M.Sc Thesis

M.Sc StudentGinzburg Yulia
SubjectPredicting of Ozone Levels Using Meteorological and Air
Quality Data in Urban Sites of Israel
DepartmentDepartment of Civil and Environmental Engineering
Supervisor PROFESSOR EMERITUS Yaacov Mamane


The central region of Israel has been suffering from exposure to high ozone concentrations. This in part is due to the strong solar radiation and large emissions of nitrogen oxides and volatile organic compounds from mainly motor vehicles in urban centers in Israel.  The purpose of this study is to develop ozone predicting models and to apply them to Israel urban sites. Existing models from other countries will be used too. For the model development we used air quality data of 2000 and 2001 from Modiin, an urban area downwind of the Tel Aviv Metropolitan region. For this site meteorological and air quality data are available. Several steps were taken in order to choose an appropriate site. Multiple regression analyses, linear and non-linear, were selected as the main statistical tool to develop an ozone forecasting model. 

Models were evaluated by comparing predicted with measured ozone concentrations of the urban site where data were obtained. The comparison methodology and evaluation process of the models are complex and not easily defined. Known statistical parameters were used in the evaluation process. The simplest and by for the best ozone prediction models were: the multiple non-linear regression (MNLR) model - based on simple input data such as today’s ozone maximum concentrations, next day’s temperature maximum, and average day time (9:00 a.m. - 4:00 p.m.) relative humidity, and the multiple linear regression (MLR) model - based on such data as today’s ozone maximum concentrations and next day’s temperature maximum only. The quality of the models was similar, both provided the best results, just the MNLR model is more complicate and require more input for computing.

The Multiple Linear Regression (MLR) model was applied to air quality data from Tivon monitoring station, where significant ozone pollution is observed. The complex terrain of the site makes the study of ozone formation and transportation more complicated. Low correlation for ozone concentrations with meteorological data is also characteristic for Tivon. These factors affect the model that shows quite poor results. Sites like Tivon need more thorough investigations to develop suitable ozone prediction models.