AIR QUALITY FORECASTING USING MATHEMATICAL MODELS FOR DELHI by ANIKENDER KUMAR Centre for Atmospheric Sciences Submitted in fulfillment of the requirements of the degree of Doctor of Philosophy to the Indian Institute of Technology Delhi February 2013
Dedicated to My Parents
Certificate This is to certify that the thesis entitled Air Quality Forecasting using Mathematical Models for Delhi being submitted by Mr. Anikender Kumar to the Indian Institute of Technology Delhi for the award of the degree of DOCTOR OF PHILOSOPHY is a record of the original bonafide research carried out by him. He has worked under my guidance and supervision and has fulfilled the requirements for the submission of thesis. The results presented in this thesis have not been submitted in part or full to any other University or Institute for award of any degree or diploma. New Delhi Professor (Mrs.) Pramila Goyal Centre for Atmospheric Sciences Indian Institute of Technology Delhi Hauz Khas, New Delhi-110 016, India
Acknowledgements This thesis is the result of my full commitment to my research work whereby I have been accompanied and supported by many people. It is a pleasant aspect that I have now the opportunity to express my gratitude for all of them. I would like to express my deepest gratitude to my thesis supervisor, Prof. Pramila Goyal, Centre for Atmospheric Sciences (CAS), IIT Delhi, for her invaluable guidance, suggestions and endless encouragement. She always gave me the freedom to pursue my own interests and provided me with insightful suggestions and support in developing independent thinking and research skills. She has been an exceptional mentor and I appreciate both our professional and personal conversations over the years. The knowledge and wisdom I have gained from her will forever guide me in education and in life. I would like to thank Prof. S. K. Dash, Head, CAS, IIT Delhi, for providing all the essential facilities in the Centre to carry out the work. I am also thankful to Prof. O.P. Sharma, former head, Centre for Atmospheric Sciences for his immense help and encouragement during the course of this work. I am grateful to Prof. M. Sharan for his invaluable suggestions throughout the course of this work. I also wish to extend my deep appreciation to my SRC members; Prof. A. D. Rao, Chairman, SRC, Prof. A. Chandra, Center for Energy Studies. I am thankful to Prof. G. Jayaraman, Prof. M. Mohan, Prof. S.K. Dube, Prof. U.C. Mohanty, Dr. H.C. Upadhyay, Dr. K. Achuta Rao and Dr. S. Dey for their fruitful suggestion, whole hearted support and encouragement. I sincerely acknowledge to IITD and MHRD for providing me financial support in the form of scholarship during my research. I would like to thank and acknowledge National Centre for Atmospheric Research (NCAR), USA and National Centers for Environmental Prediction (NCEP), USA for providing the model and reanalysis data sets. Central Pollution Control Board (CPCB) and Delhi Pollution Control Board (DPCC) are also acknowledged for providing the air pollutants concentration data. India Meteorological Department (IMD) is highly appreciated and acknowledged for providing the meteorological data sets.
I thank the whole CAS staff especially Mr. Data Ram, Mr. Kedar Nath, Mr. K.K. Sharma, Mrs. Kusum Sehrawat, Mr. L. S. Negi, Mr. Prashant, Mrs. Saroj Bala Gupta, Mr. S.S. Negi and Mr. V. K. Kaushik for their help and support. In my daily work, I have been blessed with a friendly and cheerful group of fellow students. I convey special thanks to my friends Dr. Mukesh, Dr. Pramod, Dr. Sarvesh and Dr. Sunil with whom I shared my joy and sorrows. Their warm company, unwavering support in my ups and downs, and helpful suggestions offered at various stages of my Ph.D. work, made my stay at IIT pleasant and memorable. At this happy moment, I fondly remember all my seniors, colleagues and friends especially, Dr Neeru, Dr Palash, Dr Sankalp, Dr Senthil, Dr Subrat, Dr Sujata, Dr Sushil, Dr Swagata, Dipak, Liby, Kanhu, Rani, Srinivas and Suraj for their support and nice company. I wish to express thanks to my recent friends Abhishek, Amit, Aniket, Arjya, Dhirendra, Gavendra, Himansu, Kamlesh, Mohit, Piyush, Pushp Raj, Ragi, Rajeev, Ram, Rati, Saurabh, Sathiya, Surendra, Sushant, Tarkeshwar, Vijay and many others at IIT Delhi who made my stay a memorable one. I especially thank my friends at hostel, and particularly Amit, Arun, Avnish, Gyanendra, Inam, Rohit, Sachin, Updesh for standing by me in good and bad times. I am indebted to my IIT Roorkee friends especially Amioy, Amit, Anurag, Ashwani, Jitendra, Manoj, Mohit, Nilesh, N. Srinivash, Pankaj, Ramesh and Shailendra for providing a stimulating and fun environment in which to learn and grow. Lastly, I offer my regards and blessings to all of those who supported me in any respect during the completion of the thesis. Words cannot completely express my love and gratitude to my family who have supported and encouraged me through this journey. I would like to thank my parents, brother, sister, sister-inlaw and brother-in-law for their life-long support, everlasting love, and sacrifices, which sustained my interest in research and motivated me towards the successful completion of this study. Finally, I thank the almighty God for the passion, strength, perseverance and the resources to complete this study. New Delhi Anikender Kumar
