Analysis of Oxides of Nitrogen and Ground-Level Ozone in Ambient Air Aoife Fitzgerald School of Chemical and Pharmaceutical Sciences DIT Kevin St

Analysis of Oxides of Nitrogen and Ground-Level Ozone in Ambient Air
Aoife Fitzgerald

School of Chemical and Pharmaceutical Sciences
DIT Kevin St.

May 2018
Thesis submitted in Partial Fulfilment of
Examination Requirements Leading to the Award
BSc (Chemical Sciences with Medicinal Chemistry)
Dublin Institute of Technology
Thesis Supervisors:
Dr. David O’ Connor
Dr. Jose Maria Maya ManzanoDeclarationI certify that this thesis which I now submit for examination for the award of BSc is entirely my own work and has not been taken from the work of others, save and to the extent that such work has been cited and acknowledged within the text of my work.

This thesis was prepared according to the regulations provided by the School of Chemical and Pharmaceutical Sciences, Dublin Institute of Technology and has not been submitted in whole or in part for another award in any Institute or University.
The Institute has permission to keep, lend or copy this thesis in whole or in part, on condition that any such use of the material of the thesis is duly acknowledged.

Signature
______________________
Name
Date

Acknowledgements
I would like to thank Dr. David O’ Connor for his great help and guidance throughout my thesis project.
would like to express my deepest appreciation to all those who provided me with the opportunity to complete this report. Thank you to all of the laboratory technicians for their relentless efforts to ensure our work runs smoothly, with a special thanks to Howard Wallace, Martin Kitson, Hans EckhardtI would also like to thank my parents and family for all their help and support.
I give my most sincere thanks to the teaching and technician staff within the School of Chemical and Pharmaceutical Sciences DIT.

Abstract

List of Abbreviations Used:
NOX – Oxides of nitrogen
NO2 – Nitrogen dioxide
NO – Nitric oxide
O3 – Ozone
UV – Ultra-violet
VOC – Volatile organic compound
AQI – Air Quality Index
MARS – Multivariate adaptive regression splines
EPA – Environmental Protection Agency
EEA – European Environmental Agency
TEA – TriethanolamineNEDA – N-(l-Napthyl)-ethylenediamine dihydrochloride
Table of Contents
Statement of Originality iAcknowledgements ii
Abstract iii
List of Abbreviations Used iv
Table of Contents v
Chapter 1: Introduction 1
1.1 General Background 1
1.2 NOX 1
1.2.1 What is NOX? 1
1.2.2 How are Oxides of Nitrogen (NOX) formed? 2
1.2.3 Sources of NOX 4
1.2.4 Techniques for the Removal of NOX from Exhaust Gases 5
1.2.5 Harmfulness of NOX and the Need to Monitor 5
1.2.6 Chemical Methods for Sampling NOX 6
1.3 Ozone 7
1.3.1 What is Ozone (O3)? 7
1.3.2 How is Ozone formed? 7
1.3.3 Sources of Ozone 8
1.3.4 Harmfulness of Ozone and the Need to Monitor 8
1.3.5 Chemical Methods for Sampling Ozone 9
1.4 Statistical Analysis by Using R and Data Collection 10
1.5 Project Aim 10
Chapter 2: Experimental 11
2.1 Determination of Nitrogen Dioxide in Ambient Air 11
2.1.1 Reagents 11
2.1.2 Instruments 11
2.1.3 Solutions Prepared 11
2.1.4 Procedure 12
2.2 Measurement of Ground-Level Ozone in Ambient Air 14
2.2.1 Reagents 14
2.2.2 Instruments 14
2.2.3 Solutions Provided 14
2.2.4 Procedure 14
2.3 Building NOx and Ozone Models using R Studio 17
2.3.1 Model 1 17
2.3.2 Model 2 17
2.3.3 Model 3 17
2.3.4 Model 4 18
2.3.5 Multivariate Linear Regression 18
2.3.6 Correlation19
2.3.7 Boxplots 19
2.3.8 Histograms20
2.3.9 Quantile-Quantile Plots 20
Chapter 3: Results and Discussion 21
3.1 Chemical Methods: NO2 21
3.2 Chemical Methods: Ozone 27
3.3 Data Modelling: Oxides of Nitrogen 33
3.3.1 Dublin NO2 38
3.4 Data Modelling: Ozone 57
3.4.1 Dublin Ozone 67
3.5 Comparison of Sampled NO2 and NO2 Model 93
3.6 Comparison of Sampled Ozone and Dublin Ozone Model 95
Conclusion 97
References 98
Appendix 21
Appendix 1 – Dublin NOX 21
Appendix 2 – Dublin NO 27
Appendix 3– Cork Ozone 33
Appendix 4– Cork Ozone 33
Appendix 5 – Mayo Ozone 33
Appendix 6 Scripts 33
Chapter 1: Introduction
1.1 General Background
“Pollution has been defined as the introduction of substance or energy into the environment, resulting in damaging effects of such a nature as to endanger human health, harm living resources and ecosystems, and impair or interfere with amenities and other legitimate uses of the environment.” ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “URL” : “https://www.eea.europa.eu/themes/water/wise-help-centre/glossary-definitions/pollution”, “accessed” : { “date-parts” : “2018”, “5”, “21” }, “author” : { “dropping-particle” : “”, “family” : “European Environment Agency”, “given” : “”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “id” : “ITEM-1”, “issued” : { “date-parts” : “2018” }, “title” : “Pollution u2014 European Environment Agency”, “type” : “webpage” }, “uris” : “http://www.mendeley.com/documents/?uuid=a6a5bb3a-6cef-34d0-b8d7-1032c26f9c9d” } , “mendeley” : { “formattedCitation” : “<sup>1</sup>”, “plainTextFormattedCitation” : “1”, “previouslyFormattedCitation” : “<sup>1</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }1 Consequently, all elements, either generated by human activities (traffic, industry) or as a result of other processes (agriculture, natural phenomena’s) can be defined as pollutants.

Air pollution is one of the major challenges faced nowadays, due to a combination of the ever increasing population and more awareness about environmental protection. “From a human exposure point of view, the main outdoor air pollutants are nitrogen dioxide (NO2), fine particles (PM10 and smaller size fractions such as PM2.5), carbon monoxide (CO), sulphur dioxide (SO2), ozone (O3), lead, benzene, 1,3-butadiene and certain polycyclic aromatic hydrocarbons (PAH). In practice, NO2 and PM10 are causing most of the exceedances of air quality standards in European cities, with ozone and benzene also causing some concern, mainly in southern Europe.”ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “DOI” : “10.1039/9781847559654”, “ISBN” : “978-1-84755-907-4”, “author” : { “dropping-particle” : “”, “family” : “Hertel”, “given” : “Ole”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” }, { “dropping-particle” : “”, “family” : “Salmond”, “given” : “Jennifer”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” }, { “dropping-particle” : “”, “family” : “Bloss”, “given” : “William”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” }, { “dropping-particle” : “”, “family” : “Salma”, “given” : “Imre”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” }, { “dropping-particle” : “”, “family” : “Vardoulakis”, “given” : “Sotiris”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” }, { “dropping-particle” : “”, “family” : “Maynard”, “given” : “Robert”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” }, { “dropping-particle” : “”, “family” : “Williams”, “given” : “Martin”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” }, { “dropping-particle” : “”, “family” : “Goodsite”, “given” : “Michael E”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” }, { “dropping-particle” : “”, “family” : “Harrison”, “given” : “R M”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” }, { “dropping-particle” : “”, “family” : “Hester”, “given” : “R E”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “collection-title” : “Issues in Environmental Science and Technology”, “id” : “ITEM-1”, “issued” : { “date-parts” : “2009” }, “number-of-pages” : “P001-P148”, “publisher” : “The Royal Society of Chemistry”, “title” : “Air Quality in Urban Environments”, “type” : “book” }, “uris” : “http://www.mendeley.com/documents/?uuid=f1d8e9c7-f25b-43d9-98fe-c2ededeac890” } , “mendeley” : { “formattedCitation” : “<sup>2</sup>”, “plainTextFormattedCitation” : “2”, “previouslyFormattedCitation” : “<sup>2</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }2 Due to this, this project will be focused on the study of NO2 and ozone concentrations in an Irish context, taking into account real EPA pollution data from recent years.

In this introduction, NOX and ozone will be discussed in terms of their definition, sources and the harm they cause.

1.2 NOX
1.2.1 What is NOx?
Nitrogen oxides (NOX) are major pollutants in the atmosphere, being a precursor to acid rain, photochemical smog, and ozone accumulation. The oxides are mainly nitrogen dioxide (NO2) and nitric oxide (NO), both of which are corrosive and hazardous to health. Another, less common NOX, is nitrous oxide (N2O).

Nitrogen dioxide (NO2)
Nitrogen dioxide is a major pollutant and component of smog.  Chronic exposure to NO2 can cause respiratory effects including airway inflammation in healthy people and increased respiratory symptoms in people with asthma. NO2 creates ozone which causes eye irritation and exacerbates respiratory conditions. NO2 is not very soluble and therefore, passes into the pulmonary region where tissue damage may occur. In individuals occupationally exposed to high NO2 levels, adverse effects such as pulmonary edema only manifest themselves hours after exposure has ended. Elevated NO2 exposures have been known to occur in the manufacture of nitric acid (HNO3), in farm silos, form electric arc welding, and from use of explosives in mining. ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “ISBN” : “156670586X”, “abstract” : “4th ed. The atmosphere — Atmospheric pollution and pollutants — Atmospheric dispersion, transport, and deposition — Atmospheric effects — Health effects — Welfare effects — Air quality and emissions assessment — Regulation and public policy — Control of motor vehicle emissions — Control of emissions from stationary sources — Indoor air quality — Environmental noise.”, “author” : { “dropping-particle” : “”, “family” : “Godish”, “given” : “Thad.”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “id” : “ITEM-1”, “issued” : { “date-parts” : “2004” }, “number-of-pages” : “460”, “publisher” : “Lewis Publishers”, “title” : “Air quality”, “type” : “book” }, “uris” : “http://www.mendeley.com/documents/?uuid=bbde3552-004b-3a73-ad96-19d19540bdc8” } , “mendeley” : { “formattedCitation” : “<sup>3</sup>”, “plainTextFormattedCitation” : “3”, “previouslyFormattedCitation” : “<sup>3</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }3
Nitric oxide (NO)
Nitric oxide is a colourless, odourless, tasteless, relatively non-toxic gas. It is produced naturally in soil through biological nitrification and denitrification processes and as a result of biomass burning, lightning, and oxidation of NH3 by photochemical processes. It is also transported from the stratosphere into the troposphere. Its importance lies in the fact that it is readily oxidised in the atmosphere to nitrogen dioxide, which has a much higher toxicity
Nitrous oxide (N2O)
Nitrous oxide is widely used as an anaesthetic in medicine and dentistry. Human exposure to elevated concentrations produces a kind of hysteria, and as such, it is often referred to as laughing gas. It is generally considered to be non-toxic but it does react with vitamin B12, which may be a problem for those who are deficient. It is a significant greenhouse gas, and has been defined as being 298 times as radiative forcing as CO2 because of its interaction with light, and the time taken to decompose in the atmosphere.ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “URL” : “http://clean-carbonenergy.com/nox-emissions.html”, “accessed” : { “date-parts” : “2018”, “4”, “7” }, “author” : { “dropping-particle” : “”, “family” : “Omersa”, “given” : “Ken”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “id” : “ITEM-1”, “issued” : { “date-parts” : “0” }, “title” : “NOx emissions from diesel engines”, “type” : “webpage” }, “uris” : “http://www.mendeley.com/documents/?uuid=a101dce7-341d-3350-b6d5-813c60a7b0bc” } , “mendeley” : { “formattedCitation” : “<sup>4</sup>”, “plainTextFormattedCitation” : “4”, “previouslyFormattedCitation” : “<sup>4</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }4 It is broken down in the stratosphere and catalyses the breakdown of ozone. Nitrous oxide has no known tropospheric sink. As a result, it has a very long atmospheric lifetime (~150 years).

1.2.2 How are Oxides of Nitrogen (NOX) formed?
Thermal NOX formation
Thermal NOX formation describes the process when nitrogen (in the combustion air) reacts with oxygen (in the combustion air) to produce NOX, this occurs in both spark ignition car engines and in gas turbines. This formation requires very high temperatures and is exponentially dependent on temperature. Because the process is nonlinear, there are hot spots, confined areas with a higher temperature than the average. These hot spots can have a very large effect on the amount of NOX produced, (if the fuel and air were to be premixed prior to the flame, these hotspots could be avoided due to the mixture being homogeneous.) Therefore the maximum temperature is more important than the average temperature, and the process is very difficult to model accurately because of this. Another important factor in thermal NOX formation is the residence time, which describes the length of time that the combustion gas is at the higher temperature. The turbulence and the amount of excess oxygen are also important factors.ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “abstract” : “As the regulations of pollutant emission is getting more and more strict, models of pollutant formation, as a part of system level models for control design, are needed. In this thesis NOx (nitrogen oxides) formation is modeled. A method is developed to mix equilibrium and non-equilibrium reaction equations to model NOx formation in combustion processes. This algorithm reduces high-index DAE systems byintroducireaction invariants”. Two models are built with help of this algorithm which describe NOx formation during combustion in a cylinder in a spark ignition car engine and in a gas turbine (for more information about the combustion models see 4 and 7). The models use both equilibrium and non-equilibrium/dynamical reaction equations. A rough verification of the cylinder model is also done.”, “author” : { “dropping-particle” : “”, “family” : “Schwerdt”, “given” : “Christian”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “id” : “ITEM-1”, “issued” : { “date-parts” : “2006” }, “publisher” : “Lund University”, “title” : “Modelling NOx-Formation in Combustion Processes Modeling NOx-Formation in Combustion Processes (Modellering av NOx-formation vid fu00f6rbru00e4nning)”, “type” : “thesis” }, “uris” : “http://www.mendeley.com/documents/?uuid=d31155a8-c174-371d-98c6-f729483d0e03” } , “mendeley” : { “formattedCitation” : “<sup>5</sup>”, “plainTextFormattedCitation” : “5”, “previouslyFormattedCitation” : “<sup>5</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }5 The process is ruled by the following equations, referred to as the Zeldovich mechanism:
N2 + O ? NO + N (1)
N + O2 ? NO + O (2)
N + OH ? NO + H (3)
The strong triple bond in the N2 molecule needs a high temperature to break and therefore equation (1) determines the rate of the thermal NOX formation. ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “author” : { “dropping-particle” : “”, “family” : “Hesselmann”, “given” : “Gerry”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” }, { “dropping-particle” : “”, “family” : “Rivas”, “given” : “Marta”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “container-title” : “IFRF Online Combustion Handbook”, “id” : “ITEM-1”, “issued” : { “date-parts” : “2001” }, “page” : “1-4”, “title” : “What are the main NOx formation processes in combustion plant?”, “type” : “article-journal”, “volume” : “1” }, “uris” : “http://www.mendeley.com/documents/?uuid=8f20404d-426f-42ea-9fde-1b38af38115b” } , “mendeley” : { “formattedCitation” : “<sup>6</sup>”, “plainTextFormattedCitation” : “6”, “previouslyFormattedCitation” : “<sup>6</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }6
Fuel NOX formation
Fuel NOX formation is the process where N2 in fuel oxidises with O2 in the combustion air to form NOX. Gas fuels have a moderately low amount of bound nitrogen, and therefore produce low volumes of NOX emission by this process. Coal and oil have considerably more bound nitrogen, therefore fuel NOX formation is a much larger part of the overall volume of NOX produced for these types of fuel in comparison to gaseous fuels. The mechanism for the fuel NOX formation process is not entirely understood, but is modelled by the following two equations:
NComplex + OH ? NO + X (4)
NComplex + NO ? N2 + X (5)
where X symbolizes other products where the mechanism is not fully understood.ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “abstract” : “As the regulations of pollutant emission is getting more and more strict, models of pollutant formation, as a part of system level models for control design, are needed. In this thesis NOx (nitrogen oxides) formation is modeled. A method is developed to mix equilibrium and non-equilibrium reaction equations to model NOx formation in combustion processes. This algorithm reduces high-index DAE systems byintroducireaction invariants”. Two models are built with help of this algorithm which describe NOx formation during combustion in a cylinder in a spark ignition car engine and in a gas turbine (for more information about the combustion models see 4 and 7). The models use both equilibrium and non-equilibrium/dynamical reaction equations. A rough verification of the cylinder model is also done.”, “author” : { “dropping-particle” : “”, “family” : “Schwerdt”, “given” : “Christian”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “id” : “ITEM-1”, “issued” : { “date-parts” : “2006” }, “publisher” : “Lund University”, “title” : “Modelling NOx-Formation in Combustion Processes Modeling NOx-Formation in Combustion Processes (Modellering av NOx-formation vid fu00f6rbru00e4nning)”, “type” : “thesis” }, “uris” : “http://www.mendeley.com/documents/?uuid=d31155a8-c174-371d-98c6-f729483d0e03” } , “mendeley” : { “formattedCitation” : “<sup>5</sup>”, “plainTextFormattedCitation” : “5”, “previouslyFormattedCitation” : “<sup>5</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }5
Prompt NOX formation
Prompt NOX formation is the final process describing NOX formation. In this process radical hydrocarbons are produced during the combustion of the fuel. These radicals react with the nitrogen in the combustion air to form transition compounds which then oxidise to NOX when they react with the oxygen in the combustion air. The following equation is the most significant first step in the process:

