University of Ghana http://ugspace.ug.edu.gh Numerical Simulation of Dispersion of Emissions from Tema Oil Refinery in Ghana A dissertation presented to the: Department of NUCLEAR SCIENCES AND APPLICATIONS SCHOOL OF NUCLEAR AND ALLIED SCIENCES COLLEGE OF BASIC AND APPLIED SCIENCES UNIVERSITY OF GHANA by HANNAH ASAMOAH AFFUM (ID: 10174319) BSc (KNUST), 2002 MPhil (UG), 2006 In partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY in APPLIED NUCLEAR PHYSICS 2015 University of Ghana http://ugspace.ug.edu.gh Declaration This thesis is the result of research work undertaken by Hannah Asamaoh AFFUM in the Department of Nuclear Sciences and Applications, School of Nuclear and Allied Sciences, University of Ghana, under the supervision of Prof. Emeritus E.H.K. Akaho (SNAS, UG - Legon, Ghana), Prof. Joseph J.J. Niemela (ICTP, Trieste, Italy), Prof. Vincenzo Armenio (UT, Trieste, Italy) and Dr. K. A. Danso (SNAS, UG - Legon, Ghana). Sign: ........................................... Hannah Asamaoh Affum (Student) Sign: ................................................ Sign: ................................................ Prof. Emeritus E.H.K. Akaho Prof. J.J. Niemela (Supervisor) (Co-Supervisor) Sign:................................................. Sign: ................................................ Prof. Vincenzo Armenio Dr. K.A. Danso (Co-Supervisor) (Co-Supervisor) ii University of Ghana http://ugspace.ug.edu.gh Dedicated to my amazing children, Aseyenedi Esi Tamakloe, Woedem Dzidula Tamakloe and Alesineyram Aku Tamakloe. . . iii University of Ghana http://ugspace.ug.edu.gh Acknowledgements I would like to express my greatest appreciation to Almighty God without whom the subsequent happenings would not have been possible. This PhD project was funded by the International Atomic Energy Agency (IAEA) through the International Centre for Theoretical Physics (ICTP) Sandwich Training and Exchange Programme(STEP) in Trieste, Italy. It is, as such, befitting for these kind sponsors to be the next receipient of my heart-felt gratitude. The author would like to express profound thanks to her home principal su- pervisor, Prof. Emeritus E.H.K. Akaho and home co-supervisor, Dr. K.A. Danso both of School of Nuclear and Allied Sciences(SNAS), University of Ghana. Their constant encouragement and strong belief in my ability to go through this academic exercise is truly appreciated. For agreeing to su- pervise this thesis and ensuring that I had all that was needed to carry out this research work successfully, my next thanks go to Prof. J.J. Niemela and Prof. Vincenzo Armenio of the International Centre for Theoretical Physics and the University of Trieste respectively. The great exposure ac- quired through my association is deeply cherished. The intellectual support from my entire supervisory committee is very much appreciated. Further appreciation is extended to some technical staff of the Residual Fluid Catalytic Cracking Unit of the Tema Oil Refinery, Ghana, especially iv University of Ghana http://ugspace.ug.edu.gh the manager Engineer Kwaku Darko, Randy and Priscilla Antwi for will- ingly commiting to our mini-project of estimating refinery emissions in the absence of data from flow meters. I would want to specially thank all the doctoral students and staff of the fluids laboratory of the University of Trieste whom I came into contact with and exchanged ideas during my attachment to the laboratory. Ahmad Fakari and Anna Brunetti deserve special mention for their patience they displayed in coaching me through the various codes I used at various stages in this research work. For assisting me to build the Latex code for the write-up of this thesis and others, I also say a big thanks to Christian Nuviadenu and Ernest Asare. A special thanks also goes to Dr. Clement Onime, whom I affectionately call ’Uncle Clement’, Addisu Semie and Grazianno Giancula whose instrumen- tality is helping with the set-up and running the WRF model can simply not be overemphasized. I would also like to thank Jake Doku of the Remote Sensing laboratory of the University of Ghana for willingly providing elevation and base maps of my study area. Appreciation is also extended to my colleagues at the Nuclear Application Centre of the Ghana Atomic Energy Commission for their assistance. I specifically mention Simon Adzaklo, Alexander Coleman and Godred Appiah who took turns to pick up my children from school. v University of Ghana http://ugspace.ug.edu.gh Dorotea Calligaro and all staff of the ICTP who in one way or another helped make life comfortable for me while in ICTP, Trieste are also grate- fully acknowledged. Spiritual support from members of the ICTP fellowship and the Chiesa di Evangelica especially Ettore Panizon and family, Tonino Cuschito and family is very much appreciated. Last but not least, my sincere gratitude is to all of my family especially my kind and loving husband Emmanuel for all he sacrificed to care for our chil- dren, Aseye, Woedem and Alesi, whiles I was on fellowship in Trieste, Italy. How can I forget the unparalled support from Mawuli, my parents and my in-laws. I would like to thank them for their love, care and encouragements. Truly, I could not have done this without you. All friends who gave support deserve my sincerest thanks, especially my colleagues who encouraged me during the finalisation of this PhD study. vi University of Ghana http://ugspace.ug.edu.gh Abbreviations ABL Atmospheric Boundary Layer AFWA Airforce Weather Agency AMS Accra Meteorological Station AQM Air Quality Model BID Buoyancy-Induced Dispersion CALMET California Meteorological CALPOST California Post-Processing CALPUFF California Puff CDU Crude Distillation Unit CTDM Complex Terrain Dispersion Model DEM Digital Elevation Model DOAS Differential Optical Absorption Spectrometry EIPPCB European Integration Pollution Prevention Control Bureau EPA Environmental Protection Agency FB Fractional Bias FSL Forecast Systems Laboratory GC Gas Chromatography GFS Global Forecast System GLCC Global Land Cover Characterization GM Geometric Mean GSS Ghana Statistical Service GV Geometric Variance vii University of Ghana http://ugspace.ug.edu.gh IAEA International Atomic Energy Agency ICTP International Centre for Theoretical Physics IOA Index Of Agreement IS Inertial Sub-layer ISCST Industrial Source Complex Short-Term KNUST Kwame Nkrumah University of Science and Technology LULC Land Use Land Cover MB Mean Bias NCAR National Centre for Atmospheric Research NCEP National Centres for Environmental Prediction MESOPUFF Mesoscale Puff NMSE Normalised Mean Square Error NWP Numerical Weather Prediction PBL Planetary Boundary Layer PG Pasquill-Gifford PM Particulate Matter PRF Premium Reforming R Correlation Coefficient RFCC Residual Fluid Catalytic Cracking RS Roughness Sub-layer SNAS School Of Nuclear and Allied Sciences SRTM Shuttle Radar Topography Mission TMS Tema Meteorological Station TOR Tema Oil Refinery UCL Urban Canopy Layer UG University of Ghana UNEP United Nations Environment Programme USAID United States Agency for International Development USEPA United States Environmental Protection Agency viii University of Ghana http://ugspace.ug.edu.gh USGS United States Geological Survey UT University of Trieste UTM Universal Transverse Mercaptan VOC Volatile Organic Compound WRF Weather Research and Forecasting ix University of Ghana http://ugspace.ug.edu.gh Physical Constants Constant Name Symbol = Constant Value Acceleration due to gravity g = 9.8 m/s2 Universal Gas Constant R = 8.314 kJK−1mol−1 pi π = 3.14 Von Karman constant κ = 0.41 x University of Ghana http://ugspace.ug.edu.gh Symbols Symbol Name σ Standard deviation θ Potential temperature ξ Virtual source matrix β Entrainment parameter ρ Density Uτ Friction velocity d Displacement z Roughness height u Velocity C Concentration V Volume Q Emission rate H Mixing depth rG Ground reflection coefficient D Molecular diffusivity S Source/Sink p Probability distribution function Z Terrain-following vertical coordinate ht Terrain height N Brunt-Vaisala frequency xi University of Ghana http://ugspace.ug.edu.gh k Coefficient of exponential decay Fr Local Froude number F Buoyancy Fm Momentum flux Ta Ambient Temperature P Pressure xii University of Ghana http://ugspace.ug.edu.gh Table of Contents Declaration ii Acknowledgements iv Abbreviations vii Physical Constants x Symbols xi List of Tables xvii List of Figures xviii Abstract 1 1 General Introduction 3 1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 1.2 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 1.3 Problem Statement/Research Gap . . . . . . . . . . . . . . . 5 1.4 Justification and Scope of Work . . . . . . . . . . . . . . . . 6 1.5 Objectives of the Study . . . . . . . . . . . . . . . . . . . . . 8 xiii University of Ghana http://ugspace.ug.edu.gh 1.6 Dissertation Outline . . . . . . . . . . . . . . . . . . . . . . 8 2 Literature Review 10 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 2.2 Description and Characteristics of the Atmosphere Bound- ary Layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 2.2.1 Multi-Scale Considerations . . . . . . . . . . . . . . . 13 2.3 Air Quality Models . . . . . . . . . . . . . . . . . . . . . . . 16 2.3.1 Box Models . . . . . . . . . . . . . . . . . . . . . . . 17 2.3.2 Gaussian Models . . . . . . . . . . . . . . . . . . . . 19 2.3.3 Eulerian Models . . . . . . . . . . . . . . . . . . . . 22 2.3.4 Lagrangian Models . . . . . . . . . . . . . . . . . . . 25 2.3.5 Data Requirements for Air Quality Models . . . . . . 27 2.3.5.1 Meteorological data . . . . . . . . . . . . . 27 2.3.5.2 Geophysical Data . . . . . . . . . . . . . . . 28 2.3.5.3 Emission Data . . . . . . . . . . . . . . . . 30 2.3.6 Model Evaluation . . . . . . . . . . . . . . . . . . . . 32 2.4 CALPUFF Modeling System . . . . . . . . . . . . . . . . . . 35 2.4.1 CALMET Diagnostic Meteorological Model . . . . . 36 2.4.2 CALPUFF Model . . . . . . . . . . . . . . . . . . . . 42 2.4.2.1 Dispersion . . . . . . . . . . . . . . . . . . . 43 2.4.2.2 Atmospheric Turbulence Components . . . . 46 2.4.2.3 Buoyancy-Induced Dispersion . . . . . . . . 47 2.4.3 CALPOST . . . . . . . . . . . . . . . . . . . . . . . 49 2.4.4 Weather Research and Forecasting (WRF) Model . . 51 3 Methodology 52 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 3.1.1 The Study Area . . . . . . . . . . . . . . . . . . . . . 53 xiv University of Ghana http://ugspace.ug.edu.gh 3.1.2 Tema Oil Refinery . . . . . . . . . . . . . . . . . . . 55 3.2 Tema Oil Refinery Emission Estimation . . . . . . . . . . . . 57 3.2.1 Estimation of Flue Stack Gas Rate and Composition 58 3.2.2 Estimation of Flared Gas Composition . . . . . . . . 60 3.2.3 Calculation of Flare and Flue stack Exit Gas Velocities 61 3.2.4 Modelling Period . . . . . . . . . . . . . . . . . . . . 62 3.2.5 Model Set-up . . . . . . . . . . . . . . . . . . . . . . 62 3.2.5.1 CALMET Modelling . . . . . . . . . . . . . 62 3.2.5.2 Geophysical Data Input . . . . . . . . . . . 65 3.2.5.3 CALPUFF modelling . . . . . . . . . . . . 68 3.2.5.4 Model Evaluation . . . . . . . . . . . . . . . 70 4 Results and Discussions 73 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 73 4.2 Refinery Emissions and Interannual Trends . . . . . . . . . . 73 4.3 Preliminary Dispersion Simulation . . . . . . . . . . . . . . . 76 4.3.1 Spatial Variation of Pollutants . . . . . . . . . . . . . 78 4.4 Validation of the CALPUFF Model . . . . . . . . . . . . . . 82 4.5 Validation of CALMET and WRF Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86 4.6 Spatial Distribution of Emissions . . . . . . . . . . . . . . . 105 4.7 Interannual Predicted Concentrations of Emissions at Re- ceptors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112 4.8 Seasonal Variation of Pollutants . . . . . . . . . . . . . . . . 119 5 Conclusions and Recommendations 127 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 127 5.1.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . 128 5.1.2 Recommendations . . . . . . . . . . . . . . . . . . . . 130 xv University of Ghana http://ugspace.ug.edu.gh References 133 Appendix A 146 A Estimation of Refinery Emission Rates . . . . . . . . . . . . 146 A.1 Estimation of Flue Stack Gas Rate and Composition 146 A.1.1 Combustion Air Correction to Dry Basis . . 146 A.1.2 Calculation of Flue Gas Rate . . . . . . . . 147 A.1.3 Flue Stack Gas Components . . . . . . . . . 147 A.2 Estimation of Flared Gas Composition . . . . . . . . 150 A.3 Calculation of Flare and Flue Stack Exit Gas Velocities152 xvi University of Ghana http://ugspace.ug.edu.gh List of Tables 3.1 Operational Average Flow Parameters of the RFCCU of the Tema Oil Refinery for 2008 - 2013 . . . . . . . . . . . . . . . 58 3.2 Receptor Locations in the Study Area . . . . . . . . . . . . . 69 4.1 Statistical Performance Indices of the CALPUFF model . . . 85 4.2 Statistical Performance Indices of CALMET and WRF models 87 4.3 Statistical Performance Indices of CALMET and Observa- tions from the TMS and AMS . . . . . . . . . . . . . . . . . 90 1 Flue Gas Composition . . . . . . . . . . . . . . . . . . . . . 147 2 RFCC Fuel Gas Composition . . . . . . . . . . . . . . . . . 151 3 RFCCU and (Total Refinery) Flare Stack Emission Rates(kg/hr)152 4 RFCCU and (Total Refinery) Flue Stack Emission Rates(kg/hr)152 5 RFCC Point Sources Parameters . . . . . . . . . . . . . . . 152 6 Average Exit Gas Velocities of Point Sources Used for the Simulations for 2008 - 2013 . . . . . . . . . . . . . . . . . . . 154 xvii University of Ghana http://ugspace.ug.edu.gh List of Figures 2.1 Description of the atmospheric layers of the earth (Stull, 2012) 12 2.2 Description of the atmospheric boundary layer (Stull, 2012) 12 2.3 Flow layers over an urban environment (Raupach and Thom, 1981) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 2.4 Visualization of a buoyant Gaussian air pollutant dispersion plume (Holmes and Morawska, 2006) . . . . . . . . . . . . . 19 2.5 A schematic diagram of the program elements in the CAL- MET/CALPUFF modelling (Scire et al., 2000b) . . . . . . . 50 3.1 Map of Study Area showing the Tema Oil Refinery . . . . . 54 3.2 Nested Computational domains in the WRF simulations . . 64 4.1 Interannual Variation of CO2 Emission Rates from the Tema Oil Refinery . . . . . . . . . . . . . . . . . . . . . . . . . . . 74 4.2 Interannual Variation of VOCs Emission Rates from the Tema Oil Refinery . . . . . . . . . . . . . . . . . . . . . . . 74 4.3 Interannual Variation of PM2.5 Emission Rates from the Tema Oil Refinery . . . . . . . . . . . . . . . . . . . . . . . 75 4.4 Interannual Variation of SO2 Emission Rates from the Tema Oil Refinery . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 4.5 Interannual Variation of NO2 Emission Rates from the Tema Oil Refinery . . . . . . . . . . . . . . . . . . . . . . . . . . . 76 xviii University of Ghana http://ugspace.ug.edu.gh 4.6 Terrain Map of the Study area showing receptor locations and the Refinery (red square) . . . . . . . . . . . . . . . . . 77 4.7 LandUse Map of the Study Area . . . . . . . . . . . . . . . . 77 4.8 Predicted Daily Average Concentrations of SO2, NO2 and PM2.5 at northern receptors in the Study area . . . . . . . . 79 4.9 Predicted Daily Average Concentrations of SO2, NO2 and PM2.5 at north eastern receptors in the Study area . . . . . . 79 4.10 Predicted Daily Average Concentrations of SO2, NO2 and PM2.5 at south eastern receptors in the Study area . . . . . 80 4.11 Predicted Daily Average Concentrations of SO2, NO2 and PM2.5 at south western receptors in the Study area . . . . . 80 4.12 Predicted Daily Average Concentrations of SO2, NO2 and PM2.5 at northern western receptors in the Study area . . . 81 4.13 Wind Rose Depicting Surface Winds in Tema for 2008 . . . 82 4.14 Plots of measured and modelled SO2 Concentrations . . . . 83 4.15 Plots of measured and modelled NO2 Concentrations . . . . 83 4.16 Plots of Observed and Modelled Wind Speeds . . . . . . . . 86 4.17 Plots of Observed and Modelled Wind Direction . . . . . . . 87 4.18 Wind Rose Depicting CALMET Surface Winds . . . . . . . 88 4.19 Wind Rose Depicting WRF Surface Winds . . . . . . . . . . 89 4.20 Plots of Modelled and Observed(AMS) Surface Wind direc- tion for 2008 . . . . . . . . . . . . . . . . . . . . . . . . . . . 91 4.21 Plots of Modelled and Observed(AMS) Surface Wind direc- tion for 2009 . . . . . . . . . . . . . . . . . . . . . . . . . . . 92 4.22 Plots of Modelled and Observed(AMS) Surface Wind direc- tion for 2010 . . . . . . . . . . . . . . . . . . . . . . . . . . . 92 4.23 Plots of Modelled and observed(AMS) Surface Wind direc- tion for 2011 . . . . . . . . . . . . . . . . . . . . . . . . . . . 93 xix University of Ghana http://ugspace.ug.edu.gh 4.24 Plots of Modelled and observed(AMS) Surface Wind direc- tion for 2012 . . . . . . . . . . . . . . . . . . . . . . . . . . . 93 4.25 Plots of Modelled and observed(AMS) Surface Wind direc- tion for 2013 . . . . . . . . . . . . . . . . . . . . . . . . . . . 94 4.26 Plots of Modelled and Observed(TMS) Surface Wind direc- tion for 2008 . . . . . . . . . . . . . . . . . . . . . . . . . . . 94 4.27 Plots of Modelled and Observed(TMS) Surface Wind direc- tion for 2009 . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 4.28 Plots of Modelled and Observed(TMS) Surface Wind direc- tion for 2010 . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 4.29 Plots of Modelled and observed(TMS) Surface Wind direc- tion for 2011 . . . . . . . . . . . . . . . . . . . . . . . . . . . 96 4.30 Plots of Modelled and observed(TMS) Surface Wind direc- tion for 2012 . . . . . . . . . . . . . . . . . . . . . . . . . . . 96 4.31 Plots of Modelled and observed(TMS) Surface Wind direc- tion for 2013 . . . . . . . . . . . . . . . . . . . . . . . . . . . 97 4.32 Plots of Modelled and Observed(AMS) Surface Wind Speed for 2008 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98 4.33 Plots of Modelled and Observed(AMS) Surface Wind Speed for 2009 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99 4.34 Plots of Modelled and Observed(AMS) Surface Wind Speed for 2010 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99 4.35 Plots of Modelled and Observed(AMS) Surface Wind Speed for 2011 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100 4.36 Plots of Modelled and Observed(AMS) Surface Wind Speed for 2012 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100 4.37 Plots of Modelled and Observed(AMS) Surface Wind Speed for 2013 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 xx University of Ghana http://ugspace.ug.edu.gh 4.38 Plots of Modelled and Observed(TMS) Surface Wind Speed for 2008 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102 4.39 Plots of Modelled and Observed(TMS) Surface Wind Speed for 2009 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103 4.40 Plots of Modelled and Observed(TMS) Surface Wind Speed for 2010 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103 4.41 Plots of Modelled and Observed(TMS) Surface Wind Speed for 2011 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104 4.42 Plots of Modelled and Observed(TMS) Surface Wind Speed for 2012 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104 4.43 Plots of Modelled and Observed(TMS) Surface Wind Speed for 2013 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105 4.44 2008 Annual Average Concentration contours of SO2 in the Study Area . . . . . . . . . . . . . . . . . . . . . . . . . . . 106 4.45 2008 Annual Average Concentration contours of NO2 in the Study Area . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 4.46 2008 Annual Average Concentration contours of PM2.5 in the Study Area . . . . . . . . . . . . . . . . . . . . . . . . . . . 108 4.47 2009 Annual Average Concentration contours of SO2 in the Study Area . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 4.66 Wind Rose Depicting 2013 Surface Winds in the Study area 109 4.48 2009 Annual Average Concentration contours of NO2 in the Study Area . . . . . . . . . . . . . . . . . . . . . . . . . . . 110 4.49 2009 Annual Average Concentration contours of PM2.5 in the Study Area . . . . . . . . . . . . . . . . . . . . . . . . . . . 110 4.50 Wind Rose Depicting 2009 Surface Winds in the Study area 111 4.51 2010 Annual Average Concentration contours of SO2 in the Study Area . . . . . . . . . . . . . . . . . . . . . . . . . . . 112 xxi University of Ghana http://ugspace.ug.edu.gh 4.52 2010 Annual Average Concentration contours of NO2 in the Study Area . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 4.53 2010 Annual Average Concentration contours of PM2.5 in the Study Area . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 4.54 Wind Rose Depicting 2010 Surface Winds in the Study area 114 4.55 2011 Annual Average Concentration contours of SO2 in the Study Area . . . . . . . . . . . . . . . . . . . . . . . . . . . 115 4.56 2011 Annual Average Concentration contours of NO2 in the Study Area . . . . . . . . . . . . . . . . . . . . . . . . . . . 115 4.57 2011 Annual Average Concentration contours of PM2.5 in the Study Area . . . . . . . . . . . . . . . . . . . . . . . . . . . 116 4.58 Wind Rose Depicting 2011 Surface Winds in the Study area 117 4.59 2012 Annual Average Concentration contours of SO2 in the Study Area . . . . . . . . . . . . . . . . . . . . . . . . . . . 118 4.60 2012 Annual Average Concentration contours of NO2 in the Study Area . . . . . . . . . . . . . . . . . . . . . . . . . . . 118 4.61 2012 Annual Average Concentration contours of PM2.5 in the Study Area . . . . . . . . . . . . . . . . . . . . . . . . . . . 119 4.62 Wind Rose Depicting 2012 Surface Winds in the Study area 120 4.63 2013 Annual Average Concentration contours of SO2 in the Study Area . . . . . . . . . . . . . . . . . . . . . . . . . . . 121 4.64 2013 Annual Average Concentration contours of NO2 in the Study Area . . . . . . . . . . . . . . . . . . . . . . . . . . . 