A comparison of nonlinear regression and neural network models for ground-level ozone forecasting

A hybrid nonlinear regression (NLR) model and a neural network (NN) model, each designed to forecast next-day maximum 1-hr average ground-level O3 concentrations in Louisville, KY, were compared for two O3 seasons--1998 and 1999. The model predictions were compared for the forecast mode, using forecasted meteorological data as input, and for the hindcast mode, using observed meteorological data as input. The two models performed nearly the same in the forecast mode. For the two seasons combined, the mean absolute forecast error was 12.5 ppb for the NLR model and 12.3 ppb for the NN model. The detection rate of 120 ppb threshold exceedances was 42% for each model in the forecast mode. In the hindcast mode, the NLR model performed marginally better than the NN model. The mean absolute hindcast error was 11.1 ppb for the NLR model and 12.9 ppb for the NN model. The hindcast detection rate was 92% for the NLR model and 75% for the NN model.     

An enhanced PM2.5 air quality forecast model based on nonlinear regression and back-trajectory concentrations

An enhanced PM_2.5 air quality forecast model based on nonlinear regression (NLR) and back-trajectory concentrations has been developed for use in the Louisville, Kentucky metropolitan area. The PM_2.5 air quality forecast model is designed for use in the warm season, from May through September, when PM_2.5 air quality is more likely to be critical for human health. The enhanced PM_2.5 model consists of a basic NLR model, developed for use with an automated air quality forecast system, and an additional parameter based on upwind PM_2.5 concentration, called PM24. The PM24 parameter is designed to be determined manually, by synthesizing backward air trajectory and regional air quality information to compute 24-h back-trajectory concentrations. The PM24 parameter may be used by air quality forecasters to adjust the forecast provided by the automated forecast system. In this study of the 2007 and 2008 forecast seasons, the enhanced model performed well using forecasted meteorological data and PM24 as input. The enhanced PM_2.5 model was compared with three alternative models, including the basic NLR model, the basic NLR model with a persistence parameter added, and the NLR model with persistence and PM24. The two models that included PM24 were of comparable accuracy. The two models incorporating back-trajectory concentrations had lower mean absolute errors and higher rates of detecting unhealthy PM2.5 concentrations compared to the other models.     

Accuracy and reliability of an automated air quality forecast system for ozone in seven Kentucky metropolitan areas

An automated forecast system for ozone in seven Kentucky metropolitan areas has been operational since 2004. The forecast system automatically downloads the required input data twice each day, produces next-day forecasts of metro area peak 8-h average ozone concentration using a computer coded hybrid nonlinear regression (NLR) model, and posts the results on a website. The automated models were similar to previous NLR models, first applied to forecasting ozone in the Louisville metro area. The forecast system operated reliably during the 2004 and 2005 O₃ seasons, producing at least one forecast per day better than 99% of the time. The forecast accuracy of the automated system was good. For all 2004 and 2005 forecasts, the mean absolute error was equal to 8.7 ppb, or 15.6% of the overall mean concentration. The overall detection rate of air quality standard exceedences was 56%, and the overall false alarm rate was 42%. In Louisville, the performance of the automated system was comparable to that of expert forecasters using the NLR model as a forecast tool.     

An enhanced ozone forecasting model using air mass trajectory analysis

An enhanced ozone forecasting model using nonlinear regression and an air mass trajectory parameter has been developed and field tested. The model performed significantly better in predicting daily maximum 1-h ozone concentrations during a five-year model calibration period (1993–1997) than did a previously reported regression model. This was particularly true on the 28 “high ozone” days ([O₃]>120 ppb) during the period, for which the mean absolute error (MAE) improved from 21.7 to 12.1 ppb. On the 77 days meteorologically conducive to high ozone, the MAE improved from 12.2 to 9.1 ppb, and for all 580 calibration days the MAE improved from 9.5 to 8.35 ppb. The model was field-tested during the 1998 ozone season, and performed about as expected. Using actual meteorological data as input for the ozone predictions, the MAE for the season was 11.0 ppb. For the daily ozone forecasts, which used meteorological forecast data as input, the MAE was 13.4 ppb. The high ozone days were all anticipated by the ozone forecasters when the model was used for next day forecasts.     