CH + N2 ? HCN + N (6)
where the transition compound, HCN, is converted into atomic nitrogen through a sequence of steps. At a higher temperature, the reaction C + N2 ? CN + N (7) also participates in the breaking of the N2 bond. The nitrogen atoms from these equations are then oxidised to NO. This process is observed in low temperatures at the beginning of the combustion process and is only relevant in very fuel-rich combustion. ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “abstract” : “As the regulations of pollutant emission is getting more and more strict, models of pollutant formation, as a part of system level models for control design, are needed. In this thesis NOx (nitrogen oxides) formation is modeled. A method is developed to mix equilibrium and non-equilibrium reaction equations to model NOx formation in combustion processes. This algorithm reduces high-index DAE systems byintroducireaction invariants”. Two models are built with help of this algorithm which describe NOx formation during combustion in a cylinder in a spark ignition car engine and in a gas turbine (for more information about the combustion models see 4 and 7). The models use both equilibrium and non-equilibrium/dynamical reaction equations. A rough verification of the cylinder model is also done.”, “author” : { “dropping-particle” : “”, “family” : “Schwerdt”, “given” : “Christian”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “id” : “ITEM-1”, “issued” : { “date-parts” : “2006” }, “publisher” : “Lund University”, “title” : “Modelling NOx-Formation in Combustion Processes Modeling NOx-Formation in Combustion Processes (Modellering av NOx-formation vid fu00f6rbru00e4nning)”, “type” : “thesis” }, “uris” : “http://www.mendeley.com/documents/?uuid=d31155a8-c174-371d-98c6-f729483d0e03” } , “mendeley” : { “formattedCitation” : “<sup>5</sup>”, “plainTextFormattedCitation” : “5”, “previouslyFormattedCitation” : “<sup>5</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }5
1.2.3 Sources of NOX
Fuel Combustion:
At high temperatures, oxygen and nitrogen in the air join to form nitrogen oxides.
Nitric acid manufacture:
Waste gases which can’t be economically recovered in the final absorber normally comprise of 2-3% nitrogen oxides based on the weight of acid manufactured.

Metal finishing operations:
Many metal surface treatment operations using nitrates, nitrites, or nitric acid form nitrogen oxides.
Chemical processes:
Many processes where nitric acid, nitrates, or nitrites are used as reagents also form nitrogen oxides.
High temperature processes:
Processes where materials are made at high temperatures form nitrogen oxides.

Burning plant material: Releases nitrogen oxides, because all plants contain nitrogen.ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “id” : “ITEM-1”, “issued” : { “date-parts” : “1999” }, “number-of-pages” : “57”, “title” : “Nitrogen Oxides (NOx), Why and How They Are Controlled CORRECTION NOTICE Public Access and Information Transfer”, “type” : “report”, “volume” : “800” }, “uris” : “http://www.mendeley.com/documents/?uuid=0272c39a-6a94-379e-bc69-d55cbb609e94” } , “mendeley” : { “formattedCitation” : “<sup>7</sup>”, “plainTextFormattedCitation” : “7”, “previouslyFormattedCitation” : “<sup>7</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }7
Fuel Combustion
In diesel engines, air is injected into the cylinder, where it is then compressed twice as much as in a petrol engine. This compression produces heat, so that diesel burns spontaneously when injected. In petrol engines, a mixture of fuel and air is injected into the chamber. This is then compressed and ignited by a spark plug.
Diesel engines yield more NOX than petrol engines because diesel engines run at a much higher temperature and pressure than petrol engines. These conditions favour the formation of NOX gases. The amount of NOX produced depends on the volume and duration of the hottest part of the flame.
Lowering the combustion temperature can reduce NOX emissions from a diesel engine. Some exhaust gas is cooled and then injected back into the combustion chamber. There is less oxygen in the exhaust gas because some has been consumed by previous combustion, therefore, there is not as much left to feed the flame. The exhaust gas also has a higher heat capacity than air, so it takes longer to heat up.

1.2.4 Techniques for the Removal of NOX from Exhaust Gases
NOX can be removed from exhaust gases by using various techniques, depending on the applications, although a lot of effort goes into designing burners which reduce NOX emissions in the first place.
Selective Catalytic Reduction (SCR) is the most common method for NOX removal in diesel exhausts, but it is expensive and therefore isn’t used in small cheap vehicles. There are various proprietary blends of ammonia and urea which can be injected into the exhaust flow. These react with NOX gases over a catalyst, which turns them into harmless nitrogen and water.
Selective Non-Catalytic Reduction (SNCR) – this type of reduction takes place in ducting where the temperature is about 1000°C. Urea or ammonia is injected, and the NOX gases are reduced to nitrogen without a catalyst.
On an industrial scale, exhaust gases can be scrubbed with chemicals such as sodium hydroxide, hydrogen peroxide, or a mixture of hydrogen peroxide and nitric acid. These chemicals react with the NOX gases and remove them. ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “URL” : “http://clean-carbonenergy.com/nox-emissions.html”, “accessed” : { “date-parts” : “2018”, “4”, “7” }, “author” : { “dropping-particle” : “”, “family” : “Omersa”, “given” : “Ken”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “id” : “ITEM-1”, “issued” : { “date-parts” : “0” }, “title” : “NOx emissions from diesel engines”, “type” : “webpage” }, “uris” : “http://www.mendeley.com/documents/?uuid=a101dce7-341d-3350-b6d5-813c60a7b0bc” } , “mendeley” : { “formattedCitation” : “<sup>4</sup>”, “plainTextFormattedCitation” : “4”, “previouslyFormattedCitation” : “<sup>4</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }4
1.2.5 Harmfulness of NOX and the Need to Monitor
Air pollution is the single largest environmental health risk in Europe. Air pollutants are emitted from man-made and natural sources and can be transported over long distances. Some air pollutants persist in the environment for long periods of time and they may accumulate in the environment and in the food chain, affecting humans and animals not only via air intake, but also water and food intake. ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “DOI” : “10.2800/62459”, “ISBN” : “9789292137021”, “ISSN” : “1977-8449”, “PMID” : “8875828”, “abstract” : “This report presents an updated overview and analysis of air quality in Europe. It is focused in the state in 2013 and the development from 2004 to 2013. It reviews progress towards meeting the requirements of the air quality directives. An overview of the latest findings and estimates of the effects of air pollution on health and its impacts on ecosystems is also given.”, “author” : { “dropping-particle” : “”, “family” : “Ortiz”, “given” : “Alberto Gonzu00e1lez”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “container-title” : “Report”, “id” : “ITEM-1”, “issue” : “5”, “issued” : { “date-parts” : “2015” }, “number-of-pages” : “1-64”, “title” : “Air quality in Europe u2014 2015 report”, “type” : “report” }, “uris” : “http://www.mendeley.com/documents/?uuid=1f39b288-bf0b-4923-a21d-3ce26576c390” } , “mendeley” : { “formattedCitation” : “<sup>8</sup>”, “plainTextFormattedCitation” : “8”, “previouslyFormattedCitation” : “<sup>8</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }8 Cardiovascular complications such as heart disease and stroke are the main causes of premature death attributed to air pollution.ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “author” : { “dropping-particle” : “”, “family” : “Jasarevic”, “given” : “Tarik”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” }, { “dropping-particle” : “”, “family” : “Thomas”, “given” : “Glen”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” }, { “dropping-particle” : “”, “family” : “Osseiran”, “given” : “Nada”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “container-title” : “WHO”, “id” : “ITEM-1”, “issued” : { “date-parts” : “2014” }, “publisher” : “World Health Organization”, “title” : “WHO | 7 million premature deaths annually linked to air pollution”, “type” : “report” }, “uris” : “http://www.mendeley.com/documents/?uuid=e6d86557-6de1-3eeb-82f6-5d4785c9b961” } , “mendeley” : { “formattedCitation” : “<sup>9</sup>”, “plainTextFormattedCitation” : “9”, “previouslyFormattedCitation” : “<sup>9</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }9 Air pollution itself is responsible for 80% of cases of premature death in Europe.ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “DOI” : “10.2800/62459”, “ISBN” : “9789292137021”, “ISSN” : “1977-8449”, “PMID” : “8875828”, “abstract” : “This report presents an updated overview and analysis of air quality in Europe. It is focused in the state in 2013 and the development from 2004 to 2013. It reviews progress towards meeting the requirements of the air quality directives. An overview of the latest findings and estimates of the effects of air pollution on health and its impacts on ecosystems is also given.”, “author” : { “dropping-particle” : “”, “family” : “Ortiz”, “given” : “Alberto Gonzu00e1lez”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “container-title” : “Report”, “id” : “ITEM-1”, “issue” : “5”, “issued” : { “date-parts” : “2015” }, “number-of-pages” : “1-64”, “title” : “Air quality in Europe u2014 2015 report”, “type” : “report” }, “uris” : “http://www.mendeley.com/documents/?uuid=1f39b288-bf0b-4923-a21d-3ce26576c390” } , “mendeley” : { “formattedCitation” : “<sup>8</sup>”, “plainTextFormattedCitation” : “8”, “previouslyFormattedCitation” : “<sup>8</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }8 This negative impact on human health also has a substantial negative impact on European member state economies, with working days lost, reduced productivity and increased medical costs.ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “DOI” : “10.1787/9789264257474-en”, “ISBN” : “9789264257467”, “abstract” : “This report provides a comprehensive assessment of the economic consequences of outdoor air pollution in the coming decades, focusing on the impacts on mortality, morbidity, and changes in crop yields as caused by high concentrations of pollutants. Unless more stringent policies are adopted, findings point to a significant increase in global emissions and concentrations of air pollutants, with severe impacts on human health and the environment. The market impacts of outdoor air pollution are projected to lead to significant economic costs, which are illustrated at the regional and sectoral levels, and to substantial annual global welfare costs.”, “author” : { “dropping-particle” : “”, “family” : “OECD”, “given” : “”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “id” : “ITEM-1”, “issued” : { “date-parts” : “2016” }, “number-of-pages” : “20”, “title” : “The Economic Consequences of Outdoor Air Pollution: Policy Highlights”, “type” : “report” }, “uris” : “http://www.mendeley.com/documents/?uuid=ba351d88-7f01-41ea-b52d-cea58ed8f5cc” } , “mendeley” : { “formattedCitation” : “<sup>10</sup>”, “plainTextFormattedCitation” : “10”, “previouslyFormattedCitation” : “<sup>10</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }10
Air pollution also has effects on vegetation, soil and water quality. For example NH3, NOX and SO2 contribute to the acidification of soil and water and ground?level ozone damages agricultural crops, forests and plants.
All mandatory monitoring NO2 levels are low in rural areas and smaller towns but close to, or above the limit value in Dublin and Cork city centre stations. NOX levels in urban areas are influenced by weather episodes. A period of stable airflow over a city centre can lead to build up of NO2, therefore continued increase in NOX emissions within urban centres can lead to further breaches of the limit value. Even though technological advances help to lower NOX emissions from individual cars, this is offset by the increase in the amount of vehicles on the road. Emissions from traffic are the main source of nitrogen oxides in Ireland, together with electricity generating stations and industry. ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “author” : { “dropping-particle” : “”, “family” : “O’ Dwyer”, “given” : “Dr. Micheu00e1l”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “id” : “ITEM-1”, “issued” : { “date-parts” : “2015” }, “number-of-pages” : “75”, “title” : “Air Quality in Ireland 2015 Key Indicators of Ambient Air Quality”, “type” : “report” }, “uris” : “http://www.mendeley.com/documents/?uuid=9cbcbe9a-8454-3dc2-a52f-f3e88c237db9” } , “mendeley” : { “formattedCitation” : “<sup>11</sup>”, “plainTextFormattedCitation” : “11”, “previouslyFormattedCitation” : “<sup>11</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }11 Also, these technological advances have also been used in the opposite manner, as seen in the 2015 Volkswagen emissions scandal. This was where Volkswagen installed a “defeat device” in their vehicles. This device was not a physical device but a programme in the engine software that let the car perceive if it was being driven under test conditions, and if so perform differently. “Clean diesel” engines cut emissions through techniques such as adjusting air-fuel ratios and exhaust flows. When running normally, requiring greater performance, Volkswagen’s controls would not operate in the same way. ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “author” : { “dropping-particle” : “”, “family” : “Topham”, “given” : “Gwyn”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” }, { “dropping-particle” : “”, “family” : “Clarke”, “given” : “Sean”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” }, { “dropping-particle” : “”, “family” : “Levett”, “given” : “Cath”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” }, { “dropping-particle” : “”, “family” : “Scruton”, “given” : “Paul”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” }, { “dropping-particle” : “”, “family” : “Fidler”, “given” : “Matt”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “container-title” : “The Guardian”, “id” : “ITEM-1”, “issued” : { “date-parts” : “2015”, “9”, “23” }, “title” : “The Volkswagen emission scandal explained”, “type” : “article-newspaper” }, “uris” : “http://www.mendeley.com/documents/?uuid=86f5fbd2-d8f6-4831-9375-9afb1b6aa22f” } , “mendeley” : { “formattedCitation” : “<sup>12</sup>”, “plainTextFormattedCitation” : “12”, “previouslyFormattedCitation” : “<sup>12</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }12
1.2.6 Chemical Methods for Sampling NOX
There are many instrumental techniques for measuring NOx. These include:
Chemiluminescence (CLS) analysers: These analysers exploit a chemiluminescent reaction which occurs between NO and ozone.
Infrared spectroscopic analysers (IR): These analysers include gas-filter correlation (GFC-IR) and fourier-transform (FTIR) analysers. NO, N2O and NO2 all have strong IR spectra, so this technique is suitable for measuring NOX.
Ultraviolet (UV) spectroscopic analysers: NO and NO2 also absorb electromagnetic radiation in the UV region of the electromagnetic spectrum.
Electrochemical cells: There are analysers using electrochemical cells, available for NO and NO2.ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “abstract” : “This note describes the techniques and standards required to monitor oxides of nitrogen, covering (i) definitions and sources of oxides of nitrogen, (ii) common techniques, (iii) applicable standards, (iv) reporting emissions of oxides of nitrogen, and (v) calibration functions for NOx when applying EN 14181. 2. Practical Guidance 2.1 Definition of sources of oxides of nitrogen Oxides of nitrogen (NO x) can theoretically include any gaseous compounds that consist of oxygen and nitrogen. However, this guidance note focuses on the commonest compounds in NO x , which are: Nitric oxide (NO) Nitrogen dioxide (NO 2) The dominant source of NO x is combustion and within the emissions from combustion plant the total NO x is typically dominated by NO. In most installations, the concentration of NO in total NO x is typically greater than 95% with NO 2 concentration less than 5%. In combined-cycle gas turbines (CCGTs) and some types of chemical process, emissions of NO 2 can be much higher than 5%. Some types of combustion plant emit relatively smaller proportions of nitrous oxide (N 2 O), whilst a small number of other types of installation are also known to emit N 2 O.”, “author” : { “dropping-particle” : “”, “family” : “Stewart”, “given” : “Duncan”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “chapter-number” : “5”, “container-title” : “Monitoring Quick Guide”, “id” : “ITEM-1”, “issued” : { “date-parts” : “0” }, “publisher” : “Scottish Environment Protection Agency”, “publisher-place” : “Scotland”, “title” : “Monitoring oxides of Nitrogen”, “type” : “chapter” }, “uris” : “http://www.mendeley.com/documents/?uuid=7d917756-07c1-3d3d-8447-d45a65a1fcf0” } , “mendeley” : { “formattedCitation” : “<sup>13</sup>”, “plainTextFormattedCitation” : “13”, “previouslyFormattedCitation” : “<sup>13</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }13
Differing from what was outlined above, a wet chemical method was employed for the sampling of nitrogen dioxide during this project. In this case, the triethanolamine (TEA) method, which involved air being pulled through an aqueous solution of triethanolamine; nitrogen dioxide present in the sampled air volume was absorbed by the solution. A subsequent colourimetric analysis of the exposed absorbing solution indicated the concentration of NO2 in the air sample. ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “author” : { “dropping-particle” : “”, “family” : “(EPA)”, “given” : “Environmental Protection Agency”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “id” : “ITEM-1”, “issued” : { “date-parts” : “1974” }, “number-of-pages” : “35”, “title” : “Evaluation of Triethanolamine: Procedure for Determination of Nitrogen Dioxide in Ambient Air”, “type” : “report” }, “uris” : “http://www.mendeley.com/documents/?uuid=4177c93c-cde3-4938-ad7a-41aee1d12bad” } , “mendeley” : { “formattedCitation” : “<sup>14</sup>”, “plainTextFormattedCitation” : “14”, “previouslyFormattedCitation” : “<sup>14</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }14
1.3 Ozone
1.3.1 What is Ozone?
Ozone (O3) is a colourless, reactive gas comprised of three oxygen atoms. Ozone occurs both at ground-level and in the Earth’s upper atmosphere. Ozone in the upper atmosphere is vital for absorbing UV rays; at the earth’s surface, it is harmful. Ozone found at ground-level is referred to as tropospheric or “bad” ozone. Ozone found in the Earth’s upper atmosphere is referred to as stratospheric or “good” ozone. Stratospheric ozone forms a protective layer that shields from the sun’s harmful UV rays. Manmade chemical have partially destroyed this beneficial ozone layer, causing what is referred to as a “hole in the ozone”. ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “URL” : “https://www.epa.gov/ozone-pollution/basic-information-about-ozone#regulations”, “abstract” : “Learn the difference between good (stratospheric) and bad (tropospheric) ozone, how bad ozone affects our air quality, health, and environment, and what EPA is doing about it through regulations and standards.”, “accessed” : { “date-parts” : “2018”, “4”, “25” }, “author” : { “dropping-particle” : “”, “family” : “US EPA”, “given” : “”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “id” : “ITEM-1”, “issued” : { “date-parts” : “0” }, “title” : “Basic Information about Ozone”, “type” : “webpage” }, “uris” : “http://www.mendeley.com/documents/?uuid=ec8d3714-3193-31a8-aa83-6cc9639536b8” } , “mendeley” : { “formattedCitation” : “<sup>15</sup>”, “plainTextFormattedCitation” : “15”, “previouslyFormattedCitation” : “<sup>15</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }15
1.3.2 How is Ozone Formed?
Ozone is formed throughout the atmosphere in multistep chemical processes that require sunlight.
Ground-level ozone is not emitted directly into the air, but is formed via photochemical chemical reactions between NOX and volatile organic compounds (VOCs) in the presence of sunlight. The following reaction shows one of the ways that ground-level ozone can be formed:
NO2 + O2 NO + O3
Stratospheric ozone occurs naturally, the process begins with an oxygen molecule (O2) being broken down by UV radiation from the sun to produce two oxygen atoms. Next, each of these highly reactive oxygen atoms combines with an oxygen molecule to produce an ozone molecule (O3). These reactions occur constantly, whenever solar UV radiation is present in the stratosphere. ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “URL” : “https://www.epa.gov/ozone-pollution/basic-information-about-ozone#regulations”, “abstract” : “Learn the difference between good (stratospheric) and bad (tropospheric) ozone, how bad ozone affects our air quality, health, and environment, and what EPA is doing about it through regulations and standards.”, “accessed” : { “date-parts” : “2018”, “4”, “25” }, “author” : { “dropping-particle” : “”, “family” : “US EPA”, “given” : “”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “id” : “ITEM-1”, “issued” : { “date-parts” : “0” }, “title” : “Basic Information about Ozone”, “type” : “webpage” }, “uris” : “http://www.mendeley.com/documents/?uuid=ec8d3714-3193-31a8-aa83-6cc9639536b8” } , “mendeley” : { “formattedCitation” : “<sup>15</sup>”, “plainTextFormattedCitation” : “15”, “previouslyFormattedCitation” : “<sup>15</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }15 This is represented by the following reactions:
O2 O + O
O + O2 O3
Ozone molecules can also be decomposed by UV radiation into an oxygen atom and an oxygen molecule. Therefore, ozone is continuously produced and destroyed in the atmosphere by UV radiation coming from the sun. In an unpolluted atmosphere there is a balance between the amount of ozone being produced and destroyed, hence the total concentration of ozone remains relatively constant.

