adapted from Cleveland and Costanza (1984).
................................................................................................................................... 12
Fig. 1.4 – Gross production of offshore oil and natural gas in the U.S., 1960 – 2004.
Source: Energy Information Administration. ........................................................... 15
Fig. 1.5 – Percentage of gross energy production in the U.S. GOM from depths greater
than 200 m, 1992 – 2005. Source: Energy Information Administration. ................ 16
Fig. 1.6 – Simplified model diagram. ............................................................................... 17
Fig. 1.7 – EROI of GOM offshore extraction using two alternative treatments of energy
cost (T1 and T3) and simulated technology, 1985 – 2004. ....................................... 23
Fig. 1.8 – EROI for various methods of electric power generation (Kubiszewski and
Cleveland et al., 2007). ............................................................................................. 24
Fig. 1.9 – EROI of GOM offshore extraction by water depth using two alternative
treatments of energy cost (T1 and T3) and constant 1985 technology, 1985 – 2004.
................................................................................................................................... 25
Fig. 1.10 – T3 EROI of extraction in the 0 – 60 m and 61 – 200 m regions using constant
1985 technology and simulated technology, 1985 – 2004. ....................................... 26
Fig. 1.11 – Historical GOM T3 EROI (1985 – 2004) and two alternate scenarios for the
near future (2005 – 2013).......................................................................................... 27
vii
Fig. 1.12 – Historical GOM energy production (1985 – 2004) and two alternate scenarios
for the near future (2005 – 2013). ............................................................................. 28
Fig. 2.1 – Two alternative methods of scaling: a) by “majority rule” and b) by
maintaining a percentage distribution. ...................................................................... 49
Fig. 2.2 – Simplified model diagram. ............................................................................... 49
Fig. 2.3 – 2001 land cover pattern (30 m) of the Winooski watershed. ............................ 52
Fig. 2.4 – HSG pattern (90 m) of the Winooski watershed. ............................................. 52
Fig. 2.5 – Modeled and observed runoff in the Winooski watershed, June 1 – October 31,
1992. .......................................................................................................................... 53
Fig. 2.6 – Modeled and observed runoff in the Winooski watershed, June 1 – October 31,
2001. .......................................................................................................................... 54
Fig. 2.7 – Total water regulation (thousand m3) in the Winooski watershed from June 1 –
October 31, 2001 at 330 m grain............................................................................... 56
Fig. 2.8 – Modeled runoff in the Winooski watershed June 1 – October 31, 2001 using
two alternate land cover scenarios: historical data and completely forested. ........... 58
Fig. 2. 9 Modeled runoff in the Winooski watershed June 1 – October 31, 2001 using
two alternate land cover scenarios: historical data and completely urbanized. ........ 58
Fig. 2.10 – Total water regulation (thousand m3) in the Winooski watershed, June 1 –
October 31, 2001 at 4 grain sizes: a) 450 m, b) 900 m, c) 2700 m, and d) 6000 m. . 60
Fig. 2.11 – Global land cover pattern (1 km). ................................................................... 62
Fig. 2.12 – Global HSG pattern (5 arcmin). ..................................................................... 62
viii
Fig. 2.13 – Global water regulation (1 deg) as the regulated percentage of a 2 in rainfall
event. ......................................................................................................................... 63
1
Introduction
As the scale of the human presence in the biosphere increases, Earth’s once
abundant stock of natural capital is becoming increasingly scarce. As Daly (1996) has
pointed out, “We have moved from a world relatively full of natural capital and empty of
man-made capital (and people) to a world relatively full of the latter and empty of the
former.” In the current “full world” era (that is, one full of humans and our various
artifacts), natural capital has replaced manufactured capital or human labor as the limiting
factor of economic development. Moreover, the vital life-support services that natural
capital stocks provide, such as climate and water regulation, have been degraded, and
continual degradation of these services may ultimately compromise the sustainability of
humans on Earth. Thus, there now exists an urgent and overarching need for improved
management of our environmental resources through proactive decision-making. To
inform decision-makers, an entirely new generation of computer models must be devised
and deployed; two such models are presented in this thesis.
A model is any abstract representation of reality. Based on the original insights of
Kenneth Craik (1943), recently developed “mental model” theories (Johnson-Laird,
1983; Gentner and Stevens, 1983) suggest that human reasoning and comprehension is
based on the construction and execution of “small scale models” within the human mind.
These psychological representations are derived using an understanding of the given
premises and some general knowledge and assumptions about how the world works.
Although commonly thought of as internal pictures or images, mental models are a more
2
general notion; they may also represent abstract concepts, such as happiness or value,
which cannot be arranged spatially.
Humans interact with their mental models via semantic, conceptual operations.
The inferences yielded are essential in decision situations for assessing the relative
utilities of the known alternatives. Often, a distinct mental model is conceived for each
explicit alternative scenario option, and a decision is reached based on interaction with
multiple models. Due to the limited capacity of the human working memory these
models are restricted in terms of complexity and resolution, and they are primarily based
on qualitative relationships. Nevertheless, mental models are usually sufficiently detailed
and accurate to provide reliable “ad hoc” decision support in everyday life.
In environmental management domains, however, complex systems prevail.
Economies and ecosystems are strongly linked and deviously complex systems (Costanza
et al., 1993). As such, they exhibit baffling dynamic behaviors characterized by non-
linearity, complex feedbacks, and spatial and temporal delays. Moreover, these
properties are often “emergent” and cannot be determined or explained by simply
aggregating the individual behaviors of the system’s component parts (von Bertalanffy,
1968). In order to reason about complex systems and properly anticipate their dynamic
behaviors, the modeling approach must not only be holistic and integrative (Levins,
1966), but also based on quantitative relationships between the constituent parts. Mental
models are thus only of limited insight in environmental management; informed and
dependable decision support in this domain requires an extension of our mental modeling
capabilities.
3
Mathematical models, for example, consist of mathematical expressions that
formally specify real systems. In particular, dynamic mathematical models are sets of
differential or difference equations that articulate a system’s behavior over time. These
models are potent, quantitative extensions to our mental models of dynamic systems.
However, the differential equations that describe complex, non-linear behavior are
generally very difficult, if not impossible, to solve analytically (von Bertalanffy, 1968);
their solutions can only be approximated using advanced numerical methods executed on
computers. Thus, prior to the widespread availability of computers, quantitative analysis
of complex dynamic systems was severely limited to the examination of simple, linear
dynamics at coarse resolutions. Developments in numerical analysis and continuous
improvements in computer speed and accessibility, however, have recently eroded those
limits.
Dynamic computer modeling is now a well accepted and widely used technique in
scientific analysis. The “system dynamics” methodology of Forrester (1968), for
example, has been used to study and manage a wide range of complex systems cutting
across multiple disciplines in both the natural and social sciences. In the early 1970s, the
World2 (Forrester, 1971) and World3 (Meadows et al., 1972) global models pioneered
the use of these techniques to quantify complex human-environment interactions, and
since then dynamic modeling has been effectively used in a wide variety of
environmental research and management contexts (cf. Costanza and Ruth, 1998); the
Millennium Ecosystem Assessment (MA, 2005) is a recent example.
4
In this thesis, dynamic modeling is used to build quantitative understanding of
energy resources and ecosystem services. In 2004, the U.S. Department of the Interior’s
Minerals Management Service estimated that 49% of the oil and 57% of the natural gas
yet to be discovered offshore in the United States are located in the Gulf of Mexico Outer
Continental Shelf region (MMS, 2004). While the existence of these energy resources is
critical to the nation’s future economic well-being, of equal importance is the amount of
already extracted energy that will be required to deliver the new fuel to society in a useful
form; the difference between the two quantities is the net energy supply. “Energy return
on investment” (EROI) is an indicator of the net productivity of an energy supply
process; specifically, it is the ratio of gross energy production to total, direct plus indirect,
energy cost (Cleveland et al., 1984). Chapter 1 describes a dynamic model designed to
calculate the EROI of offshore energy extraction in the U.S. Gulf of Mexico from 1985 to
2004 under differing assumptions regarding energy cost and technology. In 2004, the
EROI of the process is estimated to range from 10 to 25 depending on how
comprehensively energy costs are defined. In comparison, the EROI of U.S. onshore
petroleum extraction in the 1930s was at least 100 (Cleveland, 1992).
Ecosystem services are those functions of ecosystems that support, directly or
indirectly, human welfare (Costanza et al., 1997; Daily, 1997). Although interest in
ecosystem services has surged in recent decades and is currently still on the rise, these
phenomena have yet to be universally quantified. The current Multi-scale, Integrated
Models of Ecosystem Services (MIMES) project is an ambitious attempt to do so through
dynamic, spatially explicit modeling. As a part of this broad initiative, Chapter 2 details
5
the development and testing of a model designed to measure and map the ecosystem
service “water regulation” at multiple scales. The model is an extension of the well
known and widely used “runoff curve number” method originally developed by the U.S.
Department of Agriculture; it is applied to the Winooski watershed (Vermont, U.S.A.)
and to the entire globe.
Energy limits and loss of ecosystem services have been identified as among the
most serious environmental problems facing present and future societies (Diamond,
2005; Tainter, 1988). When faced with such issues in the past, some societies (such as
the infamous Easter Islanders) unfortunately collapsed, while others culturally evolved to
prosper for long periods of time. Diamond (2005) concludes that a key determinant of
success or failure is how a society responds to the signals of emerging environmental
problems. By supplementing our mental models with innovative computer models such
as those presented in this thesis, we can both clarify those ecological signals and better
inform our managerial responses.
6
Chapter 1.
The EROI of U.S. offshore energy extraction: A net energy analysis of the Gulf of
Mexico
Abstract
In 2004, the U.S. Department of the Interior’s Minerals Management Service estimated
that 49% of the oil and 57% of the natural gas yet to be discovered offshore in the United
States are located in the Gulf of Mexico Outer Continental Shelf region. While the
existence of these energy resources is critical to the nation’s future economic well-being,
of equal importance is the amount of already extracted energy that will be required to
deliver the new fuel to society in a useful form; the difference between the two quantities
is the net energy supply. “Energy return on investment” (EROI) is an indicator of the net
productivity of an energy supply process; specifically, it is the ratio of gross energy
production to total, direct plus indirect, energy cost. This paper describes a dynamic
model designed to calculate the EROI of offshore energy extraction in the U.S. Gulf of
Mexico from 1985 to 2004 under differing assumptions regarding energy cost and
technology. In 2004, the EROI of the process is estimated to range from 10 to 25
depending on how comprehensively energy costs are defined. In comparison, the EROI
of U.S. onshore petroleum extraction in the 1930s was at least 100.