121 4.67 Daily Average SO2 Concentrations at various receptors . . . 122 4.68 Daily Average NO2 Concentrations at various receptors . . . 122 4.69 Daily Average PM2.5 Concentration at various receptors . . . 123 4.70 2013 Monthly Average concentrations of pollutants at Tema Steelworks . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124 xxii University of Ghana http://ugspace.ug.edu.gh 4.71 2013 Monthly Average concentrations of pollutants at Tema Comm. 25 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124 4.72 2013 Monthly Average concentrations of pollutants at Kpone 125 4.73 2013 Monthly Average concentrations of pollutants at Tema Gen. Hosp . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 4.74 Variation of Ambient Temperature around the refinery . . . 126 xxiii University of Ghana http://ugspace.ug.edu.gh Abstract The petrochemical industry is a major contributor of industrial air pol- lutants which are known to have dire consequences on human health and the environment, neccesitating research into their dispersion and transport. The objective of the study, therefore, is to simulate the dispersion and trans- port of pollutants emitted during the processing of crude oil by the Tema Oil Refinery in the Greater Accra region of Ghana using the California Puff (CALPUFF) modeling system. This thesis couples the Weather Re- search and forecasting Model (WRF) with the non-steady state California Puff(CALPUFF) modelling system to simulate the dispersion and trans- port of emissions from the refinery in a coastal urban/industrial area in Ghana. The mass balance approah was employed to estimate the refinery emission rates which were used as input for the dispersion model. Emission rates of five species were estimated - SO2, NO2, PM2.5, CO2 and VOCs. The transport and dispersion of SO2, NO2 and PM2.5 were modelled over the period between 2008 - 2013 and their impact on 38 identified receptors investigated. Simulation results showed that the radius of impact of the emissions is approximately 10 km. As a result of the prevailing predomi- nant south-westerly winds in the study area, concentrations of emissions at receptors located upwind of the emission source were found to be higher as the winds carried the pollutant clouds in their direction. Conversely, south and south-western receptors, relative to the refinery, on the other hand, 1 University of Ghana http://ugspace.ug.edu.gh Abstract 2 were minimally impacted. Concentrations of SO2 and NO2 at 2 out of the 38 receptors exceeded the regulatory limit of the World Health Organisa- tion and Ghana’s Environmental Protection Agency. It can be concluded, therefore, that SO2 and NO2 emissions from the refinery do not pose any danger to the larger population and the general environment nearby. PM2.5 levels at 36 receptors however exceeded the WHO guideline value leading to the conclusion that the refinery operations could pose some dangers to the environment regarding PM2.5. The dispersion model results were compared with measurements at the same location in order to validate the model. Similarly, observations from two meteorological stations were compared with results from the meteorological model. The performance evaluation, with the aid of statistical measures revealed that the models’ performance were acceptable. University of Ghana http://ugspace.ug.edu.gh Chapter 1 General Introduction 1.1 Introduction This thesis deals with the modeling and simulation of the long-range trans- port and dispersion of refinery emissions over a defined study area. The first chapter of the dissertation provides a brief background of the thesis subject area, gives the problem statement, the justification and scope of this research and finally outlines the objectives of the study. 1.2 Background All life forms on this planet depend on clean air. Air quality not only affects human health but also components of environment such as water, soil and forests which are the vital resources for human development. A major 3 University of Ghana http://ugspace.ug.edu.gh Chapter 1. General Introduction 4 threat to the availability of clean air is urbanisation. Urbanization is a process of relative growth in a country's urban population accompanied by an even faster increase in the economic, political and cultural importance of cities relative to rural areas. As an integral part of economic development, urbanisation brings in its wake a number of challenges including the increase in urban population, industrial activities, high rise buildings and vehicular movement. All these activities contribute to air pollution (de Leeuw et al., 2001, Fenger, 1999). Exposure to such air pollutants may adversely affect human health. Short term exposure to peak levels of particulate matter has been strongly associ- ated with adverse respiratory health impacts (e.g., respiratory diseases such as asthma and pulmonary function insufficiency) (Brunekreef and Holgate, 2002, Guilbert et al., 2003, Hansard et al., 2011). Furthermore, particulate matter, hereafter referred to as PM, is known to degrade atmospheric visi- bility. High concentrations of sulphur dioxide, SO2, which are also emitted from power plants, industrial processes and during the combustion of fossil fuels, can aggravate respiratory diseases as well as cause problems such as acid rain and damaged vegetation in the form of foliar necrosis (Khamsi- mak et al., 2012). Additionally, air pollution is not only a human health problem: its effects on ecosystems and materials are well identified and documented (Fowler et al., 2009). Economic costs can also be associated with poor air quality as well as political/governmental measures taken to University of Ghana http://ugspace.ug.edu.gh Chapter 1. General Introduction 5 prevent or reduce pollution (Muller and Mendelsohn, 2007). 1.3 Problem Statement/Research Gap The USEPA, USAID and UNEP as far as July 2004, selected Accra, Ghana as one of two cities in Africa to benefit from an Air Quality Monitoring Ca- pacity Building Project with the aim to build and establish local capacity on air quality monitoring that will provide policy-makers with information on the air quality in Accra and its impacts on health. Subsequent to this, the EPA, Ghana replicated the project in a few more cities in Ghana with emphasis on PM measurements. SO2 and NO2 emissions from vehicular traffic have also been monitored in parts of Accra. Apart from the EPA’s efforts, similar monitoring activities have been carried out, especially within the Accra metropolis, by students and groups of scientists for different pe- riods of time largely on heavy metals. For example, Arku et al. (2008) conducted a study for an initial assessment of the levels and spatial and/or temporal patterns of multiple pollutants in the ambient air in two low- income neighborhoods in Accra, Ghana over a 3-week period while Ofosu et al. (2012) characterized fine particulate sources at Ashaiman in Greater Accra, Ghana. Others include the assessment of particulate matter and heavy metals along major highways and in mining areas (Affum et al., 2008, Bansah and Amegbey, 2012, Safo-Adu et al., 2014). Quite recently, Sackey University of Ghana http://ugspace.ug.edu.gh Chapter 1. General Introduction 6 (2012) employed Differential Optical Absorption Spectroscopy (DOAS) in the measurement of atmospheric constituents as a result of combustion, vehicular emissions and industrial activities around the Tema Oil refin- ery. There is, however, limited or no information about the behaviour and spread of these pollutants from specific sources in the atmosphere within the country. This is because none of the research works focussed on a spe- cific industry, its air emissions and rates and the transport of the emissions within its catchment area. A gap therefore exists in this area of research which thesis seeks to fill through atmospheric dispersion modelling. 1.4 Justification and Scope of Work The simplest technique for evaluating patterns of local-scale urban air pol- lution concentration involves the interpolation of ambient concentrations from existing monitoring networks (Ferretti et al., 2008, Perez Ballesta et al., 2008). However, the measured data from these stations or study areas are not necessarily representative of areas beyond their immediate vicinity. This is because concentrations of pollutants in urban areas may greatly vary on spatial scales that range from tens to hundreds of metres. Moreover, the establishment and operation of monitoring stations are ex- pensive and can only be expected to be established in few locations. At the same time, the temporal behaviour of primary and secondary pollutants University of Ghana http://ugspace.ug.edu.gh Chapter 1. General Introduction 7 changes considerably between day and night due to solar radiation, so that daily average measurements become unsatisfactory in determining or ex- plaining high pollution episodes. In such instances, air pollution dispersion models become necessary. Dispersion modelling studies, in combination with air quality monitoring, are essential and complementary tools for long and short term air pollu- tion control strategies in effective air quality management. Air quality and dispersion models become valid instruments for environmental managers in many activities, such as setting emission control regulations, testing the compliance of actual pollution levels, predicting the impact of new facili- ties on human health, selecting the best location for monitoring stations and assessing the impact of different emissions scenarios on selected loca- tions (Puliafito et al., 2011). Modelling studies are crucial as they provide useful data on the dynamics of pollutant dispersion and transport among others, which feed into the formulation of environmental policies as well as management processes. Furthermore, they improve the limitations of monitoring networks by providing predictions of the temporal and spatial distribution of actual pollution levels. The focus of this thesis, therefore, is to utilize air quality modeling to fill the gap that presently exists in the area of air pollution studies within the country. Due to the significant contribution of petroleum refineries to air pollution, this thesis investigates the long-range transport of emissions from Ghana’s University of Ghana http://ugspace.ug.edu.gh Chapter 1. General Introduction 8 only refinery, Tema Oil Refinery, using numerical models. 1.5 Objectives of the Study The main objective of the study is to simulate the dispersion and transport of pollutants emitted during the processing of crude oil by the Tema Oil Refinery, hereafter referred to as TOR, in the Greater Accra region of Ghana using the California Puff (CALPUFF) modeling system. This will be achieved through the following specific objectives: 1. To estimate the emission rates of CO2, SO2, NO2, volatile organic compounds (VOCs) and PM2.5 during refinery operations at the TOR. 2. To simulate the transport and dispersion of the emissions and 3. To access the impact of meteorological conditions on the dispersion of the emissions. 1.6 Dissertation Outline This thesis is organized as follows: The first chapter of the dissertation provides a brief background of the thesis area, gives the problem statement, the justification and scope of this research and finally outlines the objectives of the study. University of Ghana http://ugspace.ug.edu.gh Chapter 1. General Introduction 9 In Chapter 2, a literature review of the subject area is presented. Emissions and their sources are discussed briefly after which an overview of some air quality models and their applications as well as their strengths and limi- tations are presented. A detailed description of the CALPUFF/CALMET models used for the simulation is also included in this chapter. A short presentation of the prognostic mesoscale WRF model used to simulate the wind fields is also provided. In Chapter 3, the methodology for the simulation of the pollutants disper- sion in the study area is presented. First of all, the methodology for the estimation of pollutants from the refinery is presented. This is followed by a description of the various data resources used for the simulation. The val- idation methodology of all simulation results is also described using some statistical tools. A thorough discussion of the research results and its contribution to the wider literature are presented in Chapter 4. Finally in Chapter 5, conclusions and recommendations are presented. The limitations and challenges of the work are also discussed and future work outlined. University of Ghana http://ugspace.ug.edu.gh Chapter 2 Literature Review 2.1 Introduction This Chapter presents some general concepts and theoretical frameworks in air pollution studies. Various air quality models in literature are also reviewed. A detailed description of the numerical models used for the sim- ulations in this thesis, namely the California Puff (CALPUFF) modelling system and WRF, a description of the study area and the emission source of interest, TOR are also presented. The equations forming the basis of the models are presented as well as an overview of the Residual Fluid Catalytic Cracking (RFCC) process in petroleum refining at the TOR. 10 University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 11 2.2 Description and Characteristics of the Atmosphere Boundary Layer A discussion of the layers in the earth’s atmosphere is needed to better understand where air pollution dispersion takes place. The main layers of the earth’s atmosphere, from the surface of the ground upwards, as shown in Fig.2.1 are the troposphere (0 to 15 km), the stratosphere (15 to 50 km), the mesosphere (50 to 85 km), the thermosphere and others (more than 85 km). The lowest part of the troposphere is the Atmospheric Boundary Layer (ABL) or Planetary Boundary Layer (PBL) which extends from the earth’s surface to about 1.5 to 2.0 km in height (Stull, 2012). The ABL is made up of the mixing layer, capped by the inversion layer, and is separated by a change in temperature behaviour in the vertical direction as shown in Fig.2.2. Human activity is generally confined to the ABL such that pollutant sources are created in this area. Thus the challenge of modeling pollutant dispersion is understanding how materials are mixed by turbulence and transported from the release point to larger scales (Fernando, 2010). Fundamentally, it is characterized by a large shearing stress resulting from momentum transfer at the surface. The exact structure of the ABL is determined by both the character of the surface and the geostrophic winds aloft driving it. Geostrophic winds are a global scale phenomena derived from a basic University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 12 Figure 2.1: Description of the atmospheric layers of the earth (Stull, 2012) Figure 2.2: Description of the atmospheric boundary layer (Stull, 2012) University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 13 balance between pressure and Coriolis force which is the apparent deflection of objects moving in a straight path relative to the earth’s surface. For conditions of neutral stability, the depth of the ABL can vary from several hundred meters to over a thousand meters depending on the speed of the geostrophic wind (Pasquill and Smith, 1983). Neutral stability is often a fair assumption for the ABL over urban areas, especially at night, as physical factors like surface drag force due to roughness and heat-storage promote these conditions (Britter and Hanna, 2003). 2.2.1 Multi-Scale Considerations To do a better analysis of the fluid dynamics of the lower atmosphere, it is helpful to recognize the multi-scale phenomena present. The global processes that drive regional weather conditions and atmospheric bound- ary layer formation will not be the focus of this Thesis. Instead, flow phenomena affecting dispersion at the cities/neighbourhood scale will be considered. At the city/neighborhood scale, the ABL is further divided into regions describing the impact of the urban environment. The effect of individual buildings on the flow at this scale is conceptualized as flow over a series of roughness elements such as buildings, trees and hills. Closest to the ground, the buildings are said to reside in the canopy layer which extends to the University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 14 height of the tallest building. After the canopy layer is the Roughness Sublayer (RS) followed by the Inertial Sublayer (IS) as seen in Fig.2.3. Figure 2.3: Flow layers over an urban environment (Raupach and Thom, 1981) The RS, as defined by Raupach and Thom (1981), is the region over which mean flow and turbulence properties depend on the specific details of the roughness (i.e. the buildings) itself. As such, the exact definition of its extent varies throughout the literature. The lower boundary is sometimes considered to be the mean building height and in other cases the zero-plane displacement, d, which is defined based on a logarithmic velocity profile (Rotach, 1994). The upper boundary of the RS occurs when the horizontal variation in flow and turbulence parameters caused by the canopy subsides, though exact applications of this concept vary. University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 15 The IS, is by definition, a layer of constant shear stress where Monin- Obukhov similarity arguments apply (Monin and Obukhov, 1954). For stable conditions over a rough surface, this leads to a logarithmic velocity profile of the form as depicted by Eqn.( 2.1): ( ) uτ z − d u(z) = ln (2.1) κ z0 where uτ is the friction velocity, z0 is the characteristic roughness height d is the previously mentioned zero-plane displacement and κ is the Von Karman constant. Jackson (1981) showed that d represents the level at which the mean surface drag acts, while z0 is a measure of the magnitude of that drag. Values of z0 and d have been tabulated from experimental data for a range of surfaces from croplands and forests to concrete roads and towns (Wieringa et al., 2001). The range reported for any one given surface type illustrates the approximate nature with which z0 and d are often interpreted. The complexity in determining a precise value stems from the fact that z0 and d depend not only on the specific geometrical properties of the roughness elements, but also on the flow conditions. Nevertheless, the logarithmic University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 16 velocity profile is typically used not only in the range where it is strictly valid, but is extended into the roughness sublayer. Macdonald (2000) noted that this is done out of a lack of information about flow within the canopy and is generally an acceptable approximation for studies of larger, elevated plumes. 2.3 Air Quality Models In basic applications of air quality models, the processes of air pollution transport are considered as a distributed parameter system, which is gov- erned by a set of transport equations, along with respective boundary and initial conditions. The exact form and structure of a model usually depend on its practical application, type of the polluting compounds considered and the scale of modelling. A model usually takes into account the input data (emission field and meteorological data) as well as the main physical and chemical processes which determine the transport in the atmosphere and transformations of air pollution components (Holnicki, 2011). The characteristics of each specific problem will define the physical and chemi- cal processes involved, and consequently, the best model to use. The main criteria for choosing appropriate software are: 1. The dimension of the area under study University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 17 2. The number of pollution sources 3. The chemical species involved and 4. The time scale of the episode. The spatial and temporal scales of the environmental impact of air pollu- tion are correlated with the lifetime of a pollutant. Thus, depending on the analysis scale, there are respective categories of modelling: local, regional and global. Regarding the practical application and the scale of modelling, the most common types (implementations) of air pollution models are dis- cussed in the following section. 2.3.1 Box Models Box models are derived by simply applying a control volume-based mass conservation approach to determining concentration levels in a domain of interest. This leads to a uniform prediction of concentration without pro- viding information about spatial variation in concentration within the box (Holmes and Morawska, 2006). The domain is treated as a box into which pollutants are emitted and undergo chemical and physical processes. They require the input of simple meteorology and emissions and the movement of pollutants in and out of the box is allowed. The model also assumes that the incoming pollution is instantaneously mixed with the surrounding University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 18 air, creating a homogeneous concentration throughout the airshed (Venka- tram, 1978). The mass conservation constraint permits the construction of a mass balance equation of the form as shown in Eqn.(2.2): dcV = QA+ ucinWH − ucWH (2.2) dt where V is the volume described by the box, c is the homogeneous species concentration within the airshed, cin is the species concentration entering the airshed, Q is the emission rate per unit area of sources within the box, u is the average wind speed normal to the box, A is the horizontal area of the box (L × W ), W is the width of the box and H is the mixing depth. Integrating Eqn.