Journal of Theoretical and Applied Computer Science
ISSN 2299-2634
Vol. 7, No. 1, 2013, pp. 56-69
http://www.jtacs.org
Air quality modeling in Warsaw Metropolitan Area
Piotr Holnicki, Zbigniew Nahorski
Systems Research Institute of the Polish Academy of Sciences, Warsaw
{holnicki,nahorski}@ibspan.waw.pl
Abstract:
Decision support of air quality management needs to connect several categories of the input
data with the analytical process of air pollution dispersion. The aim of the respective model
of air pollution is to provide a quantitative assessment of environmental impact of emission
sources in a form of spatial/temporal maps of pollutants’ concentration or deposition in the
domain. These results are in turn used in assessment of environmental risk and supporting
respective planning actions. However, due to the complexity of the forecasting system and
the required input data, such environmental prognosis and related decisions contain many
potential sources of imprecision and uncertainty. The main sources of uncertainty are
commonly considered meteorological and emission input data. This paper addresses the
problem of emission uncertainty, and impact of this uncertainty on the forecasted air pollution concentrations and adverse health effects. The computational experiment implemented
for Warsaw Metropolitan Area, Poland, encompasses one-year forecast with the year 2005
meteorological dataset. The annual mean concentrations of the main urban pollutants are
computed. The impact of uncertainty in emission field inventory is also considered. Uncertainty assessment is based on the Monte Carlo technique where the regional scale CALPUFF model is the main forecasting tool used in air quality analysis.
Keywords: air quality model, urban-scale emission inventory, uncertainty analysis
1. Introduction. Modeling of air pollution transport
Forecasting models of air quality and the more complex Integrated Assessment Systems
(IAM) are recently used for supporting decisions concerning air quality management and
emission control policy [1, 2, 3, 4, 5, 13, 16]. The natural application of environmental
models is predicting dispersion of pollutants, analysis of ecological results of some specific
meteorological conditions or evaluation of environmental impact of emission sources. To
quantify possible ecological, economic or health benefits of emission abatement, there is a
need to estimate an incremental contribution of the respective group of emission sources to
ambient concentrations with a reasonable accuracy. However, due to a very complex, multidisciplinary structure of such systems as well as a sophisticated structure of the mathematical model, there exist many sources of imprecision and uncertainty [1, 5, 6, 7, 11, 19, 20] in
modeling of environmental effects of atmospheric pollution and also in the resulting regulatory decisions.
The problem related to urban air pollution is high in the priorities of environmental concern. Numerous studies of model outputs and measurement data have shown that most of air
pollution models poorly describe both temporal and spatial dependencies of pollutant concentrations [1, 2, 18, 20]. Estimation of the urban-scale pollution is a computationally sophisticated modeling problem due to complexity of emission field, but also, due to
Air quality modeling in Warsaw Metropolitan Area
57
complicated building orography and wind-field effects. Emission inventory of urban areas
usually encompasses different categories of emission sources, characterized by specific
emission parameters. Varieties of primary pollutants generate secondary compounds, by
means of chemical transformation processes, which may be even more dangerous for the
environment. Due to high population density, urban air pollution exposure is a crucial factor
associated with numerous adverse health effects. In particular, many research results indicate that a considerable harm in public health is caused by fine particulate matter, especially
PM2.5. Some results concerning of this type of environmental impact can be found in [9].
It is known that official emission data are not accurate, due to inventory uncertainties related to some categories of urban emissions. Emissions of major power plants of energy
sector can be treated as relatively accurate because of well specified parameters of combustion process as well as those of the fuel used. On the other hand, emission data that characterize residential area or transportation system are usually based on some aggregated and
averaged information related to the fuel consumption and parameters. These categories of
emission data do reflect neither the real temporal variability nor chemical constitution of
polluting compounds and are remarkably uncertain. In complex uncertainty analysis, correlation between some pollutants emitted by a source [19, 20, 22] should also be taken into
account.