7
1. Introduction
The distinction between gross and net measures of production is a fundamental
concept in standard economic theories. Conventional macroeconomic theory, for
example, provides two estimates of a nation’s total economic output: gross national
product (GNP) and net national product (NNP). GNP is the total market value of all
goods and services produced by the citizens a nation over a period of time, whereas NNP
is equal to GNP less capital depreciation. Similarly, conventional microeconomic theory
provides two measures of a firm’s production: revenue and profit. Revenue is the total
market value of a firm’s output over a period of time, while profit is equal to revenue
minus the total market cost of the factors of production.
The output of any economic production process, be it an individual firm or an
aggregate national economy, should be considered in both gross and net terms. But this
point is often missed in regards to one of the most critical economic production
processes: energy supply. The energy supply process is the “fuel line” of an economy; it
provides the low-entropy free energy required to power all of the various economic
sectors (including its own). Cottrell (1955) termed the net output of energy production
“surplus energy” and defined it as the difference between the energy delivered by a
process and the energy invested in the delivery.
Surplus energy may be the most economically relevant measure of energy
production because it represents the ability to produce final demand goods and services.
Yet in a review of the data accessible from the United States Energy Information
Administration’s official website (http://www.eia.doe.gov/) no references to surplus
8
energy could be found; all production figures, both historical and forecasted, are in gross
terms. From a net perspective, these data may look quite different.
In contrast, the academic community has recognized the importance of surplus
energy for many years. Numerous separate studies have estimated the net productivities
of a variety of energy supply processes, including ethanol production (Farrell et al., 2006;
Hammerschlag, 2006), nuclear energy generation (Tyner et al., 1988; Kidd, 2004), and
petroleum extraction (Cleveland, 1992; Cleveland, 2005), to name a few. However, there
currently is no standard protocol for net energy analysis, and so the results of these
studies are often inconsistent and difficult to interpret; the recent debate over the net
energy yield of corn ethanol production is a telling example (cf. Farrell et al., 2006;
Hagens et al., 2006; Cleveland et al., 2006).
In this paper, we develop a standard framework for net energy analysis using the
“energy return on investment” (EROI) productivity index, which is the ratio of gross
energy production to total, direct plus indirect, energy cost (Cleveland et al., 1984). We
then construct a dynamic computer model to apply this analytical framework to the
extraction of offshore energy in the U.S. Gulf of Mexico (GOM).
9
2. EROI theory
2.1. The stationary state
Fig. 1.1 – A generic energy supply process defined over the duration [0,T].
Figure 1.1 illustrates a generic energy supply process that is characterized by its
energy input and output flows. For every time t contained within the process duration,
[0,T], its state may be specified analytically by a 3-tuple as follows:
{EI(t), EO(t), EC(t)} where 0 ≤ t ≤ T (1)
The three coordinates in Eq. (1) are cumulative, non-decreasing functions of time; each
represents a distinct energy flow common to all energy supply processes. EI(t) is the raw,
“in situ” energy input, or energy in its natural form (e.g., wind, water, oil “in the
ground”). EO(t) is the transformed energy output, or energy at its social “point of use”
(e.g., electricity, gasoline “at the pump”). EC(t) is the total, direct plus indirect, energy
cost of the transformation. Collectively, these three functions provide a static “snapshot”
of the process at each time-step of its duration.
10
Given the information in Eq. (1) for a process operating at time-step t (where 0 t
T), various metrics can be derived that describe its performance up to and including
that particular moment. EROI, for example, can be formally defined based on Eq. (1) as
follows:
EROI(t) = EO(t) / EC(t) (2)
Expressed as a function of time, the EROI index is the ratio of delivered end-use energy
to the energy cost of delivery; both quantities are cumulative up to the specified time t.
EI(t) is not required to calculate EROI, however, it is used in other performance
indicators (such as “energy capture efficiency”: EI(t) / EC(t) + EO(t)).
Operationalizing this analytical model depends largely on the availability of
existing data, or the feasibility of deriving reliable data, for the two ratio components.
While the measurement of delivered end-use energy is relatively straightforward and data
are readily available, determining the energy cost of its delivery is significantly more
subjective and data are extremely limited. Energy cost data, therefore, often must be
derived essentially “from scratch”. To do so properly, the boundary of the energy supply
process must first be clearly established.
We will define the “frontier” of the energy supply process as encompassing all
anthropogenic activities (e.g., exploration, extraction or capture, processing, refinement,
storage, transportation) required to transform and deliver a particular quantity of in situ
energy to its social point of use. The process environment is the natural human
environment: a backdrop of ecosystem structures and processes that support and sustain
production. The energy cost of the process thus consists of 1) the total, direct plus
11
indirect, energy embodied in all economic feedbacks necessary to power the
transformative human activities, and 2) the energy cost of any damages to the external
environmental support structure.
These energy costs can be divided into four “tiers” of increasing
comprehensiveness. “Tier 1” (T1) energy costs include only direct fuels, which are the
already-delivered energies whose point of use is the production (or more correctly, the
reproduction) of new fuel. Consider a vehicle that transports diesel fuel to the
distribution station, while running on the same product. Indirect energy costs are more
difficult to identify and assess, but often are more significant than direct costs (cf.
Cleveland, 2005); they are distributed amongst the three upper tiers. “Tier 2” (T2)
energy costs include direct fuels as well as the energy embodied in the built capital and
other industrial goods and services used in the process. “Tier 3” (T3) energy costs
include T2 costs plus the energy equivalent of labor and government services. “Tier 4”
(T4) costs are the most comprehensive; they consist of T3 costs plus the energy cost of
any damages resulting from the depletion or pollution of the external process
environment, which can be measured in terms of ecosystem services. Figure 1.2 depicts
this energy cost tier structure, in which each tier is inclusive of everything below. In Fig.
1.3, these costs are presented in the context of a typical energy supply system.
12
Fig. 1.2 – The energy cost tier structure.
Fig. 1.3 – A typical energy supply system, adapted from Cleveland and Costanza
(1984).
13
2.2. EROI dynamics
Energy supply processes exhibit certain long run productivity dynamics. For
example, the EROI of non-renewable resource production naturally decays with the
depletion of the in situ resource stock. This is because, as Daly (1996, p.66) has pointed
out, surplus energy is a natural subsidy to the economy; it represents “value added” by
nature. Energy resources with more natural value added require less labor and capital for
delivery, sell for a lower market price, and therefore get produced and used up first.
Onshore oil from Texas, for example, was a much greater natural subsidy than is offshore
oil from the GOM. We refer to this phenomenon as the “depletion effect”.
The depletion effect does not apply to the production of renewable energy
resources because their supplies are unlimited. However, the available sites at which
renewable energies may be captured are limited, as is the maximum potential rate of
capture at each site. A long-term increase in the overall rate of renewable resource
production thus requires that a portion of the remaining available capture sites become
“saturated” with production. Because the most highly “subsidized” sites are saturated
first, EROI declines as production rates increase; we refer to this phenomenon as the
“saturation effect”.
Technological progress can offset or overcome depletion or saturation effects
through improvements in the energy efficiency of production. In the long run, a balance
of these two opposing forces determines EROI dynamics. If technology cannot
compensate for the effects of depletion or saturation, then EROI declines and the quality
14
of the resource diminishes. Conversely, if technology can overcome depletion or
saturation effects, then EROI increases and the quality of the resource is enhanced.
3. The offshore process
Gross production of U.S. offshore oil and natural gas has surged over the past
fifty years (Fig. 1.4). Historically, the GOM has been a hotspot for offshore extraction,
and in 2004 the U.S. Department of the Interior’s Minerals Management Service (MMS)
confirmed that production will continue in the region for yeas to come, estimating that
49% of the oil and 57% of the natural gas yet to be discovered offshore in the United
States are located there (MMS, 2004). While the existence of these energy resources is
critical to the nation’s future economic well-being, of equal importance is the amount of
already extracted energy that will be required to deliver the new fuel to society in a useful
form.
15
0
1
2
3
4
5
6
1960
1962
1964
1966
1968
1970
1972
1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
2002
2004
Year
Qu
ad
rillio
n B
TU
Crude oil
Natural gas
Fig. 1.4 – Gross production of offshore oil and natural gas in the U.S., 1960 – 2004.
Source: Energy Information Administration.
To derive the EROI of GOM offshore extraction, the boundaries of this process
must first be established. As defined here, the “offshore process” refers exclusively to
the “upstream” sector of the offshore industry, including all activities necessary to find
and physically extract crude oil and natural gas from the offshore environment, but
excluding any further processing, storage, refinement, marketing, distribution, etc., of the
extracted fuel. Ei is in situ offshore crude oil and natural gas; Eo is extracted crude oil
and natural gas (to be used as input to the next process in series); Ec is the total energy
cost of the extraction.
16
Figure 1.5 illustrates an important characteristic of the offshore process: over
time, the depletion effect extends production from shallow waters to more challenging
deepwater environments. GOM offshore extraction can thus be viewed at various stages
of development by disaggregating the region by water depth and treating extraction in
each depth region as an independent process; processes operating in shallow water
regions are in later stages of development relative to those in deepwater regions.
0
10
20
30
40
50
60
70
80
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
Year
Pe
rce
nt
Crude oil
Natural gas
Fig. 1.5 – Percentage of gross energy production in the U.S. GOM from depths
greater than 200 m, 1992 – 2005. Source: Energy Information Administration.
17
4. Model specification
We designed a dynamic model to calculate the EROI of GOM offshore extraction
using Simile, an object-based, graphical modeling environment (Simulistics, Ltd.;
www.simulistics.com). The model diagram in Fig. 1.6 has been simplified from the final
version to highlight the model’s major components. The two box compartments, Eo and
Ec, are the model’s “state variables”; consistent with our analytical definition of process
state in Eq. (1), these variables cumulatively increase over time and uniquely define the
model at each time-step of its duration. The other relevant features of the model are
outlined below.
Fig. 1.6 – Simplified model diagram.
4.1. Temporal and spatial articulation
The model runs for a twenty-year period (1985 – 2004) at a yearly time-step. A
longer duration would have been preferred, however, the temporal extent was constrained
18
by data availability. The GOM is spatially disaggregated into four distinct sub-regions
defined by water depth: 0 to 60 m, 61 to 200 m, 201 to 900 m, and deeper than 900 m.