( 2.2) provides a steady state estimation of species concen- tration (Venkatram, 1978) assuming the dynamics of the mixing depth to be quasi-stationary and the source emissions to be constant. In the case of reactive species, the chemical reaction dynamics can be incorporated into the mass balance equation as well as wet and dry deposition effects. University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 19 It is incapable of imparting any spatial information regarding the dispersive nature of a pollutant. This precludes the box model approach from a significant proportion of air quality modelling applications. Nevertheless, the method is computationally fast and is capable of providing satisfactory predictions, particularly for scenarios where detailed information on the domain and meteorological conditions is unavailable. 2.3.2 Gaussian Models The Gaussian model forms the basis for the majority of air pollution mod- els, and is the most well known and documented approach. The model presupposes that the dispersion associated with the polluting species can be described by a modified Gaussian or normal distribution curve as shown in Fig.2.4. Figure 2.4: Visualization of a buoyant Gaussian air pollutant dis- persion plume (Holmes and Morawska, 2006) University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 20 A three-dimensional axis system is employed to provide a downwind, cross- wind and vertical resolution. The species concentration is defined as being proportional to the emission rate of the source, diluted by the wind veloc- ity at the source of emission. The dispersion behaviour of a pollutant is determined by the standard deviations associated with the Gaussian distri- bution function. These standard deviations which are related to the turbu- lent diffusivities are typically functions of atmospheric stability, localised turbulence and distance downwind from the source. Since the turbulent diffusivity is unknown, however, the deviations are parameterized based on experimental measurements and observations. The parametrization is typically a function of conditions like atmospheric stability (Holmes and Morawska, 2006). The model is usually aligned so that the downwind axis corresponds to the direction of the prevailing wind (Collett and Oduyemi, 1997). The model equation is derived from basic considerations of the dif- fusion of gaseous matter in three-dimensional space as shown in Eqn.(2.3): ( )[ ( ) ( )] Q −y2 −(h− z)2 −(h+ z)2 C = exp exp + rGexp 2πuσyσz 2σ2 2 2y 2σy 2σy (2.3) where C is the species concentration at a location (x, y, z), Q is the source emission rate, University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 21 u is the average wind speed normal to the box, σy is the standard deviation of the horizontal crosswind distribution of the plume concentration and is a function of the downwind distance x, σz is the standard deviation of the vertical crosswind distribution of the plume concentration and is a function of the crosswind distance z, h is the effective source height to which the plume has risen, rG is the ground reflection coefficient where 0≤ rG≥ 1, y is the crosswind distance and z is the receptor height above ground. The effective source height (h) or plume rise is the height to which an emission will initially rise as a result of thermal buoyancy and vertical mo- mentum. The upward movement of the plume is retarded on mixing with ambient air reaching an equilibrium point when the internal energy of the plume is equal to that of the surrounding atmosphere. A review of vari- ous semi-empirical methods for the estimation of plume rise can be found in Zannetti (2013). Several assumptions are implied in the derivation of Eqn.(2.3), including the uniformity and time invariance of the emission characteristics of the source, the homogeniety of the meteorological condi- tions within the domain of interest and the weakness of the topography of the domain so as not to affect pollutant dispersion on plant operation. In University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 22 the light of such limitations, the Gaussian model can only be considered workable when such factors are static enough to be regarded as homoge- neous. The behaviour of the Gaussian model depends heavily upon the correct calculation of the dispersion coefficients, σy and σz. Various approaches currently exist and a summary of the methods for the estimation of σy and σz can be found in Zannetti (2013) and Boubel et al. (2013). The limitations of the Gaussian model precludes its use in cases where the short term prediction of species concentration (i.e.sub-hourly averaged values), or the prediction of species concentrations relative to complex environmental constraints are required. 2.3.3 Eulerian Models The Eulerian approach to dispersion modelling solves the conservation of mass equation for a given pollutant species of concentration c. A stationary or normal frame of reference is assumed, with the dispersion phenomena calculated as a concentration field relative to the domain. The physical conditions within the reference domain are generally regarded as turbulent (Brown, 1991). Therefore any dependent variable is composed of an average component, denoted by an overbar, and a fluctuating component, denoted University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 23 by a prime, as shown in Eqn.(2.4):. U = U + U ′ (2.4) where U is the Eulerian wind field vector U(x,y,z). Given that 〈U〉 = U and 〈U ′〉 = 0 where 〈〉 represents the ensemble average or theoretical mean and c = 〈c〉+ c′, then the general form for the equation of conservation of pollutant species c is shown in Eqn.(2.5): ∂〈ci〉 = −U · ∇〈c 〉 − ∇ · 〈c′U ′〉+D∇2i i 〈ci〉+ 〈Si〉 (2.5)∂t where ci is the concentration associated with the i th species, D is the molecular diffusivity and University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 24 Si is the sink/source term of the i th species and accounts for chemical re- actions, deposition and emission sources. The first, second and third terms on the right hand side of Eqn.(2.5) rep- resent the rate of advection, turbulent diffusion and molecular diffusion of pollutants respectively. The term 〈c′ ′iU 〉 represents the turbulent atmospheric diffusion eddies whose magnitude and effects are significantly greater than that of molecular diffu- sion. For the majority of cases it is usual to ignore the molecular diffusivity term as its overall contribution will be negligible. The eddy diffusivity term is unresolvable as c′ is unknown and a suitable description of U ′ is more often than not impossible to obtain. Consequently, the diffusivity term, 〈c′iU ′〉, has to be modelled if Eqn.(2.5) is to be closed. After selection of a suitable closure method, Eqn.(2.5) is typically solved in one, two or three dimensions on a discrete mesh architecture, using a suitable numerical method including finite difference, finite element or fi- nite volume (Versteeg and Malalasekera, 2007). Eulerian models suffer the disadvantage that their resolution is confined by the spatial and temporal discretisation of the mesh on which they are solved. The use of a mesh is computationally expensive and traditionally requires some form of op- timisation to achieve any degree of efficiency. The approach is, however, University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 25 information-rich, providing a description of the relevant transport dynamics at all defined points throughout the domain (Collett and Oduyemi, 1997). 2.3.4 Lagrangian Models The Lagrangian approach to atmospheric dispersion modelling differs from its Eulerian counterparts in that its reference system follows the prevailing vector of atmospheric motion. The term Lagrangian is applied to a wide range of models which simulate pollutant dispersal relative to a shifting reference frame. The general equation of motion describing the atmospheric dispersion of a single pollutant species is given by Eqn.(2.6) according to Zannetti (2013): ∫ t ∫ 〈c(r, t)〉 = p(r, t|r′, t′)S(r′, t′)dr′dt′ (2.6) −∞ where 〈c(r, t)〉 is the ensemble average concentration at r, at time t, S(r′, t′) is the source term and p(r, t|r′, t′) is the probability density function that an air parcel is moving from r′ at t′ to r at time t. University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 26 Lagrangian models incorporate changes in concentration due to mean fluid velocity, turbulence of the wind components and molecular diffusion. They work well both for homogeneous and stationary conditions over flat terrains and for inhomogeneous and unstable media condition for complex terrains (Tsuang, 2003, Venkatesan et al., 2002). A fundamental problem that workers have encountered with Eqn.(2.6) and in a wider sense, Lagrangian dispersion models generally, is the relative interpretation of their results. It is fair to assume that the great major- ity of real-time meteorological and air quality measurements are obtained relative to a stationary reference frame. As a consequence, the results from Lagrangian based models cannot be easily compared with observed measurements. This often presents difficulties during the initial validation and verification phases of model development and throughout the post- development period, where simulation results have to be mapped to real life scenarios if they are to be considered useful. This has resulted in a trend where those atmospheric dispersion models which have adopted a La- grangian approach have typically utilised Lagrangian methods which can be more readily compared to an Eulerian reference frame (e.g. particle/puff models). Those which have been successfully developed have been required to include some form of Eulerian mapping of their results, in order to achieve a wider application (Collett and Oduyemi, 1997). University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 27 2.3.5 Data Requirements for Air Quality Models At this stage, it will be helpful to recognise some data requirements for air quality models notably meteorological, geophysical and emission source data. 2.3.5.1 Meteorological data Meteorological data is a critical input for AQMs, as it is necessary to ob- tain accurate description of winds, turbulence fields and radiation in order to correctly describe transport, dispersion, deposition and chemical reac- tions of a released pollutant (Demuzere et al., 2009, Pearce et al., 2011, Schürmann et al., 2009). Perhaps the most important meteorological ele- ment controlling levels of atmospheric pollution is the wind, according to Abdul-Wahab (2003). Wind moves in three dimensions. However, usu- ally the horizontal component is dominant and its important properties are speed and direction. Wind speed determines the travel time from a source to a given receptor and the total area over which the plume will be dispersed. Other effects of wind speed include a dilution in the downwind direction. Usually wind speed has two opposing effects on the dispersion of pollutants. It affects both the spread of the plume (rate of dilution of pollutants) and the height to which the plume will rise. Thus, with a high wind speed, a plume will be diluted quickly, but it will not be able to rise University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 28 since higher wind speeds tend to bend a plume, retarding its vertical mo- tion. In calm winds, the dilution factor will be small, but a hot plume may be able to rise to a considerable height. The wind direction determines the course the effluents will take or the area to which the plume will be directed. The correlation of wind direction and pollution concentration at any site can therefore help to identify the sources mainly responsible for the pollution measured at that site. The effects of other meteorological elements like precipitation, temperature and relative humidity cannot be overemphazised. Meteorological data requirements for local scale models (i.e., steady-state Gaussian, Puff-models) and more complex models, vary considerably. Steady- state Gaussian plume models need data only from a single station, since they assume that meteorological conditions do not vary throughout the do- main up to the top of the boundary layer (Cimorelli et al., 1998). More advanced models (both puff and grid models) allow meteorological condi- tions to vary across the modelling domain and up through the atmosphere, thus requiring more complex meteorological data. 2.3.5.2 Geophysical Data The next data resource needed is geophysical data consisting mainly of ter- rain and land use/land cover. The elevation of a geographic location is its University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 29 height above or below a fixed reference point, most commonly a reference geoid, a mathematical model of the Earth’s sea level as an equipotential gravitational surface. Terrain features around a pollutant source can sig- nificantly affect the pattern of dispersion. Steady-state Gaussian models contain limited algorithms that include terrain effects. Advanced models contain more sophisticated procedures for modelling the effects of terrain, with a correspondingly greater effort required by the user to specify the static data. Since terrain data will be required for every receptor on the grid, there are several pre-processing tools that extract and format the Digital Elevation Model (DEM) data. Land use plays an important role in air dispersion modelling from meteoro- logical data processing to defining modelling characteristics such as urban or rural conditions. Land use data can be obtained from digital and paper land-use maps. The maps provide an indication into the dominant land use types within an area of study, such as industrial, agricultural, forest and others. This information can then be used to determine dominant disper- sion conditions and estimate values for the critical surface characteristics which are surface roughness length, albedo and the Bowen ratio. The most common global data sets are: the United States Geological Ser- vice (USGS) GTOPO30 with a horizontal grid spacing of 30 arc-seconds (approximately 1km); USGS SRTM30, with the same horizontal grid spac- ing, but covering the globe only from 60◦N latitude to 56◦S latitude, with University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 30 a seamless and uniform representation; and SRTM3 data with a horizontal grid spacing of 3 arc-seconds (about 90 m). Land Use and Land Cover (LULC) data are also available from the USGS, at the 1:250,000 scale, or in some cases at the 1:100,000 scale. The USGS Global Land Cover Char- acterization (GLCC) Database is developed on continental basis for land use, while land cover maps are classified into 37 categories, with a spatial resolution of 1 km (USGS, 2010). 2.3.5.3 Emission Data Emission inventories are also a key input for AQMs. There are numerous ways of estimating emissions of air pollutants with the popular ones being direct measurements, use of mass balances or fuel analysis, emission factors and emission models. The most accurate way of estimating emissions is directly measuring the concentration of air pollutants from the source. Source tests and continuous emission monitoring systems (CEMS) are two methods of collecting actual emission data (USEPA, 2010, 2011). A CEMS involves the installation of the monitoring equipment that accumulates data on a pre-determined time schedule in a source (for example a stack or duct). It provides a continuous record of emissions over an extended and uninterrupted period of time. Data from source-specific emission tests or continuous emission University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 31 monitors are usually preferred for estimating pollutant releases because the data provide the best representation of emissions from tested sources. However, test data from individual sources are not always available, and may not even reflect the variability of actual emissions over time. Thus, emission factors are frequently used for estimating emissions, in spite of their limitations (Puliafito et al., 2011). Emission factors are generally derived from measurements made on a num- ber of sources representative of a particular emission sector and are usually expressed as the weight of pollutant divided by a unit weight,volume, dis- tance, or activity duration that releases the pollutant (e.g., kilograms of particulate emitted per megagram of coal burned). Such factors facilitate estimation of emissions from various sources of air pollution. In most cases, these factors are simply averages of all acceptable quality data available, and are generally assumed to be representative of long-term averages for all facilities in the source category (i.e., a population average). Emission factors are founded on the premise that there exists a linear relationship between the emissions of air contaminant and the activity level pollutant (e.g., kilograms of particulate emitted per megagram of coal burned). Such factors facilitate estimation of emissions from various sources of air pol- lution. In most cases, these factors are simply averages of all acceptable quality data available, and are generally assumed to be representative of University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 32 long-term averages for all facilities in the source category (i.e., a popu- lation average). Emission factors are founded on the premise that there exists a linear relationship between the emissions of air contaminant and the activity level (Puliafito et al., 2011, USEPA, 2001). Mass balance involves the quantification of total materials into and out of a process, with the difference between inputs and outputs being accounted for in terms of releases to the environment, or as part of the facility waste. Mass balance is particularly useful when the input and output streams can be quantified, and this is most often the case for individual process units and operations. Mass balance techniques can be applied across individual unit operations, or across an entire facility. These techniques are best applied to systems with prescribed inputs, defined internal conditions and known outputs (USEPA, 2011). 2.3.6 Model Evaluation Evaluation of an AQM is the process of assessing its performance in simu- lating spatial-temporal features embedded in the air quality observations. When evaluating air quality management strategies, policy-makers need in- formation about relative risk and likelihood of success of different options. In these cases, a range of values reflecting the model uncertainties, is more important than the model best guess, or actual output. End users are University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 33 more likely to work with operational and dynamic evaluation tools, while the other two categories of evaluation are more related to model develop- ment. The kind of data needed for verifying model output, will depend on the model itself and the users needs. For models with meteorological pre- processors, like CALMET, or coupled meteorological/chemical models like WRF/Chem, atmospheric variables observation in some points of the do- main would be required in order to validate results. Observations can be made at ground level or with a vertical profile, in the case of three dimen- sional simulations. In the case of chemical species concentration, monitor- ing stations could supply data needed to check model results. Some ground or satellite instruments can also provide vertical profile for chemical species (Martin, 2008). In any case, a consistent procedure should be applied in order to evaluate the model performance. The most usual practice is to use the information content shown between the observed and the model- predicted values. In this respect, Willmott (1982) and Seigneur et al. (2000) propose some statistical performance measures namely: correlation coef- ficient(R), mean bias(MB), fractional bias(FB), normalised mean square error(NMSE), geometric mean(GM), geometric variance(GV) and index of agreement(IOA). The coefficient of correlation is the measurement of the relationship between observed and predicted values. It indicates the tendency of the predicted University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 34 values to change with a change in the observed values. A value of R close to unity implies good model performance. The NMSE measures the random spread of the values around the mean. It characterises the amount of deviation between predictions and observations. A good model will have an NMSE value of 0. The IOA reflects the degree to which the observed variable is accurately predicted. The IOA varies from 0 (the theoretical minimum for an inadequate prediction) to 1(perfect accuracy between the predicted and observed values). The FB is a measure of the systematic bias of the model. It indicates the tendency and the sign of the deviation. A negative FB value indicates model over-prediction and a positive value, an under-prediction. Air quality modelers do not agree fully upon the magnitude of standards for accepting or rejecting model performance. In most cases, a model is considered acceptable if most of its predictions are within a factor of 2 of the observations (Hanna et al., 1993, 1991). On the other hand, studies by Ahuja (1996), Kumar et al. (1993), Zawar-Reza et al. (2005) report that a model can be deemed acceptable if: NMSE ≤ 0.5, -0.5 ≤ FB ≤ +0.5, and IOA > 0.5. University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 35 2.4 CALPUFF Modeling System The model used for the simulation studies in this research, the CALPUFF modelling system, is described at this point. Many dispersion models typi- cally assume steady, horizontally homogeneous wind fields instantaneously over the entire modeling domain and are usually limited to 50 kilometers from a source. However, for applications with emission source hundreds of kilometers away, other models or modeling systems. At these distances, the transport times are sufficiently long that the mean wind fields cannot be considered steady or homogeneous. CALPUFF is one such modeling system, consisting of three components: CALMET, a meteorological pre- processor that utilizes surface, upper air, and on-site meteorological data to create a three-dimensional wind field and derive boundary layer parameters based on gridded land use data; CALPUFF, a puff dispersion model that can simulate the effects of temporally and spatially varying meteorological conditions on pollutant transport, removes pollutants through dry and wet deposition processes and transforms pollutant species through chemical re- actions; and CALPOST, a postprocessor that takes the hourly estimates from CALPUFF and generates estimates at specified hours as well as tables of maximum values (Scire et al., 2000b). CALPUFF is a transport and dispersion model that advects puffs of mate- rial released from modelled sources. It requires 3-dimensional fields of wind University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 36 and temperature, along with associated 2-dimensional fields such as mixing heights, surface characteristics and dispersion properties. To develop these fields, a deterministic meteorological processor (CALMET) was created. CALMET requires both hourly surface and twice-daily upper-air data to construct the meteorological fields. CALMET cannot forecast meteorology, but has a Diagnostic Wind Module (DWM) that adjusts wind and tem- perature fields due to the influence of terrain and vegetation. There are several switches in the CALMET model that must be set by the modeller to reflect the unique geophysical characteristics within an airshed (Scire et al., 2000b). The model accounts for a variety of effects such as spatial variability of meteorological conditions, causality effects, dry deposition and dispersion over different types of land surfaces, plume fumigation, low wind-speed dispersion, primary pollutant transformation and wet removal. CALPUFF has various algorithms to include the use of turbulence-based dispersion coefficients derived from a similarity theory or observations. The individual components of the modeling system are described in detail in the following sections. 