A Proactive Approach for Particulate Matter
Air Pollution Management
Stephanie Fincher and Krystyna Stave
Department of Environmental Studies
University of Nevada, Las Vegas
4505 S. Maryland Parkway
Box 454030
Las Vegas, NV 89154-4030
Main (702) 895-4833 - Fax (702) 895-4436
Steph_fincher@ hotmail.com and Krystyna.stave@ unlv.edu
Abstract
This paper analyzes the management approach used in the Las Vegas Valley to manage
particulate matter (PM) pollution, demonstrates that system dynamics concepts can
improve the current strategy, and proposes a more proactive approach to management.
A retroactive policy analysis, beginning in 1960, was performed to analyze the benefits
and tradeoffs of using a system dynamics approach. The analysis showed that including
a system dynamics perspective improves the utility of the model for policy analysis.
Analysis supports the hypothesis that a proactive approach to management could have
prevented PM exceedances in the Valley, and provides greater flexibility in managing the
problem, but in some cases may have prohibitively high initial and/or sustained costs.
Keywords: air quality management, proactive management, reactive management, particulate
matter, air pollution, rapidly growing urban areas, sustainable development
I. Problem Statement
Introduction
Although the Clean Air Act (CAA) of 1970 has led to improvements in air quality over
the past few decades, problems with air pollution and air quality management still exist (National
Research Council [NRC] 2004, EPA-4 2003). The concentrations of pollutants throughout the
United States on average have decreased, but in some areas concentrations remain above
standards (NRC 2004). Air quality management in the U.S. is often characterized by a short-
term perspective that focuses on meeting CAA requirements. Additionally, since the system is
constantly changing -- politically, socially, and physically -- managers often find themselves in a
situation of crisis-management. As we have learned through many applications of system
dynamics, such situations can lead to counter-intuitive behavior (Forrester 1995, Sterman 2000).
One troublesome element of current air quality management is particulate matter
pollution, ten micrometers (10m) or smaller in diameter (Environmental Protection Agency
[EPA] 1996B, EPA 2004, Department of Air Quality and Environmental Management
[DAQEM] FAQ). PM consists of extremely small solid or liquid particles, made up of a great
variety of minerals and chemicals (EPA 1996B, EPA-4 2003, EPA 2004, DAQEM). Over 300
counties did not meet PM standards when standards were first established in 1971 (Chay,
Dobkin, and Greenstone 2003). In 1992, the Environmental Protection Agency (EPA) changed
the status of eight nonattainment areas from moderate to serious (EPA-10 2007). In 2006, there
were still eight serious nonattainment areas
Figure 1 Map of Clark County, Nevada for PMj and over 75 moderate
nonattainment areas across the United
States (NRC 2004, EPA-10 2007).
Clark County
Nan-
= “eels nals In this study, we use the case of
HT particulate matter pollution in the Las
ear ~ Vegas Valley (LVV) to examine the
2 benefits of using system dynamics for
policy making. The LVV, located within
Clark County, Nevada as shown in Figure
1, may have local geologic, geographic and
meteorological characteristics reinforcing
PM pollution problems in the area, and
rapid urban development has played a key
role in creating the problem that has
plagued the area for over 30 years. The current management approach in Clark County has been
focused on responding to changes in legislation and growth in the area, leading to the trends
described in the following section.
Source: CBC 2001 (SIF)
PM Trends in the LVV
The current national standard requires that PM 9 concentrations not exceed an average of
150 micrograms per cubic meter (ug/m’) in any 24-hour period (EPA 2004, NRC 2004)!.
Monitored concentrations above this limit are called “exceedances,” (EPA-7 1999). An area
with regular exceedences is considered to be a non-attainment area (NRC 2004, Kubasek and
Silverman 2005, EPA-7 1999). In 1993, the LVV was declared a serious non-attainment area
(CCBC 2001).
Figure 2 shows the reference mode of historic and projected PM1o concentrations in the
LVV. The trends show PMipo levels exceeding standards for several years but presently on a
downward trend that should stabilize in the future. These concentrations are based on both
monitoring data as well as estimates in EPA documents (Fed. Reg. 69:54006, 2004). Even
though trends currently show a decrease and may drop below standards in the near future, the
long history of PM; management problems provides an excellent case for examining air quality
management in rapidly growing areas.
' The annual standard was recently discarded by the U.S. Environmental Protection A gency (EPA) for lack of
sufficient evidence relating long term average concentrations to significant health effects (EPA-8 2006).
Figure 2 Reference Mode for PM 24-hour Standard
Reference Mode: 24hr PM Concentration Range
3007
150>===
Bstimated yaigd \: \!
fe{\) i oe
Al ie
PM Concentration (g/m)
Time (Y ear)
PM Impacts
2005 2025
In 2003, 97 counties in the U.S. had monitored levels of PM pollution above either the
PMio or PM? 5 standards or both— this represents 62 million people exposed to very unhealthy
levels of PM pollution (EPA-4 2003). PM1o particles are inhaled into the lungs where they
accumulate in the bronchia and can cause increased incidence of coughing, painful breathing,
and decreased lung function, aggravation and increased potency of pre-existing respiratory
conditions (e.g. asthma), increased absences from work and school, area-wide increased hospital
admissions and emergency room visits, and premature death (CCBC 2001, EPA-2 2003, EPA-4
2003, Lippmann 2003).