1.3.3 Sources of Ozone
Emissions from industrial facilities, chemical plants, power plants, motor vehicles and chemical solvents are some of the major sources of NOx and VOCs. Ground-level ozone is not emitted directly into the air, but is formed via photochemical chemical reactions between NOX and VOCs in the presence of sunlight. Hence, ozone is most likely to reach unhealthy levels on hot sunny days in urban environments, but can still reach high levels during colder months. Rural areas can also experience high ozone levels as ozone can be transported long distance by wind.ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “URL” : “https://www.epa.gov/ozone-pollution”, “abstract” : “Known as tropospheric or “ground-level” ozone, this gas is harmful to human heath and the environment. Since it forms from emissions of volatile organic compounds (VOCs) and nitrogen oxides (NOx), these pollutants are regulated under air quality standards.”, “accessed” : { “date-parts” : “2018”, “4”, “25” }, “author” : { “dropping-particle” : “”, “family” : “US EPA”, “given” : “”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “id” : “ITEM-1”, “issued” : { “date-parts” : “0” }, “title” : “Ozone Pollution”, “type” : “webpage” }, “uris” : “http://www.mendeley.com/documents/?uuid=16645eb5-b9b3-3a0e-bea5-3ff464ec242b” } , “mendeley” : { “formattedCitation” : “<sup>16</sup>”, “plainTextFormattedCitation” : “16”, “previouslyFormattedCitation” : “<sup>16</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }16
1.3.4 Harmfulness of Ozone and the Need to Monitor
Ground-level ozone is a harmful pollutant, because of its effects on people and the environment, and it is the main ingredient in “photochemical smog.”  Ozone is one of the six common air pollutants identified in the Clean Air Act. The Environmental Protection Agency (EPA) calls these “criteria air pollutants” because their levels in outdoor air need to be limited based on health criteria. ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “URL” : “https://www.epa.gov/ozone-pollution”, “abstract” : “Known as tropospheric or “ground-level” ozone, this gas is harmful to human heath and the environment. Since it forms from emissions of volatile organic compounds (VOCs) and nitrogen oxides (NOx), these pollutants are regulated under air quality standards.”, “accessed” : { “date-parts” : “2018”, “4”, “25” }, “author” : { “dropping-particle” : “”, “family” : “US EPA”, “given” : “”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “id” : “ITEM-1”, “issued” : { “date-parts” : “0” }, “title” : “Ozone Pollution”, “type” : “webpage” }, “uris” : “http://www.mendeley.com/documents/?uuid=16645eb5-b9b3-3a0e-bea5-3ff464ec242b” } , “mendeley” : { “formattedCitation” : “<sup>16</sup>”, “plainTextFormattedCitation” : “16”, “previouslyFormattedCitation” : “<sup>16</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }16 People most at risk from breathing ozone polluted air are children, the elderly, people with lung problems such as asthma, and people who are active outdoors, especially those who work outside. The main health concern of exposure ozone is its effect on the respiratory system, especially on lung function. Several factors influence these health impacts, including the concentrations of ground-level ozone in the atmosphere, the duration of exposure, average volume of air breathed per minute (ventilation rate), and the length of intervals between short-term exposures. Inhaling ozone can trigger things such as chest pain, coughing, throat irritation and airway inflammation. Along with reducing lung function and harming lung tissue, exposure to ozone can also worsen bronchitis, emphysema and asthma. ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “abstract” : “Ozone (O 3) is a colorless, reactive oxidant gas that is a major constituent of atmospheric smog. Ground-level ozone is formed in the air by the photochemi-cal reaction of sunlight and nitrogen oxides (NO x), facilitated by a variety of volatile organic com-pounds (VOCs), which are photochemically reactive hydrocarbons. The relative importance of the vari-ous VOCs in the oxidation process depends on their chemical structure and reactivity. Ozone may be formed by the reaction of NO x and VOCs under the influence of sunlight hundreds of kilometers from the source of emissions. Ozone concentrations are influenced by the intensity of solar radiation, the absolute concen-trations of NO x and VOCs, and the ratio of NO x and VOCs. Diurnal and seasonal variations oc-cur in response to changes in sunlight. In addi-tion, ground-level ozone accumulation occurs when sea breezes cause circulation of air over an area or when temperature-induced air inversions trap the compounds that produce smog (Chilton and Sholtz 1989). Peak ground-level ozone con-centrations are measured in the afternoon. Mean concentrations are generally highest during the summer. Peak concentrations of ground-level ozone rarely last longer than two to three hours (WHO 1979). Registered average natural background concen-trations of ground-level ozone are around 30u2013100 micrograms per cubic meter (u00b5g/m 3). Short-term (one-hour) mean ambient concentrations in urban areas may exceed 300u2013800 u00b5g/m 3 (WHO 1979).”, “author” : { “dropping-particle” : “”, “family” : “WORLD BANK GROUP”, “given” : “”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “id” : “ITEM-1”, “issued” : { “date-parts” : “1998” }, “number-of-pages” : “4”, “title” : “Ground-Level Ozone”, “type” : “report” }, “uris” : “http://www.mendeley.com/documents/?uuid=13f3b4f6-87df-32af-919f-807ad96f140a” } , “mendeley” : { “formattedCitation” : “<sup>17</sup>”, “plainTextFormattedCitation” : “17”, “previouslyFormattedCitation” : “<sup>17</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }17
Ground-level ozone can also have harmful effects on sensitive vegetation and ecosystems, including forests, parks, wildlife refuges and wilderness areas. Ozone especially harms plantlife during the growing season.

1.3.5 Chemical Method for Sampling OzoneMicro-amounts of ozone and other oxidants liberate iodine when absorbed in a 1% solution of potassium iodide buffered at pH 6.8 ± 0.2. The iodide can be determined spectrophotometrically by measuring the absorption of triiodide ion at 352nm.

The stoichiometry is approximated by the following reaction:
O3 + 3KI + H2O ? KI3 + 2KOH + O2
O3 + 2KI + H2O ? O2 + I2 + 2KOH
O2 + I2 + KI ? KI3 + O2
KI3 is formed through the equilibrium:
I2 + KI KI3
The molar absorption coefficient of the I?3 at its wavelength of maximum absorption ( = 352 nm) is very large (? = 26,000 dm3 mol-1 cm-1) while that of the I2 at the same wavelength is comparatively small (?= 324 dm3 mol-1 cm-1). Thus the reaction of ozone with potassium iodide can be followed conveniently by measuring the absorbance at 352nm. ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “author” : { “dropping-particle” : “”, “family” : “Dublin Institute of Technology”, “given” : “”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “id” : “ITEM-1”, “issued” : { “date-parts” : “0” }, “number-of-pages” : “4”, “title” : “CHEM 4006 LAB EXPERIMENTS”, “type” : “report” }, “uris” : “http://www.mendeley.com/documents/?uuid=536e273a-0984-4c08-9ee7-d8d8dd7c24ed” } , “mendeley” : { “formattedCitation” : “<sup>18</sup>”, “plainTextFormattedCitation” : “18”, “previouslyFormattedCitation” : “<sup>18</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }18
1.4 Statistical Analysis by Using R and Data Collection
The forecast models for the different pollutants and all statistical techniques used have been studied by using R, which is a language and environment for statistical computing and graphics. It provides a wide variety of statistical and graphical techniques. The R language is widely used among statisticians and data miners for developing statistical software and data analysis. For this project, RStudio has been used. RStudio is a free and open-source integrated development environment for R. RStudio was founded by JJ Allaire in 2009.ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “URL” : “https://www.rstudio.com/about/”, “accessed” : { “date-parts” : “2018”, “5”, “6” }, “id” : “ITEM-1”, “issued” : { “date-parts” : “0” }, “title” : “About u2013 RStudio”, “type” : “webpage” }, “uris” : “http://www.mendeley.com/documents/?uuid=1ab81785-e545-3d4c-8cb5-7efcbbb9e0c9” } , “mendeley” : { “formattedCitation” : “<sup>19</sup>”, “plainTextFormattedCitation” : “19”, “previouslyFormattedCitation” : “<sup>19</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }19 Firstly, the different statistical techniques which have been used will be discussed in more detail in the Experimental section. Secondly, the data sets for previous years were obtained from the EPA website. ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “URL” : “http://www.epa.ie/irelandsenvironment/environmentalindicators/”, “abstract” : “environmental indicators for each thematic”, “accessed” : { “date-parts” : “2018”, “5”, “21” }, “author” : { “dropping-particle” : “”, “family” : “(EPA)”, “given” : “Environmental Protection Agency”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “container-title” : “Environmental Protection Agency (EPA)”, “id” : “ITEM-1”, “issued” : { “date-parts” : “0” }, “publisher” : “Environmental Protection Agency (EPA)”, “title” : “Environmental Indicators – Environmental Protection Agency (EPA”, “type” : “webpage” }, “uris” : “http://www.mendeley.com/documents/?uuid=a87a431a-1f6a-3c05-b70b-425634f9ae11” } , “mendeley” : { “formattedCitation” : “<sup>20</sup>”, “plainTextFormattedCitation” : “20”, “previouslyFormattedCitation” : “<sup>20</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }20

1.5 Project Aim
The aim of this project is to monitor NO2 and ground-level ozone in Dublin city centre by using a chemical sampling method for each pollutant, the “Evaluation of Triethanolamine Procedure for Determination of Nitrogen Dioxide in Ambient Air” for NO2 and “Evaluation of 1 Percent Neutral Buffered Potassium Iodide Procedure for Calibration of Ozone Monitors” for O3. Also, additional data was obtained from the EPA website on the NOx and ozone emissions from previous years. The data obtained was used to create different predictive models for a; studied pollutants using R.

Chapter 2: Experimental
2.1 Determination of Nitrogen Dioxide in Ambient Air
Nitrogen dioxide was sampled using the wet chemical method obtained from the EPA, “Evaluation of Triethanolamine Procedure for Determination of Nitrogen Dioxide in Ambient Air”.ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “author” : { “dropping-particle” : “”, “family” : “(EPA)”, “given” : “Environmental Protection Agency”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “id” : “ITEM-1”, “issued” : { “date-parts” : “1974” }, “number-of-pages” : “35”, “title” : “Evaluation of Triethanolamine: Procedure for Determination of Nitrogen Dioxide in Ambient Air”, “type” : “report” }, “uris” : “http://www.mendeley.com/documents/?uuid=4177c93c-cde3-4938-ad7a-41aee1d12bad” } , “mendeley” : { “formattedCitation” : “<sup>14</sup>”, “plainTextFormattedCitation” : “14”, “previouslyFormattedCitation” : “<sup>14</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }14
2.1.1 Reagents
(TEA) Triethanolamine N(CH2CH2OH)3
Hydrogen peroxide (85%) H2O2
Sulfanilamide C6H8N2O2S
(NEDA) N-(l-Napthyl)-ethylenediamine dihydrochloride C12H16Cl2N2
Sodium nitrite NaNO2
Phosphoric acid (30%) H3PO4
2.1.2 Instruments
UV spectrophotometer, Shimadzu, UV-1800
NOX absorbing apparatus
2.1.3 Solutions prepared
Absorbing solution – Triethanolamine 13.27 cm3, 100.5 mmol was dissolved in 500cm3 of deionised water. This was then diluted to one litre with deionised water.

Hydrogen peroxide solution – 30% hydrogen peroxide 0.2 cm3 0.222 g, 6.5 mmol of was diluted to 250 cm3 using deionised water. This was done in the fume hood.

Sulfanilamide solution – Sulfanilamide 20.0000 g, 0.8 mmol of was dissolved in 700 cm3 deionised water. The solution needed to be heated and stirred at the same time for the sulfanilamide to dissolve. Then phosphoric acid 50 cm3, 84.25 g, 859.8 mmol of was added while stirring, and this then diluted to one litre using deionised water. This was done in the fume hood.