2.4.1 CALMET Diagnostic Meteorological Model CALMET is a diagnostic meteorological model that develops hourly wind and temperature fields on a three-dimensional gridded modelling domain, University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 37 including two-dimensional fields such as mixing height, surface characteris- tics and dispersion properties. The CALMET model operates in a terrain- following vertical coordinate system using Eqn.(2.7): Z = z − ht (2.7) Where Z is the terrain-following vertical coordinate (m), z is the Cartesian vertical coordinate (m) and ht is the terrain height (m). The vertical velocity, W, in the terrain-following coordinate system is de- fined by Eqn.(2.8): ∂ht ∂ht W = w − u − v (2.8) ∂x ∂y Where w is the physical vertical wind component (m/s) in Cartesian coordinates and u , v are the horizontal wind components (m/s). The diagnostic wind field module in CALMET uses a two-step approach in the computation of wind fields (Scire et al., 2000a). In the first step, an initial-guess wind field is adjusted for kinematic effects of terrain, slope University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 38 flows and terrain blocking effects to produce a Step-1 wind field. CALMET parameterizes the kinematic effects of terrain using the approach of Liu and Yocke (1980). The Cartesian vertical velocity is computed by Eqn.(2.9): w = (V · ∇ht)exp(−kz) (2.9) Where V is the domain-mean wind speed, ht is the terrain height, k is a stability-dependent, coefficient of exponential decay and z is the vertical coordinate The exponential decay coefficient, k, increases with increasing atmospheric stability and is given by Eqn.(2.10): N k = (2.10) |V | Where |V | is the speed of the domain-mean wind and N is the Brunt-Vaisala frequency (1/s) and is given by Eqn.(2.11): University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 39 [ ] 1 g dθ 2 N = ( ) (2.11) θ dz Where θ is the potential temperature (K) and g is the acceleration due to gravity (m/s2) The kinematic effects of the terrain on the horizontal wind components are evaluated by applying a divergence-minimization procedure to the initial guess of the wind field. The thermodynamic blocking effects of terrain on the wind flow are param- eterized in terms of the local Froude number given by Eqn.(2.12): V Fr = (2.12) N∆ht Where Fr is the local Froude number, V is the wind speed(m/s) at the grid point, N is the Brunt-Vaisala frequency (1/s) and ∆ht is the effective obstacle height(m) and is given by Eqn.(2.13): ∆ht = (hmax)ij − (z)ijk (2.13) University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 40 Where (hmax)ij is the highest gridded terrain height within a radius of influence of the grid point (i,j) and (z)ijk is the height of level k of grid point (i,j) above the ground. The second step consists of an objective analysis procedure to introduce observational data (surface and upper-level observational) into the step 1 wind field in order to produce the final wind field. An option is provided to allow gridded prognostic wind fields to be used by CALMET, which better represents regional flows and certain aspects of sea breeze circulations. The prognostic data, as a 3D.DAT file, can be introduced into CALMET in three ways: as a replacement for the initial guess wind field, as a replacement for the step 1 field or as observations in the objective analysis procedure. In order to improve the initialization of the diagnostic model, the wind fields from the WRF prognostic model are ingested every hour by CALMET as the initial-guess wind field. This step is expected to improve the models performance by providing equally spaced data points both at the surface and upper levels within the modelling domain where observational data are not available (Chandrasekar et al., 2003). The prognostic winds are interpolated to the fine-scale CALMET grid. The diagnostic module in CALMET then adjusts the initial-guess University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 41 wind field for kinematic effects of terrain, slope flows and terrain blocking effects using fine-scale CALMET terrain data. CALMET reads hourly surface observations of wind speed, temperature, cloud cover, ceiling height, surface pressure, relative humidity, and precip- itation (only if wet removal is to be computed). The twice-daily upper air observations required by CALMET include vertical profiles of wind speed, wind direction, temperature, pressure, and elevation. CALMET also re- quires geophysical data, including gridded fields of terrain elevations and land use categories. Gridded fields of other geophysical parameters, such as the surface roughness length, albedo, Bowen ratio, soil heat flux parameter, anthropogenic heat flux and vegetation leaf area index are also required. CALMET consists of two boundary layer meteorological modules for over- land and over water applications (Scire et al., 2000b). For overland sur- faces, the energy balance method of Holtslag and Van Ulden (1983) is used to compute hourly gridded fields of the heat flux, surface friction veloc- ity, Monin-Obukhov length, and convective velocity scale. Mixing heights are determined from the computed hourly surface heat fluxes and observed temperature soundings. The model also determines the Pasquill-Gifford stability classes and optional hourly precipitation rates. University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 42 2.4.2 CALPUFF Model The basic equation for the contribution of a puff at a receptor by CALPUFF is by Eqn.(2.14): { ( ) ( )} Q −d 2 2a −d C = g exp exp c (2.14) 2πσ σ 2σ2x y x 2σ 2 y Where C is the ground-level concentration (g/m3), Q is the pollutant mass (g), σx, σy and σz are the standard deviations (m) of the Gaussian distribution in the along-wind direction, cross-wind and vertical directions respectively, da and dc are the distances (m) from the puff center to the receptor in the along-wind and cross-wind directions respectively and g is the vertical term (m) of the Gaussian equation. The vertical term, g, is given by Eqn.(2.15): ∞ 2 ∑ [ ] g = exp −(H + 2nh)21 e /(2σ 2z ) (2.15) (2π) 2σz n=−∞ Where He is the effective height (m) above the ground of the puff center and University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 43 h is the mixed-layer height (m). 2.4.2.1 Dispersion The key modelling consideration in CALPUFF is the specification of the horizontal and vertical Gaussian dispersion coefficient, σy and σz, for a puff (or each end of a slug) at the start and end of a sampling step, and also for each receptor at which the cloud has a computed contribution during the step. The coefficients for the puff location at the start of a step are equal to those found at the end of the preceding sampling step, because cloud-size is continuous between sampling steps. Those at the end of the step, or at nearby receptors during the step, are computed according to an ambient turbulence growth relationship and possibly a source-related constant variance. The growth due to ambient turbulence may be formulated as either a function of time, or as a function of distance. Therefore, a generic metric ξ which stands for either one is used (Scire et al., 2000b). The dispersion coefficients for an incremental ’position’ ∆ξ, relative to the beginning of sampling step n are given by Eqn.(2.16): σ2y,n(∆ξy) = σ 2 yt(ξyn + ∆ξy) + σ 2 2 ys + σyb (2.16) University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 44 and Eqn.(2.17): σ2z,n(∆ξz) = σ 2 2 zt(ξzn + ∆ξz) + σzb (2.17) Where ξyn and ξzn are the virtual-source metrics (time;distance) defined implicitly by the requirement that the sigmas match those at the end of the previous step when ∆ξ = 0, σy,n and σz,n are the total horizontal and vertical dispersion coefficients (m) respectively at some position during sampling step n, σyt and σzt are the functional forms of the dispersion coefficients (m) of σy and σz respectively due to atmospheric turbulence. σyb and σzb are the components (m) of σy and σz respectively due to plume buoyancy at the time of release and σys is the component of the horizontal dispersion coefficient (m) due to the lateral (cross-wind) scale of an area-source. The increment ∆ξ is positive when describing the growth of the puff during the sampling step, but can be either positive or negative for receptor-specific sigmas. For example, ∆ξ would be negative for a receptor located just upwind of the puff at the start of a sampling step. This allows CALPUFF to reproduce plume-like features during steady meteorological conditions, University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 45 using very few puffs. Negative ∆ξ could also drive the argument of σyt and σzt through zero if a lower limit on the size of the sigmas at the source were not enforced. So there is an initial ξ0 imposed, defined implicitly by Eqn.(2.18) and Eqn.(2.19): σ2 (ξ ) = σ2yt oy yo (2.18) and σ2 2zt(ξoz) = σzo (2.19) Where σyo and (ξoz) are the initial values (m) of σy and σz respectively due to the nature of the source (e.g., volume source) or the rapid initial dilution associated with building downwash of point sources and ξo is the initial virtual-source metric defined implicitly and separately for y and z. Thus, quadratic addition of initial dispersion components for buoyant rise effects and for the lateral size of an area source are assumed, but other initial cloud dimensions and the subsequent growth of the puff or slug are accomplished using the virtual-distance or virtual-transport time approach. This virtual- source approach is necessary if current puff growth is to be dependent only on the current size of the puff and not on how it reached University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 46 that size. The concept of a virtual source is particularly important when a puff can move between substantially different dispersion regimes in just one sampling step. For example, land use varies by grid cell, so a puff may go from an overwater cell with weak dispersion into an overland cell with substantial vertical convection. Or a young puff in the late afternoon mixed layer may see the turbulence decay rapidly. In both cases, CALPUFF com- putes subsequent growth during the step using the appropriate turbulence (actual or parameterized), and the growth rate appropriate to its size (Scire et al., 2000b). 2.4.2.2 Atmospheric Turbulence Components In the calculation of the atmospheric turbulence components, the basic strategy in the design of the dispersion module is to allow the use of the most refined data available in the calculation of σyt and σzt. For situa- tions in where this data is not available, backup algorithms not requiring specialized data are provided. The five dispersion options are: 1. Dispersion coefficients computed from measured values of turbulence, σv and σw. 2. Dispersion coefficients from internally calculated σv and σw using mi- crometeorological variables. University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 47 3. Pasquill-Gifford(PG) dispersion coefficients for rural areas (computed using the Industrial Source Complex Short-Term (ISCST) multi-segment approximation) and McElroy-Pooler coefficients in urban areas. 4. Same as 3 except PG coefficients are computed using the MESOPUFF II equations. 5. Complex Terrain Dispersion Model(CTDM) sigmas used for stable and neutral conditions (assumes that measured σv and σw are read). For unstable conditions, sigmas are computed as in Dispersion Option 3. 2.4.2.3 Buoyancy-Induced Dispersion The effect of plume buoyancy on the dispersion coefficients are parameter- ized in terms of the plume rise (Irwin, 1983). It is given by Eqn.(2.20): ∆H σyb = σzb = (2.20) 3.5 Where ∆H is the plume rise(m). Buoyancy-induced dispersion (BID) is automatically included for all the dispersion coefficient options described in Section 2.4.2.2. University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 48 The basic point source plume rise relationships are based on the Briggs equations (Briggs, 1975). The plume rise due to buoyancy and momentum during neutral or unstable conditions, zn is calculated using Eqn.(2.21): [ ]1/3 zn = 3Fmx/(β 2 2 2 2 3 j us ) + 3Fx /(2β1 us ) (2.21) Where Fm is the momentum flux (m 4/s2), F is the buoyancy (m4/s3), us is the stack height wind speed (m/s), x is the downwind distance (m), β1 is the neutral entrainment parameter (∼ 0.6), βj is the jet entrainment coefficient (βj = 1/3 + us/w), w is the stack gas exit speed (m/s). During stable conditions, the plume rise, zs is determined by Eqn.(2.22): [ ] z = 3F /(β 2u S1/2 1/3 s m j s ) + 6F/(β 2 2 usS) (2.22) Where β2 is the stable entrainment parameter (∼ 0.36) and University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 49 S is a stability parameter given by Eqn.(2.23): ( )( ) g dθ S = (2.23) Ta dz Where g is the acceleration due to gravity (m/s2), Ta is the ambient temperature (K) and (dθ/dz) is the potential temperature lapse rate (K/m). 2.4.3 CALPOST CALPOST is used to process output files from CALMET and CALPUFF, producing tabulations that summarize the results of the simulation, identi- fying the highest and second highest 3-hour average concentrations at each receptor for example. Fig. 2.5 is the schematic diagram of the CALPUFF system. The application of CALMET/CALPUFF modelling system is well known, and several validation tests have been published (Dios et al., 2013, Dresser and Huizer, 2011, Fishwick and Scorgie, 2011, Ghannam and El-Fadel, 2013, Hernández et al., 2014, Levy et al., 2003, Protonotariou et al., 2004). Most of them were based on specific experiments with passive tracers and a large compilation of surface and aloft meteorological measurements during University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 50 Figure 2.5: A schematic diagram of the program elements in the CALMET/CALPUFF modelling (Scire et al., 2000b) the experiments, in order to achieve the best model performance evalua- tion. However, with actual pollutants sources and limited meteorological datasets, uncertainties arise (both in measurements and models results) and worse models performance is expected. University of Ghana http://ugspace.ug.edu.gh Chapter 2. Literature Review 51 2.4.4 Weather Research and Forecasting (WRF) Model The WRF model is a fully compressible, non-hydrostatic mesoscale model with a hydrostatic option which uses a terrain-following hybrid sigma- pressure vertical coordinate in its meteorology simulation (Janjic, 2003, Janjić, 2002). It was developed in a collaborative effort by the National Center for Atmospheric Research (NCAR), the National Centers for En- vironmental Prediction (NCEP), the Forecast Systems Laboratory (FSL), the Air Force Weather Agency (AFWA), Oklahoma University (OU) and other university scientists. It is a state-of-the-art numerical weather pre- diction (NWP) and data assimilation system suitable for horizontal grid scales in the 1 to 10 km range. The model can be run in a nested way with the outer domain on a regional scale, covering distances usually between 500 1,000 km (Michalakes et al., 2004, Shamarock et al., 2008). University of Ghana http://ugspace.ug.edu.gh Chapter 3 Methodology 3.1 Introduction This Chapter describes the study area for the research and the calculations used in the estimation of the emissions from the Tema Oil Refinery. The methodology employed in the meteorological and dispersion simulations are also presented. The mass balance method is employed in the estimation of pollutants from the RFCC unit of the refinery due to the unavailability of such data from the refinery. Based on these calculated rates, emissions from the whole refinery are also derived. The flare and flue stack are also characterised by parameters such as stack/flare height and diameter, emis- sion escape velocity and discharge temperature. The methodology used in validating the CALPUFF/CALMET simulation results are also presented. 52 University of Ghana http://ugspace.ug.edu.gh Chapter 3. Methodology 53 3.1.1 The Study Area The study area, capturing Ghana’s only refinery in Tema, the emission source for this study and spaning a domain size of 60 km × 60 km, is lo- cated in the Greater Accra Region of Ghana stretching along the Ghanaian Atlantic coast and extending a bit north into Ghana's interior as shown by Fig.3.1. The Universal Transverse Mercator (UTM) coordinate of the southwestern corner of the domain in zone 30 are 795 km easting and 600 km northing. The study area covers most part of the Accra Metropoli- tan area, the Tema municipal area and extends to parts of the Eastern and Volta regions. The southern boundary of the metropolis of Accra which coincides with that of the study area is the Gulf of Guinea. Tema is the most industrialised city in Ghana on the Gulf of Guinea and Atlantic coast of Ghana. It is located 25 km east of the capital city, Accra, in the region of Greater Accra with a population of approximately 402,637 people (G.S.S, 2012). Tema is locally nicknamed the ’Harbour Town’ be- cause of its status as Ghana’s largest seaport. It is a major trading center, home to Ghana’s only oil refinery and other numerous factories, and is linked to Accra by a highway and railway. The Greater Accra Region is located in the Dahomeyan Gap, where the coast runs parallel to the prevailing moist monsoonal winds. It features a tropical savanna climate and a savanna vegetation that borders on a University of Ghana http://ugspace.ug.edu.gh Chapter 3. Methodology 54 Figure 3.1: Map of Study Area showing the Tema Oil Refinery University of Ghana http://ugspace.ug.edu.gh Chapter 3. Methodology 55 semi-arid climate. Accra experiences two rainy seasons: April to June and September to October with the former being the heaviest. The average annual rainfall is about 730 mm which falls primarily during Ghana's two rainy seasons. Variation in temperature throughout the year is recorded to be nominal with mean monthly temperature ranging from 24.7◦C in August to 28◦C in March and an annual average of 26.8◦C. Relative hu- midity is generally high, varying from 65% in the mid afternoons to 95% at night. Accra experiences a breezy dry heat during windy harmattan seasons usually in December and January (Muff and Efa, 2006). The landscape is mainly flat with terrain elevations between 200 m and 500 m, except for the low mountain system (Akropong range north of the refinery and the Shai hills) east of the refinery. The traditional land use clas- sification within the study area includes built-up/urbanisation, forestry/a- griculture, water, shoreline, protected areas and bare areas inferred from Addo (2013). 