Figure 3 Increases in Daily Mortality Based on PM Pollution
25> ’
o 4
Percentage increase
ot r 1 1 r
o 25 50 75 100 125
PM concentration (wg/m®)
150 175 © 200
=+= Mean (PM...) Upper confidence limit
Lower confidence limit
— Mean (PM,,)
Source: Schwela 2003
Significant epidemiological
evidence has demonstrated that
that the dose-response curve for
mortality is linear as shown in
Figure 3 and there is no threshold
for PMio (Schwela 2003). Less
severe health effects have an even
steeper relationship and are much
more likely to occur. Any change
in PM10 pollution poses
considerable consequences for
human health (Schwela 2003).
PM can also cause aesthetic
deterioration to an area through
haze, reduced visibility, and
physical damage to building
surfaces (EPA-4 2003). Degradation of vegetation and entire ecosystems can also be caused by
PM pollution (EPA-4 2003). Exceeding federal PM1o standards can be very costly in terms of
increased procedural burdens, potential loss of federal highway funds, and forced adoption of
increasingly expensive (some with marginal benefits) control strategies.
PMio pollution is not unique to the United States. It is a problem being faced by many
countries, especially rapidly-developing areas (McGranahan 2003). While other air pollutants
are very dangerous to human health, both cohort studies and time-series studies have concluded
that premature deaths from air pollution are caused predominately by PM as opposed to other
criteria pollutants (Molina and Molina 2004).
Current Management Strategies for PM
There has been little improvement in non-attainment areas, indicating either persistent
problems in these areas or insufficiencies in the current management strategy. The general
management strategy for air quality includes the development of a state implementation plan
(SIP) when an area exceeds standards. SIPs describe the non-attainment area’s characteristics,
present monitoring data, detail emission sources, and describe any mitigating actions or controls
an area will implement to stay below standards (EPA-3, Plater et al. 1998, NRC 2004).
Although standards will inevitably change, managers tend to respond to new regulations as they
occur instead of planning for continual air quality improvement and anticipating those changes.
Standards have typically become more stringent with time (NRC 2004, EPA 2004), yet air
quality goals in most areas are usually set at these levels and not below (NRC 2004).
Therefore, when standards are changed, a crisis-management situation is sparked—
managers rush to complete new documents and requirements while attempting to simultaneously
lower emissions. Coupling this with the fact that air quality systems are slow to change (both
due to chemical and physical inertia and to the time necessary to develop, implement, and
enforce new regulations on industry and individuals), the result is often that a given area is
classified as a non-attainment area. This paper proposes that a system dynamics approach could
help managers anticipate changes in standards, develop more proactive management strategies,
and potentially avoid non-attainment classification.
Proportional Rollback Model
The Clark County Department of Air Quality and Environmental Management
(DAQEM) developed a model in support of the 2001 SIP for demonstrating that PM1o in the
LVV would be below standards for the year 2006. The major limitations of the original format
of the model are detailed in Fincher and Stave (2006) and include a fragmented structure, a
manual and error-prone process for running policy analysis, limited policy options, static
representation of causes (usually exogenous), unclear representation of controls and other
calculation, and exclusion of several significant mechanisms.
The model is an empirical rollback model, using observed relationships between pollutant
concentrations and emissions and not representing many chemical and physical processes
causing pollutant levels (NRC 2004). Functionally, the original model consisted of a series of
independent spreadsheets that required manually copying and pasting calculations from one
sheet to another. The newer version developed in Fincher and Stave (2006) has a more user-
friendly, explicit, and integrated context, although it still relies on the original underlying
assumptions and calculation methodology to determine emissions.
To determine 24-hour emissions the model uses a “design day”. The design day is
defined as a day with normal conditions (i.e. wind speeds are assumed to be low and there is no
precipitation). The model does not calculate PM1o levels on different days and so does not
represent a continuous trend in emissions but rather shows how conditions on the one
representative day would change in response to policy changes.
Figure 4 Causal Loop Diagram (CLD) of DAQEM Proportional Rollback Model
design ~ controlled PM10
concentration + Poncentralion.
reduction in mass + controlled PM10<
issi emissions =~
emissie 44 overall control
Y reduction
PMi0emision
+ + a vacant land
duration of emissions +
construction if
vacant land
construction emissions factor
emission factors
+
a _ vacant land
acres demanded ee
construction
rat +
change in *
population
land consumption
factor (people/acre)
Figure 4 shows the causal loop diagram representing the structure of the Proportional
Rollback Model. The major driver of emissions is population (an exogenous input table).
Increases in population increase the number of acres in construction and thus raise emissions
from construction activities. Balancing emissions and construction is the depletion of vacant
land over time. As the amount of vacant land decreases, land-based emissions decrease, which
decreases total emissions.
Fincher and Stave (2006) describes how the Proportional Rollback Model structure was
converted into a system dynamics representation. The main purpose of the Proportional Rollback
Model is to determine the concentration that would result from an already designed policy.
Although alternative policies could be tested to determine the preferred choice, the model was
not designed to be used for policy development. The major limitations of this model for policy
analysis included restricted policy options, a short time horizon, high sensitivity and poor
response to extreme tests (and even many reasonable policy changes). Results do not provide a
context for understanding the given concentration and how policies are affecting pollution with
time.
This paper shifts the focus to exploring the benefits of using a system dynamics approach
for managing PMio. A system dynamics model was developed for the case of PMio in the Las
Vegas Valley, as described in the following section. It was hypothesized that a system dynamics
approach would allow a better representation of feedback, more policy testing and evaluating
assumptions, and help managers better understand the system instead of focusing on point
estimates. The paper describes the model development and hypothesis testing.
Il. Model Development
Physical Site Characteristics
The non-attainment area of the Las Vegas Valley covers roughly 4,000 km? (DRI 2002).