Figure 1. Main processes in air pollution transport (based on [24])
In basic applications of air quality models the processes of air pollution transport and
transformations (see Fig. 1) are considered as a distributed parameter system, which is governed by the set of transport equations. The exact form and structure of the model usually
depends on its practical application, type of the polluting compounds which are considered,
and the scale of modeling. The model applied usually takes into account the input data
(emission field inventory and meteorological data) as well as the main physical and chemical processes which decide on the transport in the atmosphere and transformations of air
58
Piotr Holnicki, Zbigniew Nahorski
pollution components. The mathematical description of these processes within time interval,
related to one polluting compound and a single vertical layer of the atmosphere, usually
takes a general form of an advection-diffusion equation (see e.g. [8, 12, 13, 15, 24] and
many more publications)
r
∂ci
r
+ ∇ ⋅ Uci = ∇ρ D∇(ci / ρ ) + Ri (c1 , c 2 , K , cn , t ) + S i ( x , t ),
∂t
for
(1)
i = 1, 2, K , n,
along with the respective boundary and initial conditions. Each equation of the set (1) rer
lates to i-th polluting compound, where: ci denotes concentration; U – wind field vector;
D – turbulent diffusion coefficient; Ri – chemical transformation rates of pollutants;
Si – emission/reduction rate of a specific pollutant at a given spatial and temporal location;
ρ – air density. Most of parameters depend on the actual meteorological conditions.
Implementation of the used model relates to the requirements of an application and to
the pollutants which are considered. But the basic types of air quality models can differ significantly in approach to the analysis of equations (1) and also to the scale of modeling. Spatial and temporal scales of the environmental impact of air pollution are correlated with and,
moreover, they directly depend on the lifetime of pollutant [8, 12, 15] . This parameter can
differ significantly between compounds. Thus, depending on the analysis scale, there are
respective categories of modeling: local, regional and global.
Regarding the practical application as well as the scale of modeling, the most common
types (implementations) of air pollution models are (see, for example [13, 15, 23, 24]):
− Gaussian model – based on a simplified, analytical solution of the transport equations,
originally used mainly on a local scale analysis [9, 15]. However, the new generation of
Gaussian models is now available, where variability of main meteorological fields is
taken into account, with models also used on regional scale. An example is a generalized, non-steady state Gaussian puff dispersion model CALPUFF [23]. This model –
which includes chemical transformations, wet and dry deposition, complex terrain algorithms, building downwash, and other effects – was applied in computations for Warsaw
Metropolitan Area which are described in the following sections of this paper.
− Lagrangian model – where the related trajectory of an air polluting parcel is observed
and analyzed, according to the wind field and other meteorological parameters. Mathematical description takes a form of the respective set of ordinary differential equations
(source-oriented approach). The advantage of this method is a natural ability to assess
the individual environmental impact of selected emission sources via transfer matrices.
This approach is utilized in the analysis of emission abatement strategy (compare, e.g.
[4, 8, 14, 24]).
− Eulerian model – mathematically governed by finite-dimensional approximation of
equations (1), where a modeling region is horizontally and vertically discretized into the
respective number of cells. Parameters of numerical scheme (temporal and spatial discretization steps) must be accordingly set, to satisfy stability and monotonicity conditions [8, 12, 21 24]. Implementations usually include evolution of pollutant
concentrations, including advection, diffusion, chemistry, sedimentation and deposition.
This category of models, characterized by high computing requirements, is utilized in
the most complex regional and multi-scale implementations (receptor-oriented approach).
Air quality modeling in Warsaw Metropolitan Area
59
2. Air pollution forecasts for Warsaw Metropolitan Area
The computations performed in the framework of the study relate to the forecasts and
analysis of air pollution dispersion in Warsaw agglomeration. The aim was to evaluate the
environmental impact of the main categories of emission sources as well as to estimate uncertainty of this forecast, which is related to the uncertainty of emission field inventory. The
regional scale, Gaussian puff dispersion model CALPUFF v.5 (Earth Tech, Inc.) was utilized to simulate the air pollution transport and transformations within the domain. This a
new generation, a generalized non-steady-state air quality modeling system which includes
chemical transformations, dry and wet depositions, building downwash, plume fumigation,
and other effects. The main forecasting model is integrated with meteorological module
CALMET which includes a diagnostic wind field generator containing objective analysis
and parameterized treatments of slope flows, complex terrain effects, divergence minimization procedure, and some others [23].
It is a common view in the literature, e.g. [18, 21, 22] that emission field inventory is
one of the main sources of uncertainty in modeling of air pollution dispersion. The problem
is particularly significant when analysis relates to big industrial or urban agglomerations.