Extraction in each sub-region is considered a unique instance of the general offshore
process.
4.2. EROI calculation
The model calculates the EROI of offshore extraction in each of the four depth
regions and in the entire GOM at each time-step of the simulation. Annual gross
production includes both crude oil and natural gas; those data were obtained from the
MMS online database system (www.gomr.mms.gov). Energy cost data, however, are not
available and therefore must be calculated by the model internally. The approach is to
first derive monetary costs, and then convert them to energy costs.
The MMS has identified 11 major activity phases of the GOM offshore industry
(Dismukes et al., 2003). Nine phases are used in the model to derive monetary cost
(Table 1.1); two (“gas processing and storage: O&M” and “gas processing and storage:
construction”) are not because the frontier of the offshore process as defined above
excludes them. For each phase, the MMS has determined a standard physical cost unit
and its annual monetary cost (Table 1.1), as well as a sector-specific expenditure profile
(e.g., Table 1.2). By combining this information with data obtained from the MMS
online database system, the total monetary cost of each phase by water depth is derived
and then allocated to the appropriate economic sectors.
Sector-specific monetary costs can be converted to energy units using the energy
intensity factors of the various economic sectors. The energy intensity of an individual
19
economic sector is the energy cost per dollar of sector production (BTU/$). Based on the
input-output (I-O) techniques of Leonteif (1941) and methods described by Bullard and
Herendeen (1975) and Costanza (1980), Costanza and Herendeen (1984) have derived
energy intensities using both T1 and T3 energy costs for an 87-sector breakdown of the
1972 U.S. economy. These data are adjusted for inflation (to constant 1998$) and
technology (see below), and then are used to convert sector-specific monetary costs to
their energy equivalents, which are summed across sectors to determine the total energy
cost of each activity phase by water depth.
Table 1.1 – Activity phases of the offshore process and their annual per unit costs by
water depth (Dismukes et al., 2003).
Activity Phase 0-60 m 61-200 m 201-900 m 900 m+
Exploratory and development drillinga
Cost per well drilled ($000) 4,238 3,007 5,971 10,178
Production
Cost per producing well 240,908 253,764 263,152 271,070
Platform fabrication and installationb
Cost per platform ($ Million) 19.491 75.894 127.108 133.791
Pipeline construction and installationb
Cost per mile of pipeline 541,315 891,654 1,509,284 3,850,385
Pipeline operations and maintenance
Cost per MMBTU produced .0520 .0642 .0660 .0767
20
Workoversc
Cost per workover per well 13,704 14,385 14,566 14,566
Oil spills
Cost per gallon spilled 107.19 67.88 45.45 45.45
Platform abandonment and removalb,d
Cost per platform 878,555 1,816,058 9,664,348 12,710,826
All costs in constant 1998$.
a Exploratory drilling and development drilling, two distinct activity phases, are
aggregated into one general “drilling” category due to a lack of differentiation in the data
between wells drilled for development purposes and those drilled for exploratory
purposes; the average of the two costs is used for each depth.
b Denotes a fixed cost category; fixed costs are amortized evenly over the entire process
duration.
c Workover costs are the average of three major types of workovers: recompletions,
major workovers, and wireline services. It is assumed that each producing well has a
70% probability annually of having to undergo one of these three types, regardless of
water depth.
d Abandonment costs for each water depth are the average of several removal methods for
4-pile and 8-pile platforms.
21
Table 1.2 – Expenditure profile for platform abandonment and removal (Dismukes
et al., 2003).
IMPLAN
Sector
Sector
Description
0-60 m 61-200 m 201-900 m 900+ m
38 Oil & gas operations 0.14855 0.16929 0.17309 0.20139
50 New gas utility facilities 0.03732 0.04683 0.05043 0.05972
53 Misc. natural resource
facility construction
0.05050 0.03210 0.01563 0.01667
57 Other oil & gas field services 0.17102 0.11066 0.10253 0.01389
436 Water transport 0.58534 0.62886 0.64319 0.69028
506 Environmental/engineering
services
0.00728 0.01226 0.01513 0.01806
Total 1.00000 1.00000 1.00000 1.00000
4.3. Technology
The energy intensity factors calculated by Costanza and Herendeen (1984) reflect
1972 technology and therefore are not appropriate for this model considering its temporal
extent. To resolve this issue, the 1972 intensities are scaled based on a relative index of
the overall energy intensity of the U.S. economy. The most commonly used measure of
aggregate energy intensity is the “energy to gross domestic product” (E/GDP) ratio. This
measure, however, tends to overstate the extent to which improvements in energy
22
efficiency have occurred in an economy due to a disregard for structural change (cf.
Cleveland et al., 1984).
The U.S. Department of Energy’s Office of Energy Efficiency and Renewable
Energy has recently developed a new system of energy intensity indicators, the “energy
intensity indices”, that provide a more accurate measure of changes in the energy
throughput of an economy that actually result from changes in energy efficiency (i.e.,
from technological change). We use an adjusted E/GDP ratio, the “index of aggregate
intensity”, to scale the 1972 sector intensities so that they reflect 1985 technology. The
intensities can then be varied throughout a model run based on relative changes in the
index of aggregate intensity to simulate technological change; alternatively, they can be
held constant at 1985 levels to isolate the depletion effect.
23
5. Results and conclusions
0
5
10
15
20
25
30
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Year
ER
OI
Tier 1
Tier 3
Fig. 1.7 – EROI of GOM offshore extraction using two alternative treatments of
energy cost (T1 and T3) and simulated technology, 1985 – 2004.
The EROI of GOM offshore extraction ranged from 10.1 to 25.4 in 2004
depending on how comprehensively energy costs were defined (Fig. 1.7; see also
Appendix A). This range is understated because the environmental damages incurred by
the process, which include for example the degradation of wetland ecosystems and the
storm protection services that they provide, are not accounted for. Nevertheless, the
results confirm the significance of indirect energy costs and emphasize the importance of
the energy cost tier structure (Fig. 1.2). In comparison, the EROI of U.S. onshore
24
petroleum extraction in the 1930s was at least 100 (Cleveland, 1992). More recent EROI
estimates for various methods of electric power generation are given in Fig. 1.8, however,
because energy costs are not specified, comparisons must be made with caution.
Fig. 1.8 – EROI for various methods of electric power generation (Kubiszewski and
Cleveland et al., 2007).
25
0-60 m
201-900 m
61-200 m
901 m+
201-900 m
901 m+
0-60 m
61-200 m
0
10
20
30
40
50
60
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Year
ER
OI
Tier 1
Tier 3
Fig. 1.9 – EROI of GOM offshore extraction by water depth using two alternative
treatments of energy cost (T1 and T3) and constant 1985 technology, 1985 – 2004.
The productivity dynamics of the individual depth regions are presented in Fig.
1.9. Two distinct patterns prevail; the most striking is the dramatic increase in EROI for
the two deepwater regions (the primary cause of the recent increase in aggregate GOM
EROI). Because technology is held constant in this particular model run, these results
seem counterintuitive. What is increasing EROI if not technology, and what of the
depletion effect? To resolve these apparent inconsistencies, we must recall that the
deepwater processes are in early stages of development, and that the depletion effect is
predominantly a long run phenomenon. Thus, some other economic force, perhaps
economies of scale, must temporarily be at work.
26
In contrast, EROI in the 0 – 60 m and 61 – 200 m regions has steadily declined
since 1992 and 1996, respectively (see also Appendix B), indicating that resource
depletion may have just recently begun to decrease productivity in those regions.
However, when the impacts of technological progress are considered (Fig. 1.10), these
effects are almost completely offset. We therefore conclude that depletion in the GOM
has not yet significantly influenced the overall quality of the remaining reserves; thus far,
technology has provided sufficient compensation.
61-200 m
0-60 m
0
2
4
6
8
10
12
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Year
ER
OI
Constanttechnology
Simulatedtechnology
Fig. 1.10 – T3 EROI of extraction in the 0 – 60 m and 61 – 200 m regions using
constant 1985 technology and simulated technology, 1985 – 2004.
In 2004, the MMS announced that, spurred by government incentives, crude oil
and natural gas production in the GOM were expected to increase by 43% and 13%,
27
respectively, over the following ten years (Melancon et al., 2004). Substantial increases
in depletion rates along with continuously “maturing” deepwater processes may prove to
be technologically insurmountable in the near future. To illustrate the implications of a
decline in EROI for future production in the region, we extended the computer generated
GOM T3 EROI curve to create two differing productivity scenarios based on two
alternate worldviews. The first scenario is technological optimism, represented by an
increasing EROI; the second is technological skepticism, represented by a declining
EROI (Fig. 1.11).
0
2
4
6
8
10
12
1985
1987
1989
1991
1993
1995
1997
1999
2001
2003
2005
2007
2009
2011
2013
Year
ER
OI
Technologicaloptimism
Technologicalskepticism
Fig. 1.11 – Historical GOM T3 EROI (1985 – 2004) and two alternate scenarios for
the near future (2005 – 2013).
28
0
2
4
6
8
10
12
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
Year
Qu
ad
ril
lio
n B
TU
Grossproduction
Netproduction
Technologicalskepticism
Technologicaloptimism
Fig. 1.12 – Historical GOM energy production (1985 – 2004) and two alternate
scenarios for the near future (2005 – 2013).
Figure 1.12 depicts the net production curves that correspond to the two alternate
EROI scenarios; they were calculated based on the MMS gross production forecast
(Melancon et al., 2004). An increasing EROI results in the convergence of gross and net
production, while a declining EROI causes them to diverge. This emphasizes the
fundamental difference between gross and net energy: whereas gross energy is strictly a
measure of resource quantity, net energy accounts for both quantity and quality. Net
energy is therefore a superior measure of fuel supply and should be a key driver of the
overall energy strategy of a society.
29
References
Bullard, C.W., Herendeen, R.A., 1975. The energy costs of goods and services. Energy
Policy 3, 268-278.
Cleveland, C.J., 1992. Energy quality and energy surplus in the extraction of fossil
fuels in the U.S. Ecological Economics 6,139-162.
Cleveland, C.J., 2005. Net energy from the extraction of oil and gas in the United States.
Energy 30(5), 769-782.
Cleveland, C.J., Costanza, R., 1984. Net energy analysis of geopressured gas resources
in the U.S. Gulf coast region. Energy 9, 35-51.
Cleveland, C. J., Costanza, R., Hall, C.A.S., Kaufmann, R., 1984. Energy and the United
States economy: a biophysical perspective. Science 225, 890-897.