There is great diversity of land classifications and uses in the area, which causes a variety
impacts on air quality. Valleys often have more persistent and problematic air pollution issues
than areas without mountains (CDSN and DAQEM 2003), since mountains act as physical
barriers, trapping air and thereby slowing dispersion of pollutants (Spellman 1999). DAQEM
estimates that particles are settle within four kilometers of their sources (CDSN and DAQEM
2003, EPA-4 2003). PMyp travels relatively short distances, ranging from <1 to ten kilometers
(Lippmann 2003).
The LVV has distinct seasons and strong winds. Winter and spring winds affect large
areas while summer winds have more localized effects (Gorelow 2005). Wind both removes
particulate matter from and adds it to land surfaces (CDSN and DAQEM 2003. In dry, calm
weather and without input from other sources, PM10 is balanced between suspension and settling
(Lippmann 2003). In winter months, the LVV is subject to inversions and low wind velocities,
which trap pollutants (CDSN and DAQEM 2003). PMio concentrations follow seasonal patterns
due to these annual fluctuations.
Emissions from large areas of land are a major problem for Clark County and one of the
major reasons why previous SIPs were not approved (DAQEM). When desert land is in its
natural state fugitive dust emissions are low, but disturbance to the desert crust, such as
disturbance for urban construction, results in high particulate emissions (CDSN and DAQEM
2003). Chow et al. (1999) reported that fugitive dust accounted for 80-90% of all PMio
emissions in residential areas.
Boundaries of the Model
The key variables are identified in Table 1. PM pollution varies seasonally depending on
weather components such as temperature, humidity and wind (EPA-4 2003). Although
temperature and atmospheric pressure may control how air rises and falls (Spellman 1999), the
processes controlling these conditions are quite complex and beyond the level of detail needed in
this regional policy-making model. Daily temperature fluctuations are also not represented since
night and day variations would average when looking at an entire day.
Table 1 Key variables by sector
Sector Endogenous Exogenous Omitted
PM10 Stable PM jo on surface Normal removal & Other meteorological
Unstable PM 9 on surface settling rates factors
PM in air Wind factor and Spatial
rate of disturbance and normal rain eventt dispersion/hot
stabilization Height of boundary spots
Area source emissions layert PM io characteristics
Mobile source emissions —_ Silt loading factors (e.g.,
Volume of the air shed Point sources subcomponents,
Actual removal & settling Emission factors* chemicals)
rates Control reductions*
Land Native Desert acres Designed density Spatial variation
Unstable acres Land stabilization time
Stable acres Emission factors*
Acres in construction Control reductions*
Developed/Urban area
Residential capacity*
Annual construction
demand
Disturbance rate
Population People desiring to move _ Birth rate Sensitive populations
to LVV Normal death rate Population
Population in LVV characteristics
Residential capacity* (e.g., age, sex)
actual in-migration and
out-migration
actual death rate
Attractiveness
Transportation Paved road Lanemiles normal planned Types of roadways
actual planned acres of
roads and support
Unpaved roads
Unpaved shoulders
personal trips per person
per day
Vehicle miles traveled
Effective lanemile
capacity
roadway demand
obligatory trips per
person per day
Emission factors*
and lanemiles
(e.g., freeway,
arterial)
* crosses sectors {+ seasonal
The model is not spatially distributed. It is not intended to analyze specific “hot spots”.
Its focus is regional management of particulate matter. It aggregates particulate matter across the
entire region. However, there is great variation in localized PM levels resulting from buildings
and especially nearby land use (E.g, construction sites), making some monitoring stations prone
to higher recorded levels. Chow et al. (1999) showed concentrations differing by a factor of five
for sites experiencing similar meteorological conditions but located in different areas.
Precipitation, wind, and boundary layer height were chosen as parameters representing
essential meteorological effects. The first two parameters affect PM addition and removal
processes, while the boundary layer height is vital for determining the volume of air in the valley
and thus the concentration of PMo. This model is not intended to be either a meteorological or
predictive model. Therefore, average seasonal values, based on historical trends, were used for
these factors. For precipitation, seasonal information included the maximum probability of rain
days for each month (with a minimum of zero). The height of the boundary layer depends on the
valley's depth as well as the intensity of radiative cooling (Spellman 1999). The value of this
parameter is driven by many complex meteorological processes but tends to follow a seasonal
trend. Therefore, an average boundary layer height for each month of the year was developed
based on historical experience.
Using data from a monitoring site, the minimum and maximum average daily wind speed
for each month were used to generate a random daily wind speed using a beta distribution. This
was then divided by the annual mean derived for over 40 years of NOAA data (see Gorelow
2005), giving a wind factor seasonally varying around 1.0. The wind factor modifies certain
normal emissions factors to determine an actual emission factor.
PM Sources in the LVV
Emissions are divided according to their source as area, point, or mobile emissions
(Solomon 1994). Area sources in the LVV include vacant land emissions, emissions from land
disturbed by construction activity, and minor emissions such as residential firewood burning.
Mobile sources include direct emissions from vehicles, brake dust, and particles that are
entrained (emitted) from road surfaces (DAQEM). PMio emissions from paved roads are
entrained by vehicles but the source of dust is actually nearby area sources (RTC 2004).
However EPA tracks the source as the actual physical manner of entrainment and not the source
of particles. PMio can build more rapidly on paved surfaces near construction sites or
construction- or off-road vehicles tracking dust onto surfaces (RTC 2004).