Emission field in such cases usually encompasses variety of numerous sources, point-wise,
areal and linear, with different technological parameters (spatial characteristics, stack
height, temperature and velocity of the outlet gasses), emission intensities, composition of
emitted species, and also – with different uncertainty which is introduced to the system.
This uncertainty must be taken into account in complex analysis, when the results are to be
used in supporting regulatory decisions.
In case of the discussed Warsaw study the total emission field was decomposed into
four basic categories, mainly according to emission parameters and intrinsic uncertainty.
The assumed emission categories are:
− High point sources (mainly the energy sector) – uncertainty is relatively low, since both
the combustion process and fuel parameters are well defined and stable. On the other
hand, due to the high stacks, modeling procedure should include the initial plume development near the source,
− Low point sources (other industrial sources) – higher uncertainty due to less precise
technological characteristics as well as fuel parameters,
− Area sources (residential sector and distributed industrial sources) – high uncertainty;
emission data are mainly estimated basing on fuel type use and fuel consumption,
− Linear sources (transportation system) – high uncertainty; emission data are estimated
based on several traffic parameters (traffic modes and intensity, fuel used, its quality and
consumption, age and technological parameters of cars).
Analysis covers a rectangle domain, approximately 30 km x 40 km of Warsaw Metropolitan Area (about 520 km2 within administrative borders) shown in Fig. 2. For computational purposes, domain was discretized with a homogeneous grid with the step size h = 1
km. According to the previous remarks, emission field was categorized into following four
classes: (a) 16 high point sources (power/heating plants), (b) 1002 low point sources (industry), (c) 872 area sources (residential sector), (d) 1157 area sources (transportation).
Location of the point sources is defined by their geographical coordinates, while the area
and linear sources are characterized by the respective, spatial mesh elements 1 km x 1 km
that coincide with the domain discretization grid. Computations take into account temporal
variability of the meteorological and emission input data, where the assumed step-size of
time discretization is τ = 1h.
60
Piotr Holnicki, Zbigniew Nahorski
Model:
CALPUFF
Emission sources:
high point sources
other point sources
area sources
linear sources
–
–
–
–
16
1002
872
1157
Emission & meteorological data:
year 2005, (1h time step)
Spatial discretization:
1 km x 1 km
Receptors:
563 (grid 1 km x 1 km)
Figure 2. Computational domain and the main parameters
Results contain forecasts of the annual mean concentrations (for the year 2005) of the
main urban polluting species, which are recorded at 563 fictitious receptor points, located in
the nodes of the discretization grid. The input dataset for the year 2005 is selected for analysis, mainly due to the representative (the typical) meteorological conditions as well as the
complete emission data inventory. Computational domain, receptor sites, location of the
main monitoring stations as well as the other parameters are shown in Fig. 2
Table 1. Polluting species considered (primary and secondary)
Emissions/concentrations of pollutants
SO2 – sulfur dioxide
SO =4 – sulfate aerosol (secondary pollution)
NOx – nitrogen oxides
NO 3− – nitrate aerosol (secondary pollution)
HNO3 – nitric acid (secondary pollution)
PPM10 – primary particulate matter emission, Ф ≤ 10 µm
PPM10_R –re-suspended particulate matter emission, Ф ≤ 10 µm
PM10 = PPM10 + PPM10_U + SO =4 + NO 3− – total PM10 concentration
PPM2.5 – primary particulate matter emission, Ф ≤ 2.5 µm
PPM2.5_R –re-suspended)particulate matter emission, Ф ≤ 2.5 µm
PM2.5 = PPM2.5 + PPM2.5_U + SO =4 + NO 3− – total PM2.5 concentration
Ni – nickel
Cd – cadmium
Pb – lead
BaP – benzo-alfa-pyren
PAH – polycyclic aromatic hydrocarbons
Air quality modeling in Warsaw Metropolitan Area
61
The list of the main primary and secondary pollutants considered in the study is presented below as Table 1. The composition of emitted species depends on the source category. In
particular, the point sources emit mainly sulfur oxides, nitrogen oxides, particulate matter,
PPM10 and PPM2.5 and, for certain particular sources, also BaP and some heavy metals. Area and linear sources may emit most of the listed compounds, but only linear sources are
responsible for so called secondary emission of particulate matter, PPM10_R and PPM2.5_R
(total amount of dust particulates re-suspended due to the traffic).