3.1.2 Tema Oil Refinery Tema Oil Refinery (TOR) Limited, formerly Ghanaian Italian Petroleum (GHAIP) Limited was commissioned on December 12, 1963. The Gov- ernment of Ghana became the sole shareholder in 1977 and renamed the company Tema Oil Refinery Limited in 1991. The Refinery predominantly University of Ghana http://ugspace.ug.edu.gh Chapter 3. Methodology 56 processes crude oil into refined products like LPG, Gasoline, Jet Fuel, Kerosene, Diesel and Fuel Oils. The refinery is situated on an area of 440,000 square meters located in the industrial hub of the Greater Accra Re- gion. The refinery complex, as a whole, contains the refinery itself, its tank farm containing storage tanks, wastewater treatment plant, gas turbine power generating sets and its utility system of boilers. The refinery has an installed total capacity of 65 000 barrels per stream day(bpsd) equivalent to 3,000,000 tonnes per annum (tpa) with contributions from the Crude Distil- lation Unit (CDU)- 2,000,000 tpa, Residue Fluid Catalytic Cracker(RFCC)- 685,000 tpa and the Premium Reforming unit(PRF)-315,000 tpa (Darko and Quist, 2012). The RFCC process converts heavy crude oil fractions into lighter, more valuable hydrocarbon products at high temperature and moderate pressure in the presence of finely divided silica/alumina based catalyst in crude refineries. In the course of cracking the large hydrocarbon molecules into smaller molecules, a non-volatile carbonaceous material, commonly referred to as coke, is deposited on the catalyst. The presence of coke on the catalyst deactivates the catalytic cracking activity of the catalyst by blocking access to the active catalytic sites. In order to regenerate the catalytic activity of the catalyst, the coke deposited on the catalyst is burned off with air in the regenerator vessel. The regenerator is normally operated at conditions that achieve complete combustion of C to CO2. Flue gas exits the cyclone University of Ghana http://ugspace.ug.edu.gh Chapter 3. Methodology 57 separators to minimize catalyst entrainment prior to discharge from the regenerator through the flue stack (Maya-Yescas et al., 2005). Other stack emissions include SOx, NOx and unburnt hydrocarbons. The refinery is also equipped with a flare which primarily burns off flammable gas released by pressure relief valves during unplanned over-pressuring of plant equipment. It is used both for safety reasons during process upsets (start-up, shut down, system blow-down) and for managing the disposal of waste gases with hydrocarbons from routine operations. However, through combustion, the flare also produces other pollutants including SOx, NOx, CO, CO2 and PM (both total PM and micro-particulates- ie, PM10 and PM2.5). Infact, according to USEPA (2011), it is one of the air pollution sources in a refinery emission inventory followed by the flue stack. It is for this reason that other emissions from stationary combustion sources, including process heaters, boilers, combustion turbines and similar devices, are not calculated but extrapolated, in this study. 3.2 Tema Oil Refinery Emission Estimation The air emissions from the refinery are mainly due to the combustion of fuel gas. In this work, the pollutants of interest are SOx, NOx, CO2, PM and VOCs and the point sources under consideration are the flue stack and the flare. As mentioned early on, the mass balance approach is employed University of Ghana http://ugspace.ug.edu.gh Chapter 3. Methodology 58 Table 3.1: Operational Average Flow Parameters of the RFCCU of the Tema Oil Refinery for 2008 - 2013 Year Feed Rate Airflow Rate Fuel Gas rate Initial Catalyst (m3/hr) (Nm3/hr) (kg/hr) Inventory(kg/hr) 2008 78.83 70989.00 6248.93 120000 2009 72.02 63882.88 6239.00 120000 2010 80.11 63147.37 6858.40 120000 2011 73.44 56954.83 6368.20 120000 2012 74.71 55508.67 5853.17 120000 2013 74.44 55643.80 5207.14 120000 during which the compound being combusted is examined to predict the emissions. The gases involved in the calculations were assumed to be ideal and so the idea gas laws were applied. For example, the flue stack gas contains products of the combustion of deposited carbon on the catalyst during cracking. Based on the stoichio- metric equations of the combustion reactions, products are estimated given the reactant concentrations. In the case of the flare, emissions rates are in- formed by a gas chromatographic analysis of the fuel gas exiting the flare. Table 3.1 shows the quantities of some parameters used in the calculations. 3.2.1 Estimation of Flue Stack Gas Rate and Com- position In the regeneration of the catalyst in the regenerator through combustion, air is used as a source of oxygen. The carbon reacts to form CO2 upon com- plete combustion as is the case with this refinery. Based on the regenerator University of Ghana http://ugspace.ug.edu.gh Chapter 3. Methodology 59 air flow rate, therefore, the emission rates of CO2 and carbon particulates are estimated. The air flow rate provided by the refinery is on a wet basis and was converted to dry basis prior to the calculations. The larger component of the total FCC PM emitted is as a result of cyclone inefficiency estimated at 0.25% (Harding et al., 2001, Jiménez-Garćıa et al., 2011) which is what is used in the calculations. Research also shows that PM2.5 is a large fraction of the total particulate emissions even without consideration of primary sulphate or organic emissions (England et al., 2000, Hildemann et al., 1994). Therefore, PM modelled in this study is considered to be of aerodynamic diameter of 2.5. According to Corma et al. (1997), in the FCC unit, between 45 and 55% of the total sulphur present in the feed, especially the nontiophenic com- pounds, is converted to H2S, while approximately 35 to 45% stays in the liquid products as sulphur compounds. The remaining 5% is left on the catalyst as part of coke and results in 90% SO2 and 10% SO3 during the coke combustion. The TOR processes mainly light sweet and light sour crude with % feed sulphur between 0.5 - 1.5 (Darko and Quist, 2012, Prod- ucts, 1997, Zhao et al., 1997). In this study, an average of 1% feed sulphur is used. Elemental nitrogen typically equals 0.32% of the FCC feed, 15% of which ends up on the catalyst together with the coke. Ninety percent (90%) of University of Ghana http://ugspace.ug.edu.gh Chapter 3. Methodology 60 this is converted to NO and the remaining 10% to NO2 and ends up in the flue gas (Jiménez-Garćıa et al., 2011, Zhao et al., 1997). A sample of the calculations, using flow parameters for the year 2009, are shown in Appendix A. 3.2.2 Estimation of Flared Gas Composition When fuel is burnt, carbon in the fuel reacts to form either CO2 or CO, hydrogen forms H2O, sulphur forms SO2 and SO3 and some of the nitrogen in the air forms NO2 at high temperatures (Abdul-Wahab, 2003, Corma et al., 1997). However, when flaring is done in the presence of excess air, CO is not formed which is the case in this study. Small amounts of NO2 are formed at temperatures greater than 1400◦C. Since this temperature is greater than the flare temperature, it is also assumed that no NO2 is produced in the flare. The flare emissions therefore contain H2, O2, N2, CO2, SO2 and some unburnt hydrocarbons hereafter referred to as volatile organic compounds (VOCs). VOCs are present in flare emissions as a re- sult of flare efficiency (destruction efficiency) given as 98% (TOR Manual). Flares are designed to operate at 98 - 99% (Al-Fadhli et al., 2011, Mart́ın et al., 2003, Wood et al., 2012). In this methodology, the composition of the flare gas combustion products is calculated through atomic and molec- ular balances. At constant temperature and pressure, mole % of gases is University of Ghana http://ugspace.ug.edu.gh Chapter 3. Methodology 61 equal to volume %. Therefore from the GC analysis report provided by the quality control laboratory of the refinery, the mole fractions of the flare gas components are calculated as shown in Appendix A. Given the average molecular weight of the fuel gas of 20 kg/kmol (Zadakbar et al. (2008)) and its mass flow rate, the total moles of the fuel gas mixture are also calculated. From this, the moles of the components are estimated using their respective fractions. Based on the stoichiometry of the combustion reactions, the combustion products are obtained: CO2, H2O, SO2 and N2. SO2 and VOCs are the flare pollutants considered. Again, the estimation is carried out using operational data for 2009 on an hourly basis and is shown in Appendix A. Total refinery emissions are then estimated as percentages of the RFCC emissions (EIPPCB, 2012). For the sake of simplicity, SO2 and SO3 from the flue stack are added and considered as SO2. Similarly, NO and NO2 are also lumped together and modelled as NO2. 3.2.3 Calculation of Flare and Flue stack Exit Gas Velocities The flare and flue gas velocities are calculated by dividing their volumetric flow rates by the cross-sectional areas (obtained from their diameters) of University of Ghana http://ugspace.ug.edu.gh Chapter 3. Methodology 62 the flare stack and flue stack respectively. The calculations are shown in Appendix A. 3.2.4 Modelling Period A minimum of a year is necessary to capture the time-varying emissions and meteorology during a full cycle of seasons. However, to be able to do a trend analysis to inform future predictions, a six-year period from January 2008 to December 2013, was considered in the study. 3.2.5 Model Set-up As stated earlier in this section, the key program elements and data in- put requirements for CALPUFF are meteorological wind fields, emissions data, receptor locations (points in the modeling domain where the model will predict concentrations), and CALPUFF modeling parameters. These elements are described below. 3.2.5.1 CALMET Modelling Due to the unavailability of complete observational meteorological datasets from the local weather stations, the CALMET model was initialised with WRF data to develop the meteorological field for CALPUFF. This option is University of Ghana http://ugspace.ug.edu.gh Chapter 3. Methodology 63 available in the CALMET model. Because the prognostic models like WRF are usually run over very large domains with much coarser resolutions, they are not suitable for direct use in CALPUFF. In this case, CALMET will interpolate the gridded prognostic data to develop a 3-D fine scale field of wind speeds, directions and other parameters. The WRF model runs were performed at the International Centre for The- oretical Physics (ICTP), Trieste, Italy. Two one-way nested computational domains were set with horizontal grid spacing of 24 and 12 km represent- ing horizontal grid dimensions of outer and inner domains respectively as shown in Fig.3.2. The vertical grid contained 29 full sigma levels. The inner domain covered the whole of Ghana within which the study area is located. The model was run with Global Forecast System (GFS) 1 degree reanalysis as initial and boundary conditions with full physics processes included in order to reproduce real meteorological phenomena. Other options include WRF Single Momentum (WSM) 6 Class Graupel for Microphysics Scheme, Rapid Radiative Transfer Model (RRTM) scheme for long wave radiation, Dudhia scheme for short wave radiation, Monin-Obukhov scheme for sur- face layer, and Noah model (5 soil layers) for land-surface interactions, while the dynamic options were left as default to produce hourly averages (Skamarock et al., 2005). University of Ghana http://ugspace.ug.edu.gh Chapter 3. Methodology 64 Figure 3.2: Nested Computational domains in the WRF simulations CALMET-compatible 3D.DAT files containing data of horizontal and ver- tical velocity components, pressure, temperature and precipitation among others were then created by running CALWRF on the WRF output data. The CALMET meteorological processor was then used to interpolate the 12 km resolution WRF winds to the CALPUFF domain. University of Ghana http://ugspace.ug.edu.gh Chapter 3. Methodology 65 A horizontal grid spacing of 1 km was selected for the CALMET simula- tion; the study area therefore corresponds to 60 rows by 60 columns. With this grid spacing, it was possible to maximize run time and file size efficien- cies while still capturing large-scale terrain feature influences on wind flow patterns. CALMET uses fewer vertical layers than WRF, in part because air pollution modeling does not require detailed information on the upper atmosphere. Therefore to adequately characterize the vertical structure of the atmospheric boundary layer (i.e., the layer within about 2000 metres above the Earth's surface), ten vertical layers with cell face heights at 0, 20, 40, 80, 160, 320, 640, 1000, 1500, 2200 and 3000 m were used in the CALMET runs. CALMET defines a vertical layer as the midpoint between two faces (i.e., 11 faces correspond to 10 layers, with the lowest layer always being ground level or 10 m). CALMET automatically interpolated from the WRF model grid system to the CALMET grid (Scire et al., 2000a, Zhou et al., 2003). 3.2.5.2 Geophysical Data Input Prior to running CALMET, a geophysical file containing information about the land use type, elevation and surface parameters about the domain un- der study was created. Due to the unavailability of terrain elevation and land use land cover (LULC) data in the required format for Ghana from the local agency in the required format, data from the United States Geological University of Ghana http://ugspace.ug.edu.gh Chapter 3. Methodology 66 Survey (USGS) web site at http://edcwww.cr.usgs.gov/doc/edchome/ndcdb/ndcdb.html was used. Detailed terrain features were incorporated using the USGS global 3 arc-sec SRTM3 data (∼ 90m resolution). Elevations are in meters relative to mean sea level. The spacing of the elevations along each profile is 3 arc-seconds, which corresponds to a spacing of approximately 90 meters. A vertical and horizontal grid spacing of 1km was selected to adequately represent the important terrain features. The raw terrain data were processed into each gridded field. These terrain fields effectively resolve major land features in the model domain. Land use and land cover (LULC) data were downloaded from the USGS at the 1:250,000-scale with file names corresponding to the 1:250,000-scale map names. Land use data were processed to produce a 1-km resolution gridded field of fractional land use categories and land use weighted values of surface and vegetation properties for each CALMET grid cell. The 14 default CALPUFF land use categories were used (10, 20, 30, 40, 51, 54, 55, 60, 61, 62, 70, 80, and 90). Each land use type has six assigned values, one each for surface roughness, albedo, Bowen ratio, soil heat flux parameter, anthropogenic heat flux, and leaf area index. The surface roughness is a measurement of the average vertical relief and small-scale irregularities of the terrain surface. The surface roughness is low for the areas in the study area classified as water bodies and is high for the areas classified as urban or University of Ghana http://ugspace.ug.edu.gh Chapter 3. Methodology 67 built-up land and forest. The albedo is the fraction of light that is reflected by a given surface. For instance, the albedo of a snow-covered area is much higher than that of a paved parking lot. The leaf area index is the ratio of the total area of all leaves on a plant to the area of ground covered by the plant. All areas that have the potential to have plant growth have leaf area indexes assigned to them. A value of zero is assigned to the water bodies. The leaf area index is as high as seven (7) in the areas identified as forest land. Terrain and land Use data for the modelling domain were then processed through the TERREL and CTGPRO CALMET pre-processors respectively. Since the modeling domain spans both over-water and over-land regions, the global self-consistent hierarchical high-resolution shoreline (GSHHS) was used to define the coastal lines. The CALMET GEO file was then cre- ated using the MAKEGEO pre-processor which combines data from TER- REL and CTGPROG. CALMET was configured with the default wind field options and parame- ters. University of Ghana http://ugspace.ug.edu.gh Chapter 3. Methodology 68 3.2.5.3 CALPUFF modelling Even though emissions of five species were estimated, only three are mod- elled, namely SO2, NO2 and PM2.5. The CALPUFF model predicts concen- trations at specific points, or receptors, which are established by the user within the modelling domain. Thirty-eight (38) receptors, whose locations are given in Table 3.2 representing populated areas, schools, a hospital and universities were identified within the domain. In addition, receptors were also located at one ambient monitoring site in order to facilitate the model- to-monitor comparisons that are part of the model validation process. For modelling purposes, constant emission rates of the pollutants were as- sumed. The refinery was assumed to process crude throughout the whole year. The model determines average concentration at each of the approxi- mately 38 modelled receptors for each of the 366/365 days within the years considered in this research. Due to data file size limitations, CALMET was executed for four quaters of each year thus generating four CALMET.dat files per year. The outputs of the CALMET simulation were used as inputs for CALPUFF. CALPUFF was also then executed for all the years under consideration. The PRTMET, a post-processor of CALMET, was used to analyze and display of meteorological parameters from the binary CALMET output file. The CALPOST was used to process the files from CALPUFF, producing University of Ghana http://ugspace.ug.edu.gh Chapter 3. Methodology 69 Table 3.2: Receptor Locations in the Study Area Receptor Code Easting(km) Northing(km) Tema Oil Refinery TOR 831.699 627.210 Tema Steelworks TSW 832.215 629.955 Ashiaman ASH 830.287 630.477 Central University CU 841.410 638.219 Dawenya DAW 837.148 637.054 Sebrepor SEB 831.172 633.778 Prampam PRAM 843.381 632.814 Tema Newtown TNT 835.168 625.508 Kpone KP 837.970 630.162 Tema Fishing Harbour TFH 834.227 624.685 Tema General Hospital TGH 829.622 628.685 Tema Community 4 TC4 831.463 625.952 Tema Community 7 TC7 831.175 626.749 Tema Community 10 TC10 829.664 626.435 Afienya AFNY 832.542 641.534 Akropong AKR 822.313 661.283 Dodowa DDW 821.429 651.073 Oyibi OYB 819.252 643.347 New Ningo NNG 850.317 635.186 Old Ningo ONG 852.617 636.861 Kpeshie KPSH 820.829 619.060 Teshie TSH 820.860 617.597 Adenta AD 814.697 632.409 Madina MD 813.895 628.998 University of Ghana UG 811.117 625.613 Sakumono SAK 826.715 623.704 Sege SEG 866.159 651.299 Shai Hills ShHL 839.477 654.0509 Odumasi Krobo OdKR 837.889 659.386 Larteh LT 823.382 656.945 Volta River VR 869.021 669.86 Osu OSU 813.070 614.740 Tema Community 20 TC20 824.261 625.726 Tema Community 19 TC19 825.599 625.599 Tema Community 18 TC18 824.329 625.120 Tema Community 25 TC25 834.031 633.978 Kotoka Int. Airport KIA 813.748 620.858 Heavy Industrial Area HIA 835.469 628.393 University of Ghana http://ugspace.ug.edu.gh Chapter 3. Methodology 70 a summary of simulation results in tabulated as well as in gridded form CALPOST was also used to produce wind roses from the CALMET and WRF output data. 3.2.5.4 Model Evaluation The most commonly used measures of model performance are the statistics recommended in the U.S.EPA 1992 modeling guidance. To determine the reliability of the simulation data, verification of simulated values using the WRF and CALMET models was conducted for surface wind speed and direction. Datasets from the Tema and Accra meteorological stations were used. Even though there are other stations located within the study do- main, many of them are not functional. Therefore, only the two stations mentioned were used. Due to the large number of missing data of the other surface wind parameters, they could not be used for validation. Only wind speed and direction were validated. To validate the dispersion model, CALPUFF, measured data on SO2, NO2 emissions made from the 10 - 23 January, 2008 were used. The Differen- tial Optical Absorption System (DOAS) used for the measurements was installed on the premises of the Tema Oil refinery to monitor and measure University of Ghana http://ugspace.ug.edu.gh Chapter 3. Methodology 71 the concentration of gases emitted into the atmosphere due to its activi- ties and also due to the operations of other industries within its environs. Details of the set-up and results can be obtained in Sackey (2012). The statistical verification of model performance in this study was per- formed using four statistical indicators including the Index of Agreement (IOA), Fractional Bias (FB), Normalized Mean Square Error (NMSE), and Pearson correlation coefficient (R). The formulas used to derive these five indicators are given in :Eqn. 3.1 - Eqn. 3.3. ∑N ( )( ) Coi − Co Cpi − Cp R = i=1 √ (3.1) σoσp Co − Cp FB = ( ) (3.2) 0.5 Co + Cp ( )2 Co − Cp NMSE = (3.3) CoCp ∑N ( )2 Cpi − Coi IOA = 1−∑(∣ i=1∣ ∣∣ ∣∣ ∣∣ ) (3.4)N 2Cpi − Co + Coi − Cp i=1 University of Ghana http://ugspace.ug.edu.gh Chapter 3. Methodology 72 where Co is the observed quantity, Cp the predicted quantity, Coi and Cpi are the observed and predicted quantities respectively for N cases. University of Ghana http://ugspace.ug.edu.gh Chapter 4 Results and Discussions 4.1 Introduction This Chapter presents results of the estimated refinery emission rates and other parameters used for the dispersion simulations. Both the meteoro- logical and dispersion simulation results are also presented and discussed as well as their validation. 