The major PM1o sources in the LVV for 2001 are shown in Figure 5 and include vacant
land dust (36%), construction activities (27%), paved road dust (26%), unpaved road dust (8%),
point and other area sources (2%), and mobile source exhaust (1%). The major industries
(tourism, gaming, defense, chemical manufacture, sand and gravel operations, utilities, and
construction; DAQEM) are not pollution intensive, except as indirectly encouraging longer
commute distances (DRI 2002). Although managers realize that the distribution of sources will
change, plans and control strategies are implemented following this static breakdown of sources.
Figure 5 Emissions for Clark County
PM,, Emissions Inventory (2001)
Paved Road.
‘Dust
26%
Constmetion
Point & Area
Sources
2%
Mobile Source
Exhaust
1%
Vacant Land
——Dust
36%
Unpaved Road
Dust
8%
Model Structure
Representing PMjo in the atmosphere at its simp!
lest requires tracking addition and
removal processes, as shown graphically in Figure 6. PM1o is added by emissions processes of
direct human disturbance or secondary entrainment from wind. Pollution in the air eventually
returns to the surface through deposition. The two primary deposition processes are washout and
rainout, with particles attaching to water droplets, or dry
deposition, commonly referred to as
settling or fallout (Spellman 1999, Society for Risk Analysis). Another removal and addition
process includes PM pollution transported in or out of the area. This is considered a minor
source because PM 1p does not travel far in suspension.
Figure 6 Stock and Flow Diagram of Simple PMi» model.
Oo | PM10 in
particles air particles
added removed
Additions Processes (Emissions) Removal Processes
Anthropogenic physical disturbances Dry deposition
to surface (construction, paved road Washout
entrainment, off-road activity, concrete and
gravel operations)
Fires (human structures and wild/land)
Industrial activities/ release
Wind events
Transport (only changes
location of pollution)
In general terms, the structure of the model is similar to that of the Proportional Rollback
model: growing population drives PM-emitting activities, which increase the amount of
pollutants in the air. However, rather than using an average conversion factor to convert mass
PM 0 (tons) to a concentration (ug/m*) as was done in the Proportional Rollback model, the CLD
shown in Figure 7 represents concentration as a function of the mass of PM 1p particles in the air
and the volume of air in the LVV.
Figure 7 CLD of SD model structure
boundary layer
height
surface area of NE
LW. ¥ volume of air in
LV
~ PM10
concentration
+
average wind 6 Se
speed PM emission __ a
factors ~ mass PM10 10} .
f inair __ ‘
4+; _—»jobs~ R3. xe
. Zz Oe PSs
controls PM-emitiing, —» attfactiveness of
Mig _
_ activities “7 ——~ scresfa available + the LVV
at ~ construction, ¥ services :
4 R2
Bl) w \
oe — Population in t-
vacentland — LW }
B3(.) roadway
* capacity
ss _ 1 RL
vehivle miles ———— w
traveled .
+ OP, ongestion
trips per _ ——
person ~—_—
The PMjo concentration depends on the mass PMjo in air and volume of air in LVV,
which depends on the boundary layer height and the surface area of LVV. The mass particulate
matter is a function of emission activities, which are reduced by controls. PM-emitting activities
and PM emission factors also depend on wind speeds. The majority of emissions are from
vacant land and acres in construction. However, as acres in construction increases, they
decrease vacant land forming a balancing loop (B1). Acres constructed depend on the
population which is driven by attractiveness factors. There is a dotted line connecting PM10
concentration to the attractiveness (B2) because increased pollution does not necessarily slow
growth (by reducing attractiveness), although it would for some.
As the population grows, there are increased vehicle miles traveled (VMT). These miles
add to PM-emitting activities, which increase the concentration of particulate matter, decrease
the attractiveness and therefore decrease the population change, forming the next balancing loop
(B3). As VMT increases, congestion also increases, which reduces attractiveness, population
change, and VMT, thereby forming balancing loop B4. City planners recognize the impact of
congestion and so as the population grows, there are more acres in construction for roads which
increases road capacity, decreases congestion and increases attractiveness, forming the first
reinforcing loop (R1).
Planners are not the only ones who respond to congestion. As congestion increases,
individuals reduce the number of unnecessary trips decreasing trips per person which decreases
VMT and therefore relieves traffic, (B5). Additionally, growth is also driven by availability of
services as development progresses, which increases the attractiveness, further increasing
population and leading to more acres constructed. Likewise, construction and increased urban
10
development leads to more jobs which increases attractiveness and leads to more in-migration
and further development.
The model is simulated in days. The time horizon chosen for this model is from 1960 to
2025 to incorporate the historic trends of development and management that lead to the current
situation for particulate matter pollution. The major sectors of the model include land
development, population, and PMi. The major processes that occur in each of these sectors are
shown in Figure 8. Land is goes through a process of development changing it from vacant, to
under construction, and finally developed. Particles are either on the surface or suspended in the
air. Population in the LVV grows when people desiring to move to the Valley migrate, which
depends on development of space. The population affects transportation (which crosses both the
population and land sectors) and the desiring population drives demand for construction.
Figure 8 Major sectors in model
The land and demand for land sectors of the model are subscripted according to
construction project type (such as airports, commercial, residential homes, and so forth). Figure
9 shows the stock-and-flow representation of the land sector of the model. Vacant land is
represented as either “Native Desert”, “Stable Land” or “Unstable Land.” “Annual acres
constructed” is determined by factors such as acres of services required per capita and grows
with time. These acres are allocated across the three land stocks and flow into the “Acres
ordered backlog” stock where they await construction.