The resulting concentrations of particulate matter (PM10 or PM2.5) are calculated as a
sum (compare Table 1) of: primary pollution emitted by all categories of sources (PPM10 or
PM2.5), pollution related to secondary emission of the linear sources (PPM10_R or
PPM2.5_R) and concentrations of sulfate and nitrate aerosols, which are also components of
particulate matter.
In order to assess uncertainty of the concentration forecasts related to emission uncertainty, Monte Carlo analysis was applied (compare [1, 6, 7, 9, 11, 17, 20, 22]. For each
source and source category, 100, 1000 and 2000 randomly generated sets of emission values
were preprocessed, according to the assumed normal distribution and uncertainty range. To
avoid generating unrealistic emission episodes, the correlation was established between the
emission of individual compounds in a source (compare, e.g. [19, 21]). The details concerning correlation assumptions for the emission classes can be found in the report [10]. Table 2
presents the assumed emission uncertainty intervals for 4 emission classes.
Table 2. Emission uncertainty range (for normal distribution and 95% confidence interval)
Pollutant
SO2
NOx
PPM10
PPM2.5
PPM10_R
High point
± 15%
± 20%
± 25%
± 25%
–
PPM2.5_R
BAP
Ni
Cd
Pb
WWA
–
± 30%
± 30%
± 30%
± 30%
–
Emission sources
Other point
Area
± 20%
± 30%
± 30%
± 40%
± 40%
± 40%
± 40%
± 40%
–
–
–
± 40%
± 40%
± 40%
± 40%
–
Linear
± 30%
± 40%
± 40%
± 40%
± 40%
± 40%
–
± 50%
± 50%
± 50%
± 50%
± 50%
± 50%
± 50%
± 50%
± 50%
± 50%
3. Discussion of the selected results
In order to assess accuracy of CALPUFF model forecasts, the calculated year averaged
concentrations were compared (including the inflow from the surrounding areas) with
measurements of several monitoring stations (locations of the main stations are seen in Fig.
1). Selected results of this evaluation, respectively for PM10 and NOx (the sufficient number
of the measurements available), are shown in Fig. 3.
Figure 4 presents exemplary maps of annual average concentrations of SO2 and PM10 at
receptor points. The height of the traditional “box plot” relates to the mean concentration
value, while the top section represents (according to the Legend below) uncertainty range
62
Piotr Holnicki, Zbigniew Nahorski
resulting from emission uncertainty. This type of graphical presentation of uncertainty has
rather qualitative and approximate character.
a)
b)
Figure 3. Calculated vs. measured concentrations [µg/m3]: (a) PM10, (b) NOx
The more precise extension of uncertainty characteristics for any selected receptor is also available. This possibility is illustrated in Fig. 4 for receptor 273 (SO2 concentration) and
receptor 275 (PM10 concentration). In this case one can obtain graphs of the experimental
distribution curve, the density function and the standard “box plot” for uncertainty distribution. It can be seen by comparing graphs Fig. 4a and Fig. 4b that the mean concentrations of
PM10 at the same receptors is about 4–5 times higher than the respective value for SO2 (in
µg/m3). But moreover, the resulting uncertainty range of PM10 forecast is much higher than
the respective value for SO2 pollution. Explanation of this fact can be found in the next figures.
Figure 5 illustrates another possible form of data presentation – except the total concentration at the receptor point – is the relative share of emission categories in summary concentration of a given pollutant. The enclosed figures present a comparison of the relative
contribution of emission categories to the total concentrations of SO2 and PM10 measured at
receptors located in the downtown districts. In the case of sulfur dioxide (Fig. 5a) one can
see a balanced contribution of all four classes of emission sources, while for particulate matter (PM10 in the presented example) the share of linear sources (transportation system) definitely dominates (Fig. 5b). This effect is mainly due to very high PM concentrations in
central districts (high traffic intensity), because the linear sources constitute the only emission category, which is responsible for secondary emission (re-suspended particulate matter). On the other hand, this category generates respectively high uncertainty (compare
Table 2) in the resulting concentrations. Investigating Fig. 5 one can observe that the share
of emission categories presented in maps slightly changes depending on receptor location;
but definitely more significant variability is seen when considering the total area of agglomeration.