Cleveland, C.J., Hall, C.A.S., Herendeen, R., 2006. Energy returns on ethanol
production. Science 312,1746.
Costanza, R., 1980. Embodied energy and economic valuation. Science 210, 1219-1224.
Costanza, R., Herendeen, R.A., 1984. Embodied energy and economic value in the
United States economy: 1963, 1967, and 1972. Resources and Energy 6,129-164.
Cottrell, W.F., 1955. Energy and Society: The Relation between Energy, Social Change
and Economic Develolpment. McGraw-Hill, New York.
Daly, H.E., 1996. Beyond Growth. Beacon Press, Boston, Massachusetts.
Dismukes, D.E., Olatubi, W.O., Mesyanzhinov, D.V., Pulsipher, A.G., 2003. Modeling
the Economic Impacts of Offshore Oil and Gas Activities in the Gulf of Mexico:
Methods and Applications. Prepared by the Center for Energy Studies, Louisiana
State University, Baton Rouge, La. U.S. Department of the Interior, Minerals
Management Service, Gulf of Mexico OCS Region, New Orleans, La. OCS Study
MMS 2003-018.
Farrell, A.E., Plevin, R.J., Turner, B.T., Jones, A.D., O’Hare, M., Kammen, D.M., 2006.
Ethanol can contribute to energy and environmental goals. Science 311, 506-508.
Hagens, N., Costanza, R., Mulder, K., 2006. Energy returns on ethanol production.
Science 312, 1746.
30
Hammerschlag, R., 2006. Ethanol's energy return on investment: A survey of the
literature 1990 - Present. Environmental Science & Technology 40(6), 1744-1750.
Kidd, S., 2004. Nuclear: is there any net energy addition? Nuclear Engineering
International 49(603), 12-13.
Kubiszewski, I., Cleveland, C.J., Endres, P.K., 2007. Energy return on investment
(EROI) for wind energy. In: Cleveland, C.J., editor. Encyclopedia of Earth.
Leontief, W.W., 1941. The structure of American economy, 1919, 1929: an empirical
application of equilibrium analysis. Harvard University Press, Cambridge,
Massachusetts.
Melancon, M.J., Baud, R.D., Boice, A.G., Bongiovanni, R., DeCort, T., Desselles, R.,
Kazanis, E.G., 2004. Gulf of Mexico Oil and Gas Production Forecast: 2004-2013.
U.S. Department of the Interior, Minerals Management Service, Gulf of Mexico
OCS Region, New Orleans, La. OCS Study MMS 2004-065.
Minerals Management Service (MMS), 2004. Assessment of undiscovered technically
recoverable oil and gas resources of the nation’s outer continental shelf, 2003
update. MMS Fact Sheet RED-2004-01.
Tyner, G., Costanza, R., Fowler, R.G., 1988. The net-energy yield of nuclear power.
Energy 13, 73-81.
31
Appendix A
The EROI of GOM offshore extraction using two alternative treatments of energy cost
and simulated technology.
0 – 60 m 61 – 200 m 201 – 900 m 900 m+ GOM
Year T1 T3 T1 T3 T1 T3 T1 T3 T1 T3
1985 20.7 8.52 21.1 8.744 12.91 3.987 0 0 19.4 7.67
1986 22.2 9.251 21.6 8.992 13.58 4.251 0 0 20.4 8.17
1987 22.9 9.528 21.7 9.08 14.77 4.653 0 0 20.9 8.38
1988 23 9.548 21.7 9.078 15.08 4.743 0 0 21 8.39
1989 23 9.557 21.7 9.08 15.36 4.829 0 0 21 8.39
1990 23.3 9.657 21.5 8.981 16.63 5.225 0 0 21.2 8.45
1991 23.4 9.732 21.6 9.016 18.71 5.908 0 0 21.4 8.54
1992 23.5 9.82 21.7 9.112 21.25 6.73 0 0 21.7 8.68
1993 23.5 9.805 21.7 9.085 23.58 7.487 0 0 21.8 8.71
1994 23.6 9.824 21.7 9.094 25.99 8.26 0.203 0.06 21.9 8.78
1995 23.7 9.841 21.7 9.077 28.52 9.061 0.644 0.18 22.1 8.85
1996 23.7 9.857 21.8 9.122 31.5 10 1.226 0.34 22.4 8.94
1997 23.8 9.88 21.9 9.169 34.27 10.86 2.708 0.76 22.7 9.04
1998 23.9 9.912 22 9.21 37.72 11.94 5.041 1.41 23.1 9.19
1999 23.9 9.922 22.1 9.23 43.22 13.67 7.901 2.2 23.6 9.37
2000 23.8 9.874 22 9.216 47.89 15.13 11.45 3.18 23.9 9.51
32
2001 23.8 9.865 22.1 9.242 51.91 16.38 15.5 4.3 24.4 9.68
2002 23.7 9.835 22 9.192 55.01 17.36 20.46 5.66 24.8 9.83
2003 23.6 9.796 21.8 9.13 56.89 17.98 26.26 7.25 25.2 9.97
2004 23.5 9.73 21.6 9.043 57.95 18.32 31.71 8.75 25.4 10.1
33
Appendix B
The EROI of GOM offshore extraction using two alternative treatments of energy cost
and constant 1985 technology.
0 – 60 m 61 – 200 m 201 – 900 m 900 m+ GOM
Year T1 T3 T1 T3 T1 T3 T1 T3 T1 T3
1985 20.74 8.52 21.13 8.74 12.91 3.99 0.00 0.00 19.40 7.67
1986 21.83 9.10 21.21 8.84 13.37 4.19 0.00 0.00 20.09 8.04
1987 22.43 9.35 21.26 8.90 14.49 4.57 0.00 0.00 20.51 8.22
1988 22.49 9.35 21.23 8.89 14.77 4.65 0.00 0.00 20.55 8.21
1989 22.48 9.34 21.18 8.87 15.02 4.72 0.00 0.00 20.53 8.20
1990 22.64 9.40 20.91 8.74 16.18 5.08 0.00 0.00 20.60 8.22
1991 22.69 9.44 20.91 8.74 18.15 5.73 0.00 0.00 20.74 8.29
1992 22.75 9.49 21.00 8.80 20.53 6.50 0.00 0.00 20.92 8.38
1993 22.67 9.45 20.90 8.75 22.74 7.22 0.00 0.00 20.97 8.40
1994 22.62 9.42 20.82 8.72 24.93 7.92 0.19 0.05 21.04 8.42
1995 22.58 9.40 20.69 8.66 27.23 8.65 0.61 0.17 21.13 8.45
1996 22.56 9.37 20.72 8.68 29.94 9.51 1.16 0.33 21.29 8.50
1997 22.50 9.34 20.72 8.67 32.34 10.26 2.55 0.71 21.44 8.54
1998 22.42 9.30 20.65 8.64 35.28 11.17 4.71 1.31 21.64 8.61
1999 22.26 9.23 20.52 8.58 40.05 12.66 7.29 2.03 21.90 8.71
2000 21.98 9.10 20.32 8.50 43.97 13.89 10.44 2.90 22.06 8.75
34
2001 21.76 9.01 20.19 8.44 47.13 14.86 13.92 3.85 22.27 8.82
2002 21.51 8.91 19.89 8.33 49.50 15.62 18.13 5.01 22.43 8.88
2003 21.24 8.80 19.57 8.21 50.77 16.04 23.02 6.35 22.57 8.93
2004 20.91 8.66 19.21 8.06 51.26 16.21 27.49 7.57 22.60 8.94
35
Chapter 2.
On semantics and scale: Developing a dynamic multi-scale model of the ecosystem
service water regulation
Abstract
Interest in ecosystem services (i.e., the set of ecosystem functions that contribute to
human welfare) has surged in recent decades and is currently still on the rise. Although
by now these phenomena have become relatively well established and understood from a
qualitative perspective, they have yet to be universally quantified. This paper details the
initial development and testing of a dynamic, spatially explicit, computer simulation
model of the ecosystem service “water regulation”, which has been created as a part of
the broader Multi-scale, Integrated Models of Ecosystem Services (MIMES) initiative.
We discuss: 1) the conceptualization of this service as the process of hydrologic
abstraction; 2) its formalization, which extends the conventional functionality of the well
known and widely used “runoff curve number” method of the United States Department
of Agriculture; and 3) the application of the formal model to the Winooski watershed
(Vermont, U.S.A.) and to the entire globe.
1. Introduction
Ecosystem services are defined as the set of ecosystem functions that support,
directly or indirectly, human welfare (Costanza et al., 1997; Daily, 1997). Table 2.1 is a
list of ecosystem services and their corresponding ecosystem functions. The services
36
listed are fundamentally diverse in several ways. For example, some are critical to
human wellbeing (e.g., climate regulation, food production), while others simply enhance
it (e.g., recreation, aesthetics). Furthermore, ecosystem services, like ecological patterns
and processes in general, operate on a wide range of characteristic scales, from local and
regional scales (e.g., water regulation, soil formation) up to the global scale (e.g., climate
regulation). Despite these inherent dissimilarities, all ecosystem services do share an
important unifying feature: they represent benefits that bypass the monetary economy and
accrue directly to the general public, as opposed to any single private entity. Thus, they
have long been considered “free gifts of nature” according to the major doctrines of
economic value. Unfortunately, this situation has the potential to result in their
undervaluation, abuse, and subsequent destruction, i.e., the “tragedy of the commons”
(Hardin, 1968).
Table 2.1 – A list of ecosystem services and their corresponding ecosystem functions
(Costanza et al., 1997).
Number Ecosystem Service Ecosystem Functions
1 Gas regulation Regulation of atmospheric composition.
2 Climate regulation Regulation of global temperature, precipitation, and
other biologically mediated climatic processes at
global or local levels.
3 Disturbance regulation Capacitance, damping and integrity of ecosystem
response to environmental fluctuations.
37
4 Water regulation Regulation of hydrological flows.
5 Water supply Storage and retention of water.
6 Erosion control and
sediment retention
Retention of soil within an ecosystem.
7 Soil formation Soil formation processes.
8 Nutrient cycling Storage, internal cycling, processing and acquisition
of nutrients.
9 Waste treatment Recovery of mobile nutrients and removal or
breakdown of excess or xenic nutrients and
compounds.
10 Pollination Movement of floral gametes.
11 Biological control Trophic-dynamic regulations of populations.
12 Refugia Habitat for resident and transient populations.
13 Food production That portion of gross primary production
extractable as food.