4.2 Refinery Emissions and Interannual Trends Plots of estimated emission rates of CO2, VOCs, PM2.5, SO2 and NO2 and their averages for 2008 - 2013 are shown in Figs.4.1 - Fig.4.5 respectively. It can be noticed that total CO2 emissions were the highest of the pollutants estimated from the Tema Oil Refinery ranging from about 130 to 165 t/hr. 73 University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 74 Figure 4.1: Interannual Variation of CO2 Emission Rates from the Tema Oil Refinery Figure 4.2: Interannual Variation of VOCs Emission Rates from the Tema Oil Refinery The VOCs are next in line with emission rates ranging between 2083 and 2500 kg/hr. Particulate matter rates were observed to decline marginally over the years with an average rate of 1032 kg/hr followed by SO2 and NO2 with average rates of 776 and 371 kg/hr respectively. CO2 emissions were highest in 2008, dipped in 2009, increased again in 2010 after which a steady decrease over the remaining years is observed. The highest emissions of PM2.5 also occured in 2008 followed by a marginal University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 75 Figure 4.3: Interannual Variation of PM2.5 Emission Rates from the Tema Oil Refinery Figure 4.4: Interannual Variation of SO2 Emission Rates from the Tema Oil Refinery steady decline over the years. However, the other emissions peaked in 2010 and declined steadily over the following years. The decline in emission rates was a result of a reduction in production due to the frequent shortfalls in the refinery’s crude oil supply, consequently leading virtually to a halt of refinery operations in 2014 preceded by intermittent shutdowns. University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 76 Figure 4.5: Interannual Variation of NO2 Emission Rates from the Tema Oil Refinery 4.3 Preliminary Dispersion Simulation Geophysical inputs, terrain data (with receptor locations relative to the refinery) and Land Use data, are given Figs.4.6 and 4.7 respectively. The most prominent topographical feature is the Gulf of Guinea that forms the southern boundary of the study area. The other topographical feature with elevated terrain, between 300 - 600 m, is the Akropong-range located at the north-western corner of the study area as shown in Fig.4.6. Land use within the study area is fairly complex, varying from open ocean to the south, marshy and wetlands along the coast to the urban areas of Osu (Fig.4.7). The built-up Tema community is not captured probably due to poor resolution. In addition, there are large tracts of agricultural land also present in the study area. University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 77 Figure 4.6: Terrain Map of the Study area showing receptor loca- tions and the Refinery (red square) Figure 4.7: LandUse Map of the Study Area University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 78 4.3.1 Spatial Variation of Pollutants Results of the preliminary simulations of the dispersion of the refinery emis- sions in the study domain are presented in the form of daily average con- centrations of the modelled emissions at the indicated receptors respec- tively. Considering the receptor locations within the study area, as shown in Fig.4.6, it was observed that predicted daily average pollutant concen- trations were high at receptors closest to the refinery and low at receptors far away from the refinery. This can be explained by the dispersion equa- tion, Eqn.(2.14), which expresses an exponential decay of emissions with distance from emission source. Considering receptor locations relative to the refinery, predicted concen- trations were noted to be high at northern receptors as seen in Fig.4.8. These receptors include Tema Steelworks, Ashiaman and Sebrepor. This was followed by receptors on the north-eastern side up to about 10 km of the refinery including Heavy Industrial Area, Dawenya, Kpone, Prampam, Old and New Ningo and the Central University as shown in Fig.4.9. Tema Newtown and the Tema Fishing Harbour located south east of the refin- ery also receive significant emissions followed by south western receptors including Tema Communities 4, 7 and 10 and the Tema General Hospi- tal which are located within a distance of about 4 km of the refinery as shown in Figs.4.10 and 4.11. Receptors located on the north western part University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 79 Figure 4.8: Predicted Daily Average Concentrations of SO2, NO2 and PM2.5 at northern receptors in the Study area Figure 4.9: Predicted Daily Average Concentrations of SO2, NO2 and PM2.5 at north eastern receptors in the Study area of the refinery are the least impacted by the refinery emissions as they showed very low emission concentrations. These include Madina, Adenta, the University of Ghana, Oyibi, Larteh, Dodowa and Akropong. University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 80 Figure 4.10: Predicted Daily Average Concentrations of SO2, NO2 and PM2.5 at south eastern receptors in the Study area Figure 4.11: Predicted Daily Average Concentrations of SO2, NO2 and PM2.5 at south western receptors in the Study area University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 81 Figure 4.12: Predicted Daily Average Concentrations of SO2, NO2 and PM2.5 at northern western receptors in the Study area The prevailing wind directions during the year under study were south- westerly (87%) of total winds as depicted by Fig.4.13. Therefore, receptors located on the north and north-eastern parts of the refinery show relatively higher emission concentrations than other parts of the study domain be- cause the south-westerly winds carry the pollutant clouds from the refinery towards them. Approximately 77% of winds in the study area are clustered in 3rd wind speed classes (3.3 - 5.4 m/s) with winds in the next upper class of (5.4 - 8.5 m/s) being 19%. Winds from 1.8 - 3.3 m/s account for 8% of total winds. The average wind speed is estimated at 3.45 m/s with calm winds forming 0.057% of total winds. Based on these preliminary results, it can be concluded that emissions from the refinery impact communities located within a radius of about 10-15 km around the refinery. Consequently, subsequent simulations of emissions of University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 82 Figure 4.13: Wind Rose Depicting Surface Winds in Tema for 2008 the other years will focus on the receptors which fall within this radius of influence. 4.4 Validation of the CALPUFF Model At this point, results of the performance evaluation of CALPUFF, are presented and discussed. Plot of the comparison between measured and modelled SO2 and NO2 are seen in Fig.4.14 and Fig.4.15 respectively. The performance assessment of the model, based on direct quantitative comparisons of observed and predicted mean concentrations as seen in University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 83 Measured Modelled 50 45 40 35 30 25 20 15 10 5 0 10 11 12 13 14 15 16 17 18 19 20 21 Day of Month Figure 4.14: Plots of measured and modelled SO2 Concentrations Measured Modelled 25 20 15 10 5 0 10 11 12 13 14 15 16 17 18 19 20 21 Day of Month Figure 4.15: Plots of measured and modelled NO2 Concentrations Concentration (µg/m3) Concentration (µg/m3) University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 84 Figs.4.14 and 4.15, reveals that CALPUFF predictions for SO2 were bet- ter than NO2 concentrations. From the 10 - 12, 17 - 19 and 21 January, predicted SO2 concentrations closely approached measurements. Quanti- tative agreement between predicted and modelled NO2 concentrations is excellent on 13 January and good for 19 - 21. However, from 13-16 Jan- uary, CALPUFF significantly under-predicted the measured values. De- spite these differences between predicted and measured values on some days, the trends in the measurements are accurately predicted especially for SO2. The simulation results did not take into account SO2 and NO2 background concentrations because this data was not available. This may explain the reason for this under-prediction. The measured concentrations are likely to include emissions from other sources which are not considered in the simulations. Another important reason could be the assumption made that emission rates from the refinery are constant. This is unlikely to be representative of actual operating conditions at the refinery. It was, however, necessary to make this assumption using a constant estimated average emission rates for modeling purposes. Additionally, possible errors associated with the measurements of SO2 and NO2 at the measurement location cannot be overlooked. Results of model performance evaluation after applying the USEPA guide- lines pertaining to model evaluation protocol are presented in Table 4.1. The indices are the correlation coefficient (R), the index of agreement University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 85 (IOA), the normalised mean square error (NMSE) and the fractional bias (FB). Table 4.1: Statistical Performance Indices of the CALPUFF model Pollutant R NMSE IOA FB SO2 0.66 0.73 0.39 0.41 NO2 -0.25 0.36 1.34 0.36 The index of agreement between predicted concentrations and measure- ments is better for SO2 and NO2. The correlation coefficient of SO2 with observations is satisfactory but that for NO2 is negatively weak as shown in Table 4.1. The NO2 emissions from automobiles make up a large frac- tion of ambient NO2 levels. It is therefore probable that a large fraction of the measurements consists of contributions from the vehicular emission, re- vealed in the weak correlation and the poor index of agreement. CALPUFF under-predicts NO2 by a smaller factor than SO2 as revealed by their FB values. The NMSE values for SO2 is acceptable whiles that of NO2 is not. Therefore, based on these indices, the performance of CALPUFF can be described as good. University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 86 4.5 Validation of CALMET and WRF Models Results obtained from CALMET and WRF simulations are compared with available surface observations (wind speed and direction) for January 2008 from the Tema meteorological station (TMS) as shown in Figs.4.16 and 4.17. Wind roses generated from the models for the same period are also presented in Figs.4.18 and 4.19. Figure 4.16: Plots of Observed and Modelled Wind Speeds It is apparent from Fig.4.16 that WRF and CALMET models generally under-predicted wind speeds from January 9 - 14 and over-predicted speeds on January 16 - 24. However, on days 11, 16, 21 and 23, CALMET predic- tions closely approach observations whiles WRF predicts similar speed on University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 87 Figure 4.17: Plots of Observed and Modelled Wind Direction the 15th January. Fig. 4.17 reveals the over-prediction of wind direction by both models for most part of the period considered. From the other sta- tistical performance measures given in Table 4.2, it can be observed that both CALMET and WRF share similar indices rating their performances on the same scale with CALMET indicating slightly higher correlation co- efficients for both wind direction and speed than WRF. Under-prediction by CALMET is greater than by WRF as seen in the FB values. Table 4.2: Statistical Performance Indices of CALMET and WRF models CALMET WRF Parameter R NMSE IOA FB R NMSE IOA FB Speed 0.58 0.66 0.24 0.34 0.57 0.65 0.09 0.15 Direction 0.29 0.66 0.02 -0.10 0.28 0.66 0.02 -0.10 University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 88 Figure 4.18: Wind Rose Depicting CALMET Surface Winds The wind rose generated from CALMET (Fig.4.18) indicated the predom- inant prevailing wind is south south-westerly (26%) whiles southerly and southwesterly represent about 26%. The remaining winds are spread across a wide spectrum ranging from westerly to north north-easterly. WRF wind rose (Fig.4.19) predicted an increased percentage of south south-westerly winds (31%) than CALMET. It also indicates 12% of southerly winds and 12% of south-westerly winds which is lower than CALMET. It is clear that the percentages of the other winds are also lower in the WRF wind rose than University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 89 Figure 4.19: Wind Rose Depicting WRF Surface Winds the CALMET. The predominant wind speed class predicted by both models is 3.3 - 5.4 m/s with CALMET predicting a higher percentage (75%) than WRF (64%). Thirty percent (30%) of the total winds have speeds between 5.4 - 8.5 m/s according to WRF whiles CALMET predicts only 3% of that class. Generally, average wind speeds is higher for WRF than CALMET. Calm winds in both modelled wind (>0.5 m/s) account for only 0.28% of the data. Based on the statistics, it appears that CALMET predictions better approximates observations as it is able to capture small scale effects University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 90 which cannot be accounted for by WRF. Numerical meteorological models tend to have spatial and temporal phase errors in simulating surface wind. This is because the lowest-level WRF winds must be extrapolated down to ground-level to initialize CALMET. Thus unless the vertical resolution is quite fine, it would be expected that the near-surface CALMET output winds will be biased towards the trends seen in the lowest level of the WRF data, which has a higher elevation in the region compared to the surface stations (Protonotariou et al., 2004). On this basis, the remaining valida- tion results for only CALMET with data from TMS and AMS for 2008 - 2013 are presented in Table 4.3. Table 4.3: Statistical Performance Indices of CALMET and Obser- vations from the TMS and AMS TMS AMS Year Parameter R NMSE IOA FB R NMSE IOA FB 2008 Speed 0.58 0.24 0.66 0.34 0.56 0.88 0.72 0.08 Direction 0.29 0.02 0.66 -0.10 0.57 0.02 0.71 -0.14 2009 Speed 0.48 0.12 0.62 0.17 0.52 0.99 0.72 0.85 Direction 0.07 0.02 0.72 -0.12 -0.01 0.02 0.72 -0.14 2010 Speed 0.32 0.17 0.60 0.24 0.64 0.80 0.72 0.77 Direction 0.06 0.01 0.68 -0.09 0.03 0.02 0.69 -0.11 2011 Speed 0.31 0.10 0.54 0.09 0.59 0.78 0.72 0.76 Direction -0.02 0.01 0.69 -0.09 0.22 0.01 0.71 -0.11 2012 Speed 0.34 0.12 0.60 0.20 0.47 1.40 0.73 0.98 Direction 0.03 0.01 0.72 -0.10 0.35 0.02 0.72 -0.11 2013 Speed 0.45 0.16 0.64 0.28 0.61 1.33 0.73 0.96 Direction 0.03 0.01 0.70 -0.1 -0.01 0.03 0.69 -0.14 From Figs.4.20 - 4.31, it can be seen that the observed direction by both University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 91 stations was almost constant for 2008 - 2010 and showed only slight vari- ations for 2011 - 2013. Observed speeds, on the other, appeared highly variable. It is also clear that the quantitative agreement between wind speeds observed at the TMS and CALMET predictions for all the years was higher than from the AMS. Figure 4.20: Plots of Modelled and Observed(AMS) Surface Wind direction for 2008 The correlation coefficient values between modelled and observed speed for the six-year period were higher for AMS (0.47 - 0.64) than for TMS (0.31 - 0.58) indicating a stronger relationship of the modelled with the observed winds from AMS. The highest R of 0.58 was existed between modelled speeds and TMS speed in 2008 whiles a weakest correlation of 0.31 was found between 2011 modelled and observed data. The highest correlation of 0.64 was obtained with 2010 data at AMS whiles the least value of 0.47 University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 92 Figure 4.21: Plots of Modelled and Observed(AMS) Surface Wind direction for 2009 Figure 4.22: Plots of Modelled and Observed(AMS) Surface Wind direction for 2010 University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 93 Figure 4.23: Plots of Modelled and observed(AMS) Surface Wind direction for 2011 Figure 4.24: Plots of Modelled and observed(AMS) Surface Wind direction for 2012 University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 94 Figure 4.25: Plots of Modelled and observed(AMS) Surface Wind direction for 2013 Figure 4.26: Plots of Modelled and Observed(TMS) Surface Wind direction for 2008 University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 95 Figure 4.27: Plots of Modelled and Observed(TMS) Surface Wind direction for 2009 Figure 4.28: Plots of Modelled and Observed(TMS) Surface Wind direction for 2010 University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 96 Figure 4.29: Plots of Modelled and observed(TMS) Surface Wind direction for 2011 Figure 4.30: Plots of Modelled and observed(TMS) Surface Wind direction for 2012 University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 97 Figure 4.31: Plots of Modelled and observed(TMS) Surface Wind direction for 2013 was found for 2012. Generally, R values for wind direction were lower between model and observed data from both stations. NMSE values were encouragingly low for wind direction for both stations (0.01 - 0.2) reflecting a low scatter in the entire data set. However, nmse values obtained for speed for AMS were high (0.88 - 1.4) indicating a large scatter in the speed data. NMSE value for speed for TMS were satisfactory (0.1 - 0.24). IOA between modelled and observed data for all the years was above 0.60 for both variables from both stations except for 2011 speeds. The negative FB values for direction are indicative of the overpredictions University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 98 by CALMET of the observed wind direction from both meteorological sta- tions as shown in Figs.4.20 - 4.25 for AMS and Figs.4.26 - 4.31 for TMS. CALMET clearly under-predicts observed speeds as is seen in the positive FB values. Whiles CALMET overpredicts surface wind direction, it under- estimates wind speeds as shown in Figs.4.32 - 4.37 for AMS and Figs.4.38 - 4.43 for TMS. Figure 4.32: Plots of Modelled and Observed(AMS) Surface Wind Speed for 2008 University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 99 Figure 4.33: Plots of Modelled and Observed(AMS) Surface Wind Speed for 2009 Figure 4.34: Plots of Modelled and Observed(AMS) Surface Wind Speed for 2010 University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 100 Figure 4.35: Plots of Modelled and Observed(AMS) Surface Wind Speed for 2011 Figure 4.36: Plots of Modelled and Observed(AMS) Surface Wind Speed for 2012 University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 101 Figure 4.37: Plots of Modelled and Observed(AMS) Surface Wind Speed for 2013 University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 102 Figure 4.38: Plots of Modelled and Observed(TMS) Surface Wind Speed for 2008 These challenges can be attributed to the quality of the initial winds. The quality of the initial wind fields of the diagnostic model (from WRF) is of vital importance for the resultant wind fields. While CALMET is able to add value to even 5 km WRF fields, a resolution reduction of WRF to 10 km grid spacing induces errors with respect to unresolved air flows which CALMET is unable to correct in most cases (Truhetz et al., 2007). These statistical measures, generally, suggest a good performance of CAL- MET . University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 103 Figure 4.39: Plots of Modelled and Observed(TMS) Surface Wind Speed for 2009 Figure 4.40: Plots of Modelled and Observed(TMS) Surface Wind Speed for 2010 University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 104 Figure 4.41: Plots of Modelled and Observed(TMS) Surface Wind Speed for 2011 Figure 4.42: Plots of Modelled and Observed(TMS) Surface Wind Speed for 2012 University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 105 Figure 4.43: Plots of Modelled and Observed(TMS) Surface Wind Speed for 2013 4.6 Spatial Distribution of Emissions Concentrations contours of annual averages of the modelled emissions for the 6-year period are presented in Figs.4.44 - 4.46. The plots reflect the spread of the pollutants and also shows the likely wind direction preva- lent in the study area for 2008. The spread of the SO2 plume is larger due to its high diffusivity and low density compared with NO2 and PM2.5. Consequently, it impacts a larger part of the study area with maximum contours of 610 µg/m3 near the souce of emission and minimum contours of 10 µg/m3. Due to the predominant south westerly winds of the year University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 106 Figure 4.44: 2008 Annual Average Concentration contours of SO2 in the Study Area (windrose shown earlier), receptors located upwind are the most hit. Ac- cording to Chan and Kwok (2000), fine and especially ultra-fine particles are expected to disperse in the air like gases. The larger-sized particles, however, are greatly affected by gravity and thus have a shorter residence time in the air. This explains the fairly elevated PM2.5 levels. Even though the emission rates for 2009 is similar to that of 2008, the concentration contours of pollutants are lower with maximum SO2, NO2 and PM2.5 contours of 75 µg/m 3, 72 µg/m3 and 62 µg/m3 respectively as shown in Figs.4.47, 4.48 and 4.49 respectively. Since emission rates are assumed to be constant, the differences could be attributed mainly to University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 107 Figure 4.45: 2008 Annual Average Concentration contours of NO2 in the Study Area meteorological conditions. The wind rose for 2009 is shown in Fig.4.50. The predominant south westerly winds of the year represent 90% of total winds. Wind speed class of 3.3-5.4 m/s constitutes 73% followed by 22% of speeds between 5.4 - 8.5 m/s and 3% of total winds falling into class 1.8 - 3.3 m/s . Total calm winds are nil. A comparison of wind roses for 2009 and 2008 shows that the average wind speed was higher in 2009 than in 2008. Strong winds tend to increase pollutant mixing and dispersion and hence decreasing ground level concentrations. The year 2008 experienced more low wind speeds and calm periods. This meteorological condition favours increased ground level concentrations of pollutants as evidenced by the 2008 contour plots. Low synoptic winds create a well-known meteorological University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 108 Figure 4.46: 2008 Annual Average Concentration contours of PM2.5 in the Study Area situation that favours air pollution built-up in urban areas (Jones et al., 2000, Qin and Kot, 1993). Figs.4.51 - 4.53 present concentration contours of the pollutants for 2010 with its accompanying wind rose in Fig. 4.54. Emission rate of the pol- lutants for 2010 was the highest. However, the positive influence of the prevailing winds for the year was seen in the generally low concentrations similar to those in 2009. The statistics of the 2010 surface winds are quite University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 109 Figure 4.47: 2009 Annual Average Concentration contours of SO2 in the Study Area similar with 2009 winds. Concentration contours for the remaining years and their corresponding windroses are shown by Figs. 4.55 - 4.66. It can be seen from Fig. 4.58, Fig.4.62 and Fig. 4.66 that the percentage of the upper wind speed class, 5.4 - 8.5 m/s, of total winds decreased giving rise to an increase in low speed percentages. Infact, 2013 recorded a low Figure 4.66: Wind Rose