Ground-level ozone forecasting model considering precursor emissions and heat-island effect

In the United States (U.S.), a number of regional episodic ground-level Ozone (O3) control programs have been established nationwide in response to the persistence of the O3 air pollution problem. Key components of these programs are the O3 monitoring and forecasting of probable high O3 days, and subsequent promulgation of an Ozone Action Day in the community. Air quality monitoring stations are located in numerous cities around the U.S., however, many small cities and rural areas do not have monitoring stations. Therefore, reliable air quality models are needed to serve as a decision-making tool in city planning and air quality management programs. For O3 forecasting programs, statistical models are often used due to the ease of use and fairly good forecasts. The existing models generally require prior day's O3 values.

The objectives of this study are: (1) to develop a daily maximum 8-hour average O3 concentration prediction model considering O3 precursor emissions and heat-island effect using the monitoring data for rural areas of Tupelo and Hernando in Northern Mississippi, (2) to validate the model using historical measured O3 data from the monitoring sites, and (3) to implement the model in Oxford, Mississippi, where there is no monitoring station. A multiple linear regression model for predicting daily maximum 8-hour average O3 concentrations was developed using 1996-2000 data for rural areas (Tupelo and Hernando) in Northern Mississippi. The developed model not only contains key meteorological parameters affecting O3 levels but also includes O3 precursor as well as surface temperature variables.

It has been found that the long-term change in traffic is highly associated with the change in O3 concentrations. However, none of the previously developed multiple regression and Artificial Neural Network models for predicting O3 concentrations incorporate precursor emissions into the model. Therefore, in this study, the effects of daily Volatile Organic Compounds (VOC) and Oxides of Nitrogen (NOx) emissions from highway motor vehicles and point sources are considered in the development of the new O3 prediction model.

An additional key factor affecting O3 concentrations is the “heat-island” effect, which can be represented by the surface temperature of an area. On warm summer days the air in a city can be 3-4 °C (6-8 °F) hotter than its surrounding areas. The increase in air temperature results in higher O3 concentrations. In many areas of the nation, a warming of 2.2 °C (4 °F) could increase O3 concentrations by about 5%. A preliminary study of the surface temperature in Oxford shows that the “heat-island” effect is present even for small rural towns. During the hottest hour in 2001, the weighted average surface temperature in the city area of Oxford was about 5 °C (9 °F) higher than that of surrounding areas, and about 15 °C (27 °F) higher than the air temperature.

In this study, surface temperatures were predicted using a heat-conduction model, calibrated for asphalt and concrete surfaces in earlier studies. The model predicts a temperature at the surface using local meteorological data and appropriately selected thermal properties of surface materials. Surface temperatures of eight surface classes; asphalt, concrete, grass, trees, soil, buildings, water, and unknown were predicted. The percentage areas of all surface classes in the defined 8x8 sq km study area of Oxford were obtained from surface classification results using high-resolution satellite imagery. Weighted average surface temperatures then were calculated based on the percentage areas of these surface classes. The calculated daily maximum weighted average surface temperatures for Tupelo during a typical week of summer 2001 show a similar trend as the levels of O3 concentration. Its Pearson correlation value (R) with the daily maximum 8-hour average O3 concentration is +0.911 indicating a stronger correlation with O3 concentrations than the R values of the maximum and minimum air temperatures (+0.716 and –0.841). Hence, the surface temperature variable is included in this new O3 model.

The O3 concentration model developed in this study is a function of air temperature, surface temperature, wind speed and direction, precipitation, cloud cover, solar radiation, vehicular traffic and emissions, point source emissions, aircraft operations, and day of the year. The model has a multiple correlation R of +0.73 and standard error of estimate of 0.012 parts per million (ppm). The O3 predictions for the hottest day in Tupelo and Hernando in 2001 differ by 10% and 2%, respectively, from the measured values. The model is implemented for Oxford and predictions for selected days in 2001 are in the range of 0.016-0.048 ppm, which are reasonably accurate considering the expected background level of O3. These results show reasonableness of the model for predicting O3 concentrations in the rural areas of Mississippi. The model will need to be calibrated for use in urban and metropolitan areas.

The O3 model, validated for Tupelo and Hernando, gave reasonably accurate prediction results on the selected days in 2001. The model was implemented in Oxford showing reasonably accurate prediction results. This model is applicable to locations that do not have monitoring stations since it does not require the prior day's O3 value to predict O3 concentrations. On the other hand, it can be used to identify candidate locations for establishing an air quality monitoring station. The developed O3 model can also be used in air quality management programs such as air pollution control programs, and/or used in air quality trend analysis. In addition, it can serve as a decision-making tool in landuse and transportation planning for local agencies. The inclusion of the surface temperature variable in the model helps to evaluate the adverse impact of built-up areas on air quality degradation, and lead to improved decision-making for air quality management strategies, such as providing more trees, parks, lakes, and light-color roofing materials.