14 Raw materials That portion of gross primary production
extractable as raw materials.
15 Genetic resources Sources of unique biological materials and
products.
16 Recreation Providing opportunities for recreational activities.
17 Cultural Providing opportunities for non-commercial uses.
38
Fortunately, following the recent introduction of the concept there has been a
significant burst of interest in ecosystem services. The vast majority of the initial
research has been of a subjective nature, focusing mainly on their description and
classification (e.g., MA, 2003), and especially their valuation. For example, based
primarily on preference-oriented, or “willingness-to-pay”, valuation techniques, the non-
market values of the various individual ecosystem services listed in Table 2.1 have been
estimated in hundreds of separate studies, for a range of different ecosystems. A
comprehensive synthesis of this research has suggested that the total value of the world’s
ecosystem services far exceeds the value of conventional marketed economic goods and
services (Costanza et al., 1997).
In comparison to this relative abundance of primarily qualitative information,
quantitative knowledge of ecosystem services is currently lacking. Kremen (2005) has
pointed out that measurement of the underlying biophysical role of ecology in providing
ecosystem services has generally been neglected, and that quantifying this role is an
important pending research agenda. The current Multi-scale, Integrated Models of
Ecosystem Services (MIMES) project is an ambitious attempt to build such quantitative
understanding through dynamic, spatially explicit, computer simulation modelling. By
formally representing ecosystem function in this manner, the resulting suite of models (in
combination with geographic information system (GIS) technology and web-based
facilities) will provide land managers, policymakers, and the general public with a
sophisticated, standard toolset to analyze the dynamics of the ecosystem services in any
39
particular area of interest and estimate how their value and function may change under
various management and policy scenarios.
As a part of the larger MIMES initiative, we have developed a model to measure
and map the ecosystem service “water regulation” at multiple scales. In this paper, we
discuss: 1) the conceptualization of that service as the process of hydrologic abstraction;
2) its formalization, which extends the familiar “runoff curve number” method originally
developed by the United States Department of Agriculture (USDA); and 3) the
application of the formal model to the Winooski watershed (Vermont, U.S.A.) and to the
entire globe.
2. Semantics
Quantifying an ecosystem service requires clear semantics. For example, what
exactly is the meaning of “water regulation”? Costanza et al. (1997) describe its
underlying ecological function generally as “Regulation of hydrological flows” (Table
2.1), while Farber et al. (2006) describe it simply as “Flow of water across the planet
surface”. As examples of the human-derived benefits that result from this function,
Costanza et al. provide “Provisioning of water for agricultural (such as irrigation) or
industrial (such as milling) processes or transportation”, whereas Farber et al. offer
“Modulation of the drought-flood cycle; purification of water”. While these qualitative
descriptions and examples provide simple and concise introductions to the concept, they
offer little immediate insight on how it might be measured, mapped, or formally modeled.
40
A more meaningful and unambiguous conceptualization of water regulation is
needed; we develop one here based on the above citations. First, we assume that the
“flow” referred to in both functional descriptions pertains only to freshwater hydrologic
flows occurring at the watershed scale. Second, we assume that the examples given by
Farber et al. (namely, the “modulation of the drought-flood cycle” and the “purification
of water”) are the two primary benefits to human populations that this service embodies.
It follows that direct runoff is a key biophysical force involved in water
regulation. Direct runoff is the result of rainfall excess; it occurs, for example, when the
rate of rainfall exceeds the infiltration capacity of a soil (Hortonian overland flow), or
when rainfall continues after a soil profile has become completely saturated (saturation
overland flow). The magnitude of direct runoff is dependent on a variety of physical and
meteorological factors, some of which (e.g., land cover characteristics) are highly
susceptible to anthropogenic change. In general, more direct runoff increases the peak
discharge rates of rivers and decreases groundwater recharge, thus increasing the relative
frequency of floods and droughts in a watershed. Conversely, less direct runoff decreases
peak discharge rates and increases groundwater recharge, thus tempering, or
“modulating”, the drought-flood cycle.
In addition, there is a strong connection between direct runoff and water
purification. Runoff water bypasses the natural filtration system that soils provide.
Moreover, as the water moves, it picks up various natural and human-made pollutants,
which is the primary cause of problematic nonpoint source pollution.
41
A simple expression of conservation of mass summarizes the rainfall to runoff
conversion:
Q = P – L (1)
where Q = direct runoff; P = rainfall; and L = hydrologic abstractions, which in the short
term consist primarily of interception storage, surface storage, and infiltration to the
subsurface. Based on this expression and the foregoing discussion, we conceptualize
water regulation as follows: “unregulated water” is the quantity Q of direct runoff that
results from a rainfall event, whereas “regulated water” is the quantity L, i.e., the
abstracted portion of that event. The meaning of water regulation is now clear; it is the
process of hydrologic abstraction, which can be measured universally in terms of
regulated water volume. Based on these simple, yet powerful semantics, a formal method
of measurement can be devised.
3. Scale
Scale refers to the temporal or spatial dimension of a particular phenomenon, and
scaling to the transfer of knowledge across and between scales (Turner et al., 1989).
Most ecological patterns and processes occur at certain “characteristic scales”, or
characteristic ranges of scales, which best reflect their intrinsic behavior (Clark, 1985).
Ecosystem services, for example, operate at their own distinctive scales, or ranges of
scales, at which their dynamics can be most accurately described and easily understood.
Characteristic scale is often described in terms of a typical spatial extent or
temporal frequency. However, for many natural phenomena this information is generally
42
not known a priori, and must be either detected (e.g., through fractal analysis) or
described by subjective observation (Wu, 1999). Based on the conceptualization of water
regulation presented above, it may be inferred that the process inherently operates within
the watershed range of spatial extents and at a daily event frequency.
There also exist several types of scale that may be associated with, but are not
inherent to, an ecosystem service. For example, the scale at which an ecosystem service
is modeled (i.e., the “modeling scale”) may be dependent on the scale of the policy-
making that the model is intended to inform (i.e., the “policy scale”). In general, laws
and regulations concerning a particular ecosystem service should be implemented at a
scale that is commensurate with the intrinsic scale of that service. In practice, however,
this may not always be the case. For example, environmental policies are rarely
implemented at the watershed scale; more often, they are implemented at the state,
national, or even global scales. Thus, when modeling a service such as water regulation
scaling may be required to bridge “scale gaps”.
Scaling involves altering one or both of the two primary components that
characterize scale: resolution (or grain, in spatial contexts) and extent. Resolution is the
precision of measurement, and extent is the total spatial or temporal expanse under
consideration (Tuner et al., 1989). “Upscaling”, for example, is translating information
from a finer resolution or smaller extent (i.e., a local scale) to a coarser resolution or
larger extent (i.e., a broader scale), whereas “downscaling” is translation in the opposite
direction.
43
Spatial scaling is a core problem in ecosystem service modeling because
ecosystem services are dependent on the spatial patterns of land use and land cover,
which are heterogeneous across most landscapes. Spatial scaling with dynamic models
typically consists of two distinct steps: 1) correctly defining and quantifying the fine
scale heterogeneity, and 2) properly aggregating the heterogeneous information to obtain
predictions at a broader target scale (King, 1991). Using GIS and remote sensing
technologies, spatial heterogeneity can be explicitly quantified in the form of digital maps
that become model parameters and inputs. Aggregation of this heterogeneity generally
poses a greater challenge as it involves the alteration of those parameters and inputs. Wu
and Li (2006) provide a detailed summary and comparison of a variety of dynamic
modeling spatial aggregation schemes.
4. A formal model of water regulation
Formal rainfall-runoff modeling has a relatively lengthy history. A large variety
of models and methods spanning wide ranges of conceptual, temporal, and spatial
complexities have been developed, from early single-event empirical methods, such as
the “rational method” of Mulvaney (1850), to recent continental-scale macro-models,
such as those of Vörösmarty et al. (1989) and Arnell (1999). It is notable that models and
methods at all points along this spectrum are currently in use. No single model is
appropriate for every possible task, its usefulness must be judged with the specific task at
hand in mind.
44
For the present modeling purpose, we have adopted the USDA Soil Conservation
Service (SCS) “runoff curve number” method, which is detailed in the National
Engineering Handbook Section 4: Hydrology (NEH-4) (USDA-SCS, 1985). The curve
number method is a well known and widely used empirical procedure for the estimation
of direct runoff from storm rainfall; it is based on the following rainfall-runoff relation:
Q = (P – Ia)2 / (P – Ia) + S if P > Ia; Q = 0 otherwise (2)
where Q = direct runoff (in); P = rainfall (in); S = maximum potential retention (i.e.,
rainfall not converted to runoff) that could possibly occur if rainfall were to continue
without limit, which consists mainly of the infiltration that happens after runoff begins
(in); and Ia = initial abstraction, which consists mainly of the interception, infiltration,
and surface storage that occur before runoff begins (in).
In order to remove the independent parameter Ia from Eq. (2), the SCS has
suggested a linear relationship between Ia and S:
Ia = S (3)
where = initial abstraction ratio, which has been assigned a default value of .2 based on
empirical findings ( = .2). This value, however, can vary as figures ranging from 0.0 to
0.3 have been reported (Ponce and Hawkins, 1996).
Assuming = .2 and substituting Eq. (3) into Eq. (2) yields the more familiar
version of the rainfall-runoff expression that is most commonly applied in practice:
Q = (P – 0.2S)2 / (P + 0.8S) if P > 0.2S; Q = 0 otherwise (4)
For convenience in application, the SCS has mapped S, which ranges from 0 to , to the
dimensionless parameter CN, which ranges from 0 to 100:
45
S = (1000 / CN) – 10 0 CN 100 (5)
where CN is considered constant for a particular storm event and is determined based on
the “soil cover complex”, a combination of the existing soil and land cover conditions.
The SCS has estimated CN for a variety of soil-cover complexes based on: 1) hydrologic
soil group (HSG) classified as A, B, C, or D according to the soil’s minimum infiltration
rate; 2) land cover classified according to Anderson et al. (1976); 3) hydrologic surface
condition; and 4) antecedent moisture condition (USGS-SCS, 1985).
The curve number method is thus a simple, predictable, and stable empirical
procedure for estimating the direct runoff that results from a given rainfall event.