Depicting 2013 Surface Winds in the Study area University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 110 Figure 4.48: 2009 Annual Average Concentration contours of NO2 in the Study Area Figure 4.49: 2009 Annual Average Concentration contours of PM2.5 in the Study Area University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 111 Figure 4.50: Wind Rose Depicting 2009 Surface Winds in the Study area speed class (1.8 - 33) m/s of about 18%. Low wind speed conditions are typ- ically associated with the worst air pollution episodes in cities (Di Sabatino et al., 2003). Therefore, predicted concentrations at the various receptors should have been higher for 2011 - 2013. This, however, was not observed. It should be recalled that emission rates from the refinery saw a steady decline from 2011 to 2013 as a result of shortfalls in crude supply. This is the reason for the low concentration contours observed within the study area. University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 112 Figure 4.51: 2010 Annual Average Concentration contours of SO2 in the Study Area 4.7 Interannual Predicted Concentrations of Emissions at Receptors As was mentioned earlier in this chapter, receptors located within 10 - 15 km of the refinery are the most impacted by the refinery emissions. For this reason, this subsection takes another look at the annual variations in the modelled daily average pollutant concentrations at these receptors using the different emission rates estimated for the years under consideration. Results of this investigation are presented in Figs. 4.67, 4.68 and 4.69. The plots reflect the year-to-year variability in weather conditions. University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 113 Figure 4.52: 2010 Annual Average Concentration contours of NO2 in the Study Area Figure 4.53: 2010 Annual Average Concentration contours of PM2.5 in the Study Area University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 114 Figure 4.54: Wind Rose Depicting 2010 Surface Winds in the Study area A look at Fig. 4.67 reveals that the highest daily SO2 concentration occured at Tema Steelworks area followed by Tema Comm 25 for all the years. Con- centrations at the Tema Steelworks area for years ranged from about 220 - 280 µg/m3 with 2010 concentrations being the highest. Concentrations at Tema Comm 25 ranged between 60 - 100 µg/m3 with the highest occur- ing in 2010. Other receptors recording concentrations above 50 µg/m3 are Ashiaman and Sebrepor. The other receptors showed concentrations lower than 50 µg/m3 except the Tema General Hospital which recorded about 70 µg/m3 in 2013. Comparing these concentration values with WHO daily average guideline of 125 µg/m3 for SO2 (WHO (2000)), it is obvious that University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 115 Figure 4.55: 2011 Annual Average Concentration contours of SO2 in the Study Area Figure 4.56: 2011 Annual Average Concentration contours of NO2 in the Study Area University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 116 Figure 4.57: 2011 Annual Average Concentration contours of PM2.5 in the Study Area concentrations at the Tema Steelworks area well exceeded the regulatory limit. These concentrations also exceeded the Ghana EPA guideline values of 150 µg/m3 for industrial areas and 100 µg/m3 for residential areas. From Fig. 4.68, maximum NO2 concentration of about 78 µg/m 3 occured again at the Tema Steelworks after which TOR, Tema Comm 25 and Tema General Hospital follow. A comparison of the concentrations at the recep- tors with the WHO guideline value for NO (50 µg/m32 ) and the Ghana EPA values of 60 µg/m3 for residential and 150 µg/m3 for industrial areas, it can be concluded concentrations fall below these regulatory limits except the Tema Steelworks area. As far as the Ghana guideline values are con- cerned, the conclusion can be made that the refinery is not polluting the University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 117 Figure 4.58: Wind Rose Depicting 2011 Surface Winds in the Study area environment in respect of NO2. Since Ghana does not have PM2.5 guideline values, a comparison was made with only the WHO values of 25 µg/m3. Apart from Old and New Ningo where concentrations for all the years did not exceed this guideline value, it is quite clear from Fig. 4.69 that PM2.5 concentrations at all other receptors exceeded the WHO guideline value. In 2013, for example, the Tema Steelworks area exceeded this value by a factor of 9 in 2013. In 2009, 2010 and 2013, concentrations at Afienya, Shai Hills, Tema Newtown, Tema Comms. 4 and 7, Tema Fishing harbour and Prampam were below University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 118 Figure 4.59: 2012 Annual Average Concentration contours of SO2 in the Study Area Figure 4.60: 2012 Annual Average Concentration contours of NO2 in the Study Area University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 119 Figure 4.61: 2012 Annual Average Concentration contours of PM2.5 in the Study Area this guideline value. Generally, however, it is quite obvious that PM2.5 pollution by the refinery is substantial. 4.8 Seasonal Variation of Pollutants Results of seasonal variability investigations are presented in monthly aver- age concentration plots of pollutants in 2013 at Tema Steelworks and Tema Comm. 25, Kpone and the Tema General hospital as seen in Figs. 4.70 - 4.73. Predicted monthly concentrations of the pollutants for the entire year show that highest values occured in August at Tema Steelworks and Tema Comm 25. At Kpone, highest values occured in October and in December University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 120 Figure 4.62: Wind Rose Depicting 2012 Surface Winds in the Study area at the hospital. Further, it can be seen that, generally, the second half of the year during which the dry season occurs, shows slightly higher concen- trations than the first half. The major wet season in the first half of the year (April-June) could be responsible for the lower concentrations of the pollutants. The effect of precipitation on atmospheric pollutants are well researched (Aleksandropoulou and Lazaridis, 2004, Davies, 1976, Pandey et al., 1992, Ravindra et al., 2003). A particular factor that can affect pollutant distribution is the type of sta- bility condition which is a function of temperature. During the dry and warmer months of January, February and March, the atmosphere is more University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 121 Figure 4.63: 2013 Annual Average Concentration contours of SO2 in the Study Area Figure 4.64: 2013 Annual Average Concentration contours of NO2 in the Study Area University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 122 Figure 4.67: Daily Average SO2 Concentrations at various receptors Figure 4.68: Daily Average NO2 Concentrations at various recep- tors unstable because of the elevated temperatures. This increases the level of turbulence, which, in turn, increases the dispersive capability of the atmo- sphere. As a consequence, pollutants disperse relatively rapidly in both the horizontal and vertical directions, and its ground level concentration decays over relatively short distances. The atmosphere is more stable during the cooler months of July and August due to the slightly reduced temperatures University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 123 Figure 4.69: Daily Average PM2.5 Concentration at various recep- tors causing a reduction in the dispersive ability of the atmosphere. As a result, ground level concentrations of pollutants could increase as seen generally in Figs. 4.70 - 4.73. A plot showing the monthly variation of temperature over the years around the refinery is shown in Fig. 4.74. University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 124 Figure 4.70: 2013 Monthly Average concentrations of pollutants at Tema Steelworks Figure 4.71: 2013 Monthly Average concentrations of pollutants at Tema Comm. 25 University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 125 Figure 4.72: 2013 Monthly Average concentrations of pollutants at Kpone Figure 4.73: 2013 Monthly Average concentrations of pollutants at Tema Gen. Hosp University of Ghana http://ugspace.ug.edu.gh Chapter 4. Results and Discussions 126 Figure 4.74: Variation of Ambient Temperature around the refinery University of Ghana http://ugspace.ug.edu.gh Chapter 5 Conclusions and Recommendations 5.1 Introduction The overriding purpose of this study was to simulate the dispersion and transport of pollutants emitted during the processing of crude oil by TOR, in the Greater Accra Region of Ghana using the California Puff (CALPUFF) modeling system. To accomplish this goal, it became necessary to reach some prerequisite goals. The importance of air quality models in under- standing air pollution and their effects was emphazised during the literature review conducted for this dissertation. Subsequent to this, it was neces- sary to estimate daily average emission rates from the refinery for the years under review as an important input for CALPUFF. The WRF model was 127 University of Ghana http://ugspace.ug.edu.gh Chapter 5. Conclusions and Recommendations 128 run to generate the initial meteorological data necessary to initialize CAL- MET model and to drive the dispersion model, CALPUFF. To validate the CALPUFF modelling system, comparisons were made between some measured and modelled data. Once these prerequisite goals were achieved, the main purpose of the research was realised. This chapter reports the conclusions and recommendations that resulted from this study. 5.1.1 Conclusions The mass balance approach, which was employed to estimate the refinery pollutant emissions revealed that total CO2 emissions were the highest, peaking in 2008 and followed by VOCs. Particulate matter rates were ob- served to be fairly constant over the years under review. The SO2 and NO2 emission rates were next with NO2 being the least. A steady decline in emissions was observed over the years as a result of the reduction in production due to the frequent shortfalls in the refinery’s crude oil sup- ply consequently leading virtually to a halt in refinery operations in 2014 preceded by intermittent shutdowns. To determine the radius of impact of the refinery emissions, the dispersion of SO , NO and PM was modelled over a space of 60 km22 2 centred nearly around the refinery was carried out. Within this space, receptor locations representing residential and health facilities, schools and industries were University of Ghana http://ugspace.ug.edu.gh Chapter 5. Conclusions and Recommendations 129 identified and the maximum daily average concentrations of the pollutants at these location calculated using the CALPUFF. Meteorological data used to drive CALMET, the meteorological model, was generated by running WRF. It was concluded, based on the preliminary simulation results, that the impact of the refinery emissions was felt at receptor locations within a radius of about 20 km around the refinery. Furthermore, receptors located on the north and north-eastern parts of the refinery showed relatively higher pollutant concentrations than other parts of the study domain as a result of the predominant south-westerly winds in the study area. Conversely, south- western receptors were the least impacted by refinery emissions. A check with ambient pollutant guidelines of WHO and EPA Ghana showed that the concentrations of SO2 and NO2 at 36 receptors did not exceed regulatory limits. Two receptors, however, exceeded regulatory limits. Ground level concentrations of PM, however, exceeded the regulatory limits at almost all receptors. In order to validate the modelling system, results of the meteorological and dispersion simulation were compared with measurements and observa- tions using statistical measures. The statistical measures revealed a good agreement between model and observations leading to the conclusion that CALPUFF can be used satisfactorily for air pollution studies and for check- ing compliance of industrial set-ups with standards. University of Ghana http://ugspace.ug.edu.gh Chapter 5. Conclusions and Recommendations 130 It should be highlighted that the utilization of CALPUFF for the sim- ulation of dispersion of refinery emissions and for the description of air quality issues over the Tema Metropolitan Area and parts of the Accra Metropolitan area represents a useful instrument that may contribute to the establishment of environmental managing policies and to regulatory purposes. Even though the CALPUFF modelling system has been used to successfully simulate the dispersion of refinery emissions, it is not without limitations for developing countries. The major limitation is related to the meteorological data requirements of CALMET. Calmet has switches for the use of either observations or meteorological data from a numerical weather prediction model or a combination of both, which is usually preferred. However, when observations (surface and upper air) are to be used, hourly to 3-hourly data is required. This specification of data is not available in most developing countries hence limiting its use. For example, the Ghana meteorological service could provide daily surface data and not upper air data. 5.1.2 Recommendations The following recommendations are offered for related research and relevant institutions: University of Ghana http://ugspace.ug.edu.gh Chapter 5. Conclusions and Recommendations 131 1. Even though the refinery emissions in almost all the receptors do not generally exceed regulatory limits, it is obvious that the inclusion of all emittors in the model will result in increased levels of pollutants which could have serious health implications on the exposed popula- tion. To be able to quantify the contributions of the other emittors within the industrial complex, their emission rates are necessary. This can be obtained through an emissions inventory campaign within the Tema Industrial area by EPA, Ghana in collaboration with all the in- dustries. Closely linked to this, emission factors for various pollutants can also be developed. 2. The effects of refinery emissions on critical receptors, receptors lo- cated in the immediate north and north east of the refinery should be investigated through a health assessment to determine which diseases affect the populations in these receptors. 3. Despite the serious health and environmental implications of popula- tion exposure to VOCs and their strong association with ozone (O3), their dispersion and transport could not be modelled. This is be- cause of its composite nature. CALPUFF models single entities and not composites. It is therefore recommended that further research be carried out to characterise the VOCs for modelling purposes. University of Ghana http://ugspace.ug.edu.gh Chapter 5. Conclusions and Recommendations 132 4. SO2 and NO2 are easily transformed into sulphates and nitrates in the atmosphere which are implicated in many environmental challenges and would therefore be interesting to look at closely. The effect of chemical transformation of these pollutants can be investigated using the mesopuff module in CALPUFF. 5. Real time applications using CALPUFF is recommended as this will be useful in real time analysis and prediction of plume transport and diffusion during accidental releases. 6. Since results of this research clearly indicates PM2.5 pollution from TOR, it is important that EPA Ghana conducts regular monitoring exercises at all industries in order to enforce compliance to guidelines. 7. To reduce the error associated with emission rates input into air qual- ity models, it is essential to use data from measurements made di- rectly from the point source. It is therefore recommended that TOR puts in measures to ensure that flow meters are installed at various exit points. University of Ghana http://ugspace.ug.edu.gh References 133 References Abdul-Wahab, S. A. (2003). SO2 dispersion and monthly evaluation of the industrial source complex short-term (ISCST32) model at Mina Al-Fahal refinery, Sultanate of Oman. Environmental management, 31(2):0276– 0291. Addo, K. A. (2013). Shoreline morphological changes and the human factor. Case study of Accra, Ghana. Journal of Coastal Conservation, 17(1):85– 91. Affum, H., Oduro-Afriyie, K., Nartey, V., Adomako, D., and Nyarko, B. B. (2008). Biomonitoring of airborne heavy metals along a major road in Accra, Ghana. Environmental monitoring and assessment, 137(1-3):15– 24. Ahuja, S. (1996). Evaluation of MESOPUFF-II-SOx Transport and Depo- sition in the Great Lakes Region. PhD thesis, University of Toledo. Al-Fadhli, F. M., Kimura, Y., McDonald-Buller, E. C., and Allen, D. T. (2011). Impact of flare destruction efficiency and products of incom- plete combustion on ozone formation in Houston, Texas. Industrial & Engineering Chemistry Research, 51(39):12663–12673. Aleksandropoulou, V. and Lazaridis, M. (2004). Spatial distribution of gaseous and particulate matter emissions in Greece. Water, air, and soil pollution, 153(1-4):15–34. Arku, R. E., Vallarino, J., Dionisio, K. L., Willis, R., Choi, H., Wilson, J. G., Hemphill, C., Agyei-Mensah, S., Spengler, J. D., and Ezzati, M. (2008). Characterizing air pollution in two low-income neighborhoods in Accra, Ghana. Science of the total environment, 402(2):217–231. University of Ghana http://ugspace.ug.edu.gh References 134 Bansah, K. J. and Amegbey, N. (2012). Ambient Particulate Matter Monitoring-A Case Study at Tarkwa. Research Journal of Environmental and Earth Sciences, 4(4):419–423. Boubel, R. W., Vallero, D., Fox, D. L., Turner, B., and Stern, A. C. (2013). Fundamentals of air pollution, pages 291–320. Elsevier. Briggs, G. A. (1975). Plume rise predictions, lectures on air pollution and environment impact analysis. Am. Meteorol. Soc., Boston, USA, 10. Britter, R. and Hanna, S. (2003). Flow and dispersion in urban areas. Annual Review of Fluid Mechanics, 35(1):469–496. Brown, R. A. (1991). Fluid Mechanics of the Atmosphere, volume 47, pages 197–206. Academic Press. Brunekreef, B. and Holgate, S. T. (2002). Air pollution and health. The lancet, 360(9341):1233–1242. Chan, L. and Kwok, W. (2000). Vertical Dispersion of Suspended Par- ticulates in urban area of Hong Kong. Atmospheric Environment, 34(26):4403–4412. Chandrasekar, A., Philbrick, C. R., Clark, R., Doddridge, B., and Geor- gopoulos, P. (2003). Evaluating the performance of a computationally efficient MM5/CALMET system for developing wind field inputs to air quality models. Atmospheric Environment, 37(23):3267–3276. Cimorelli, A. J. et al. (1998). AERMOD–Description of model formulation. Collett, R. S. and Oduyemi, K. (1997). Air quality modelling: a tech- nical review of mathematical approaches. Meteorological Applications, 4(03):235–246. University of Ghana http://ugspace.ug.edu.gh References 135 Corma, A., Palomares, A., Rey, F., and Márquez, F. (1997). Simultaneous Catalytic Removal of SOx and NOx with Hydrotalcite-Derived Mixed Oxides Containing Copper, and Their Possibilities to Be Used in FCC Units. Journal of catalysis, 170(1):140–149. Darko, E. and Quist, M. (2012). Low Sulphur Diesel and Refining Process, Technical and Financial Implications for Tema Oil Refinery (TOR) Lim- ited. In Workshop organized by National Petroleum Authority in Part- nership for Clean fuels and Vehicles under the auspices of United Nation Environmental Programme (UNEP) in Accra, Ghana from September 13-14, 2012. Davies, T. (1976). Precipitation Scavenging of Sulphur Dioxide in an In- dustrial Area. Atmospheric Environment (1967), 10(10):879–890. de Leeuw, F. A., Moussiopoulos, N., Sahm, P., and Bartonova, A. (2001). Urban Air Quality in Larger Conurbations in the European Union. En- vironmental Modelling & Software, 16(4):399–414. Demuzere, M., Trigo, R., Vila-Guerau de Arellano, J., and Van Lipzig, N. (2009). The impact of weather and atmospheric circulation on o3 and pm10 levels at a rural mid-latitude site. Atmospheric Chemistry and Physics, 9(8):2695–2714. Di Sabatino, S., Kastner-Klein, P., Berkowicz, R., Britter, R., and Fe- dorovich, E. (2003). The modelling of turbulence from traffic in urban dispersion modelspart i: theoretical considerations. Environmental Fluid Mechanics, 3(2):129–143. Dios, M., Souto, J., and Casares, J. (2013). Experimental development of CO2, SO2 and NO emission factors for mixed lignite and sub-bituminous coal-fired power plant. Energy, 53:40–51. University of Ghana http://ugspace.ug.edu.gh References 136 Dresser, A. L. and Huizer, R. D. (2011). CALPUFF and AERMOD model validation study in the near field: Martins Creek revisited. Journal of the Air & Waste Management Association, 61(6):647–659. EIPPCB (2012). Best Available Techniques (BAT) Reference Document for the Refining of Mineral Oil and Gas. Technical report. England, G. C., Zielinska, B., Loos, K., Crane, I., and Ritter, K. (2000). Characterizing PM2.5 emission profiles for stationary sources: compar- ison of traditional and dilution sampling techniques. Fuel Processing Technology, 65:177–188. Fenger, J. (1999). Urban air quality. Atmospheric environment, 33(29):4877–4900. Fernando, H. (2010). Fluid dynamics of urban atmospheres in complex terrain. Annual Review of Fluid Mechanics, 42:365–389. Ferretti, M., Andrei, S., Caldini, G., Grechi, D., Mazzali, C., Galanti, E., and Pellegrini, M. (2008). Integrating monitoring networks to obtain esti- mates of ground-level ozone concentrationsA proof of concept in Tuscany (central Italy). Science of the Total Environment, 396(2):180–192. Fishwick, S. and Scorgie, Y. (2011). Performance of CALPUFF in predict- ing time-resolved particulate matter concentrations from a large scale surface mining operation. In Proceedings of CASANZ Conference from 30 July2 August,2011, Auckland, pages 647–659. Fowler, D., Pilegaard, K., Sutton, M., Ambus, P., Raivonen, M., Duyzer, J., Simpson, D., Fagerli, H., Fuzzi, S., Schjørring, J. K., et al. (2009). Atmospheric composition change: ecosystems–atmosphere interactions. Atmospheric Environment, 43(33):5193–5267. University of Ghana http://ugspace.ug.edu.gh References 137 Ghannam, K. and El-Fadel, M. (2013). Emissions characterization and regulatory compliance at an industrial complex: an integrated MM5/CALPUFF approach. Atmospheric Environment, 69:156–169. G.S.S (2012). 