NEDA solution – NEDA 0.5003 g, 2.7 mmol of was dissolved in 500 cm3 deionised water.

Standard nitrite solution – one litre of 100 ?gNO2/cm3 was prepared for a stock nitrite solution. The amount of NaNO2 was calculated using the following equation:
G=1.500A×100%10where,
G = amount of NaNO2 in grams1.500 = gravimetric factor in converting NO2 into NaNO2 A = assay per cent (97%)
10 = dilution factor
G=1.50097×100%10G = 0.1546 g 1.8 mmol
2.1.4 Procedure
Sampling – 10 cm3 of absorbing solution was placed in the glass impinger of the absorbing apparatus. The flow rate of the sampling pump was adjusted to 1.0 L/min. Ambient air was sampled for two hours. The two hour sampling period was repeated twice a day for three days.

5518153577480Figure 2.1.4.1: NO2 Absorbing Apparatus
020000Figure 2.1.4.1: NO2 Absorbing Apparatus
55245073850500During the sampling period, the NO2 values were recorded from the World Air Quality website.

Analysis – 5 cm3 of the collected sample was pipetted into a test tube, 0.5 cm3 of the peroxide solution was added, along with 5 cm3 sulfanilamide solution and 0.7 cm3 NEDA solution, while thoroughly mixing after addition of each of these reagents. A blank was prepared for use as a reference in the same manner using 5 cm3 of unexposed absorbing reagent. The colour was allowed to develop for 10 minutes. The absorbance of the sample was then measured against the blank at 530 nm.
Calibration curve – The stock standard nitrite solution was diluted 50:1 with TEA absorbing solution to prepare a working standard solution containing 2.0 ?g NO2 / cm3. A range of calibrations were prepared by individually adding 0.1, 0.5, 1.0, 2.0, 3.0, 4.0 and 5.0 cm3 of the stock standard nitrite solution to 5 cm3 using the absorbing solution. These standards were prepared in the same manner as the sample above and then run on the UV. The absorbances were obtained at 530 nm. This was done in triplicate to get an average, and therefore more accuracy was obtained. These absorbance values were then plotted against concentration of NO2 in ?g/cm3 to produce a calibration curve on excel. The line of best fit of this plot was then used to find the concentration of NO2 in each sample of ambient air. ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “author” : { “dropping-particle” : “”, “family” : “(EPA)”, “given” : “Environmental Protection Agency”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “id” : “ITEM-1”, “issued” : { “date-parts” : “1974” }, “number-of-pages” : “35”, “title” : “Evaluation of Triethanolamine: Procedure for Determination of Nitrogen Dioxide in Ambient Air”, “type” : “report” }, “uris” : “http://www.mendeley.com/documents/?uuid=4177c93c-cde3-4938-ad7a-41aee1d12bad” } , “mendeley” : { “formattedCitation” : “<sup>14</sup>”, “plainTextFormattedCitation” : “14”, “previouslyFormattedCitation” : “<sup>14</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }14

2.2 Measurement of Ground-Level Ozone in Ambient Air
Ground-level ozone was sampled using the wet chemical method obtained from the EPA, “Evaluation of 1 Percent Neutral Buffered Potassium Iodide Procedure for Calibration of Ozone Monitors”.ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “author” : { “dropping-particle” : “”, “family” : “(EPA)”, “given” : “Environmental Protection Agency”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “id” : “ITEM-1”, “issued” : { “date-parts” : “1977” }, “number-of-pages” : “47”, “title” : “Evaluation of 1 Percent Neutral Buffered Potassium Iodide Procedure for Calibration of Ozone Monitors”, “type” : “report” }, “uris” : “http://www.mendeley.com/documents/?uuid=3b6699c8-8ef0-4fa6-8f02-1d0286964de1” } , “mendeley” : { “formattedCitation” : “<sup>21</sup>”, “plainTextFormattedCitation” : “21”, “previouslyFormattedCitation” : “<sup>21</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }21
2.2.1 Reagents
Potassium iodide
Potassium dihydrogen phosphate
Disodium hydrogen phosphate
Sodium hydroxide Resublimed iodine
0.025 M sodium thiosulphate2.2.2 Instruments
UV spectrophotometer, Shimadzu, UV-1800
Ozone absorbing apparatus
2.2.3 Solutions provided
Absorbing solution (1% KI in 0.1 M phosphate buffer) – Potassium dihydrogen phosphate 13.6 g 102.1 mmol, disodium hydrogen phosphate 14.2 g, 100.0 mmol and potassium iodide 10.0 g, 60.2 mmol diluted to 1 L with deionised water. pH adjusted to 6.8 with sodium hydroxide.

Stock solution (0.025 M I2) – Potassium iodide 16.0 g, 96.4 mmol and resublimed iodine 3.17 g, 1.2 mmol made up to 500 cm3 with deionised water.

0.001M I2 solution – 4.0 cm3 of the 0.025 M stock solution diluted to 100 cm3 with absorbing solution.

2.2.4 Procedure
Sampling – 10 cm3 of absorbing solution was placed in the glass impinger of the absorbing apparatus. The flow rate of the sampling pump was adjusted to 1.0 litre per minute. Ambient air was sampled for two hours. The two hour sampling period was repeated twice a day for three days.

During the sampling period, the O3 values were recorded from the World Air Quality website.

2470153070942Figure 2.2.4.1: Ozone Absorbing Apparatus
020000Figure 2.2.4.1: Ozone Absorbing Apparatus
24066527178000
Analysis – When the two hour sampling period had elapsed, the absorbance of the sampling solution was recorded at 352 nm in quartz cuvettes with absorbing solution as the reference.
Calibration Curve – While this sampling was taking place, a set of I2 calibrationsolutions were prepared as follows.

5.0 cm3 of the 0.001 M I2 solution provided was diluted to 100 cm3 using the absorbing solution. This is the equivalent to 2.4 g O3 per cm3. A range of calibrations were prepared by individually adding 0.2, 0.4, 0.6, 0.8, 1.0 and 2.0 cm3 of the I2 calibration solution to 10 cm3 volumetric flasks. Each was brought up to the mark using the absorbing solution provided. The spectrometer was zeroed and baselined using the absorbing solution as the reference in two quartz cells. The UV spectrums were recorded between 300 and 800 nm. Each calibration solutions were run on the UV, with the absorbing solution in the reference cell. The absorbance of each of these was measured at 352 nm. This was done in triplicate to get an average, and therefore more accuracy was obtained. These absorbance values were then plotted against concentration of ozone in mol/10cm3 to produce a calibration curve on excel. The line of best fit of this plot was then used to find the concentration of ozone in each sample of ambient air. ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “author” : { “dropping-particle” : “”, “family” : “(EPA)”, “given” : “Environmental Protection Agency”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “id” : “ITEM-1”, “issued” : { “date-parts” : “1977” }, “number-of-pages” : “47”, “title” : “Evaluation of 1 Percent Neutral Buffered Potassium Iodide Procedure for Calibration of Ozone Monitors”, “type” : “report” }, “uris” : “http://www.mendeley.com/documents/?uuid=3b6699c8-8ef0-4fa6-8f02-1d0286964de1” } , “mendeley” : { “formattedCitation” : “<sup>21</sup>”, “plainTextFormattedCitation” : “21”, “previouslyFormattedCitation” : “<sup>21</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }21

2.3 Building NOx and Ozone Models using R Studio
R (using the interface RStudio) was used to build and run several different things forecasting models with NOX and ozone data obtained from the Irish EPA website. The NOX data obtained was the concentration of NOx, NO2 and NO in ?g/m3 for Dublin for 2016. The ozone data obtained was for the ozone concentration in ?g/m3 for Dublin, Cork, Galway and Mayo during the period 2014 – 2016. The weather parameters (maximum temperature, minimum temperature, rain, pressure, wind speed, wind direction, sun duration and, in Dublin only, global radiation) were also taken into account in these scripts. Moreover, all the models have been trained and calibrated with varying percentages of the dataset (70-80%) and later been validated and checked with the remaining data. This was done to increase the fit of the model.

2.3.1 Model 1
The first script was to produce a linear model of the concentration variables and weather variables using multi-variate linear regression. The models plotted the observed values vs the predicted values. A logarithmic regression model was added to the script to produce a smoothed version of the model which worked much better once the adjustment was made.

2.3.2 Model 2
The second script produced a correlation plot for each concentration variable versus its corresponding weather variables. This was done using the Corrplot package in R Studio.

2.3.3 Model 3
The third script used the same model as the first script and added to it. It was used to compare the normality of the observed and predicted values. It was also used to produce a histogram for each to show the normality of the observed and predicted values. This can also be seen in the qqnorm plots produced by the script for the observed and predicted values. This script also produces a box plot for the observed and predicted, which shows any outliers present.

2.3.4 Model 4
The fourth script was used to improve the r2 values obtained from model one. This script worked with the MARS technique, and was performed using the Earth package in R Studio. This produced better models and r2 values by choosing the best fit along the slope by creating sub groups, and selecting the best parameters to suit the model.

2.3.5 Multivariate Linear Regression
Multivariate linear regression has been used to produce the models for NOX and ozone data. Linear regression modelling is an important tool in many statistical analyses for the studying the relationships among variables. Multivariate linear regression has been used to develop forecast models for ambient air pollution such as NOX and ozone. Regression analysis is primarily used for predicting values of the response variable (the pollutant in this case) at interesting values of the predictor variables (usually weather), discovering which weather predictors are associated with the response variable, and estimating how changes in the predictor variables affect the response variable.
When using linear regression, the predicted values should be close to the observed values. Therefore displaying good fit, the model should resemble as closely as possible a straight line. Ideally, all of the points should be close to this line of best fit. R2 shows the proportion of the total variance explained by the regression model, and also how much of the linear variation in the observed values is explained by the variation in the predicted values. ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “DOI” : “10.1016/j.ecolmodel.2008.05.006”, “ISBN” : “0304-3800”, “ISSN” : “03043800”, “PMID” : “258557400006”, “abstract” : “A common and simple approach to evaluate models is to regress predicted vs. observed values (or vice versa) and compare slope and intercept parameters against the 1:1 line. However, based on a review of the literature it seems to be no consensus on which variable (predicted or observed) should be placed in each axis. Although some researchers think that it is identical, probably because r2is the same for both regressions, the intercept and the slope of each regression differ and, in turn, may change the result of the model evaluation. We present mathematical evidence showing that the regression of predicted (in the y-axis) vs. observed data (in the x-axis) (PO) to evaluate models is incorrect and should lead to an erroneous estimate of the slope and intercept. In other words, a spurious effect is added to the regression parameters when regressing PO values and comparing them against the 1:1 line. Observed (in the y-axis) vs. predicted (in the x-axis) (OP) regressions should be used instead. We also show in an example from the literature that both approaches produce significantly different results that may change the conclusions of the model evaluation. u00a9 2008 Elsevier B.V. All rights reserved.”, “author” : { “dropping-particle” : “”, “family” : “Piu00f1eiro”, “given” : “Gervasio”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” }, { “dropping-particle” : “”, “family” : “Perelman”, “given” : “Susana”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” }, { “dropping-particle” : “”, “family” : “Guerschman”, “given” : “Juan P.”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” }, { “dropping-particle” : “”, “family” : “Paruelo”, “given” : “Josu00e9 M.”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “container-title” : “Ecological Modelling”, “id” : “ITEM-1”, “issue” : “3-4”, “issued” : { “date-parts” : “2008” }, “page” : “316-322”, “title” : “How to evaluate models: Observed vs. predicted or predicted vs. observed?”, “type” : “article-journal”, “volume” : “216” }, “uris” : “http://www.mendeley.com/documents/?uuid=c7ec7656-e0ed-4b57-b654-019884048c77” } , “mendeley” : { “formattedCitation” : “<sup>22</sup>”, “plainTextFormattedCitation” : “22”, “previouslyFormattedCitation” : “<sup>22</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }22 If the model has a high R2 value, all the points would be close to this line of best fit. The lower the R2 value, the weaker the predictive power of the model, and the more dispersed the points are. Points to the left or right of the plot, furthest from the mean, have the most leverage, and effectively try to pull the fitted line toward the point. Points that are vertically distant from the line represent possible outliers. Both types of points can adversely affect the fit of the model. More advanced techniques such multivariate adaptive regression (MARS, Friedman 1991), have been used also. It works in a more flexible way than traditional multivariate regression models, because the MARS model works by automatically selecting which variables to use (which variables are important and which are not), ie, the positions of the kinks in the hinge functions, and how the hinge functions are combined. As a consequence, the robustness of the model has been improved.

However, it should be noted that according to George Box, a British statistician, “Essentially, all models are wrong, but some are useful”. ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “author” : { “dropping-particle” : “”, “family” : “Statwing”, “given” : “”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “id” : “ITEM-1”, “issued” : { “date-parts” : “0” }, “title” : “Interpreting residual plots to improve your regression | Statwing Documentation”, “type” : “report” }, “uris” : “http://www.mendeley.com/documents/?uuid=401b3883-fe64-3965-9a16-1f9d32782c0b” } , “mendeley” : { “formattedCitation” : “<sup>23</sup>”, “plainTextFormattedCitation” : “23”, “previouslyFormattedCitation” : “<sup>23</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }23
2.3.6 Correlation
RStudio will also be used to construct a correlation matrix for all of the independent variables and the dependent variable from the observed data. The correlation value gives an idea about which variable is significant and by what factor.

Correlation is a statistical analysis technique that measures the strength of association between two variables, and the direction of the relationship. In relation to the strength of the relationship, the correlation coefficient values vary between +1 and -1. A correlation coefficient of ±1 means that there is a perfect degree of association between the two variables. As the correlation coefficient value goes toward 0, the relationship between the two variables becomes weaker. A positive correlation value means that there is a positive correlation between the two variables, if one increases, so does the other. A negative correlation value means that there is a negative correlation between the two variables, if one increases, the other decreases.ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “URL” : “http://www.statisticshowto.com/what-is-correlation/”, “abstract” : “Probability and Statistics Topic Index”, “accessed” : { “date-parts” : “2018”, “5”, “10” }, “author” : { “dropping-particle” : “”, “family” : “Statistics How TO”, “given” : “”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “id” : “ITEM-1”, “issued” : { “date-parts” : “2013” }, “title” : “What is Correlation in Statistics? Correlation Analysis Explained”, “type” : “webpage” }, “uris” : “http://www.mendeley.com/documents/?uuid=c4fb68b3-9871-3717-8db6-a1fc7fe69e60” } , “mendeley” : { “formattedCitation” : “<sup>24</sup>”, “plainTextFormattedCitation” : “24”, “previouslyFormattedCitation” : “<sup>24</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }24
2.3.7 Boxplots
Boxplots are a measure of how well distributed the data is in a data set. They are useful for identifying outliers and for comparing distributions. Boxplots display the distribution of data based on the five number summary: minimum, lower quartile, median, upper quartile, and maximum. The median marks the mid-point of the data and is shown by the line that divides the box into two parts. Thus half the sample values are greater than or equal to this value and half are less. The middle box represents the middle 50% of sample values for the group. The range of sample values from lower to upper quartile is referred to as the inter-quartile range (IQR). The middle 50% of sample values fall within the inter-quartile range. Seventy-five percent of the sample values fall below the upper quartile. Twenty-five percent of scores fall below the lower quartile.ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “URL” : “https://www.wellbeingatschool.org.nz/information-sheet/understanding-and-interpreting-box-plots”, “accessed” : { “date-parts” : “2018”, “5”, “10” }, “id” : “ITEM-1”, “issued” : { “date-parts” : “0” }, “title” : “Understanding and interpreting box plots | [email protected]”, “type” : “webpage” }, “uris” : “http://www.mendeley.com/documents/?uuid=3b308058-94b2-381d-9a12-d97f3a49a4b7” } , “mendeley” : { “formattedCitation” : “<sup>25</sup>”, “plainTextFormattedCitation” : “25”, “previouslyFormattedCitation” : “<sup>25</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }25
There are two main types of outliersADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “URL” : “http://www.physics.csbsju.edu/stats/box2.html”, “accessed” : { “date-parts” : “2018”, “5”, “10” }, “id” : “ITEM-1”, “issued” : { “date-parts” : “0” }, “title” : “Box Plot: Display of Distribution”, “type” : “webpage” }, “uris” : “http://www.mendeley.com/documents/?uuid=76d4e56a-a136-3942-943e-ab0f67992136” } , “mendeley” : { “formattedCitation” : “<sup>26</sup>”, “plainTextFormattedCitation” : “26”, “previouslyFormattedCitation” : “<sup>26</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }26:
Outliers are either 3×IQR or more above the third quartile or 3×IQR or more below the first quartile.