This effect is illustrated in Fig. 6, which shows the spatial variability of SO2 (Fig. 6a)
and PM10 (Fig. 6b) concentrations as well as the emission category contribution, depending
on the receptor location, observed in the whole urban area. Each figure is composed of five
map sectors: central district of the city and four peripheral (border) districts, situated N-W,
N-E, S-W, S-E, respectively.
In case of SO2, spatial changes of concentration are minor within the domain, while
more substantial are changes of emission categories share depending on receptor location.
The greater contribution of high point sources is seen in particular in Northern districts (NW and N-E), situated near the administrative city borders. It follows, first of all, from the
stack height of the main Warsaw power/heating plants (100 – 300 m), whose impact is ob-
Air quality modeling in Warsaw Metropolitan Area
63
served in some distance from the source. Another reason relates to the dominating wind
directions in the year 2005. As shown in Fig. 7 (according to [25]), S and S-W winds dominated in the period considered which resulted in higher impact of high sources in the Northern districts. The other related effect is the relatively higher contribution of low and area
sources in the Southern quarters of Warsaw.
a)
b)
Figure 4. Annual mean concentrations at receptors and uncertainty distribution for:
a) SO2 (receptor 273) and b) PM10 (receptor 275)
64
Piotr Holnicki, Zbigniew Nahorski
a)
b)
Figure 5. Relative share of emission categories in concentration (central districts):
a) SO2 and b) PM10
Air quality modeling in Warsaw Metropolitan Area
a)
b)
Figure 6. Relative share of emission categories in concentration depending on receptor location:
a) SO2 and b) PM10
65
66
Piotr Holnicki, Zbigniew Nahorski
Figure 7. Wind rose and wind velocity distribution for Warsaw (2005)
A similar effect, with respect to the impact of high point sources, is also observed in
case of NOx pollution (which is not presented in this paper), however for this pollutant there
are very significant differences in concentration between central (very high values) and peripheral districts (relatively low values). It is, first of all, the effect of the dominating contribution of linear sources (mainly car traffic) to nitrogen oxides pollution.
In case of particulate matter (it relates both PM10 as well as PM2.5 pollution) the spatial
variability of concentrations within the domain is very significant. In particular, the downtown PM10 concentrations are several times higher than in peripheral districts, while the
high concentrations are also observed along the arterial streets (compare, e.g. Fig. 6b). At
the same time, the share of emission categories varies spatially, and the contribution of area
sources as well as local point sources considerably increases in the peripheral districts. This
is rather a local effect and is recorded mainly in the neighborhood of the S_W city borders
of Warsaw. A small (several per cent) contribution of the low point sources can be also noticed in this category of air pollution. On the other hand it is evident that the impact of high
point sources is practically negligible within the whole domain. It is explained by the fact
that the main sources in this emission category are the heating plants which are all equipped
with the effective filtering installations. Moreover, the dominating contribution to particulate matter comes from the car traffic, including heavy truck traffic also in central districts
of the city. This situation is a result of deficiency of the ring roads and other arterial roads
for transit transportation in Warsaw.
One of the main objectives of this study was an evaluation of uncertainty level of air
pollution forecast which is related to the emission data uncertainty, when the urban area is
domain under question. Some selected, exemplary results of this type assessment, resulting
from the applied Monte Carlo procedure, are presented in Figure 8. Results shown in this
figure relate to one selected receptor (No 275 – Trasa Łazienkowska str.) and to the polluting species: SO2, PM10, NOx, where for each of the above kinds of air pollution the following measures are shown:
a) uncertainty level of the forecast for 95-percentile confidence interval,
b) relative share of the main emission categories in the resulting concentration,
c) uncertainty distribution presented as the standard “box plot”.