2010 Population & Housing Census of Ghana: GA-MY, page 5. Ghana Statistical Service. Guilbert, J. et al. (2003). The World Health Report 2002-reducing Risks, promoting healthy life. Education for health-Abingdon-Carfax Publishing Limited, 16(2):230–230. Hanna, S., Chang, J., and Strimaitis, D. (1993). Hazardous gas model evaluation with field observations. Atmospheric Environment. Part A. General Topics, 27(15):2265–2285. Hanna, S. R., Strimaitis, D. G., and Chang, J. C. (1991). Evaluation of fourteen hazardous gas models with ammonia and hydrogen fluoride field data. Journal of hazardous materials, 26(2):127–158. Hansard, R., Maher, B., and Kinnersley, R. (2011). Biomagnetic monitor- ing of industry-derived particulate pollution. Environmental Pollution, 159(6):1673–1681. Harding, R., Peters, A., and Nee, J. (2001). New developments in fcc catalyst technology. Applied Catalysis A: General, 221(1):389–396. Hernández, A., Saavedra, S., Rodŕıguez, A., Souto, J. A., and Casares, J. J. (2014). Coupling WRF and CALMET models: Validation during primary pollutants glc episodes in an Atlantic coastal region. In Air Pollution Modeling and its Application XXII, pages 681–684. Springer. Hildemann, L. M., Klinedinst, D. B., Klouda, G. A., Currie, L. A., and Cass, G. R. (1994). Sources of urban contemporary carbon aerosol. En- vironmental science & technology, 28(9):1565–1576. University of Ghana http://ugspace.ug.edu.gh References 138 Holmes, N. S. and Morawska, L. (2006). A review of dispersion modelling and its application to the dispersion of particles: an overview of differ- ent dispersion models available. Atmospheric Environment, 40(30):5902– 5928. Holnicki, P. (2011). Uncertainty in integrated modelling of air quality. Holtslag, A. and Van Ulden, A. (1983). A simple scheme for daytime estimates of the surface fluxes from routine weather data. Journal of Climate and Applied Meteorology, 22(4):517–529. Irwin, J. S. (1983). Estimating plume dispersion-a comparison of several sigma schemes. Journal of Climate and Applied Meteorology, 22(1):92– 114. Jackson, P. (1981). On the displacement height in the logarithmic velocity profile. Journal of Fluid Mechanics, 111:15–25. Janjic, Z. (2003). A nonhydrostatic model based on a new approach. Me- teorology and Atmospheric Physics, 82(1-4):271–285. Janjić, Z. I. (2002). Nonsingular implementation of the mellor–yamada level 2.5 scheme in the ncep meso model. NCEP office note, 437:61. Jiménez-Garćıa, G., Aguilar-López, R., and Maya-Yescas, R. (2011). The fluidized-bed catalytic cracking unit building its future environment. Fuel, 90(12):3531–3541. Jones, S., Fisher, B. E., Gonzalez-Flesca, N., and Sokhi, R. (2000). The use of measurement programmes and models to assess concentrations next to major roads in urban areas. Environmental Monitoring and Assessment, 64(2):531–547. Khamsimak, P., Koonaphapdeelert, S., and Tippayawong, N. (2012). Sim- ulated Effect of Terrain Change on Dispersion of SO2 Emission Around University of Ghana http://ugspace.ug.edu.gh References 139 Mae Moh Basin. Recent Advances in Energy,Environment and Economic Development, pages 200–204. Kumar, A., Luo, J., and Bennett, G. F. (1993). Statistical evaluation of lower flammability distance (lfd) using four hazardous release models. Process Safety Progress, 12(1):1–11. Levy, J. I., Wilson, A. M., Evans, J. S., and Spengler, J. D. (2003). Estimation of primary and secondary particulate matter intake frac- tions for power plants in Georgia. Environmental science & technology, 37(24):5528–5536. Liu, M.-K. and Yocke, M. A. (1980). Siting of wind turbine generators in complex terrain. Journal of Energy, 4(1):10–16. Macdonald, R. (2000). Modelling the mean velocity profile in the urban canopy layer. Boundary-Layer Meteorology, 97(1):25–45. Mart́ın, J., Lumbreras, J., and Rodŕıguez, M. (2003). Testing flare emission factors for flaring in refineries. In US EPA 12th International emissions inventory conference, San Diego. Martin, R. V. (2008). Satellite remote sensing of surface air quality. Atmo- spheric Environment, 42(34):7823–7843. Maya-Yescas, R., Villafuerte-Maćıas, E., Aguilar, R., and Salazar-Sotelo, D. (2005). Sulphur oxides emission during fluidised-bed catalytic crack- ing. Chemical Engineering Journal, 106(2):145–152. Michalakes, J., Dudhia, J., Gill, D., Henderson, T., Klemp, J., Skamarock, W., and Wang, W. (2004). The Weather Research and Forecast Model: Software Architecture and Performance. In Proceedings of the 11th ECMWF Workshop on the Use of High Performance Computing In Me- teorology, volume 25, page 29. World Scientific. University of Ghana http://ugspace.ug.edu.gh References 140 Monin, A. and Obukhov, A. (1954). Basic laws of turbulent mixing in the surface layer of the atmosphere. Contrib. Geophys. Inst. Acad. Sci. USSR, 151:163–187. Muff, R. and Efa, E. (2006). Ghana-Germany technical cooperation project: environmental and engineering geology for urban planning in the Accra- Tema area. Explanatory notes for the geological map for urban planning, 1(50):000. Muller, N. Z. and Mendelsohn, R. (2007). Measuring the damages of air pollution in the united states. Journal of Environmental Economics and Management, 54(1):1–14. Ofosu, F. G., Hopke, P. K., Aboh, I. J., and Bamford, S. A. (2012). Char- acterization of fine particulate sources at Ashaiman in Greater Accra, Ghana. Atmospheric Pollution Research, 3(3). Pandey, J., Agrawal, M., Khanam, N., Narayan, D., and Rao, D. (1992). Air pollutant concentrations in Varanasi, India. Atmospheric Environment. Part B. Urban Atmosphere, 26(1):91–98. Pasquill, F. and Smith, F. (1983). Atmospheric diffusion.: Study of the dis- persion of windborne material from industrial and other sources. JOHN WILEY & SONS, 605 THIRD AVE., NEW YORK, NY 10016, USA. 1983. Pearce, J. L., Beringer, J., Nicholls, N., Hyndman, R. J., Uotila, P., and Tapper, N. J. (2011). Investigating the influence of synoptic-scale meteo- rology on air quality using self-organizing maps and generalized additive modelling. Atmospheric Environment, 45(1):128–136. University of Ghana http://ugspace.ug.edu.gh References 141 Perez Ballesta, P., Field, R., Fernandez-Patier, R., Galan Madruga, D., Connolly, R., Baeza Caracena, A., and De Saeger, E. (2008). An ap- proach for the evaluation of exposure patterns of urban populations to air pollution. Atmospheric Environment, 42(21):5350–5364. Products, U. O. (1997). Fluid catalytic cracking process manual - process calculations,. pages 21–33, 64–68. Protonotariou, A., Bossioli, E., Athanasopoulou, E., Dandou, A., Tombrou, M., Assimakopopoulos, V., Flocas, H., and Chelmis, C. (2004). Valida- tion and inter-comparison of CALPUFF regulatory model to Eulerian models and measurements. An application over the greater Athens area, Greece. In Proceedings of the 9th International Conference on Harmoni- sation within Atmospheric Dispersion Modelling for Regulatory Purposes, pages 131–135. Puliafito, S. E., Allende, D., Fernández, R., Castro, F., and Cremades, P. (2011). New approaches for urban and regional air pollution modelling and management. Air Pollution/Book, 1. Qin, Y. and Kot, S. (1993). Dispersion of vehicular emission in street canyons, Guangzhou City, South China (PRC). Atmospheric Environ- ment. Part B. Urban Atmosphere, 27(3):283–291. Raupach, M. and Thom, A. S. (1981). Turbulence in and above plant canopies. Annual Review of Fluid Mechanics, 13(1):97–129. Ravindra, K., Mor, S., Kamyotra, J., Kaushik, C., et al. (2003). Variation in spatial pattern of criteria air pollutants before and during initial rain of monsoon. Environmental monitoring and assessment, 87(2):145–153. Rotach, M. (1994). Determination of the zero plane displacement in an urban environment. Boundary-Layer Meteorology, 67(1-2):187–193. University of Ghana http://ugspace.ug.edu.gh References 142 Sackey, S. S. (2012). Development of DOAS remote sensing system for ground-based measurements of trace gas emissions in an industrial area. PhD thesis. Safo-Adu, G., Ofosu, F., Carboo, D., and Armah, Y. (2014). Heavy metals and black carbon assessment of PM10 particulates along Accra–Tema highway in Ghana. International Journal of Science and Technology, 3(8). Schürmann, G., Algieri, A., Hedgecock, I., Manna, G., Pirrone, N., and Sprovieri, F. (2009). Modelling local and synoptic scale influences on ozone concentrations in a topographically complex region of Southern Italy. Atmospheric Environment, 43(29):4424–4434. Scire, J. S., Robe, F. R., Fernau, M. E., and Yamartino, R. J. (2000a). A User's guide for the CALMET Meteorological Model. Earth Tech, USA, 37. Scire, J. S., Strimaitis, D. G., and Yamartino, R. J. (2000b). A User's guide for the CALPUFF Dispersion Model. Earth Tech, Inc, 521:1–521. Seigneur, C., Pun, B., Pai, P., Louis, J.-F., Solomon, P., Emery, C., Morris, R., Zahniser, M., Worsnop, D., Koutrakis, P., et al. (2000). Guidance for the performance evaluation of three-dimensional air quality modeling systems for particulate matter and visibility. Journal of the Air & Waste Management Association, 50(4):588–599. Shamarock, W., Klemp, J., Dudhia, J., Gill, D., Barker, D., Duda, M., Huang, X., Wang, W., and Powers, J. (2008). A Description of the Advanced Research WRF version 3. NCAR technical note NCAR/T- N/u2013475. University of Ghana http://ugspace.ug.edu.gh References 143 Skamarock, W. C., Klemp, J. B., Dudhia, J., Gill, D. O., Barker, D. M., Wang, W., and Powers, J. G. (2005). A description of the advanced research wrf version 2. Technical report, DTIC Document. Stull, R. B. (2012). An introduction to boundary layer meteorology, vol- ume 13, page 2. Springer Science & Business Media. Truhetz, H., Gobiet, A., and Kirchengast, G. (2007). Evaluation of a dynamic-diagnostic modelling approach to generate highly resolved wind fields in the Alpine region. Meteorologische Zeitschrift, 16(2):191–201. Tsuang, B.-J. (2003). Quantification on the Source/Receptor Relationship of Primary Pollutants and Secondary Aerosols by a Gaussian Plume Tra- jectory Model: Part I Theory. Atmospheric Environment, 37(28):3981– 3991. USEPA (2001). Preferred and alternative methods for estimating air emis- sions from boilers. 2. USEPA (2010). Available and emerging technologies for reducing green- house gas emissions from the petroleum refining industry. 2.1.1. USEPA (2011). Emission estimation protocol for petroleum refineries. 2.1.1. USGS (2010). The Global Land Cover Characterization (GLCC)Database. NCAR technical note NCAR/TN/u2013475. Venkatesan, R., Mathiyarasu, R., and Somayaji, K. (2002). A Study of Atmospheric Dispersion of Radionuclides at a Coastal site using a mod- ified Gaussian Model and a Mesoscale Sea Breeze model. Atmospheric Environment, 36(18):2933–2942. Venkatram, A. (1978). An examination of box models for air quality sim- ulation. Atmospheric Environment (1967), 12(12):2243–2249. University of Ghana http://ugspace.ug.edu.gh References 144 Versteeg, H. K. and Malalasekera, W. (2007). An Introduction to Computa- tional Fluid Dynamics: The Finite Volume Method, pages 1–6. Pearson Education Limited, England. WHO (2000). Air quality guidelines for Europe, volume 91, pages 1–287. Second Edition,Copenhagen, WHO Regional Office for Europe. Wieringa, J. et al. (2001). New revision of davenport roughness classifica- tion. Proc., 3EACWE, Eindhoven, The Netherlands, pages 285–292. Willmott, C. J. (1982). Some comments on the evaluation of model per- formance. Bulletin of the American Meteorological Society, 63(11):1309– 1313. Wood, E. C., Herndon, S. C., Fortner, E. C., Onasch, T. B., Wormhoudt, J., Kolb, C. E., Knighton, W. B., Lee, B. H., Zavala, M., Molina, L., et al. (2012). Combustion and destruction/removal efficiencies of in-use chemical flares in the greater Houston area. Industrial & Engineering Chemistry Research, 51(39):12685–12696. Zadakbar, O., Vatani, A., and Karimpour, K. (2008). Flare gas recovery in oil and gas refineries. Oil & Gas Science and Technology-Revue de l’IFP, 63(6):705–711. Zannetti, P. (2013). Air Pollution Modeling: Theories, Computational Methods and Available Software, pages 107–130. Springer Science & Busi- ness Media, New York. Zawar-Reza, P., Kingham, S., and Pearce, J. (2005). Evaluation of a year- long dispersion modelling of PM10 using the mesoscale model TAPM for Christchurch, New Zealand. Science of the Total Environment, 349(1):249–259. University of Ghana http://ugspace.ug.edu.gh References 145 Zhao, X., Peters, A., and Weatherbee, G. (1997). Nitrogen chemistry and NOx control in a fluid catalytic cracking regenerator. Industrial & engi- neering chemistry research, 36(11):4535–4542. Zhou, Y., Levy, J. I., Hammitt, J. K., and Evans, J. S. (2003). Estimating population exposure to power plant emissions using CALPUFF: a case study in Beijing, China. Atmospheric Environment, 37(6):815–826. University of Ghana http://ugspace.ug.edu.gh Appendix A A Estimation of Refinery Emission Rates A.1 Estimation of Flue Stack Gas Rate and Compo- sition A.1.1 Combustion Air Correction to Dry Basis To correct the combustion air to dry basis, the combined gas law equation, Eqn.(1) was used: P1V1 P2V2 = (1) T1 T2 Given the following parameters (2009): P1= 1 atm, V1 = 63882.88 Nm 3/hr, T1= 273 K, P2 = 1 atm, V2 =? T2 = (120 + 273)K V = 91963.27 m32 A Relative Humidity (RH) of 80% for Accra during the study period im- plies: Moisture content = (0.017 kg H2O/ 1 kg of dry air) from the Psychometric chart 146 University of Ghana http://ugspace.ug.edu.gh Appendix. Estimation of Refinery Emission Rates 147 Therefore, the average Volumetric Air Flow rate (Dry Basis) into regener- ator = 90426.04 m3/hr Dry air density at 373 K = 0.9 kg/m3 Therefore, the average mass Air Flow rate (Dry Basis) into regenerator = 81383.421 kg = 2809.61 kmol/hr with dry air molar mass of 28.966 kg/kmol A.1.2 Calculation of Flue Gas Rate The flue gas rate can be calculated from the regenerator air rate. These two streams are related by the inert (N2 + Ar) content which remains constant through the catalyst regeneration. From a nitrogen balance: kmol (N2 + Ar) in dry air = kmol (N2 + Ar) in flue gas (2809.61 × 0.79) kmol = Flue Gas rate × (0.82) Therefore Flue gas molar flowrate = 2706.83 kmol/hr A.1.3 Flue Stack Gas Components From the flue gas composition provided in Table 1, the CO2 emission from the flue stack can be calculated. Table 1: Flue Gas Composition Component CO2 CO N2 SO2 NO2 O2 % Composition 15.50 0 82 0 0 2.5 1. CO2 From the GC in the Table 1, CO2 molar flow rate = 0.155(2706.83) = 419.56 kmol/hr University of Ghana http://ugspace.ug.edu.gh Appendix. Estimation of Refinery Emission Rates 148 Molar mass of CO2 = 44 kg/kmol is Mass flow rate of CO2 = 18460.56 kg/hr 2. SOx If the Density of feedstock = 907 kg/m3 and the volumetric flow rate of feedstock(2009) = 72.02 m3 Then the Mass flow rate of feedstock = 65322.14 kg/hr With 1% wt of feed = sulphur content Sulphur content = 653.22 kg 5% of feed sulphur is combusted in regenerator Mass of combusted sulphur = 32.66 kg Molar mass of Sulphur = 32 kg/kmol 90% of combusted sulphur reacts to form SO2 Mass of combusted sulphur to SO2 = 0.9(32.66) = 29.39 kg Moles of combusted sulphur to SO2 = 0.92 kmol Based on the following stoichiometry of the reaction: S + O2 −→ SO2 ..................... (1) Moles of SO2 formed = 0.92 kmol Molar mass of SO2 = 64 kg/kmol Therefore Mass of SO2 = 0.92(64) = 58.88 kg 10% of combusted sulphur reacts to form SO3 Mass of combusted sulphur to SO3 = 0.1(32.66) = 3.27 kg Moles of combusted sulphur to SO3 = 0.10 kmol Based on the following stoichiometry of the reaction: 2S + 3O2 −→ 2SO3 ..................... (2) University of Ghana http://ugspace.ug.edu.gh Appendix. Estimation of Refinery Emission Rates 149 Moles of SO3 formed = 0.92 kmol Molar mass of SO3 = 80 kg/kmol Therefore Mass of SO3 produced = 0.10(80) = 8.00 kg 3. NOx If 0.32% of feed = elemental nitrogen Mass of Nitrogen = 0.0032(65322.14) = 209.03 kg 15% of feed nitrogen combusted with coke on catalyst = 0.15(209.03) = 31.35 kg Molar mass of elemental nitrogen = 14 kg/kmol Moles of feed nitrogen combusted = 2.24479 kmol 90% of combusted nitrogen reacts to form NO Based on the following stoichiometry of the reaction: 2N + O2 −→ 2NO Moles of NO produced = 0.9(2.24479 kmol) = 2.02 kmol Molar mass of NO = 30 kg/kmol Mass of NO produced = 60.47 kg 10% of combusted nitrogen reacts to form NO2 Based on the following stoichiometry of the reaction: N + O2 −→ NO2 Moles of NO2 produced = 0.1(2.24479 kmol) = 0.22 kmol Molar mass of NO2 = 46 kg/kmol Mass of NO2 produced = 10.30 kg 4. (Coke + Catalyst fines) University of Ghana http://ugspace.ug.edu.gh Appendix. Estimation of Refinery Emission Rates 150 (a) Coke 1 kg of solid particulates is produced for every 1000 kg of coke- on-catalyst (COC) combusted.(TOR Manual) From the stoichiometry of the COC combustion kinetics below: 2C + 11/2O2 + H2 + S + N −→ CO2 + CO + H2O + SO2 + NO2 + O2 .........(3) Moles of CO2 produced = moles of CO2 in Flue gas = 419.56 kmol Hence moles of Coke combusted = 2(419.558) = 839.12 kmol Molar mass of coke = 12 kg/kmol Mass of coke combusted = 10069.40 kg Therefore Mass of coke particulates produced = 10.07 kg (b) Catalyst fines At a cyclone efficiency of 99.75% and a catalyst inventory of 120000 kg (TOR Manual) Catalyst fines lost to the air = 0.0025(120000) = 300 kg Total Particulates = 300 + 10.07 = 310.07 kg A.2 Estimation of Flared Gas Composition 1. Volatile Organic Compounds (VOCs) With Fuel Gas mass flow rate = 6239.0 kg and a Flare efficiency = 98% (TOR Manual) Total VOCs in flare exit = 0.02(6239.0)kg = 124.78 kg Average Fuel gas composition (Vol%) by GC method and volume by calculations is given in Table 2. Then based on the stoichiometry of the reactions below, the moles and masses of the product gases are derived. University of Ghana http://ugspace.ug.edu.gh Appendix. Estimation of Refinery Emission Rates 151 Table 2: RFCC Fuel Gas Composition Component Vol% C1 28.0319 C2 10.46 C3 0.25 Ethylene 12.37 Propylene 1.86 Butene 1.56 CO 1.03 H2 25.85 CO2 2.197 H2S 0.4 N2 14.58 From the following stoichiometric combustion reactions: CH4 + 2O2 −→ CO2 + 2H2O ..............(1) C2H6+ 7/2O2 −→ 2CO2 + 3H2O .........(2) C3H8 + 5O2 −→ 3CO2 + 4H2O ..........(3) C2H4+ 3O2 −→ 2CO2 + 2H2O ..........(4) C3H6+ 9/2O2 −→ 3CO2 + 3H2O ........(5) C4H6+ 11/2O2 −→ 4CO2 + 3H2O ........(6) CO + 1/2O2 −→ CO2 ..........................(7) 2H2S + 3O2 −→ 2SO2 + 2H2O .............(8) Total moles of CO2 produced = 283.147 kmol Mass flowrate of CO2 = 283.147(44) = 12458.49 kg 2. Total SO2 Total moles of SO2 from H2S combustion = 1.35 kmol Total mass of SO2 = 86.45 kg Tables 3 and 4 present the calculated emission rates for the years under consideration. University of Ghana http://ugspace.ug.edu.gh Appendix. Estimation of Refinery Emission Rates 152 Table 3: RFCCU and (Total Refinery) Flare Stack Emission Rates(kg/hr) Emission Rate 2008 2009 2010 2011 2012 2013 SO2 87 87 95 88 81 72 (433) (432) (475) (441) (406) (361) CO2 12478 12458 13695 12716 11688 10397 (62391) (62292) (68478) (63582) (58440) (51989) VOCs 125 125 137 127 117 104 (2500) (2496) (2743) (2547) (2341) (2083) Table 4: RFCCU and (Total Refinery) Flue Stack Emission Rates(kg/hr) Emission Rate 2008 2009 2010 2011 2012 2013 SO2 73 64 74 68 69 69 (366) (335) (372) (341) (347) (346) CO2 20514 18460 18248 16458 16040 16079 (102570) (92302) (91240) (82292) (80203) (80398) NO2 77 71 79 72 73 73 (387) (354) (394) (361) (367) (366) PM2.5 311 310 310 309 309 309 (1037) (1034) (1033) (1030) (1029) 1029) A.3 Calculation of Flare and Flue Stack Exit Gas Ve- locities Table 5 provides the parameters used for the calculations. Table 5: RFCC Point Sources Parameters Point Source Diameter (m) Height (m) Temperature (K) Flue Stack 1.28 60 513 Flare 0.6 55 1273 Given the flare gas exit temperature of 1273 K and obtaining the volumetric flow rate from the given mass rate and the cross-sectional area for the flare University of Ghana http://ugspace.ug.edu.gh Appendix. Estimation of Refinery Emission Rates 153 diameter, the flare gas velocity is calculated as follows using flow parameters for the year 2009: Given the flare cross-sectional area of 0.287 m2, flare gas mass flow rate of 6239 kg/hr and fuel gas density of 0.8 kg/m3, the volumetric flow rate can be calculated using Eqn.(2): ṁ V̇ = (2) ρ Where V̇ is the volumetric flow rate, ṁ the mass flow rate and ρ the gas density Volumetric flow rate = 7798.85 m3 From Eqn.(3), the gas velocity can be obtained. V̇ v = (3) A Where v is the velocity and A the cross-sectional area of the flare. Velocity = 7.66 m/s The flue gas volume is determined from its moles using the ideal gas equa- tion given by Eqn.(4). PV = nRT (4) Where P is the pressure, V is the volume, n is the number of moles, R is the universal gas constant and T the temperature of the gas. University of Ghana http://ugspace.ug.edu.gh Appendix. Estimation of Refinery Emission Rates 154 Using a pressure of 101.423 kPa and the moles of flue gas of 2706.828 kmols which was calculated early on, the flue gas volume is given as: Flue Gas volumetric flow rate = 112345.895m3/hr = 31.207m3/s Given the Flue stack area of 1.127 m2, Flue Gas velocity = 24.25 m/s Table 6: Average Exit Gas Velocities of Point Sources Used for the Simulations for 2008 - 2013 Exit Gas Velocity(m/s) Point Source 2008 2009 2010 2011 2012 2013 Flare 7.67 7.66 8.42 7.82 7.19 6.39 Flue Stack 26.94 24.25 23.97 21.62 21.07 21.12