Suspected outliers are slightly more central versions of outliers: either 1.5×IQR or more above the third quartile or 1.5×IQR or more below the first quartile.

2.3.8 Histograms
A histogram is a visual representation of the distribution of a dataset. The shape of a histogram is its most obvious and informative characteristic, it gives an estimate as to where values are concentrated, what the extremes are and whether there are any gaps or unusual values.ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “URL” : “https://www.r-bloggers.com/how-to-make-a-histogram-with-basic-r/”, “accessed” : { “date-parts” : “2018”, “5”, “10” }, “author” : { “dropping-particle” : “”, “family” : “R Bloggers”, “given” : “”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “id” : “ITEM-1”, “issued” : { “date-parts” : “0” }, “title” : “How to Make a Histogram with Basic R | R-bloggers”, “type” : “webpage” }, “uris” : “http://www.mendeley.com/documents/?uuid=bd2d993d-77d2-3cff-b1ab-381b63268490” } , “mendeley” : { “formattedCitation” : “<sup>27</sup>”, “plainTextFormattedCitation” : “27”, “previouslyFormattedCitation” : “<sup>27</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }27 In short, the histogram consists of an x-axis, a y-axis and various bars of different heights. The y-axis shows how frequently the values on the x-axis occur in the data, while the bars group ranges of values on the x-axis.ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “URL” : “https://datavizcatalogue.com/methods/histogram.html”, “accessed” : { “date-parts” : “2018”, “5”, “10” }, “author” : { “dropping-particle” : “”, “family” : “The Data Visualisation Catalogue”, “given” : “”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “id” : “ITEM-1”, “issued” : { “date-parts” : “0” }, “title” : “Histogram – Learn about this chart and tools to create it”, “type” : “webpage” }, “uris” : “http://www.mendeley.com/documents/?uuid=697a50bf-6a8b-3ce1-ad97-5574b1c9fb47” } , “mendeley” : { “formattedCitation” : “<sup>28</sup>”, “plainTextFormattedCitation” : “28”, “previouslyFormattedCitation” : “<sup>28</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }28 If a p value of less than or equal to 0.05 is obtained, that the test rejects the hypothesis of normality. Failing this normality test allows it to be stated with 95% confidence that the data does not fit the normal distribution. ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “URL” : “http://www.variation.com/da/help/hs140.htm”, “accessed” : { “date-parts” : “2018”, “5”, “11” }, “id” : “ITEM-1”, “issued” : { “date-parts” : “0” }, “title” : “Anderson-Darling Normality Test”, “type” : “webpage” }, “uris” : “http://www.mendeley.com/documents/?uuid=b982c1d7-a175-3646-9250-361a3684eadf” } , “mendeley” : { “formattedCitation” : “<sup>29</sup>”, “plainTextFormattedCitation” : “29”, “previouslyFormattedCitation” : “<sup>29</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }29
2.3.9 Quantile-Quantile Plot
The quantile-quantile (Q-Q) plot is a graphical technique for determining if two data sets come from populations with a normal distribution. It is a way of testing the normality of the quantiles. A Q-Q plot is a scatterplot created by plotting two sets of quantiles against one another. If both sets of quantiles came from the same distribution, the line formed should be roughly straight. Q-Q plots take your sample data, sort it in ascending order, and then plot them versus quantiles calculated from the predicted distribution. ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “URL” : “http://data.library.virginia.edu/understanding-q-q-plots/”, “accessed” : { “date-parts” : “2018”, “5”, “10” }, “author” : { “dropping-particle” : “”, “family” : “Ford”, “given” : “Clay”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “container-title” : “University of Virginia Library”, “id” : “ITEM-1”, “issued” : { “date-parts” : “2015” }, “title” : “Understanding Q-Q Plots | University of Virginia Library Research Data Services + Sciences”, “type” : “webpage” }, “uris” : “http://www.mendeley.com/documents/?uuid=919c042f-1995-38d2-a7f9-847d3e2b47e4” } , “mendeley” : { “formattedCitation” : “<sup>30</sup>”, “plainTextFormattedCitation” : “30”, “previouslyFormattedCitation” : “<sup>30</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }30 If a p value of less than or equal to 0.05 is obtained, that the test rejects the hypothesis of normality. Failing this normality test allows it to be stated with 95% confidence that the data does not fit the normal distribution. ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “URL” : “http://www.variation.com/da/help/hs140.htm”, “accessed” : { “date-parts” : “2018”, “5”, “11” }, “id” : “ITEM-1”, “issued” : { “date-parts” : “0” }, “title” : “Anderson-Darling Normality Test”, “type” : “webpage” }, “uris” : “http://www.mendeley.com/documents/?uuid=b982c1d7-a175-3646-9250-361a3684eadf” } , “mendeley” : { “formattedCitation” : “<sup>29</sup>”, “plainTextFormattedCitation” : “29”, “previouslyFormattedCitation” : “<sup>29</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }29

Chapter 3: Results and Discussion
3.1 Chemical Methods: NO2
In this section, the results from the NO2 chemical sampling method will be shown and discussed. Shown below in Table 3.1.1, is the calibration curve data for NO2, and the calibration curve is shown in Figure 3.1.1. A calibration curve is used as a method for determining the concentration of NO2 in the absorbing solution which has been left to absorb ambient air for two hours, by comparing the absorbance of the absorbing solution to the set of standard samples of known concentration and absorbance.

Table 3.1.1: NO2 Calibration Curve
x-axis Run 1 Run 2 Run 3 y-axis
NO2 Concentration (?g/cm3) Absorbance Absorbance Absorbance Average Absorbance
0.04 0.049 0.014 0.010 0.024
0.20 0.090 0.100 0.113 0.101
0.40 0.183 0.187 0.180 0.183
0.80 0.378 0.406 0.396 0.393
1.20 0.594 0.580 0.603 0.592
1.60 0.780 0.697 0.744 0.740
2.00 0.968 0.938 0.957 0.954

Figure 3.1.1: NO2 Calibration Curve
Shown below is a sample of the calculations used to find the NO2 concentration in the absorbing solution in ?g/cm3 by using the above calibration curve.

Sample Calculation:
y = 0.4722 x + 0.0058
y = 0.035
0.035 = 0.4722 x + 0.0058
0.035 – 0.0058 = 0.4722 x
0.0292 = 0.4722 x
x = 0.0292/0.4722
x = 0.062 ?g/cm3
Table 3.1.2: NO2 Absorbances and Concentrations Obtained
Date Time Absorbance NO2 Concentration (?g/cm3) NO2 Concentration (?g/m3)
24/04/2018 9.28 am – 11.28 am 0.035 0.062 5.17
24/04/2018 2.21 pm – 4.21 pm 0.060 0.115 9.58
25/04/2018 9.36 am – 11.36 am 0.073 0.142 11.83
25/04/2018 2.14 pm – 4.14 pm 0.092 0.183 15.25
26/04/2018 9.17 am – 11.17 am 0.085 0.168 14.0
26/04/2018 2.08 pm – 4.08 pm 0.041 0.0745 6.21
02/05/2018 9.17 am – 11.17 am 0.077 0.151 12.58
02/05/2018 2.08 pm – 4.08 pm 0.063 0.121 10.08
03/05/2018 9.17 am – 11.17 am 0.061 0.117 9.75
03/05/2018 2.08 pm – 4.08 pm 0.045 0.083 6.92
The calculations shown below demonstrate how to convert the NO2 concentration from ?g/cm3 to ?g/m3, ?g/m3 to ppm and ppm to ppb. This was done because the concentration of the AQI values for the same time period as the sampling method was in ppb. Therefore the collected sample values were converted into ppb for easier comparison between the two methods.

Sample Calculation:
NO2 Concentration ?g/m3= NO2 Concentration ?g/cm3×Sampling Solution Volume cm3Volume of Air Sampled m3
NO2 Concentration ?g/m3= 0.062 ?g/cm3×10 cm30.12 m3
NO2 Concentration ?g/m3= 5.17
5.17 ?g/m3 5.32 10-4 = 2.75 10-3 ppm
2.75 10-3 ppm 1000 = 2.75 ppb
The following table contains the sampled and AQI NO2 concentrations for each sampling period, along with the sunshine hours for the corresponding days. The two concentrations are then both plotted in Figure 3.1.2 for comparison.
Table 3.1.3: NO2 Concentrations and Corresponding Sunshine Hours
Date Time Sampled NO2 Concentration (ppb) AQI NO2 Concentration (ppb)ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “URL” : “http://aqicn.org/city/ireland/rathmines/”, “accessed” : { “date-parts” : “2018”, “5”, “22” }, “id” : “ITEM-1”, “issued” : { “date-parts” : “0” }, “title” : “Rathmines, Ireland Air Pollution: Real-time PM2.5 Air Quality Index (AQI)”, “type” : “webpage” }, “uris” : “http://www.mendeley.com/documents/?uuid=f1fedb70-52ec-39b3-aa34-c6c4c21db718” } , “mendeley” : { “formattedCitation” : “<sup>31</sup>”, “plainTextFormattedCitation” : “31”, “previouslyFormattedCitation” : “<sup>31</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }31 Sunshine (hours)
24/04/2018 9.28 am – 11.28 am 2.75 4.5 0.9
24/04/2018 2.21 pm – 4.21 pm 5.10 7.0 0.9
25/04/2018 9.36 am – 11.36 am 6.30 14.5 1.1
25/04/2018 2.14 pm – 4.14 pm 8.11 15.0 1.1
26/04/2018 9.17 am – 11.17 am 7.45 13.0 7.1
26/04/2018 2.08 pm – 4.08 pm 3.30 7.0 7.1
02/05/2018 9.17 am – 11.17 am 6.70 8.5 7.9
02/05/2018 2.08 pm – 4.08 pm 5.40 8.5 7.9
03/05/2018 9.17 am – 11.17 am 5.20 13.0 0.1
03/05/2018 2.08 pm – 4.08 pm 3.70 13.0 0.1

Figure 3.1.2: NO2 Concentrations Comparison Plot
Above is a plot of the concentrations of NO2 in pbb in the samples taken and the concentration from the AQI website during the same sampling period. As seen above in Figure 3.1.2 the NO2 concentrations obtained from the chemical sampling method are lower than the Air Quality Index values from the same time period, but for the most part, they seem to follow the same pattern, with certain days having higher or lower values than others. Any discrepancies between the sampling values obtained and the AQI values is most likely due to the fact that different methods were being used, and also that the sampling was carried out in two different sites, Kevin St and Rathmines. From Table 3.1.3 it can be seen that the days that have the longer sunshine hours seem due have lower NO2 concentrations. This is most likely due to the fact that NO2 and O2 are converted into O3 and NO in the presence of sunlight. Any discrepancies between the sampling values obtained and the AQI values is most likely due to the fact that different methods were being used, and also that the sampling was carried out in two different sites, Kevin St and Rathmines.

Figure 3.1.3: NO2 Concentrations R 2 Plot
Above the concentrations of NO2 sampled and the concentrations from the AQI website have been plotted against each other so that an R2 value can be obtained. As seen above in Figure 3.1.3, when the sampled and AQI NO2 concentrations are plotted against each other, and R2 value of 0.4078 is obtained. This is relatively accurate when it is taken into account that only 10 samples were taken overall.

3.2 Chemical Methods: Ozone
In this section, the results from the ozone chemical sampling method will be shown and discussed. Shown below in Table 3.2.1, is the calibration curve data for ozone, and the calibration curve is shown in Figure 3.2.1. A calibration curve is used as a method for determining the concentration of ozone in the absorbing solution which has been left to absorb ambient air for two hours, by comparing the absorbance of the absorbing solution to the set of standard samples of known concentration and absorbance.

Table 3.2.1: Ozone Calibration Curve Table
x-axis Run 1 Run 2 Run 3 y-axis
Ozone Concentration (mol/10cm3) Absorbance Absorbance Absorbance Average Absorbance
1.0 10-8 0.018 0.014 0.014 0.015
2.0 10-8 0.037 0.042 0.037 0.039
3.0 10-8 0.058 0.056 0.053 0.056
4.0 10-8 0.073 0.083 0.071 0.076
5.0 10-8 0.089 0.108 0.097 0.098
1.0 10-8 0.220 0.206 0.212 0.213

Figure 3.2.1: Ozone Calibration Curve
Shown below is a sample of the calculations used to find the ozone concentration in the absorbing solution in mol/10cm3 by using the above calibration curve. This is then converted to ?g/cm3 so that it can be converted to ppb later on.

Sample Calculation:
y = 2 106 x – 0.0085
y = 0.156
0.156 = 2 106 x – 0.0085
0.156 + 0.0085 = 2 106 x
0.1645 = 2 106 x
x = 0.1645/2 106
x = 8.2 10-8 mol/10cm3
8.2 10-8 mol/10cm3 100 = 8.2 10-6 mol/L
8.2 10-6 mol/L 48 g/mol = 3.94 10-4 g/L
3.94 10-4 g/L 1000 = 0.394 mg/L
0.394 mg/L 1000 = 394.0 ?g/L
394.0 ?g/L 1000 = 0.394 ?g/cm3
Table 3.2.2: Ozone Absorbances and Concentrations Obtained
Date Time Absorbance Ozone Concentration (mol/10cm3) Ozone Concentration (?g/cm3) Ozone Concentration (?g/m3)
17/04/2018 9.44 am – 11.44 am 0.156 8.2 10-8 0.394 32.83
17/04/2018 2.10 pm – 4.10 pm 0.132 7.0 10-8 0.336 28.0
18/04/2018 9.13 am – 11.13 am 0.068 3.8 10-8 0.1836 15.3
18/04/2018 2.10 pm – 4.10 pm 0.143 7.5 10-8 0.3636 30.3
19/04/2018 9.17 am – 11.17 am 0.178 9.3 10-8 0.448 37.3
19/04/2018 2.15 pm – 4.15 pm 0.147 7.7 10-8 0.373 31.1
30/04/2018 9.43 am – 11.43 am 0.148 7.8 10-8 0.3756 31.3
30/04/2018 2.15 pm – 4.15 pm 0.189 9.8 10-8 0.474 39.5
01/05/2018 9.19 am – 11.19 am 0.200 10.4 10-8 0.5004 41.7
01/05/2018 2.15 pm – 4.15 pm 0.1680 8.8 10-8 0.4236 35.3
The calculations shown below demonstrate how to convert the ozone concentration from ?g/cm3 to ?g/m3, ?g/m3 to ppm and ppm to ppb. This was done because the concentration of the AQI values for the same time period as the sampling method was in ppb. Therefore the collected sample values were converted into ppb for easier comparison between the two methods.