Air quality modeling in Warsaw Metropolitan Area
67
SO2
14%
Uncertainty range
for 95% of data:
26%
AREA
±6%
LIN
32%
28%
LOW
HIGH
PM10
Uncertainty range
for 95% of data:
2%
0%
8%
AREA
LIN
±18%
LOW
HIGH
90%
NOx
Uncertainty range
for 95% of data:
3% 1% 2%
AREA
±21%
LIN
LOW
94%
HIGH
Figure 8. Contribution of emission categories and related uncertainty ranges (receptor 275)
Uncertainty distributions for all pollutants presented below are symmetric (normal distribution), however forecasts related to SO2 are relatively precise (uncertainty range for 95%
confidence interval is about ±6%). Low uncertainty in this case is due to very substantial (or
dominating) share of point sources (especially high point sources) in related emission field,
because this category of emission data generates relatively low uncertainty in pollution forecasts (compare Table 2). High uncertainty, on the other hand, characterizes the forecasted
concentrations of PM10 (uncertainty range about ±18%), where dominating contribution of
emission field comes from the very uncertain linear sources of transportation system; compare Table 2). Similar situation is in the case of NOx pollution (uncertainty range about
±21%), where contribution of linear sources of road transportation system is even higher
than in the previous case.
The presented in Fig. 8 assessments of uncertainty in air pollution forecasting relate to
one, particular receptor site, however the estimates of uncertainty range for selected polluting species shown above have a more general character.
68
Piotr Holnicki, Zbigniew Nahorski
4. Summary
This study presents selected results of computations related to modeling and analysis of
air pollution dispersion in Warsaw Metropolitan Area. Analysis, dealing with the main urban-type polluting species, is based on the real meteorological data and emission field inventory for the year 2005. The main forecasting tool used in simulation of air pollution
dispersion is the known, regional scale transport model CALPUFF (Scire et al., 2000). Calculations and the related results comprise the year averaged concentrations of several polluting compounds, primary and secondary (compare Table 1) which are characteristic for
urban atmospheric environment. Some selected, exemplary results enclosed in the paper
encompass mainly sulfur and nitrogen pollutants, as well as very harmful particulate matter,
PM10 and PM2.5.
The annual mean concentrations of sulfur dioxide in the urban domain are relatively
low, approximately 8–9 µg/m3 (critical level, 20 µg/m3). The main sources, responsible for
this type pollution are power/heating plants which are equipped with very high stacks (pollutants are mainly exported outside the urban area) and desulfurization installations. On the
other hand, particulate matter and nitrogen oxides have a substantial and negative impact on
urban environment. Concentrations of PM10 and NOx are spatially strongly diversified, and
reach very high values (especially in downtown districts), which often exceed the critical
values (40 µg/m3 for both of these pollutants). In this case, the main source of adverse environmental impact is transportation system (including transit traffic). New circle roads and
displacement of the transit and heavy truck transport outside the central districts of the city
would improve the situation.
Uncertainty analysis shows the relationship of air pollution forecast of two main factors: (a) the kind of polluting compound and (b) emission source category which is the dominating contributor. In the case of air pollutants considered in this paper, relatively low
uncertainty PM2.5 applies to the forecasts of SO2, while very substantial uncertainties relate
to PM10, and NOx forecasts, which all strongly depend on very uncertain linear sources. The
recent research concentrates on assessment of health effects of urban air pollution (mainly
particulate matter and nitrogen oxides). Results will be presented the forthcoming publications.
Acknowledgements
This project was supported by the by the Polish Ministry of Science and Higher Education under Grant NN519316735. We also thank Wojciech Trapp and the EKOMETRIA team
for their assistance in emission data preparation.
References
[1] ApSimon H.M., Warren R.F., Kayin S. Addressing uncertainty in environmental modeling: a
case study of integrated assessment of strategies to combat long-range transboundary air pollution. Atmospheric Environment, 36, 2002, 5417–5426.
[2] Calori G., Clemente M., De Maria R., Finardi S., Lollobrigida F., Tinarelli G. Air quality integrated modelling in Turin urban area. Environmental Modelling & Software, 21, 2006, 468–
476.
[3] Carnevale C., Finzi G., Pisoni E., Volta M., Guariso G., Gianfreda R., Maffeis G., Thunis P.,
White L., Triacchini G. An integrated assessment tool to define effective air quality policies at
regional scale. Environmental Modelling & Software, 38, 2012, 306–315.
Air quality modeling in Warsaw Metropolitan Area
69
[4] Cofała J., Amann M., Asman W., Bertok I., Heyes C., Hoeglund I.L., Klimont Z., Schoepp W.,
Wagner F. Integrated assessment of air pollution and greenhouse gasses mitigation in Europe.