From “Acres ordered backlog,” acres are either limited by “acres of construction
permitted” or simply remain backlogged before moving into “acres in primary (active)
construction,” defined as the disturbance-intensive part of construction activities with major
earth-moving operations. The duration spent in this stock depends on the level of disturbance of
the construction project and the total duration of the project. A similar flow moves land into
“acres in secondary construction” where emissions from construction activities are greatly
11
reduced. Acres then finish construction and become part of the completed “Urban/developed
area” stock. A percentage of annual construction is reconstruction of built land which takes
“Urban/developed area” acres and puts them back into the “Acres ordered backlog” stock, where
they begin the construction cycle again. It is assumed that acres are only reconstructed for the
same type of project (i.e. from commercial to commercial acres), based on land use zoning.
Emissions are based on acres in each of these stocks, with the exception of “Urban/Developed
Area” for which only highway acres are used.
Figure 9 Land Sector of SD model
Native Desert x construction
e acres to be permitted
actiffies Kg ‘ounce onND |] actual % acres
onND
Initial acres ordered
vosllou total developed
Stable Land mae ame
senna “a constructed on SL Ase >) aE seconary
| onlerer rag construction
Sing backlog | ar gerng ffs | "Tees phase = aes stn | |
ins design % [a comemuton. “> construction
actual% |] construction daysin
acreson Sl SL secondary | UsbanDeveloped
Unstable Land average planting and days active constuction Area
acres to be peo tite construction
constructed onUL design 9
actual % acres constructiod [UL
onUL
reconstruction rate—we, 2°78 Se for
Teconstruction
affect of vacant land on daily demand for
acres constructed land by type
4 Sg % construction
| NOT vacant land
\ % construction
affect of vacant land on anmal acres: on Vacant Land
acres constructed constructed
LOOKUP
Validation
Model results are shown in Figure 10 and replicate the behavior shown in the reference
mode in Figure 2. Fluctuations come mostly from weather factors including wind and
precipitation. Variables were compared to historic and estimated data from local planning and
management entities to check validity. Figure 11 shows the model output for vacant land since
the 1960’s. These curves are relatively close for all historic data but begin to level off for future
estimates.
12
Figure 10 Output from Base Run of Model
24-hour PM10 Concentration
300
lil
1960 1970 1980 1990 2000 2010 2020
Year
"24hr PM10 concentration": Base —*—a——+— __ug/(day*m*m*m)
Figure 11 Validation of Vacant Land
Vacant Land
300,000
225,000 Ps
150,000
a
75,000 Ny
0 Ph
1960 1970 1980 1990 2000 2010 2020
Year
vacant land : Base Acres
Estimates Comp. Planning 2-2-2 BB BB ACTS
Sensitivity
The model was tested for sensitivity to certain parameters as well. Since most variables
driving the removal of PMio from the air were based on estimates this was the first test
performed. The results for this test are shown in Figure 12 and show that these variables may
greatly influence levels and that it would be worthwhile to investigate specific rates of settling,
washout, and transport. However, because it is accepted that the majority of particles settle
within the LVV, the higher estimates are unlikely because they assume all of the lowest settling
rates at one time and presume that around 60 to 80 percent of emissions stay in the air at all
times.
13
Figure 12 Sensitivity of PM) in the air to removal rates
base concent
50% 75% (NNN 95% I 100%
Mass PM10 inair
2,000
1,500
1,000 ills
500
5936 11871 17807 23742
Time (day)
Population trends were also analyzed to determine their dependence on the socio-
economic factors driving in- and out-migration. The results for this analysis are shown in
Figure 13 Sensitivity analysis of population based Figure 14 Sumulagve Cost sensitivity
. 9 on attractiveness factors
50% 75% 95% IN 100% on
Population in LVV 50% 75% (95% 100% (a
4M cumulative costs
2B
3M 15B
2M 1B
500M.
1M
0
0 5936 11871 17807 237.
0 Time (day)
0 3936 T1871 17807 73742
‘Time (day)
. Population follows the same trend for the majority of the cases, but does level off at different
points. Again, many of the lower estimates for attractiveness effects factors could be removed
since they would not be able to replicate the population trends that were seen historically.
14
Figure 13 Sensitivity analysis of population based
50% 75% NN 95% 100% [IN
Population in LVV
4M
3M
2M
1M
5936 11871 17807 23742
Time (day)
Figure 14 Cumulative cost sensitivity
on attractiveness factors
base
50% 75% [NF95% 100%
cumulative costs
2B
15B
0 5936
11871 17807 237.
Time (day)
Another important area for determining sensitivity is costs. The range of costs for each control
method comes from the 2001 State Implementation Plan (CCBC 2001). The sensitivity of costs
is shown in
and shows the upper and lower limits of costs. The high and low estimates of
costs will give a range of costs, but when an average of all costs is chosen the simulation results
are basically in the center of the range. Therefore, the average value was set for all cost
variables, although policy-makers may be interested in knowing the maximum possible value
they may have to pay which can vary up to around an extra $200 M.
Ill. Results
One of the major benefits of the SD model is that it allows for policy analysis dating back
to 1960 and projecting to 2025. This gives more perspective in determining the effects of
controls and development strategies on PMio concentrations in the Valley. Although several
tests were performed using the model, only a selected few are presented here.
The first policy test was to keep the same policies that were set in 2001 as a result of the
SIP process, but set the policy implementation year 10 years earlier. This test was performed
three times, each time implementing the policy a decade earlier back to 1970. The result of these
tests are shown in Figure 15. The results show not only a decrease in the overall length of time
that standards were exceeded but also a reduction in the overall magnitude of the problem.