Moreover, the relatively few required environmental inputs (HSG, land cover, rainfall)
are ordinarily available for most watersheds. Thus, the method is extremely popular,
albeit somewhat criticized (cf. Ponce and Hawkins, 1996), and is featured in many of the
hydrologic computer models currently in use, such as AGNPS (Young et al., 1987), EPIC
(Williams, 1995), and SWAT (Arnold et al., 1996).
We modified and extended the traditional curve number method in several
different ways:
1. Average hydrologic surface and antecedent moisture conditions are assumed; CN is
thus uniquely determined by the combination of HSG and land cover. Within the
MIMES framework, land cover is classified according to 11 biomes, or ecosystem
types, which together encompass the entire surface of the planet: Open Ocean,
Coastal Ocean, Forests, Grasslands, Wetlands, Lakes/Rivers, Deserts, Tundra,
Ice/rock, Croplands, and Urban. This ecological classification is an aggregation of
46
the16-biome scheme used by Costanza et al. (1997). Table 2.2 is the initial “default”
CN table for this 11-biome scheme; it was derived based on the CN tables originally
published by the SCS for the Anderson et al. classification (USGS-SCS, 1985).
Table 2.2 – Default CN parameterization for the 11-biome MIMES land cover
scheme and HSG.
A B C D
Open ocean NA NA NA NA
Coastal ocean NA NA NA NA
Forests 36 60 73 79
Grasslands 49 69 79 84
Wetlands 65 65 65 65
Lakes/Rivers 98 98 98 98
Deserts 55 72 81 86
Tundra 55 71 81 89
Ice/Rock 98 98 98 98
Croplands 67 78 85 89
Urban 89 92 94 95
2. After solving Eq. (5) and then Eq. (4) for direct runoff (Q), regulated water (L) is also
determined by subtracting Q from the given precipitation (P) based on Eq. (1). Q and
L are then converted from depths to volumes based on the given area of study.
3. The extended method is embedded in a dynamic system and run as a continuous
simulation driven by time-varying daily rainfall data, where each day represents a
47
unique storm event. Daily volumes of direct runoff and regulated water are
determined at each time-step and cumulatively summed over time. Simile, a
declarative modeling a simulation environment (Simulistics Ltd.;
www.simulistics.com) was the design tool.
4. The dynamic model is spatially distributed and augmented with GIS and remote
sensing technologies. Simile, billed as “system dynamics plus objects”, allows for
spatially explicit modeling through object-based representation. The dynamic “unit
model” is instantiated multiple times across a spatially disaggregated landscape
composed of either square unit cells or polygons. The individual instances are all
created from the same model object, but each is uniquely parameterized according to
spatially referenced data (HSG, land cover, rainfall) that correspond to its associated
geographic location. At each time-step, unique volumes of direct runoff and
regulated water are calculated across the modeled landscape and are summed to
determine area totals. Using this spatially explicit approach, the model can be
directly extrapolated (cf. King, 1991) to any spatial extent where the required data are
available.
In addition, this approach extends the format of a GIS, which can be used process,
organize, and store all spatially referenced data to be used in the model. For example,
digital land cover and HSG maps of matching homogeneous spatial units can be
overlain in a GIS to generate the soil-cover complex map, which in the model
determines CN for each spatial unit.
48
5. The spatially explicit model is adjusted to run at multiple spatial resolutions. Coarse
gaining the model (i.e., upscaling by increasing grain size) increases the size of each
spatial unit and decreases their overall number. If a larger spatial unit becomes
heterogeneous in terms of soil-cover complex, then the direct runoff for each existing
complex can be individually calculated and then weighted to determine a “composite”
estimate of runoff for the entire spatial unit:
Qcomp = (Qi Ai) / Ai (6)
where Qcomp = composite runoff (in); i = index for each soil-cover complex that exists
in the spatial unit; Ai = area of complex i, expressed as a percentage of the total unit
area (%); and Qi = estimated runoff for complex i (in). (CN weighting may also be
performed, however, because the weighting occurs prior to the runoff calculation,
aggregation error is introduced (USDA-SCS, 1985).)
Solving Eq. (6) at the broader scale requires that the fine scale soil-cover complex
information is aggregated and maintained as a percentage distribution of the total
area. Each broad scale spatial unit must be associated with an array of percentage
values, not a single scalar value as in simpler “lumping” aggregation schemes, such
as “majority rule” (Fig. 2.1). Eq. (6) thus minimizes aggregation error and can be
distributed across at any spatial resolution, from the maximum resolution (limited
only by computer resources) to a single aggregated spatial unit; only proper
aggregation and adjustment of the soil-cover complex input parameter is required.
Figure 2.2 is a simplified graphical representation of the final multi-scale model.
49
Fig. 2.1 – Two alternative methods of scaling: a) by “majority rule” and b) by
maintaining a percentage distribution.
Fig. 2.2 – Simplified model diagram.
50
5. Results and discussion
5.1. Calibration and validation
When applied to gauged watersheds, the model can be calibrated and validated by
comparing the modeled runoff to observed gage data. Because the primary purpose of
the model is to estimate the total volume of water regulated by an area over a given time
period (rather than the specific pattern of regulation within that time period), its
performance can be evaluated at the conclusion of each model run by calculating the
relative error between the total modeled runoff and the corresponding observed total.
Model calibration thus involves minimizing this error by first “coarse tuning” the model
by adjusting (Eq. (3)), and then “fine tuning” it by tweaking the various CN parameters.
We applied the model to the Winooski watershed (Vermont, U.S.A.; area ≈ 2,712
km2). The land cover pattern for this area has been mapped via remote sensing circa
1992 and 2001; the 1992 data were used for model calibration and the 2001 data for its
validation. Because the model does not yet account for the direct runoff that results from
snowmelt, the temporal extent was limited to June 1 through October 31 of each year
(153 daily time-steps). The analysis was conducted at a 330 m gain (56,100 spatial
units); the input data are briefly summarized below:
Daily mean streamflow data (ft3/s) for 1992 and 2001 were obtained from the
USGS via the National Water Information System: Web Interface
(http://waterdata.usgs.gov/nwis). They were recorded at the furthest downstream
gauging station on the Winooski River (near Essex Junction), to which the
majority of the watershed drains. The open-source Hydrograph Separation
51
Program (HYSEP) (Sloto and Crouse, 1996) was used to separate each
streamflow hydrograph into its baseflow and direct runoff components, and the
resulting direct runoff portion was used to evaluate the accuracy of the modeled
runoff.
Daily rainfall data (in) for 1992 and 2001 were obtained from the National
Climatic Data Center (NCDC). Although spatially distributed rainfall data would
have been ideal, these data are not easily obtainable for many watersheds, and
they are computationally impractical for most daily analyses. Thus, point
measurements from seven weather observation stations located throughout the
watershed were used to derive daily watershed averages, which were assumed to
fall uniformly across the modeled landscape.
Land cover data (30 m) for 1992 and 2001 were obtained from the University of
Vermont (UVM) Spatial Analysis Lab (SAL). The “Land Use/Land Cover for the
Lake Champlain Basin, Circa 1992 and Circa 2001” raster layers were previously
derived using the National Land Cover Database (NLCD 2001), ancillary GIS
data, and Landsat satellite imagery. These data were reclassified into the 11-
biome MIMES land cover scheme; the reclassified 2001 layer is displayed in Fig.
2.3.
HSG data (90 m) were also obtained from the UVM SAL. The HSG raster layer
was constructed based on the information currently contained in the “Hydrologic
Group” column of the component table of the NRCS SSURGO-2 dataset. This
layer was used for both 1992 and 2001; it is displayed in Fig. 2.4.
52
Fig. 2.3 – 2001 land cover pattern (30 m) of the Winooski watershed.
Fig. 2.4 – HSG pattern (90 m) of the Winooski watershed.
53
The results of the analysis are presented in Figs. 2.5, 2.6, and Table 2.3. The
model was most sensitive to changes in and, because the Winooski watershed is
predominantly forested, the forest cover curve numbers. These were the only parameters
that required calibration (Table 2.4).
0.00E+00
1.00E+06
2.00E+06
3.00E+06
4.00E+06
5.00E+06
6.00E+06
7.00E+06
8.00E+06
6/1/
01
6/8/
01
6/15
/01
6/22
/01
6/29
/01
7/6/
01
7/13
/01
7/20
/01
7/27
/01
8/3/
01
8/10
/01
8/17
/01
8/24
/01
8/31
/01
9/7/
01
9/14
/01
9/21
/01
9/28
/01
10/5
/01
10/1
2/01
10/1
9/01
10/2
6/01
Day
Dir
ect
ru
no
ff (
m3
/d
ay)
Modeled runoff
Observed runoff
Fig. 2.5 – Modeled and observed runoff in the Winooski watershed, June 1 –
October 31, 1992.
54
0.00E+00
1.00E+06
2.00E+06
3.00E+06
4.00E+06
5.00E+06
6.00E+06
7.00E+06
8.00E+06
6/1/
01
6/8/
01
6/15
/01
6/22
/01
6/29
/01
7/6/
01
7/13
/01
7/20
/01
7/27
/01
8/3/
01
8/10
/01
8/17
/01
8/24
/01
8/31
/01
9/7/
01
9/14
/01
9/21
/01
9/28
/01
10/5
/01
10/1
2/01
10/1
9/01
10/2
6/01
Day
Dir
ect
ru
no
ff (
m3
/d
ay)
Modeled runoff
Observed runoff
Fig. 2.6 – Modeled and observed runoff in the Winooski watershed, June 1 –
October 31, 2001.
Table 2.3 – Results of model calibration (1992) and validation (2001).
Year Grain (m) Total modeled
runoff (m3)
Total observed
runoff (m3)
Relative
error (%)
Total regulated
water (m3)
1992 330 9.58e+07 1.01e+08 -4.92 9.78e+08
2001 330 6.71e+07 6.40e+07 4.92 7.57e+08
55
Table 2.4 – Calibrated parameters.
Parameter Default value Calibrated value
.2 .044
CN Forests A 36 36
B 60 54
C 73 68
D 79 75
The resulting modeled and observed hydrographs are in reasonable agreement for
both years with a few exceptions, which are assumed to be artifacts of spatially lumping
the rainfall data. More importantly, the relative errors between total modeled and
observed runoff for both years are well below 10%, which is an acceptable error
threshold for this analysis. Based on these results, we report that the Winooski watershed
regulated 9.78e+08 m3 of water from June 1 through October 31 in 1992, and 7.57e+08
m3 of water for the same time period in 2001.
Figure 2.7 displays the spatial distribution of the total regulated volume for 2001.