Archives of Environmental Protection, 36, 2010, 29–39.
[5] Fisher B. Fuzzy environmental decision-making: applications to air pollution. Atmospheric
Environment, 37, 2003, 1865–1877.
[6] Hanna S.R., Chang J.C., Fernau M.E. Monte Carlo estimates of uncertainties in predictions by
photochemical grid model (UAM-IV) due to uncertainties in input variables. Atmospheric Environment, 32, 1998, 3619–3628.
[7] Hanna S.R., Lu Z., Frey H.C., Wheeler N., Vukovich J., Arunachalam S., Fernau M., Hansen
D.A. Uncertainties in predicted ozone concentrations due to input uncertainties for the UAM-V
photochemical grid model applied to the July 1995 OTAG domain. Atmospheric Environment,
35, 2001, 891–903.
[8] Holnicki P. Air Pollution Transport Models in Air Quality Control (in Polish). EXIT Publishers, 2006, Warsaw, Poland.
[9] Holnicki P., Nahorski Z., Tainio M. Uncertainty in air quality forecasts caused by emission
uncertainty. Proceedings of HARMO 13th Conference on Harmonisation within Atmospheric
Dispersion Modelling, 2010, 119–123.
[10] Holnicki P., Nahorski Z. Uncertainty in Integrated Systems of Air Quality Assessment (in
Polish). Report RB/6/2011, 2011, Systems Research Institute PAS, Warsaw.
[11] Holnicki P. Uncertainty in Integrated Modeling of Air Quality. In: Advenced Air Pollution.
INTECH Publishers, 2011, Rijeka, 239–260.
[12] Jacobson M.Z. Fundamentals of Atmospheric Modeling. Cambridge University Press, 2005,
Cambridge, U.K.
[13] Juda-Rezler K. New Challenges in Air Quality and Climate Modeling. Archives of Environmental Protection, 36, 2010, 3–28.
[14] Kelly J.A. An Overview of the RAINS Model. Environmental Research Centre Report, Environmental Protection Agency, 2006, Dublin, Ireland.
[15] Markiewicz T.M. Podstawy modelowania rozprzestrzeniania się zanieczyszczeń w powietrzu
atmosferycznym. Oficyna Wyd. Politechniki Warszawskiej, 2004, Warszawa.
[16] Mediavilla-Sahagún A., ApSimon H.M. Urban scale integrated assessment for London: Which
emission reduction strategies are more effective in attaining prescribed PM10 air quality standards by 2005? Environmental Modelling & Software, 21, 2006, 501–513.
[17] Moore G.E., Londergan R.J. Sampled Monte Carlo uncertainty analysis for photochemical grid
models. Atmospheric Environment, 35, 2001, 4863–4876.
[18] Oxley T., Valiantis M., Elskkaki A., ApSimon H.M. Background, Road and Urban Transport
modeling of Air quality Limit Values (The BRUTAL model). Environmental Modelling & Software, 24, 2009, 1036–1050.
[19] Page T., Whyatt J.D., Beven K.J., Metcalfe S.E. Uncertainty in modeled estimates of acid deposition across Wales: a GLUE approach. Atmospheric Environment, 38, 2003, 2079–2090.
[20] Park S.-K., Cobb C.E., Wade K., Mulholland J., Hu Y., Russel A.G.) Uncertainty in air quality
model evaluation for particulate matter due to spatial variations in pollutant concentrations.
Atmospheric Environment, 40, 2006, S563–S573.
[21] Russel A., Dennis D. NASTRO critical review of photochemical models and modeling. Atmospheric Environment, 34, 2000, 2283–2324.
[22] Sax T., Isakov V. A case study for assessing uncertainty in local-scale regulatory air quality
modeling applications. Atmospheric Environment, 37, 2003, 3481–3489.
[23] Scire J.S., Strimaitis D.G., Yamartino R.J. A User’s Guide for the CALPUFF Dispersion Model. Earth Technology Inc., 2000.
[24] Sportisse B. A review of current issues in air pollution modeling and simulation. Computational Geosciences, 11, 2007, 159–181.
[25] www.windandpower.com