15
Figure 15 Results of Implementing Same Controls Earlier
24-hour PM10 Exceedances
nhl OE ental
100 ais
pie
1960 1970 1980 1990 2000
Year
2010 2020
“annual #of 24hr exceedances" : 1970 11111 violations
“annual #of 24hr exceedances" : 1980 2 z 2 z 2 violations
“annual #of 24hr exceedances" : 1990 violations
2001 (base) 4-444 violations
The impacts of implementing the strategy in 1970 would have kept PMio levels below
standards, thereby reducing the number of deaths resulting from PMio exposure by a couple
hundred thousand individuals as represented in Figure 16.
However, these decreases in PMo levels and cumulative deaths come at a fairly high cost
as demonstrated in Figure 17 and Figure 18. Policy enforcement and strategy begins earlier
thereby increasing overall costs, while daily costs of implementing policies increase with time
until they are about the same as the current policy.
Figure 16 Cumulative Deaths from PM, Exposure
Cumulative Deaths from PM10 Exposure
; Lee
1960 1970 1980 1990 2000 2010 2020
current year
cumulative deaths from PM10 exposure :1970 people
cumulative deaths from PM10 exposure : base people
16
Figure 17 Cumulative Costs of Implementing Figure 18 Daily Costs of Implementing Controls in 1971
Policies in 1971
Cumulative Costs Daily Costs
28 150,000
158 112,500
1B 75,000
500M ae ail 37,500 ae all
i Ler et ;
1960 1970 1980 1990 2000 2010 2020 19601970 1980 1990 20002010 2020
current year current year
cumulative costs : 1970 $ daily control costs : 1970 ———________________._ $day
cumulative costs : base $ daily control costs : base ——-—-__ $day
Another major policy test is to examine the major sources of particulate matter as a trends
over time. Figure 19 shows how sources of emissions can change. In contrast to the static
distribution of sources as seen earlier in Figure 5, different sources may dominate at different
times. Unstable land emissions account for the vast majority of emissions (using the emission
factors provided by the DAQEM) until vacant land decreases in later years and then mobile
emissions are higher.
Figure 19 Major Sources of Emissions Through Time
Major Emissions Contributors
1960 1970 1980 1990 2000 2010 2020
Time (Year)
unstable land emissions : base t 5 t t tons/day
const. wind erosion 2 2. 2. 2 2 tons/day
const. activity 3 3 3 3 3 tons/day
stable land emissions :base —4 4 4 4 tons/day
Total Mobile sources 5 5 5 5 tons/day
Additionally, including seasonal factors allows managers to test the effects of
implementing extra controls during these seasonably high concentration times. Figure 20 shows
the results of introducing a seasonal control.
17
Figure 20 PM from Seasonal Controls
24-hour PM10 Concentration
300
225
150 | .
1960 1970 1980 1990 2000 2010 2020
Year
"24hr PM10 concentration" : seasonalcontrol —————_ ug/(day*m*m*m)
Status Quo ug/(day*m*m*m)
Combination policies were also tested to determine whether it would be possible to have
a proactive policy that could have avoided the magnitude of the peak in emissions without
dramatically increasing costs. Figure 21 shows the concentration resulting from a combination
policy, giving levels below standards from the 1980s onward. The resulting decrease in deaths is
shown in Figure 22, and reduced costs— both annual and cumulative— in Figure 23.
Figure 21 Concentration of combination policy Figure 22 Cumulative Deaths Status Quo v.
24-hour PM10 Concentration
Combination
Cumulative Deaths from PM10 Exposure
am
3M
2M Loe
1M : ee
é cee
19601970 19801990 2000 2010 2020
Time (Year)
Cumulative deaths from PM10 exposure : combo1-re people
1960 1970 1980 1990 2000 2010 2020 Status Quo people
Year
"24hr PM10 concentration" : combol-re ———————_ug/(day*m*m*m)
Status Quo ug/(daytm*m*m)
18
Figure 23 Annual and Cumulative Costs Comparison
Annual Costs Cumulative Costs
80M 28
60M 15B
40M 1B EE ie
20M ape 500M ee
rr N Bae
0 | 0 Pal
19601970 1980 1990 2000 2010 2020 1960 1970 1980 1990 2000 2010 2020
current year current year
Annual costs mem : combol-re ——________________ § Cumulative costs : combol-re $
Status Quo $ Status Quo §
IV. Discussion
Introducing systems concepts into the PMio management decision support system
provides more flexibility for policy testing, incorporates feedbacks and consequences of policies
for both PMjo concentrations and development of the LVV. Results have greater utility than the
Proportional Rollback Model (PRM) when used for policy testing. Including seasonal controls
allows for testing a variety of policies as well as better isolating problems. Although costs were
not tied in specifically to the seasonal control test, this would be a beneficial area to explore.
The narrow scope of the PRM important feedbacks which could potentially lead to the
kind of policy resistance described by Sterman (2000) and policies further exacerbating the
problem. One example is the interplay of vacant land and emissions. In the PRM representation,
increasing the rate of vacant land development speeds the transition to the “Urban/Developed
Area” and reduces emissions from vacant and constructed land. This appears to solve the
problem of fugitive dust from vacant land areas. However, in reality, there are a host of other
problems associated with rapid conversion of vacant to built land that keep this from being an
ideal strategy. Sprawl leads to greater distances traveled per vehicle-trip, increased congestion
and time in traffic, increasing the total vehicle-miles traveled and vehicular emissions. These
include two other pollutants which the LVV is currently listed as non-attainment status: carbon
monoxide and ozone.
The SD model gives managers information for comparing costs and effectiveness of
control strategies. It provides more information than the Proportional Rollback model provides.