Abstract Air quality forecasting is a topic of great interest in air quality research due to direct association with health effects. Air quality forecasts provide pre-information to the general public about the status of air environment on which they can take precautionary steps to take their health. A short term forecast of air quality is required to take preventive and evasive action during episodes of airborne pollution. Delhi, the capital of India, is considered in one of the most polluted cities in India. The air quality of Delhi includes major air pollutants namely Respirable Suspended Particulate Matter (RSPM), Sulfur dioxide (SO 2 ), Oxides of Nitrogen (NO x ), Suspended Particulate Matter (SPM), Particles smaller than 2.5 micrometer (PM 2.5 ), Carbon monoxide (CO) and Ozone (O 3 ) etc. The primary objective of the thesis is to develop and evaluate the statistical and analytical models for forecasting the air quality. The thesis is divided into six chapters. The first chapter, an introductory chapter, comprises of the fundamental knowledge of air quality forecasting, air quality modelling etc. A brief description of the mathematical models and a literature review of various statistical and dispersion models studies have been included in this chapter. The daily forecasting of air quality index (AQI), indicator of the status of the regional air quality is made through three regression techniques namely Multiple Linear Regression (MLR) (Model 1), Principal Component Regression (PCR) (Model 2) and combination of PCR model with Auto-Regressive Integrated Moving Average (ARIMA) (PCR-ARIMA) (Model 3) through out the year on seasonal basis in chapter 2. The previous day s AQI and meteorological variables are used as the predictor variables. The application of PCA with regression model aims to reduce the collinearity in the datasets, which leads to the worst predictions/forecast and also determine the relevant independent variables for the prediction of air quality. The idea of combining two
models is to use each model s unique features to capture different patterns or features in the data set. In the present chapter, an attempt has been made to improve the forecast of PCR model by combining it with time series forecasting model i.e., ARIMA model. The coefficients involved in each of these models are estimated using the data of the years 2000-2005. The performance of the model has been evaluated for different seasons of the year 2006 through statistical measures. The models, discussed above, use a known relationship between the AQI and meteorological variables, which is often not known a priori. Therefore, an approach based on Neural Network (NN) with back propagation learning algorithm has been used in the present chapter 3 for forecasting the AQI of air pollutants. Although, NN has certain limitations, which have been circumvented by introducing the PCA to NN. It is always noticeable that forecast from two different models is more accurate than individual model s forecast. A three layered feed forward fully connected PCA-NN with ARIMA is used to forecast daily AQI in all four seasons using the same data of the years 2000-2005. The model performance has been evaluated with observed data set of the year 2006. The above discussion reflects that the chapters 2 and 3 are based on statistical techniques, which need the past observed time series data of many years, which is always not available and statistical techniques also don t include dispersion characteristics of the atmosphere. In order to overcome these limitations, the deterministic models can be used as an alternative to predict the concentration of pollutant in different meteorological conditions. In chapter 4, the analytical solution of advection-diffusion equation using Neumann (total reflection) boundary condition for a bounded domain and wind speed as a power law profile of vertical height above the ground is derived for point, line and area sources. The downwind and vertical eddy diffusivities are considered as an explicit function of downwind distance and vertical height. The source strength
or emission rate of air pollutants is one the important parameter of dispersion models. Therefore, emission s estimation of criteria pollutants namely CO, NOx and RSPM from various anthropogenic sources namely domestic, industries, power plants and vehicular, in the study area of Delhi for the year 2008-09 have been made in the present chapter. Analytical models require steady and horizontally homogeneous hourly surface and upper air meteorological observations, which are not easily available for most locations in India. Therefore, the planetary boundary layer and surface layer parameters are computed through Weather Research and Forecast (WRF) model (version 3.1.1), developed by National Center for Atmospheric Research (NCAR). The performance of the present analytical model is evaluated using observed concentrations of RSPM for the months of December, 2008 (representative of winter) and May, 2009 (representative of summer). The observed data of RSPM is obtained from Central Pollution Control Board (CPCB) at different locations in Delhi. In the previous chapters, air quality has been forecasted/predicted by statistical and analytical dispersion models. Although, dispersion models are calibrated in order to predict the concentrations in certain specific conditions. Since, the analytical dispersion models are based on idealized conditions, their performance becomes limited in real-time situations. To overcome this limitation, neural network model is coupled with the dispersion model in order to improve the air quality forecast and better reproduction of the real situation in chapter 5. The hourly concentration of PM 2.5 and RSPM pollutants from dispersion models and meteorological variables are used as input parameters to the neural network. This last chapter yields the final conclusions drawn from the study undertaken in the thesis. In addition, the limitations of the study and an outline for future work are included in this chapter.