The areas that provided more regulation are lightly shaded, while those that provided
relatively less are more heavily shaded. This information could be aggregated in various
ways (e.g., by town or county) and could certainly be of practical use in a variety of land
management and environmental policy contexts; for example, it could be used as input to
a “payment for ecosystem services” (PES) system.
56
Fig. 2.7 – Total water regulation (thousand m3) in the Winooski watershed from
June 1 – October 31, 2001 at 330 m grain.
5.2. Scenario analysis
With the model calibrated, a wide variety of land cover and rainfall scenarios are
possible. We ran two general land cover scenarios for 2001 with the goal of determining
a relative scale by which the watershed’s regulatory performance could be rated. A
completely forested scenario determined the scale’s upper bound (i.e., maximum total
regulation, minimum total runoff), and a completely urbanized scenario determined its
lower bound (i.e., minimum total regulation, maximum total runoff). The results (Table
2.5) indicate that from June 1 – October 31, 2001 the regulatory performance of the
57
Winooski watershed was near full potential; it regulated 86.42% of the total portion of
rainfall that was available for either regulation or runoff during that time. Using the
hydrograph corresponding to the historical land cover data as a baseline scenario, Figs.
2.8 and 2.9 emphasize the benefits that were provided in terms of the modulation of the
drought-flood cycle. This relative rating system could be effectively used to track the
regulatory performance of a single watershed over time or to compare the performance of
multiple watersheds.
Table 2.5 – Scenario analysis.
Year Land cover scenario Total regulated
water (m3)
Water regulation
service rating (%)
2001 100% forest 7.88e+08 100.00
2001 Historical pattern 7.57e+08 86.42
2001 100% urban 5.58e+08 0.00
58
0.00E+00
1.00E+06
2.00E+06
3.00E+06
4.00E+06
5.00E+06
6.00E+06
7.00E+06
8.00E+06
6/1/
01
6/8/
01
6/15
/01
6/22
/01
6/29
/01
7/6/
01
7/13
/01
7/20
/01
7/27
/01
8/3/
01
8/10
/01
8/17
/01
8/24
/01
8/31
/01
9/7/
01
9/14
/01
9/21
/01
9/28
/01
10/5
/01
10/1
2/01
10/1
9/01
10/2
6/01
Day
Dir
ect
ru
no
ff (
m3
/d
ay)
100% forested
Historical data
Fig. 2.8 – Modeled runoff in the Winooski watershed June 1 – October 31, 2001
using two alternate land cover scenarios: historical data and completely forested.
0.00E+00
5.00E+06
1.00E+07
1.50E+07
2.00E+07
2.50E+07
6/1/
01
6/8/
01
6/15
/01
6/22
/01
6/29
/01
7/6/
01
7/13
/01
7/20
/01
7/27
/01
8/3/
01
8/10
/01
8/17
/01
8/24
/01
8/31
/01
9/7/
01
9/14
/01
9/21
/01
9/28
/01
10/5
/01
10/1
2/01
10/1
9/01
10/2
6/01
Day
Dir
ect
ru
no
ff (
m3
/d
ay)
100% urban
Historical data
Fig. 2. 9 Modeled runoff in the Winooski watershed June 1 – October 31, 2001
using two alternate land cover scenarios: historical data and completely urbanized.
59
5.3. Scale effects
To investigate scale effects, i.e., changes in the results of an analysis caused by
changing the scale at which the analysis is conducted, the model was coarse-grained and
executed multiple times using resampled 2001 data (Fig. 2.10). The scale effects (Table
2.6) were essentially negligible. Although grain was increased to almost 20 times its
initial size (resulting in a significantly greater decrease in the number of spatial units), the
model lost very little precision; relative error changed by a total of only .10%. Moreover,
the model did not require re-parameterization at each new spatial resolution.
These results represent a very favorable tradeoff in terms of computer resources.
Spatially distributed models often suffer from excessive computational demand, which
increases exponentially with the number of spatial units. A very large number of spatial
units can be a major problem even for the most advanced computing facility. We have
shown that coarse graining this model can alleviate its computational demand at the
expense of very little precision. Thus, the model can be applied at whatever resolution is
appropriate for the location at hand and the computer resources available.
60
a) b)
c) d)
Fig. 2.10 – Total water regulation (thousand m3) in the Winooski watershed, June 1
– October 31, 2001 at 4 grain sizes: a) 450 m, b) 900 m, c) 2700 m, and d) 6000 m.
Table 2.6 – Scale effects.
Grain
(m)
Number of
spatial units
Total modeled
runoff (m3)
Total observed
runoff (m3)
Relative
error (%)
Total regulated
water (m3)
330 56,100 6.71e+07 6.40e+07 4.92 7.57e+08
450 30,098 6.71e+07 6.40e+07 4.91 7.57e+08
900 7,575 6.71e+07 6.40e+07 4.90 7.57e+08
2700 850 6.71e+07 6.40e+07 4.84 7.57e+08
6000 192 6.71e+07 6.40e+07 4.82 7.57e+08
61
5.4. The global application
Although there may be several conceptual and empirical problems inherent in
applying this model at the global scale, we proceeded with caution for two primary
reasons: 1) to emphasize the need for mapping and modeling ecosystem services at all
scales; and 2) to provide an initial “default” estimate (however crude it might be) of the
global pattern of water regulation. The analysis was performed at a 1 deg grain (50,400
spatial units); the data that were used are as follows:
A global land cover (1 km) raster layer, classified according to the 11-biome
MIMES classification scheme, was obtained from the UVM SAL. This layer
represents a reclassification of the Global Land Cover 2000 (GLC2000) dataset.
Because a direct mapping from the 22-class GLC2000 scheme did not exist, a
series of geo-processing tasks and ancillary data sets (e.g., ETOPO2 global digital
elevation model, World Wildlife Fund Terrestrial Ecoregions) were employed; the
resulting layer is displayed in Fig. 2.11.
A global HSG (5 arcmin) raster layer was also obtained from the UVM SAL. To
generate this layer, an intermediate global dataset indicating the percent
composition of clay in topsoil, relative to sand and silt, was derived from the Food
and Agriculture Organization (FAO) “Digital Soil Map of the World”. These
intermediate data were then reclassified according to the following rules: 1 – 10%
= A; 11 – 20% = B; 21 – 40% = C; and 41 – 100% = D. Global HSG is displayed
in Fig. 2.12.
62
Fig. 2.11 – Global land cover pattern (1 km).
Fig. 2.12 – Global HSG pattern (5 arcmin).
Global precipitation data were not used in this analysis. Instead, the model was
run in a static context assuming 2 in of uniform rainfall across the entire surface of the
planet. Water regulation in each spatial unit is expressed as the regulated percentage of
the 2 in event (Fig. 2.13). These results are not calibrated and were obtained using the
63
model’s default parameter settings (Table 2.2). However, they do represent a first
approximation of global water regulation and can be used to make general comparisons
across the globe.
Fig. 2.13 – Global water regulation (1 deg) as the regulated percentage of a 2 in
rainfall event.
6. Conclusion
As the concept of ecosystem services continues to gain momentum, the need to
develop a standard quantitative understanding of these services is becoming more critical.
While the model presented in this paper is certainly in need of additional refinement,
testing, and review, it is a successful “proof of concept” and represents an important
initial step towards building such an understanding. It also highlights the importance of
the broader MIMES initiative and demonstrates the large amount of research and
development that will be needed to bring that project to fruition. The significance of such
64
research is profound, as it may largely determine the precision with which we can
manage and conserve ecosystem services in the future.
65
References
Anderson, J.R., Hardy, E.E., Roach, J.T., Whitmer, R.E., 1976. A land use and land
cover classification system for use with remote sensor data. Geological Survey
Professional Paper 964. U.S. Government Printing Office, Washington D.C.
Arnell, N.W., 1999. A simple water balance model for the simulation of streamflow over
a large geographic domain. Journal of Hydrology 217, 314-335.
Arnold, J.G., Williams, J.R., Srinivasan, R., King, K.W., 1996. SWAT: Soil and Water
Assessment Tool. USDA-ARS, Grassland, Soil and Water Reseach Laboratory,
Temple, TX.
Clark, W.C., 1985. Scales of climate impacts. Climatic Change 7, 5-27.
Costanza, R., d'Arge, R., de Groot, R., Farber, S., Grasso, M., Hannon, B., Limburg, K.,
Naeem, S., O'Neill, R. V., Paruelo, J., Raskin, R. G., Sutton, P., Belt, M. van den,
1997. The Value of the World's Ecosystem Services and Natural Capital. Nature 38,
253-260.
Daily, G.C., 1997. Nature's Services: Societal Dependence on Natural Ecosystems.
Island Press, Washington D.C.
Farber, S., Costanza, R., Childers, D.L., Erickson, J., Gross, K., Grove, M., Hopkinson,
C.S., Kahn, J., Pincetl, S., Troy, A., Warren, P., Wilson, M., 2006. Linking ecology
and economics for ecosystem management. Bioscience 56 (2), 117-129.
Hardin, G., 1968. The tragedy of the commons. Science 162, 1243-1248.
King, A.W., 1991. Translating models across scales in the landscape. In: Turner, M.G.
and Gardner, R.H., editors. Quantitative Methods in Landscape Ecology. Springer-
Verlag, New York, 479-517.
Kremen, C., 2005. Managing ecosystem services: What do we need to know about their
ecology? Ecology Letters 8, 468-479.
MA (Millennium Ecosystem Assessment), 2003. Ecosystems and Human Well-Being: A
Framework for Assessment. Island Press, Washington, D.C.
Mulvaney, T.J., 1850. On the use of self-registering rain and flood gauges in making
observations on the relations of rainfall and flood discharges in a given catchment.
Transactions, Institution of Civil Engineers of Ireland 4 (2), 18-31.
66
Ponce, V.M., Hawkins, R.H., 1996. Runoff curve number: Has it reached maturity?
Journal of Hydrologic Engineering 1(1), 11-19.
Sloto, R.A., Crouse, M.Y., 1996. HYSEP: A computer program for streamflow
hydrograph separation and analysis. U.S. Geological Survey Water-Resources
Investigations Report 96-4040, 46.
Turner, M.G., Dale, V.H., Gardner, R.H., 1989. Predicting across scales: theory
development and testing. Landscape Ecology 3, 245-252.
USDA-SCS, 1985. National Engineering Handbook: Section 4: Hydrology. Soil
Conservation Service, USDA, Washington, D.C.