It includes a variety of policies that can be tested. Additionally, the explicit representation of
the causal structure makes it easy for policies requiring structural changes to be easily added to
the model. The SD model also allows for learning about how changes to the system influence a
variety of variables, hopefully improving the understanding of managers and allowing for better
questions to be asked of the model.
The graph showing the major contributors confirms that unstable or disturbed land is the
major reason why PMyo levels were so high historically. It also points to the major leverage
points in the system at different stages of development. Since unstable land is so important,
determining how the number of developed acres grew helps show what could be done to avoid
problems caused by rapid development. A major particulate matter contributor was residential
19
disturbance of vacant land (through offroad vehicle use, for example). Tests changing the
development rate showed a strong leverage point.
The SD model also makes assumptions and relationships between variables explicit.
These assumptions and relationships can be easily updated to incorporate improved
understanding or parameter values. The model’s flexibility in what can be tested, the ease with
which this can be done, and the ability to represent a variety of variable types also make these
models useful for managers. The ability to examine different time horizons is also useful feature
of the SD model. As the retroactive policy analysis of the Las Vegas Valley shows, there may be
ways in which a proactive strategy can improve air quality and prevent exceedances.
While this analysis demonstrates considerable benefits of a proactive systems-based
management approach, several barriers to the use of causal models in air pollution management
exist. First, because managers must meet regulatory requirements in a timely manner, it may be
difficult to find the time or support to embrace and begin new techniques for decision-making.
There may also be an additional burden of proof that areas must undergo to demonstrate the
method as valid. Secondly, inclusion of soft variables into models is still not widely accepted
despite the significant uncertainty in readily accepted meteorological data. Nevertheless, there is
strong support of benefits, even from a retroactive application. Current non-attainment areas, or
those that may soon become non-attainment areas, stand to gain the most from a proactive
approach. A system dynamics approach can help focus the problem, examine major
assumptions, and develop policies that will help improve or avoid future problems.
20
V. References
Aragon-Correa J.A. 1998. Strategic Proactivity and Firm Approach to the Natural Environment.
Academy of Management J ournal 41(5): 556-567.
Barrow, Christopher J. 1999. Environmental Management: Principles and Practice. London,
UK: Routledge, p 4.
Ben-Ami D., D. Ramp, and D. Croft. 2006. Population viability assessment and sensitivity
analysis as a management tool for the peri-urban environment. Urban Ecosyst. 9: 227-
241.
Brewer, Garry D.(Editor). 2005. Decision Making for the Environment: Social and Behavioral
Science Research Priorities. Washington, DC, USA: National Academies Press.
Brown, R.S. and K. Marshall. 1996. Ecosystem Management in State Governments. Ecological
Applications 6(3):721-723.
Chow, J., J.G. Watson, M.C. Green, D.H. Lowenthal, D.W. DuBois, S.D. Kohl, R.T. Egami, J.
Gillies, C.F. Rogers, and C.A. Frazier. 1999 Jun. “Middle and Neighborhood Scale
variations of PM-10 Source contributions in Las Vegas, Nevada.” Journal of Air and
Waste Management Association. 49:641-655.
Clark County Board of Commissioners (CCBC), Office of the County Manager, Department of
Comprehensive Planning, Clark County Health District, Air Quality Planning Committee.
2001 Jun. PM-10 State Implementation Plan for Clark County.
Clark County Department of Air Quality and Environmental Management (DAQEM) FAQ
http://www.co.clark.nv. Accessed: 10/15/05.
Clark County Department of Air Quality and Environmental Management (DAQEM). 2001.
2001 NAMS/SLAMS Network Review Report.
Clark County Department of Air Quality and Environmental Management (DAQEM). 2005.
2005 NAMS/SLAMS Report.
Committee on Air Quality Management in the United States. 2004. Air Quality Management in
the United States. Washington DC: National Academies Press.
Conservation District of Southern Nevada (CDSN) and Clark County Department of Air Quality
and Environmental Management (DAQEM). 2003. Help Keep Our Air Clean: Las Vegas
Valley Resident’s Guide to Improving Our Air Quality, 2" Edition. Ed. W. Daniels.
DAQEM-see Clark County Department of Air Quality and Environmental Management.
Dietz, T. 2003. “What is a Good Decision? Criteria for Environmental Decision Making.”
Human Ecology Review 10 (1): 33-39.
EPA-1- US Environmental Protection A gency - Region 9. 2002. “Technical Support Document:
Proposing Approval of the PM-10 State Implementation Plan for the Clark County
Serious PM-10 Non-attainment Area Annual and 24-Hour PM-10 Standards.”
EPA-2- Environmental Protection Agency - Region 9. 2003 Jan. “FACT SHEET: Proposed
Approval of Clark County Serious Area PM o Plan for the Las Vegas Metropolitan Non-
attainment A rea”.
EPA-3- US. Environmental Protection Agency. 1993. “The Plain English Guide to the Clean Air
Act”. Available: http://www.epa.gov/oar/oaqps/peg_caa/pegcaain.html#index. Accessed:
11/2/05.
EPA-4 US Environmental Protection A gency. 2004. The Particle Pollution Report: Current
Understanding of Air Quality and Emissions through 2003.
21
EPA-5 US Environmental Protection A gency. 2006 Dec 05. “Classifications of Particulate
Matter (PM-10) Non-attainment Areas.” The Green Book.
EPA-6 US Environmental Protection A gency. 2003. Paved Roads (section 13.2.1). [draft] AP-42.
Pp. 13.2.1-1 - 13.2.1-15.