Sample Calculation:
Ozone Concentration ?g/m3= Ozone Concentration ?g/cm3×Sampling Solution Volume cm3Volume of Air Sampled m3
Ozone Concentration ?g/m3= 0.394 ?g/cm3×10 cm30.12 m3
Ozone Concentration ?g/m3= 32.83
32.83 ?g/m3 5.10 10-4 = 0.0167 ppm
0.0167 ppm 1000 = 16.7 ppb
The following table contains the sampled and AQI ozone concentrations for each sampling period, along with the sunshine hours for the corresponding days. The two concentrations are then both plotted in Figure 3.2.2 for comparison.
Table 3.2.3: Ozone Concentrations and Corresponding Sunshine Hours
Date Time Sampled Ozone Concentration (ppb) AQI Ozone Concentration (ppb)ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “URL” : “http://aqicn.org/city/ireland/rathmines/”, “accessed” : { “date-parts” : “2018”, “5”, “22” }, “id” : “ITEM-1”, “issued” : { “date-parts” : “0” }, “title” : “Rathmines, Ireland Air Pollution: Real-time PM2.5 Air Quality Index (AQI)”, “type” : “webpage” }, “uris” : “http://www.mendeley.com/documents/?uuid=f1fedb70-52ec-39b3-aa34-c6c4c21db718” } , “mendeley” : { “formattedCitation” : “<sup>31</sup>”, “plainTextFormattedCitation” : “31” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }31 Sunshine (hours)
17/04/2018 9.28 am – 11.28 am 16.7 24.5 6.8
17/04/2018 2.21 pm – 4.21 pm 14.3 24.5 6.8
18/04/2018 9.36 am – 11.36 am 7.8 18.0 7.3
18/04/2018 2.14 pm – 4.14 pm 15.45 19.0 7.3
19/04/2018 9.17 am – 11.17 am 19.0 34.0 10.6
19/04/2018 2.08 pm – 4.08 pm 16.0 16.0 10.6
30/04/2018 9.17 am – 11.17 am 16.0 36.5 11.5
30/04/2018 2.08 pm – 4.08 pm 20.1 30.5 11.5
01/05/2018 9.17 am – 11.17 am 21.3 27.0 0.7
01/05/2018 2.08 pm – 4.08 pm 18.0 22.0 0.7

Figure 3.2.2: Ozone Concentrations Comparison Plot
Above is a plot of the concentrations of ozone in ppb in the samples taken and the concentration from the AQI website during the same sampling period. As seen above in Figure 3.2.2, the ozone2 concentrations obtained from the chemical sampling method are lower than the Air Quality Index values from the same time period, but for the most part, they seem to follow the same pattern, with certain days having higher or lower values than others. Any discrepancies between the sampling values obtained and the AQI values is most likely due to the fact that different methods were being used, and also that the sampling was carried out in two different sites, Kevin St and Rathmines. From Table 3.2.3 it can be seen that the days that have the longer sunshine hours seem due have higher ozone concentrations. This is most likely due to the fact that NO2 and O2 are converted into O3 and NO in the presence of sunlight. The only day that does not follow this trend is the 01/05/2018, because according to Met Eireann there were only 0.7 hours of sunshine. This must be incorrect as this was a very sunny day, hence the larger concentrations of ozone obtained from the sampling.

Figure 3.2.3: Ozone Concentrations R 2 Plot
Above the concentrations of ozone sampled and the concentrations from the AQI website have been plotted against each other so that an R2 value can be obtained. As seen above in Figure 3.2.3, when the sampled and AQI ozone concentrations are plotted against each other, and R2 value of 0.2446 is obtained. This is not very accurate even when it is taken into account that only 10 samples were taken overall.

3.3 Data Modelling: Oxides of Nitrogen in Dublin
In this section, the models and data used for those models will be discussed mainly for Dublin NO2, although the graphs and plots produced using the same models for Dublin NOX and Dublin NO are in Appendix 1 and Appendix 2 respectively.

Below is a table of the R2 values for Dublin’s NO2, NOX and NO for the multivariate linear regression model and the MARS model.

Table 3.3.1: Comparison of R2 values obtained from different models
Pollutant R2 – Multivariate Linear Regression Model R2 – MARS Model
NO2 0.6963 0.8321
NOX 0.5302 0.84
NO 0.3854 0.8614
As seen above in Table 3.3.1, these new adjusted MARS (multivariate adaptive regression splines) model are much more robust. This adjusted model was made by using a code that selectively works with the normalised slope of the previous values creating and forecasting separate sub-groups in all of the calibration and validation dataset, and chooses the best intersection between significant variables, which improves the model. Using the MARS models has greatly improved the R2 values obtained for the models, even bringing the value for NO from 0.3854 to 0.8614.

Figure 3.3.1: Concentration vs Time Plot for Oxides of Nitrogen in Dublin
Above is a plot of the concentrations of NO2, NOX and NO at different times of the day. As can be seen from Figure 3.3.1, the concentrations of NOX, NO and NO2 peak at around 10 am and again at around 8 pm. These times are both just after rush hour, which leads to the assumption that there is a lot of NOX, NO and NO2 produced during rush hour which peaks just after rush hour and then decreases. This is most likely due to the fuel combustion from diesel engines.

Figure 3.3.2: Oxides of Nitrogen in Dublin – Concentration vs Day Plot for 2016
Above is a plot of the concentrations of NO2, NOX and NO every day for the year 2016. As can be seen from Figure 3.3.2, the concentrations of NOX, NO and NO2 are highest at the beginning of the year and the end of the year. This corresponds with the Winter season, as can be seen in Figure 3.3.3. Then the lowest concentrations are in the middle of the year which corresponds with the Summer season, which can also be seen in Figure 3.3.3. This seasonal trend is most likely due to the fact that more fossil fuels are burned during the winter due to the lower temperature. Also, more people would choose to drive in the winter rather than walk or cycle, which could also lead to an increase in the production of NOX, NO and NO2.

Figure 3.3.3: Oxides of Nitrogen in Dublin – Concentration vs Season Plot in Dublin
Above is a plot of the concentrations of NO2, NOX and NO during each of the seasons. As can be seen from Figure 3.3.3, the concentrations of NOX, NO and NO2 are highest during the Winter season. This corresponds with Figure 3.3.3. Then the lowest concentrations are in summer which also corresponds with Figure 3.3.3. As stated above, this seasonal trend is most likely due to the fact that more fossil fuels are burned during the winter due to the lower temperature. Also, more people would choose to drive in the winter rather than walk or cycle, which could also lead to an increase in the production of NOX, NO and NO2.

Shown below are the correlation plot and table for NOX, NO2 and NO, along with the corresponding weather parameters. A correlation plot shows how much one variable is affected by another. A correlation coefficient measures the extent to which the variables affect each other. The coefficient describes both the strength and the direction of the relationship. Anything with a correlation coefficient between 0.70 and 1.0, and -0.70 and -1.0, is said to have a strong correlation. In this project, the spearman correlation method was used. The Spearman correlation evaluates the monotonic relationship between two continuous or ordinal variables. In a monotonic relationship, the variables tend to change together, but not necessarily at a constant rate. The Spearman correlation coefficient is based on the ranked values for each variable rather than the raw data. ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “URL” : “http://support.minitab.com/en-us/minitab-express/1/help-and-how-to/modeling-statistics/regression/supporting-topics/basics/a-comparison-of-the-pearson-and-spearman-correlation-methods/”, “accessed” : { “date-parts” : “2018”, “5”, “22” }, “author” : { “dropping-particle” : “”, “family” : “Minitab”, “given” : “”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “id” : “ITEM-1”, “issued” : { “date-parts” : “2017” }, “title” : “A comparison of the Pearson and Spearman correlation methods”, “type” : “webpage” }, “uris” : “http://www.mendeley.com/documents/?uuid=e6d37a6a-8136-3249-8d53-a2f8b82b49d4” } , “mendeley” : { “formattedCitation” : “<sup>32</sup>”, “plainTextFormattedCitation” : “32”, “previouslyFormattedCitation” : “<sup>32</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }32 The other correlation option is the Pearson correlation, this correlation evaluates the linear relationship between two continuous variables. A relationship is linear when a change in one variable is associated with a proportional change in the other variable. ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “URL” : “http://support.minitab.com/en-us/minitab-express/1/help-and-how-to/modeling-statistics/regression/supporting-topics/basics/a-comparison-of-the-pearson-and-spearman-correlation-methods/”, “accessed” : { “date-parts” : “2018”, “5”, “22” }, “author” : { “dropping-particle” : “”, “family” : “Minitab”, “given” : “”, “non-dropping-particle” : “”, “parse-names” : false, “suffix” : “” } , “id” : “ITEM-1”, “issued” : { “date-parts” : “2017” }, “title” : “A comparison of the Pearson and Spearman correlation methods”, “type” : “webpage” }, “uris” : “http://www.mendeley.com/documents/?uuid=e6d37a6a-8136-3249-8d53-a2f8b82b49d4” } , “mendeley” : { “formattedCitation” : “<sup>32</sup>”, “plainTextFormattedCitation” : “32”, “previouslyFormattedCitation” : “<sup>32</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }32
Figure 3.3.4: Oxides of Nitrogen in Dublin – Correlation Plot
Table 3.3.2: Oxides of Nitrogen in Dublin – Correlation Table
NOX (?g/m3) NO (?g/m3) NO (?g/m3) Max_Temp Min_Temp Rain Pressure Wind_Speed Wind_Direction Sun_Duration G_Rad
NOX (?g/m3) 1.00 0.96 0.88 -0.44 -0.49 -0.03 0.14 -0.46 -0.21 -0.07 -0.33
NO (?g/m3) 0.96 1.00 0.72 -0.37 -0.41 -0.03 0.10 -0.35 -0.09 -0.09 -0.33
NO2 (?g/m3) 0.88 0.72 1.00 -0.48 -0.54 0.01 0.16 -0.57 -0.38 -0.04 -0.28
Max_Temp -0.44 -0.37 -0.48 1.00 0.86 -0.07 0.16 -0.11 0.01 0.19 0.59
Min_Temp -0.49 -0.41 -0.54 0.86 1.00 0.05 0.06 0.04 0.05 -0.10 0.30
Rain -0.03 -0.03 0.01 -0.07 0.05 1.00 -0.37 0.14 -0.02 -0.25 -0.22
Pressure 0.14 0.10 0.16 0.16 0.06 -0.37 1.00 -0.36 -0.16 0.13 0.12
Wind_Speed -0.46 -0.35 -0.57 -0.11 0.04 0.14 -0.36 1.00 0.29 -0.08 -0.16
Wind_Direction -0.21 -0.09 -0.38 0.01 0.05 -0.02 -0.16 0.29 1.00 -0.06 -0.07
Sun_Duration -0.07 -0.09 -0.04 0.19 -0.10 -0.25 0.13 -0.08 -0.06 1.00 0.74
G_Rad-0.33 -0.33 -0.28 0.59 0.30 -0.22 0.12 -0.16 -0.07 0.74 1.00
As seen in Figure 3.3.4 and Table 3.3.1, there is a relatively strong negative correlation between NOX and max Temperature, minimum temperature and wind Speed, as seen from the corresponding correlation coefficients of -0.44, -0.49 and -0.46. This means that as these variables increase, the concentration of NOX decreases. Global radiation (-0.33) and wind direction (-0.21) also have a negative correlation with NOX concentrations. Pressure then has quite a small positive correlation with NOX, at 0.14.This means that as pressure increases, so does NOX concentration.

There is also a relatively strong negative correlation between NO and max temperature and minimum temperature, as seen from the corresponding correlation coefficients of -0.41 and -0.37. This means that as these variables increase, the concentration of NO decreases. Wind speed (-0.35) and global radiation (-0.33) also have a negative correlation with NO concentrations. Pressure also has quite a small positive correlation with NO, at 0.10.This means that as pressure increases, so does NO concentration.

There is a relatively strong negative correlation between NO2 and max temperature, minimum temperature and wind speed, as seen from the corresponding correlation coefficients of -0.48, -0.54 and -0.57. This means that as these variables increase, the concentration of NO2 decreases. Global radiation (-0.28) and wind direction (-0.38) also have a negative correlation with NO2 concentrations. Pressure then has quite a small positive correlation with NO2, at 0.16.This means that as pressure increases, so does NO2 concentration.

Because none of the correlation coefficients between NOX, NO2, NO and the weather variables is above 0.70 or below -0.70, there is not a particularly strong correlation between any of the concentrations with the weather variables.

3.3.1 Dublin NO2
Figure 3.3.1.1: Dublin NO2 Multivariate Regression Model
Above is a multivariate regression model with the predicted values on the y axis and the observed values on the x axis. As seen in Figure 3.3.1.1, the model shown is quite a robust model, although there are some minor outliers. The R2 value for this model is 0.6963. This means that there is quite a strong correlation between the model’s predictions and the observed results. The data points appear to fit the line of best fit relatively well. This is the most accurate model from the original multivariate models for NOX, NO2 and NO.

Figure 3.3.1.2: Dublin NO2 Logarithmic Regression Model – Smooth and Confidence Intervals
Above is a semi-logarithmic regression model with log of the predicted values on the y axis and the observed values on the x axis. As seen in Figure 3.3.1.2, a smoothing approach was taken using logarithmic regression to help the model fit the data points more effectively. A confidence interval was also added to the model, and it can be seen that this interval is narrower where the model fits better, and wider where there are outliers in the model. A confidence interval is used to express the degree of uncertainty associated with a sample statistic. It is an interval estimate combined with a probability statement. An interval estimate is defined by two numbers, between which a population parameter is said to lie. Confidence intervals are preferred to interval estimates, because only confidence intervals indicate both the precision and the uncertainty of the estimate.ADDIN CSL_CITATION { “citationItems” : { “id” : “ITEM-1”, “itemData” : { “URL” : “http://stattrek.com/statistics/dictionary.aspx?definition=confidence_interval”, “accessed” : { “date-parts” : “2018”, “5”, “22” }, “container-title” : “Stat Trek”, “id” : “ITEM-1”, “issued” : { “date-parts” : “2018” }, “title” : “Confidence Interval: Definition”, “type” : “webpage” }, “uris” : “http://www.mendeley.com/documents/?uuid=1d4631ef-6769-3213-8ca5-34c818d1f50e” } , “mendeley” : { “formattedCitation” : “<sup>33</sup>”, “plainTextFormattedCitation” : “33”, “previouslyFormattedCitation” : “<sup>33</sup>” }, “properties” : { “noteIndex” : 0 }, “schema” : “https://github.com/citation-style-language/schema/raw/master/csl-citation.json” }33
Figure 3.3.1.3: Dublin NO2 – Concentration vs Time Plot
Above is a plot of the observed and predicted concentrations over time, one year to be specific. As seen in Figure 3.3.1.3, the observed and predicted values match each other quite well, given that it has an R2 of 0.6963 which is strong. The observed values have a few outliers but these outliers are not too far outside of the normal concentrations. The NO2 concentrations seem to range from around 5 – 40 ?g/m3, with some outliers going as far as around 60 ?g/m3.

Figure 3.3.1.4: Dublin NO2 – Boxplots
Above are boxplots to help to show the outliers for the observed and predicted models. As seen in Figure 3.3.1.4, the scale for the observed boxplot is from 10 – 60 ?g/m3 and the scale for the predicted boxplot is 0 – 40 ?g/m3 The median for the observed NO2 concentrations is around 17 ?g/m3, with a range of 2 – 52 ?g/m3. There are only four outliers present above 52 ?g/m3. The median for the predicted NO2 concentrations is around 20 ?g/m3, with a range of -6 – 43 ?g/m3. There are no outliers present. The predicted values give an ideal boxplot appearance.

Figure 3.3.1.5: Dublin NO2 – Histograms
Above are histograms to show whether or not the observed and predicted models have a normal distribution. As seen in Figure 3.3.1.5, the observed NO2 concentrations do not appear to have a normal distribution, they do not fit the expected bell curve shape. This is confirmed by the fact that the observed data has a p value of 4.12 10-12, which is less than 0.05. Therefore the assumption of normality is rejected for the observed NO2 concentrations. The predicted NO2 concentrations appear to have a relatively normal distribution, they fit the expected bell curve shape. This is confirmed by the fact that the predicted data has a p value of 0.085, which is more than 0.05. Therefore the assumption of normality can be accepted for the predicted NO2 concentrations. Because the observed data has a non-normal distribution, the Spearman test was used when calculating the correlation.

Figure 3.3.1.6: Dublin NO2 – Qqnorm Plots
Above are quantile-quantile plots to show whether or not the observed and predicted models have a normal distribution. As seen in Figure 3.3.1.6, the observed NO2 concentrations do not appear to have a normal distribution, because the line formed is not straight. This is confirmed by the fact that the observed data has a p value of 4.12 10-12, which is less than 0.05. Therefore the assumption of normality is rejected for the observed NO2 concentrations. The predicted NO2 concentrations appear to have a relatively normal distribution, the line formed is quite straight. This is confirmed by the fact that the predicted data has a p value of 0.085, which is more than 0.05. Therefore the assumption of normality can be accepted for the predicted NO2 concentrations. Because the observed data has a non-normal distribution, the Spearman test was used when calculating the correlation.

Figure 3.3.1.7: Dublin NO2 MARS Model – Variable Importance Estimates
The above images are variable importance estimates, they help to determine which of the variables have a strong correlation with the NO2 concentration. As seen in Figure 3.3.1.7, the variables that have a large importance in terms of the NO2 concentrations when producing the model are maximum temperature, minimum temperature, wind speed and wind direction. The variables that have a strong correlation with each other are maximum temperature and minimum temperature, maximum temperature and sun duration, maximum temperature and global radiation, and, wind speed and global radiation.