Vörösmarty, C.J., Moore, B., Gildea, M.P., Peterson, B., Melillo, J., Kicklighter, D.,
Raich, J., Rastetter, E., Steudler, P., 1989. A continental-scale model of water
balance and fluvial transport: Application to South America. Global Biogeochemical
Cycles 3, 241-265.
Williams, J.R., 1995. The EPIC model. In: Singh, V.P., editor. Computer Models of
Watershed Hydrology. Water Resources Publications, Highlands Ranch, CO, 909-
1000.
Wu, J., 1999. Hierarchy and scaling: extrapolating information along a scaling ladder.
Canadian Journal of Remote Sensing 25, 367-380.
Wu, J., Li, H., 2006. Perspectives and methods of scaling. In: Wu, J., Jones, K.B., Li,
H., Loucks, O.L., 2006. Scaling and Uncertainty Analysis in Ecology: Methods and
Applications. Springer-Verlag, New York, 17-44.
Young, R.A., Onstad, C.A., Bosch, D.D., Anderson, W.P., 1987. AGNPS, Agricultural
Non-Point Source Pollution Model: A Watershed Analysis Tool. USDA
Conservation Report 35. USDA-ARS, Washington, D.C.
67
Literature Cited
Anderson, J.R., Hardy, E.E., Roach, J.T., Whitmer, R.E., 1976. A land use and land
cover classification system for use with remote sensor data. Geological Survey
Professional Paper 964. U.S. Government Printing Office, Washington D.C.
Arnell, N.W., 1999. A simple water balance model for the simulation of streamflow over
a large geographic domain. Journal of Hydrology 217, 314-335.
Arnold, J.G., Williams, J.R., Srinivasan, R., King, K.W., 1996. SWAT: Soil and Water
Assessment Tool. USDA-ARS, Grassland, Soil and Water Reseach Laboratory,
Temple, TX.
Bullard, C.W., Herendeen, R.A., 1975. The energy costs of goods and services. Energy
Policy 3, 268-278.
Clark, W.C., 1985. Scales of climate impacts. Climatic Change 7, 5-27.
Cleveland, C.J., 1992. Energy quality and energy surplus in the extraction of fossil
fuels in the U.S. Ecological Economics 6,139-162.
Cleveland, C.J., 2005. Net energy from the extraction of oil and gas in the United States.
Energy 30(5), 769-782.
Cleveland, C.J., Costanza, R., 1984. Net energy analysis of geopressured gas resources
in the U.S. Gulf coast region. Energy 9, 35-51.
Cleveland, C. J., Costanza, R., Hall, C.A.S., Kaufmann, R., 1984. Energy and the United
States economy: a biophysical perspective. Science 225, 890-897.
Cleveland, C.J., Hall, C.A.S., Herendeen, R., 2006. Energy returns on ethanol
production. Science 312,1746.
Costanza, R., 1980. Embodied energy and economic valuation. Science 210, 1219-1224.
Costanza, R., Herendeen, R.A., 1984. Embodied energy and economic value in the
United States economy: 1963, 1967, and 1972. Resources and Energy 6,129-164.
Costanza, R., Ruth, M., 1998. Using dynamic modeling to scope environmental
problems and build consensus. Environmental Management 22, 183-195.
68
Costanza, R., Wainger, L., Folke, C., Maler, K. G., 1993. Modeling complex ecological
economic systems: toward an evolutionary, dynamic understanding of people and
nature. BioScience 43, 545-555.
Costanza, R., d'Arge, R., de Groot, R., Farber, S., Grasso, M., Hannon, B., Limburg, K.,
Naeem, S., O'Neill, R. V., Paruelo, J., Raskin, R. G., Sutton, P., Belt, M. van den,
1997. The Value of the World's Ecosystem Services and Natural Capital. Nature 38,
253-260.
Cottrell, W.F., 1955. Energy and Society: The Relation between Energy, Social Change
and Economic Develolpment. McGraw-Hill, New York.
Craik, K., 1943. The Nature of Explanation. Cambridge University Press, Cambridge.
Daily, G.C., 1997. Nature's Services: Societal Dependence on Natural Ecosystems.
Island Press, Washington D.C.
Daly, H.E., 1996. Beyond Growth. Beacon Press, Boston, Massachusetts.
Diamond, J., 2005. Collapse: How Societies Choose to Fail or Succeed. Viking Press,
New York.
Dismukes, D.E., Olatubi, W.O., Mesyanzhinov, D.V., Pulsipher, A.G., 2003. Modeling
the Economic Impacts of Offshore Oil and Gas Activities in the Gulf of Mexico:
Methods and Applications. Prepared by the Center for Energy Studies, Louisiana
State University, Baton Rouge, La. U.S. Department of the Interior, Minerals
Management Service, Gulf of Mexico OCS Region, New Orleans, La. OCS Study
MMS 2003-018.
Farber, S., Costanza, R., Childers, D.L., Erickson, J., Gross, K., Grove, M., Hopkinson,
C.S., Kahn, J., Pincetl, S., Troy, A., Warren, P., Wilson, M., 2006. Linking ecology
and economics for ecosystem management. Bioscience 56 (2), 117-129.
Farrell, A.E., Plevin, R.J., Turner, B.T., Jones, A.D., O’Hare, M., Kammen, D.M., 2006.
Ethanol can contribute to energy and environmental goals. Science 311, 506-508.
Forrester, J., 1968. Principles of Systems. MIT Press, Cambridge, MA.
Forrester, J., 1971. World dynamics. MIT Press, Cambridge, MA.
Gentner, D., Stevens, A.L. (Eds.), 1983. Mental Models. Lawrence Erlbaum Associates,
Hillsdale, NJ.
69
Hagens, N., Costanza, R., Mulder, K., 2006. Energy returns on ethanol production.
Science 312, 1746.
Hammerschlag, R., 2006. Ethanol's energy return on investment: A survey of the
literature 1990 - Present. Environmental Science & Technology 40(6), 1744-1750.
Hardin, G., 1968. The tragedy of the commons. Science 162, 1243-1248.
Johnson-Laird, P.N., 1983. Mental Models: Toward a Cognitive Science of Language,
Inference and Consciousness. Harvard University Press, Cambridge, MA.
Kidd, S., 2004. Nuclear: is there any net energy addition? Nuclear Engineering
International 49(603), 12-13.
King, A.W., 1991. Translating models across scales in the landscape. In: Turner, M.G.
and Gardner, R.H., editors. Quantitative Methods in Landscape Ecology. Springer-
Verlag, New York, 479-517.
Kremen, C., 2005. Managing ecosystem services: What do we need to know about their
ecology? Ecology Letters 8, 468-479.
Kubiszewski, I., Cleveland, C.J., Endres, P.K., 2007. Energy return on investment
(EROI) for wind energy. In: Cleveland, C.J., editor. Encyclopedia of Earth.
Leontief, W.W., 1941. The structure of American economy, 1919, 1929: an empirical
application of equilibrium analysis. Harvard University Press, Cambridge,
Massachusetts.
Levins, R., 1966. The strategy of model building in population biology. American
Scientist 54, 421-431.
MA (Millennium Ecosystem Assessment), 2003. Ecosystems and Human Well-Being: A
Framework for Assessment. Island Press, Washington, D.C.
MA (Millennium Ecosystem Assessment), 2005. Ecosystems and Human Well-Being:
Synthesis. Island Press, Washington, D.C.
Meadows, D.H., Meadows, D.L., Randers, J., Behrens, W.W., 1972. The Limits to
Growth. Universe, New York.
Melancon, M.J., Baud, R.D., Boice, A.G., Bongiovanni, R., DeCort, T., Desselles, R.,
Kazanis, E.G., 2004. Gulf of Mexico Oil and Gas Production Forecast: 2004-2013.
U.S. Department of the Interior, Minerals Management Service, Gulf of Mexico OCS
Region, New Orleans, La. OCS Study MMS 2004-065.
70
Minerals Management Service (MMS), 2004. Assessment of undiscovered technically
recoverable oil and gas resources of the nation’s outer continental shelf, 2003
update. MMS Fact Sheet RED-2004-01.
Mulvaney, T.J., 1850. On the use of self-registering rain and flood gauges in making
observations on the relations of rainfall and flood discharges in a given catchment.
Transactions, Institution of Civil Engineers of Ireland 4 (2), 18-31.
Ponce, V.M., Hawkins, R.H., 1996. Runoff curve number: Has it reached maturity?
Journal of Hydrologic Engineering 1(1), 11-19.
Sloto, R.A., Crouse, M.Y., 1996. HYSEP: A computer program for streamflow
hydrograph separation and analysis. U.S. Geological Survey Water-Resources
Investigations Report 96-4040, 46.
Tainter, J., 1988. The Collapse of Complex Societies. Cambridge University Press,
Cambridge.
Turner, M.G., Dale, V.H., Gardner, R.H., 1989. Predicting across scales: theory
development and testing. Landscape Ecology 3, 245-252.
Tyner, G., Costanza, R., Fowler, R.G., 1988. The net-energy yield of nuclear power.
Energy 13, 73-81.
USDA-SCS, 1985. National Engineering Handbook: Section 4: Hydrology. Soil
Conservation Service, USDA, Washington, D.C.
von Bertalanffy, L., 1968. General System Theory: Foundations, Development,
Applications. George Braziller, New York.
Vörösmarty, C.J., Moore, B., Gildea, M.P., Peterson, B., Melillo, J., Kicklighter, D.,
Raich, J., Rastetter, E., Steudler, P., 1989. A continental-scale model of water
balance and fluvial transport: Application to South America. Global Biogeochemical
Cycles 3, 241-265.
Williams, J.R., 1995. The EPIC model. In: Singh, V.P., editor. Computer Models of
Watershed Hydrology. Water Resources Publications, Highlands Ranch, CO, 909-
1000.
Wu, J., 1999. Hierarchy and scaling: extrapolating information along a scaling ladder.
Canadian Journal of Remote Sensing 25, 367-380.
71
Wu, J., Li, H., 2006. Perspectives and methods of scaling. In: Wu, J., Jones, K.B., Li,
H., Loucks, O.L., 2006. Scaling and Uncertainty Analysis in Ecology: Methods and
Applications. Springer-Verlag, New York, 17-44.
Young, R.A., Onstad, C.A., Bosch, D.D., Anderson, W.P., 1987. AGNPS, Agricultural
Non-Point Source Pollution Model: A Watershed Analysis Tool. USDA
Conservation Report 35. USDA-ARS, Washington, D.C.