EPA-7 US Environmental Protection A gency. 1999. Guideline on Data Handling Conventions
for the PM NAAQS. Office of Air Quality Planning and Standards.
EPA-8. US Environmental Protection A gency. 2006. “PM1o Standards Revision— 2006.”
Available online: http://epa.gov/pm/naagsrev2006.html Accessed: Jan 2007.
EPA-9. US Environmental Protection Agency. Office of Air Quality Planning and Standards.
1995 (with electronic updates through 2006). A P-42: Compilation of Air Pollutant
Emission Factors, Volume 1: Stationary Point and Area Sources. Fifth Edition.
EPA-10. U.S. Environmental Protection A gency. 2007 (Accessed Dec 05-May 07).
“Classifications of Particulate Matter (PM-10) Non-attainment Areas.” The Green Book.
Federal Register. "Approval and Promulgation of Implementation Plans; Nevada-Las V egas
Valley PM-10 Non-attainment A rea; Serious Area Plan for A ttainment of the Annual and
24-Hour PM-10 Standards, Final Rule." Federal Register 69 (9 July 2004): 32273-32277.
Federal Register. “Approval and Promulgation of Implementation Plans; New Source Review;
State of Nevada, Clark County Department of Air Quality and Environmental
Management.” Federal Register. 69 (7 September 2004): 54006-54019.
Fincher, S. and K. Stave. 2006. Managing PM1o in the Las Vegas Valley. Paper presented at the
international conference of the System Dynamics Society, July 23-27, in Nijmegen, The
Netherlands.
Forrester, J.W. 1995. “Counterintuitive Behavior of Social Systems.” Originally inJ.W.
Forrester, Collected papers of Jay W. Forrester. Cambridge, Mass. : Wright-Allen, 1975:
211-244.
Foundations of Success (FOS), Wildlife Conservation Society, and Conservation International.
2004 Jun 15. A synthesis of the evolution of MGE in conservation: Results of the
Measuring Conservation Impact Initiative: Discipline-Specific Results of the Measuring
Conservation Impact Initiative. Foundations of Success: Bethesda, MD, USA.
Gorelow, A. 2005. Climate of Las Vegas, Nevada. Revision to 1999 P. Skrbac Version.
Henriques, I. and P. Sadorsky. 1999. The Relationship Between Environmental Commitment
and Managerial Perceptions of Stakeholder Importance. Academy of Management
Journal. 42 (1): 87-99.
Kriebe, D., J. Tickner, P. Epstein, J. Lemons, R. Levins, E.L. Loechler, M. Quinn, R. Rudel, T.
Schettler, M. Stroto. 2001. The Precautionary Principle in Environmental Science.
Environmental Health Perspective. 109(9): 871-876.
Kubasek, Nancy and Silverman, Gary. 2005. Environmental Law, 5" Edition. Upper Saddle
River, NJ: Prentice Hall.
Lee, K. 1993. Compass and Gyroscope: Integrating Science and Politics for the Environment.
Washington, D.C.: Island Press. Chs. 1, 3, and 4.
Lippmann, M.. 2003. “Air Pollution and Health— Studies in the Americas and Europe.” In Air
Pollution and Health in Rapidly Developing Countries. ed. G. McGranahan, 35-48.
Toronto, Canada: Earthscan Canada.
McGranahan, G., Ed. 2003. Air Pollution and Health in Rapidly Developing Countries. Toronto,
Canada: Earthscan Canada.
22
Meeting Notes. November 2, 2005. Source: John Koswan, Department of Air Quality and
Environmental Management.
Miles, Raymond E. 2003. Organizational Strategy, Structure, and Process. Palo Alto, CA,
USA: Stanford University Press.
Molina M. and L. Molina. 2004 “Megacities and Atmospheric Pollution” Journal of Air and
Waste Management Association. 54:644-680.
Plater, Zygmunt, R. Abrams, W. Goldfarb, R. Graham Esq. 1998. Environmental Law and
Policy: Nature, Law, and Society, 2™ Edition. St Paul, MN: West Group.
Reichman, 0.J. and H.R. Pulliam. 1996. The Scientific Basis for Ecosystem Management.
Ecological Applications, 6(3):694-696.
Schwela, D. 2003. “Local Ambient Air Quality Management.” In Air Pollution and Health in
Rapidly Developing Countries. ed. G. McGranahan, 68-88. Toronto, Canada: Earthscan
Canada.
Sharma, S. and Vredenburg H. 1998. Proactive Corporate Environmental Strategy and the
Development of Competitively Valuable Organizational Capabilities. Strategic
Management J ournal. 19 (8): 729-753
Society for Risk Analysis. Webpage. www.sra.org accessed Feb. 2, 2007.
Solomon, P.A., Ed. 1994, Planning and Managing Regional Air Quality: Modeling and
Measurement Studies. Pub by Lewis Publishers and Pacific Gas & Electric Co.
Spellman, F.R. 1999. The Science of Environmental Pollution. Lancaster, PA.: Technomic
Publishing.
Sterman, J. 2000. Business Dynamics: Systems Thinking and Modeling for a Complex World.
Boston: Irwin/McGraw Hill.
Sterman, J. 2001. System Dynamics Modeling: Tools for Learning in a Complex World.
California Management Review. 43 (4): 8-25.
System Dynamics Society. Homepage: Systemdynamics.org.
United States Census Bureau. 2006.
Van den Belt, M. 2004. Mediated Modeling: A System Dynamics Approach to Environmental
Consensus Building. Island Press: Washington, DC.
Ventana Systems, Inc. Vensim Modeling Guide. 2004. Available online or with software.
23