Figure 3.3.1.8: Dublin NO2 MARS Model
Above is a MARS model with the predicted values on the y axis and the observed values on the x axis. As seen in Figure 3.3.1.8, this new adjusted MARS (multivariate adaptive regression splines) model is much more robust. This adjusted model was made by using a code that selectively works with the normalised slope of the previous values creating and forecasting separate sub-groups in all of the calibration and validation dataset, and chooses the best intersection between significant variables, which improves the model. The R2 value for this model is 0.8321, this is much more robust than the previous model. This can be seen by the fact that the data points fit the line of best fit quite well, with practically no outliers. Therefore, there is little variance, which is explained by the model because there is a strong correlation between the predicted and observed results.

3.4 Data Modelling: Ozone
In this section, the models and data used for those models will be discussed mainly for Dublin ozone, although the graphs and plots produced using the same models for Cork ozone, Galway ozone and Mayo ozone are in Appendix 3, Appendix 4 and Appendix 5 respectively.

Below is a table of the R2 values for ozone in Cork, Dublin, Galway and Mayo for the multivariate linear regression model and the MARS model.

Table 3.4.1: Comparison of R2 values obtained from different models
Location R2 – Multivariate Linear Regression Model R2 – MARS Model
Cork 0.3469 0.7182
Dublin 0.5603 0.6507
Galway 0.2633 0.7190
Mayo 0.3827 0.7187
As seen above in Table 3.4.1, these new adjusted MARS (multivariate adaptive regression splines) model are much more robust. Using the MARS models has greatly improved the R2 values obtained for the models, even bringing the value for Galway ozone from 0.2633 to 0.7910.

3.4.1 Dublin

Figure 3.4.1.1: Dublin Ozone – Concentration vs Time Plot
Above is a plot of the concentrations of ozone at different times of the day. As seen in Figure 3.4.1.1, the concentration of ozone peaks during the day time. This is most likely due to the fact that NO2 and O2 react in the presence of sunlight to form O3 and NO. Therefore, more ozone would be present during the daylight hours.

Figure 3.4.1.2: Dublin Ozone – Concentration vs Season Plot
Above is a plot of the concentrations of ozone during each of the seasons. As can be seen from Figure 3.4.1.2, the concentration of ozone is highest during the Spring/Summer season. This seasonal trend is most likely due to the fact that NO2 and O2 react in the presence of sunlight to form O3 and NO. Therefore, more ozone would be present during the seasons that tend to have more sunlight.

Figure 3.4.1.3: Dublin Ozone Multivariate Regression Model
Above is a multivariate linear regression model with the predicted values on the y axis and the observed values on the x axis. As seen in Figure 3.4.1.3, the model shown is quite a strong model, even though there are some outliers. The R2 value for this model is 0.5603. This means that there is a strong correlation between the model’s predictions and the observed results. This is the most accurate multivariate regression ozone model out of Cork, Dublin, Galway and Mayo.

Figure 3.4.1.4: Dublin Ozone Logarithmic Regression Model – Smooth and Confidence Intervals
Above is a semi-logarithmic regression model with the log of the predicted values on the y axis and the observed values on the x axis. As seen in Figure 3.4.1.4, a smoothing approach was taken using logarithmic regression to help the model fit the data points more effectively. A confidence interval was also added to the model, and it can be seen that this interval is narrower where the model fits better, and wider where there are outliers in the model. Although the original model was relatively strong, and therefore even the wider parts of the confidence interval are quite small.

Figure 3.4.1.5: Dublin Ozone – Concentration vs Time Plot
Above is a plot of the ozone concentration over time. As seen in Figure 3.4.1.5, the observed and predicted values match each other quite well, given that it has an R2 value of 0.5603. The observed values have a few outliers but these outliers are not too far below the normal concentrations. The ozone concentrations seem to range from around 15 – 80 ?g/m3, with some outliers towards the end the year going as far as around 5 ?g/m3.

Shown below are the correlation plot and table for Dublin ozone, along with the corresponding weather parameters. A correlation plot shows how much one variable is affected by another. A correlation coefficient measures the extent to which the variables affect each other. The coefficient describes both the strength and the direction of the relationship. Anything with a correlation coefficient between 0.70 and 1.0, and -0.70 and -1.0, is said to have a strong correlation.
Figure 3.4.1.6: Dublin Ozone – Correlation Plot
Table 3.4.1.1: Dublin Ozone – Correlation Table
Ozone Max_Temp Min_Temp Rain Pressure Wind_Speed Wind_Direction Sun_Duration G_
Rad
Ozone 1.00 -0.11 -0.02 0.01 -0.31 0.69 0.19 0.08 0.07
Max_Temp-0.11 1.00 0.79 -0.07 0.22 -0.21 -0.09 0.25 0.61
Min_Temp-0.02 0.79 1.00 0.07 0.03 0.03 -0.03 -0.12 0.26
Rain 0.01 -0.07 0.07 1.00 -0.32 0.16 -0.02 -0.27 -0.24
Pressure -0.31 0.22 0.03 -0.32 1.00 -0.44 -0.13 0.21 0.26
Wind_Speed0.69 -0.21 0.03 0.16 -0.44 1.00 0.29 -0.15 -0.24
Wind_Direction0.19 -0.09 -0.03 -0.02 -0.13 0.29 1.00 -0.07 -0.11
Sun_Duration0.08 0.25 -0.12 -0.27 0.21 -0.15 -0.07 1.00 0.77
G_Rad0.07 0.61 0.26 -0.24 0.26 -0.24 -0.11 0.77 1.00
As seen in Figure 3.4.1.6 and Table 3.4.1.1, there is a relatively strong positive correlation between ozone and wind speed, as seen from the corresponding correlation coefficient of 0.69. This means that as wind speed increases, the concentration of ozone also increases. wind direction (0.19) also has a positive correlation with ozone concentration. Pressure then has a negative correlation with ozone, at -0.31.This means that as pressure increases, ozone concentration decreases. Sun duration would be expected to have a very strong positive correlation with ozone but that does not seem to be the case. Because none of the correlation coefficients between ozone and the weather variables is above 0.70 or below -0.70, there is not a particularly strong correlation between ozone concentration and the weather variables.

Figure 3.4.1.7: Dublin Ozone – Boxplots
Above are boxplots which help to determine the outliers for the observe and predicted models. As seen in Figure 3.4.1.7, the scale for the observed boxplot is from 0 – 80 ?g/m3 and the scale for the predicted boxplot is 10 – 80 ?g/m3. The median for the observed ozone concentrations is around 40 ?g/m3, with a range of 1 – 80 ?g/m3 There are no outliers present. The observed values give an ideal boxplot appearance. The median for the predicted ozone concentrations is around 39 ?g/m3, with a range of 12 – 70 ?g/m3. There are outliers present above 70 ?g/m3.
Figure 3.4.1.8: Dublin Ozone – Histograms
Above are histograms which help to determine whether or not the observed and predicted models have a normal distribution. As seen in Figure 3.4.1.8, the observed ozone concentrations appear to have a normal distribution, they fit the expected bell curve shape. This is then contradicted by the fact that the observed data has a p value of 0.0091, which is less than 0.05. Therefore the assumption of normality is rejected for the observed ozone concentrations. The predicted ozone concentrations also appear to have quite a normal distribution, they seem to fit the expected bell curve shape. This is then also contradicted by the fact that the predicted data has a p value of 1.58 10-7, which is less than 0.05. Therefore the assumption of normality is also rejected for the predicted ozone concentrations. Because the observed and predicted data have a non-normal distribution, the Spearman test was used when calculating the correlation.

Figure 3.4.1.9: Dublin Ozone – Qqnorm Plots
Above are quantile-quantile plots which help to determine whether or not the observed and predicted models have a normal distribution. As seen in Figure 3.4.1.9, the observed ozone concentrations appear to have a normal distribution, the line formed is quite straight. This is then contradicted by the fact that the observed data has a p value of 0.0091, which is less than 0.05. Therefore the assumption of normality is rejected for the observed ozone concentrations. The predicted ozone concentrations also appear to have quite a normal distribution, the line formed is less straight than that of the observed concentrations. This is then also contradicted by the fact that the predicted data has a p value of 1.58 10-7, which is less than 0.05. Therefore the assumption of normality is also rejected for the predicted ozone concentrations. Because the observed and predicted data have a non-normal distribution, the Spearman test was used when calculating the correlation.

Figure 3.4.1.10: Dublin Ozone MARS Model – Variable Importance Estimates
The above images are variable importance estimates, they help to determine which of the variables have a strong correlation with the ozone concentration. As seen in Figure 3.4.1.10, the variables that have a large importance in terms of the ozone concentrations when producing the model are maximum temperature, wind speed and wind direction. The variables that have a strong correlation with each other are minimum temperature and wind speed, minimum temperature and wind direction, wind speed and sun duration, wind speed and global radiation, and, wind direction and global radiation.

Figure 3.4.1.11: Dublin Ozone MARS Model
Above is a MARS model with the predicted values on the y axis and the observed values on the x axis. As seen in Figure 3.4.1.11, this new adjusted MARS (multivariate adaptive regression splines) model is much more accurate than the previous model, although still not the most accurate model. This adjusted model works with the normalised slope of the previous values and selectively chooses the best intersection between significant variables. The R2 value for this model is 0.7182, this is much more accurate than the previous model. This can be seen by the fact that the data points fit the line of best fit quite well, with only a few outliers. Therefore, there is little variance in the model because there is a strong correlation between the predicted and observed results.
3.5 Comparison of Sampled NO2 and NO2 Model

Figure 3.5.1: NO2 Chemical Sampling Method Concentration Plot

Figure 3.5.2: Dublin NO2 MARS Model
As seen from Figure 3.5.1, the NO2 concentrations obtained from the chemical sampling method range from roughly 5 – 15 ?g/m3. Then in Figure 3.5.2, it can be seen that the NO2 concentrations in the model range from around 5 – 55 ?g/m3, with the largest concentration range being from around 10 – 25 ?g/m3. This means that the values obtained from the sampling are slightly less than those predicted in the NO2 model. This is similar to the fact that NO2 values obtained from the chemical sampling method being lower than those obtained from the Air Quality Index. Possibly due to the absorbing solution not absorbing as much NO2 as it should have.

Figure 3.1.3: NO2 Concentrations R 2 Plot
When the R2 value obtained above (0.4078) is compared to that of the model in Figure 3.5.2 (0.8321), it is not nearly as accurate. This is most likely due to the fact that there were only 10 data points for the sampled NO2 and there were 366 taken for the NO2 model. Also, the model is a forecasting model, whereas the plot in Figure 3.1.3 is a plot of two different samples, with two different sampling methods and two different sampling locations.

3.6 Comparison of Sampled Ozone and Dublin Ozone Model

Figure 3.6.1: Ozone Chemical Sampling Method Concentration Plot

Figure 3.6.1: Dublin Ozone MARS Model
As seen from Figure 3.6.1, the ozone concentrations obtained from the chemical sampling method range from roughly 15 – 42 ?g/m3. Then in Figure 3.6.2, it can be seen that the ozone concentrations in the model range from around 10 – 70 ?g/m3, with the largest concentration range being from around 20 – 60 ?g/m3. This means that the values obtained from the sampling are slightly less than those predicted in the ozone model, but still fairly accurate. The ozone model appears to be more accurate than the NO2 model shown in 3.5, most likely due to the fact that there was three years of data used for ozone, but only one year of data was used for NO2. This is similar to the fact that ozone values obtained from the chemical sampling method being lower than those obtained from the Air Quality Index. Possibly due to the absorbing solution not absorbing as much ozone as it should have.

Figure 3.2.3: Ozone Concentrations R 2 Plot
When the R2 value obtained above (0.2446) is compared to that of the model in Figure 3.6.1 (0.7182), it is not nearly as accurate. This is most likely due to the fact that there were only 10 data points for the sampled ozone and there were 3 years of data taken for the ozone model. Also, the model is a forecasting model, whereas the plot in Figure 3.2.3 is a plot of two different samples, with two different sampling methods and two different sampling locations.

Conclusion

ADDIN Mendeley Bibliography CSL_BIBLIOGRAPHY 1European Environment Agency, Pollution — European Environment Agency, https://www.eea.europa.eu/themes/water/wise-help-centre/glossary-definitions/pollution, (accessed 21 May 2018).

2O. Hertel, J. Salmond, W. Bloss, I. Salma, S. Vardoulakis, R. Maynard, M. Williams, M. E. Goodsite, R. M. Harrison and R. E. Hester, Air Quality in Urban Environments, The Royal Society of Chemistry, 2009.

3T. Godish, Air quality, Lewis Publishers, 2004.

4K. Omersa, NOx emissions from diesel engines, http://clean-carbonenergy.com/nox-emissions.html, (accessed 7 April 2018).

5C. Schwerdt, Lund University, 2006.

6G. Hesselmann and M. Rivas, IFRF Online Combust. Handb., 2001, 1, 1–4.

7Nitrogen Oxides (NOx), Why and How They Are Controlled CORRECTION NOTICE Public Access and Information Transfer, 1999, vol. 800.

8A. G. Ortiz, Air quality in Europe — 2015 report, 2015.

9T. Jasarevic, G. Thomas and N. Osseiran, WHO | 7 million premature deaths annually linked to air pollution, World Health Organization, 2014.

10OECD, The Economic Consequences of Outdoor Air Pollution: Policy Highlights, 2016.

11D. M. O’ Dwyer, Air Quality in Ireland 2015 Key Indicators of Ambient Air Quality, 2015.

12G. Topham, S. Clarke, C. Levett, P. Scruton and M. Fidler, Guard., 2015.

13D. Stewart, in Monitoring Quick Guide, Scottish Environment Protection Agency, Scotland.

14E. P. A. (EPA), Evaluation of Triethanolamine: Procedure for Determination of Nitrogen Dioxide in Ambient Air, 1974.

15US EPA, Basic Information about Ozone, https://www.epa.gov/ozone-pollution/basic-information-about-ozone#regulations, (accessed 25 April 2018).

16US EPA, Ozone Pollution, https://www.epa.gov/ozone-pollution, (accessed 25 April 2018).

17WORLD BANK GROUP, Ground-Level Ozone, 1998.

18Dublin Institute of Technology, CHEM 4006 LAB EXPERIMENTS, .

19About – RStudio, https://www.rstudio.com/about/, (accessed 6 May 2018).

20E. P. A. (EPA), Environmental Indicators – Environmental Protection Agency (EPA, http://www.epa.ie/irelandsenvironment/environmentalindicators/, (accessed 21 May 2018).

21E. P. A. (EPA), Evaluation of 1 Percent Neutral Buffered Potassium Iodide Procedure for Calibration of Ozone Monitors, 1977.

22G. Piñeiro, S. Perelman, J. P. Guerschman and J. M. Paruelo, Ecol. Modell., 2008, 216, 316–322.

23Statwing, Interpreting residual plots to improve your regression | Statwing Documentation, .

24Statistics How TO, What is Correlation in Statistics? Correlation Analysis Explained, http://www.statisticshowto.com/what-is-correlation/, (accessed 10 May 2018).

25Understanding and interpreting box plots | [email protected], https://www.wellbeingatschool.org.nz/information-sheet/understanding-and-interpreting-box-plots, (accessed 10 May 2018).

26Box Plot: Display of Distribution, http://www.physics.csbsju.edu/stats/box2.html, (accessed 10 May 2018).

27R Bloggers, How to Make a Histogram with Basic R | R-bloggers, https://www.r-bloggers.com/how-to-make-a-histogram-with-basic-r/, (accessed 10 May 2018).

28The Data Visualisation Catalogue, Histogram – Learn about this chart and tools to create it, https://datavizcatalogue.com/methods/histogram.html, (accessed 10 May 2018).

29Anderson-Darling Normality Test, http://www.variation.com/da/help/hs140.htm, (accessed 11 May 2018).

30C. Ford, Understanding Q-Q Plots | University of Virginia Library Research Data Services + Sciences, http://data.library.virginia.edu/understanding-q-q-plots/, (accessed 10 May 2018).

31Rathmines, Ireland Air Pollution: Real-time PM2.5 Air Quality Index (AQI), http://aqicn.org/city/ireland/rathmines/, (accessed 22 May 2018).

32Minitab, A comparison of the Pearson and Spearman correlation methods, http://support.minitab.com/en-us/minitab-express/1/help-and-how-to/modeling-statistics/regression/supporting-topics/basics/a-comparison-of-the-pearson-and-spearman-correlation-methods/, (accessed 22 May 2018).

33Confidence Interval: Definition, http://stattrek.com/statistics/dictionary.aspx?definition=confidence_interval, (accessed 22 May 2018).