A unique feature of this study is that the performances of all the proposed statistical and analytical models have evaluated through real observations, which have been analyzed quantitatively through statistical measures and qualitatively by, scatter diagrams.
Contents Certificate Acknowledgements Abstract Contents (i) List of Figures..(v) List of Tables (x) Chapter 1 General Introduction.1 1.1 Introduction..2 1.2 Air Quality Modelling..5 1.2.1 Gaussian Models..7 1.2.2 Eulerian Models.12 1.2.3 Lagrangian Models 13 1.2.4 Statistical Models...13 1.3 Air Quality Forecast...14 1.3.1 Health Alerts..14 1.3.2 Supplementing Existing Emission Control Programs...14 1.3.3 Operational Planning.15 1.3.4 Emergency Response.15 1.4 Literature Review on Air Quality Forecasting...15 1.5 Air Quality Forecasting Techniques..18 1.5.1 Climatology...18
1.5.2 Statistical Techniques 18 1.5.3 Three Dimensional (3-D) Air Quality Models...21 1.6 Organization of the Thesis.26 Chapter 2 Air Quality Forecasting using Regression Technique..32 2.1 Introduction 33 2.2 Description of Meteorological and Air Quality Data 35 2.3 Methodology..36 2.3.1 Model 1: MLR Model...38 2.3.2 Model 2: PCR Model...40 2.3.3 Model 3: ARIMA-PCR Model..43 2.4 Results and Discussion..44 2.5 Conclusions 71 Chapter 3 Air Quality Forecasting using Neural Network Based on Principal Component Analysis 73 3.1 Introduction 74 3.2 Methodology..76 3.2.1 Model s Characteristics.77 3.3 Results and Discussion..79 3.3.1 Air Quality Forecasting through Neural Network Model..80 3.3.2 Air Quality Forecasting through PCA-Neural Network Model.85 3.3.3 Air Quality Forecasting through PCA-Neural Network-ARIMA Model.91 3.4 Conclusions..100 ii
Chapter 4 Development of Analytical Model for Dispersion of Air Pollutants..102 4.1 Introduction..103 4.2 Study Area of Delhi.107 4.3 Air Quality Modelling System.108 4.3.1 Analytical Dispersion Model...108 4.3.2 Emission Inventory of Air Pollutants..118 4.3.3 WRF Model Configuration and Initialization..125 4.4 Results and Discussion 129 4.4.1 Estimated Emission Rate in Study Area..129 4.4.2 Validation of WRF Model...137 4.4.3 Model Simulations of RSPM...148 4.4.4 Validation of the Model...151 4.5 Conclusions..154 Chapter 5 Coupling of Air Dispersion Model and Neural Network for Air Quality Forecasting...157 5.1 Introduction..158 5.2 Methodology 162 5.2.1 Analytical Air Dispersion Model.162 5.2.2 Input Data of Emission and Meteorological Variables 166 5.2.3 PCA-Neural Network Model...168 5.2.4 Integrated Modelling Approach...169 5.3 Results and Discussion 170 5.4 Conclusions..180 iii
Chapter 6 Conclusions and Future Perspectives...182 6.1 Conclusions..183 6.2 Limitations and Future Perspectives 186 References...188 Appendix.211 Bio-Data..214 iv