Appendix 1 – Dublin NOX
Figure 1.1: Dublin NOX Multivariate Regression Model
Figure 1.2: Dublin NOX Logarithmic Regression Model – Smooth and Confidence Intervals
Figure 1.3: Dublin NOX– Concentration vs Time Plot
Figure 1.4: Dublin NOX – Boxplots
Figure 1.5: Dublin NOX – Histograms
Figure 1.6: Dublin NOX – Qqnorm Plots

Figure 1.7: Dublin NOX MARS Model – Variable Importance Estimates

Figure 1.8: Dublin NOX MARS Model

Appendix 2 – Dublin NO
Figure 2.1: Dublin NO Multivariate Regression Model
Figure 2.2: Dublin NO Logarithmic Regression Model – Smooth and Confidence Intervals
Figure 2.3: Dublin NO – Concentration vs Time Plot
Figure 2.4: Dublin NO – Boxplots
Figure 2.5: Dublin NO – Histograms
Figure 2.6: Dublin NO – Qqnorm Plots

Figure 2.7: Dublin NO MARS Model – Variable Importance Estimates

Figure 2.8: Dublin NO MARS Model

Appendix 3 – Cork Ozone

Figure 3.1: Cork Ozone – Concentration vs Time Plot

Figure 3.2: Cork Ozone – Concentration vs Season Plot
Figure 3.3: Cork Ozone Multivariate Regression Model

Figure 3.4: Cork Ozone Logarithmic Regression Model – Smooth and Confidence Intervals

Figure 3.5: Cork Ozone – Concentration vs Time Plot
Figure 3.6: Cork Ozone – Correlation Plot
Table 3.1: Cork Ozone – Correlation Table
Ozone Max_Temp Min_Temp Rain Pressure Wind_Speed Wind_Direction Sun_Duration
Ozone 1.00 0.02 0.03 0.19 -0.23 0.56 -0.06 0.05
Max_Temp0.02 1.00 0.85 -0.08 0.11 -0.18 -0.04 0.28
Min_Temp0.03 0.85 1.00 0.00 0.05 -0.07 -0.16 -0.03
Rain 0.19 -0.08 0.00 1.00 -0.43 0.35 -0.13 -0.33
Pressure -0.23 0.11 0.05 -0.43 1.00 -0.37 -0.03 0.20
Wind_Speed0.56 -0.18 -0.07 0.35 -0.37 1.00 0.07 -0.19
Wind_Direction-0.06 -0.04 -0.16 -0.13 -0.03 0.07 1.00 0.13
Sun_Duration0.05 0.28 -0.03 -0.33 0.20 -0.19 0.13 1.00
Figure 3.7: Cork Ozone – Boxplots
Figure 3. 8: Cork Ozone – Histograms
Figure 3. 9: Cork Ozone – Qqnorm Plots

Figure 3. 10: Cork Ozone MARS Model – Variable Importance Estimates

Figure 3. 11: Cork Ozone MARS Model

Appendix 4 – Galway Ozone

Figure 4.1: Galway Ozone – Concentration vs Time Plot

Figure 4.2: Galway Ozone – Concentration vs Season Plot
Figure 4.3: Galway Ozone Multivariate Regression Model

Figure 4.4: Galway Ozone Logarithmic Regression Model – Smooth and Confidence Intervals
Figure 4.5: Galway Ozone – Concentration vs Time Plot
Figure 4.6: Galway Ozone – Correlation Plot
Table 4.1: Galway Ozone – Correlation Table
Ozone Max_Temp Min_Temp Rain Pressure Wind_Speed Wind_Direction Sun_Duration
Ozone 1.00 -0.36 -0.36 0.18 -0.17 0.32 0.26 0.07
Max_Temp-0.36 1.00 0.82 -0.17 0.24 -0.14 -0.03 0.31
Min_Temp-0.36 0.82 1.00 -0.03 0.07 0.07 0.04 -0.04
Rain 0.18 -0.17 -0.03 1.00 -0.43 0.41 0.01 -0.29
Pressure -0.17 0.24 0.07 -0.43 1.00 -0.44 -0.02 0.22
Wind_Speed0.32 -0.14 0.07 0.41 -0.44 1.00 0.15 -0.21
Wind_Direction0.26 -0.03 0.04 0.01 -0.02 0.15 1.00 0.13
Sun_Duration0.07 0.31 -0.04 -0.29 0.22 -0.21 0.13 1.00
Figure 4.7: Galway Ozone – Boxplots
Figure 4.8: Galway Ozone – Histograms
Figure 4.9: Galway Ozone – Qqnorm Plots

Figure.4.10: Galway Ozone MARS Model – Variable Importance Estimates

Figure 4.11: Galway Ozone MARS Model

Appendix 5 – Mayo Ozone

Figure 5.1: Mayo Ozone – Concentration vs Time Plot

Figure 5.2: Mayo Ozone – Concentration vs Season Plot
Figure 5.3: Mayo Ozone Multivariate Regression Model

Figure 5.4: Mayo Ozone Logarithmic Regression Model – Smooth and Confidence Intervals
Figure 5.5: Mayo Ozone – Concentration vs Time Plot
Figure 5.6: Mayo Ozone – Correlation Table
Table 5.1: Mayo Ozone – Correlation Table
Ozone Max_Temp Min_Temp Rain Pressure Wind_Speed Wind_Direction Sun_Duration
Ozone 1.00 -0.07 -0.12 0.26 -0.21 0.58 0.20 0.07
Max_Temp-0.07 1.00 0.87 -0.16 0.30 -0.20 -0.08 0.31
Min_Temp-0.12 0.87 1.00 -0.08 0.22 -0.14 -0.03 0.00
Rain 0.26 -0.16 -0.08 1.00 -0.46 0.46 0.04 -0.32
Pressure -0.21 0.30 0.22 -0.46 1.00 -0.41 -0.01 0.27
Wind_Speed0.58 -0.20 -0.14 0.46 -0.41 1.00 0.11 -0.25
Wind_Direction0.20 -0.08 -0.03 0.04 -0.01 0.11 1.00 0.02
Sun_Duration0.07 0.31 0.00 -0.32 0.27 -0.25 0.02 1.00
Figure 5.7: Mayo Ozone – Boxplots
Figure 5.8: Mayo Ozone – Histograms
Figure 5.9: Mayo Ozone – Qqnorm Plots

Figure 5.10: Mayo Ozone MARS Model – Variable Importance Estimates

Figure 5.11: Mayo Ozone MARS Model

Appendix 6 – Scripts
Below are examples of the four different scripts used in this project. The examples below are all for Dublin NO2, but the same scripts were used for Dublin NOX, Dublin NO, Cork ozone, Dublin ozone, Galway ozone and Mayo ozone. Only minor details were different, such as file names, column names and header names.

Table 6.1: Dublin NO2 Pollutants and Regression Script
setwd(‘C:/Data/Data2/NOx’)
#install (readxl) #you have to install this package
library(readxl)
Dublin<-read.csv(file=”Dublin_NOx.csv”, header = T) #to read by using csv
#Dublin <- read_excel(“C:/data/Dublin.xlsx”)
Dublin$Date<-as.Date(Dublin$Date,”%d/%m/%Y”,tz=”Europe/London”)
Dublin$Year<-as.factor(format(Dublin$Date,’%Y’))
head(Dublin)
str (Dublin)
Dublin<-Dublinorder(Dublin$Date),
data_NO2<-Dublin,1:13
summary(data_NO2)
porc<-80 #for extracting a percentage of data to calibrate
nfilas<-floor(nrow(data_NO2)*porc*0.01) #random sampling
samp<-sample(nrow(data_NO2),size=nfilas)
data.calibration<-data_NO2samp, #80% of data are used to adjust or calibrate the model
data.validation<-data_NO2-samp, #20% of data are used to check the robustness of the model
answer<-lm(NO2_ug.m3~Max_Temp+Min_Temp+Rain+Pressure+Wind_Speed+Wind_Direction+Sun_Duration+G_Rad,data=data_NO2)
summary(answer) #This is a first approach to the model, with all the parameters included
answer_refined<-lm(NO2_ug.m3~Min_Temp+Wind_Speed+Wind_Direction+Sun_Duration+G_Rad,data=data_NO2)
summary(answer_refined) #This is a final version of the model, with only the parameters which have significance
mod.pred<-predict(answer_refined,data.validation)
obs.pred<-cbind(data.validation$NO2_ug.m3,mod.pred)
colnames(obs.pred)<-c(“observed”,”predicted”)
head(obs.pred)
plot(obs.pred, main=”NO2 Dublin”)
abline(coef(lm(mod.pred~data.validation$NO2_ug.m3)),col=”blue”)
Table 6.2: Dublin NO2 Regression and Normality Script
setwd(‘C:/Data/Data2/NOx’)
#install (readxl) #you have to install this package
library(readxl)
Dublin<-read.csv(file=”Dublin_NOx.csv”, header = T) #to read by using csv
#Dublin <- read_excel(“C:/data/Dublin.xlsx”)
Dublin$Date<-as.Date(Dublin$Date,”%d/%m/%Y”,tz=”Europe/London”)
Dublin$Year<-as.factor(format(Dublin$Date,’%Y’))
head(Dublin)
str (Dublin)
Dublin<-Dublinorder(Dublin$Date),
data_NO2<-Dublin,1:13
colnames(data_NO2) <- c(“Date”,”NOx_ug.m3″,”NO_ug.m3″,”NO2_ug.m3″,”Max_Temp”,”Min_Temp”,”Rain”,”Pressure”, “Wind_Speed”,”Wind_Direction”,”Sun_Duration”,”G_Rad”)
View(data_NO2)
summary(data_NO2)
porc<-1 #for extracting a percentage of data to calibrate
nfilas<-floor(nrow(data_NO2)*porc*0.01) #random sampling
samp<-sample(nrow(data_NO2),size=nfilas)
data.calibration<-data_NO2samp, #80% of data are used to adjust or calibrate the model
data.validation<-data_NO2-samp, #20% of data are used to check the robustness of the model
answer<-lm(NO2_ug.m3~Max_Temp+Min_Temp+Rain+Pressure+Wind_Speed+Wind_Direction+Sun_Duration+G_Rad,data=data_NO2)
summary(answer)
mod.pred<-predict(answer,data.validation)
obs.pred<-cbind(data.validation$NO2,mod.pred)
colnames(obs.pred)<-c(“observed”,”predicted”)
head(obs.pred)
plot(obs.pred, main=”NO2″)
abline(coef(lm(mod.pred~data.validation$NO2)),col=”red”)
adjustement<-data.frame(obs.pred)
observed<-adjustement$observedpredicted<-adjustement$predicted#install.packages(“ggplot2″)
library(ggplot2) #logarithmmic transformation
#qplot(observed,predicted,data=adjustement,log=”x”,xlab=”Log-observed”,ylab=”predicted”,geom=c(“point”,”smooth”),main=”Log-observed vs predicted”)
#qplot(observed,predicted,data=adjustement,log=”y”,ylab=”Log-Predicted”,xlab=”observed”,geom=c(“point”,”smooth”),main=”observed vs Log-predicted”)
#qplot(observed,predicted,data=adjustement,log=”xy”,xlab=”Log-observed”,ylab=”Log-predicted”,geom=c(“point”,”smooth”),main=”Log-observed vs Log-predicted”)
#qplot(log(observed),log(predicted),data=adjustement,xlab=”Log-observed”,ylab=”Log-predicted”,geom=c(“point”,”smooth”),main=”Log-observed vs Log-predicted”)
qplot(observed,logpredicted,data=adjustement,geom=c(“point”,”smooth”),main=”NO2 observed vs logpredicted”)
#now, we are going to check the normality. We use the Liliefors test for it, which is an improvement in the Kolmogorov-smirnov test.

#install.packages(“nortest”)
library(nortest)
norm.predicted<-lillie.test(adjustement$predicted)
print(norm.predicted)
norm.observed<-lillie.test(adjustement$observed)
print(norm.observed)
#we represent the two lines along the time, observed values vs. predicted values.
plot(observed, type=”overplotted”,
pch=1, col=”blue”, xlab=”Time (days)”,
ylab=”NO2 concentration (microgrames/m3)”,
main=”Concentration vs Time”,
ylim=c(0,100),xlim =c(0,400))
lines(predicted,type=”overplotted”,pch=1,col=”red”)
legend(“topleft”,legend=c(“observed”,”predicted”),
pch=c(1,1),col=c(“blue”,”red”))
test <- cor.test(adjustement$observed,adjustement$predicted, method=”spearman”,exact=FALSE) #to calculate spearman test between observed and predicted
print(test)
png(‘NO2 hist.png’)
par(mfrow=c(1,2))
hist(adjustement$observed,breaks=15,main=”concentrations NO2 observed”,col=”blue”,border=”white”,xlabel=”concentrations NO2 observed”,ylim=c(0,90),cex.main=1)
hist(adjustement$predicted,breaks=15,main=”concentrations NO2 predicted”,col=”blue”,border=”white”,xlabel=”concentrations NO2 predicted”,ylim=c(0,90),cex.main=1)
dev.off()
png(‘NO2 boxplot.png’)
par(mfrow=c(1,2))
boxplot(adjustement$observed,main=”concentrations NO2 observed”,col=”red”,cex.main=1,ylab=”NO2 concentration (microgrames/m3)”)
boxplot(adjustement$predicted,main=”concentrations NO2 predicted”,col=”red”,cex.main=1,ylab=”NO2 concentration (microgrames/m3)”)
dev.off()
png(‘NO2 qqnorm.png’)
par(mfrow=c(1,2))
qqnorm(adjustement$observed,main=”concentrations NO2 observed”,col=”red”,cex.main=1)
qqline(adjustement$observed,col=”black”)
qqnorm(adjustement$predicted,main=”concentrations NO2 predicted”,col=”red”,cex.main=1)
qqline(adjustement$predicted,col=”black”)
dev.off()
Table 6.3: Dublin NO2 MARS Regression Script
setwd(‘C:/Data/Data2/NOx/’)
library(datasets)
library(boot)
library(readxl)
data <- read.csv(“C:/Data/Data2/NOx/NO21.csv”, header=T, sep=”,”)
data<-datacomplete.cases(data),
View(data)
#install.packages(“earth”)
library(earth)
modelNO2<-earth(data,-c(1:6),data$NO2,degree = 2,ncross = 3,nfold = 2,varmod.method = “lm”,keepxy = T)
summary(modelNO2)
modelNO2$rsq
evimp(modelNO2)
plotmo(modelNO2,pt.col = 1,level=0.90,ylim = c(-50,100))
porc<-70 #para extraer el porcentaje de calibraci?n que queramoSnfilas<-floor(nrow(data)*porc*0.01) #Muestreo aleatoriosamp<-sample(nrow(data),size=nfilas)
database.calibration<-datasamp, #se utilizar?a el 80% de los datos para calibrar o ajustar el modelodatabase.validation<-data-samp, #se utilizar?a el 20% de los datos para validar el modelorespuesta_lm<-earth(database.calibration,-c(1:2,12),database.calibration$NO2,degree = 2,ncross = 3,nfold = 2,varmod.method = “lm”,keepxy = T)
summary(respuesta_lm)
library(mgcv) #to check with other distribution model, gam (Generalized additive model)
respuesta_GAM<-earth(database.calibration,-c(1:2,14),database.calibration$NO2,degree = 2,ncross = 3,nfold = 2,varmod.method = “gam”,keepxy = T)
summary(respuesta_GAM)
respuesta_GAM$rsqmod.pred<-predict(respuesta_GAM,database.validation)
obs.pred<-cbind(database.validation$NO2,mod.pred)
colnames(obs.pred)<-c(“micrograms/m3 (observed)”,”micrograms/m3 (predicted)”)
head(obs.pred)
plot(obs.pred, main=”NO2 concentrations”)
abline(coef(lm(mod.pred~database.validation$NO2)),col=”red”)
Table 6.4: Dublin NO2 Correlation Script
setwd(‘C:/Data/Data2/NOx’)
Dublin_Corrplot<-read.csv(file=”Dublin.csv”, header = T) #to read by using csv
head(Dublin_Corrplot)
M<-cor(Dublin_Corrplot)
head(round(M,2))
library(corrplot)
corrplot(M,type=”upper”)
mc_data <- Dublin_Corrplot,2:length(Dublin_Corrplot)
round(cor(Dublin_Corrplot),2)