Trends in Meteorologically Adjusted Ozone Concentrations in Six Kentucky Metro Areas, 1998–2002

Abstract

Meteorologically adjusted ozone (O₃) concentrations during five recent O₃ seasons (1998-2002) were computed for six Kentucky metro areas using a nonlinear regression model originally developed for forecasting ground-level O₃ concentrations. The meteorological adjustment procedure was based on modifying actual measured O₃ concentrations according to model-predicted responses to climate departures with respect to a reference year. For all six Kentucky metro areas, meteorologically adjusted O₃ concentrations declined over the five-year period. The linear best-fit rate of decline in mean adjusted O₃ concentrations ranged from 0.9 to 2.6 ppb/yr for these metro areas; the average rate of decline was 1.6 ppb/yr. The rates of decline in meteorologically adjusted extreme value (e.g., 95th percentile) concentrations were approximately the same, but there is greater statistical uncertainty concerning the extreme value trends.     

Fuzzy system models combined with nonlinear regression for daily ground-level ozone predictions

This study focuses on applying a Takagi–Sugeno fuzzy system and a nonlinear regression (NLR) model for ozone predictions in six Kentucky metropolitan areas. The fuzzy “c-means” clustering technique coupled with an optimal output predefuzzification approach (least square method) was used to train the Takagi–Sugeno fuzzy system. The fuzzy system was tuned by specifying the number of rules and the fuzziness factor. The NLR models were based in part on a previously reported, trajectory-based hybrid NLR model that has been used for years for forecasting ground-level ozone in Louisville, KY. The NLR models were each composed of an interactive nonlinear term and several linear terms. Using a common meteorological parameter set as input variables, the NLR models and the Takagi–Sugeno fuzzy systems model exhibited equivalent forecasting performance on test data from 2004. For all 2004 ozone season forecasts for the six metropolitan areas, the mean absolute error was 8.1 ppb for the NLR model and 8.0 ppb for the Takagi–Sugeno fuzzy model. When a nonlinear term (which was part of the NLR model) was included in the fuzzy model, the combined NLR–fuzzy model had slightly better performance than the original NLR model. For all 2004 metropolitan area forecasts, the mean absolute error of the NLR–fuzzy model forecasts was 7.7 ppb. These small differences may be statistically significant, but for practical purposes the performance of the fuzzy models was equivalent to that of the NLR models.     

Neural Network Model for Predicting Peak Photochemical Pollutant Levels

ABSTRACT

In this paper, an attempt is made for the 24-hr prediction of photochemical pollutant levels using a neural network model. For this purpose, a model is developed that relates peak pollutant concentrations to meteorological and emission variables and indexes. The analysis is based on measurements of O₃ and NO₂ from the city of Athens. The meteorological variables are selected to cover atmospheric processes that determine the fate of the airborne pollutants while special care is taken to ensure the availability of the required input data from routine observations or forecasts. The comparison between model predictions and actual observations shows a good agreement. In addition, a series of sensitivity tests is performed in order to evaluate the sensitivity of the model to the uncertainty in meteorological variables. Model forecasts are generally rather insensitive to small perturbations in most of the input meteorological data, while they are relatively more sensitive in changes in wind speed and direction.     

Forecasting Peak Daily Ozone Levels: Part 2—A Regression with Time Series Errors Model Having a Principal Component Trigger to Forecast 1999 and 2002 Ozone Levels

Abstract

A modified time series approach, a Box-Jenkins regression with time series errors (RTSE) model plus a principal component (PC) trigger, has been developed to forecast peak daily 1-hr ozone (O₃) in real time at the University of Wisconsin-Milwaukee North (UWM-N) during 1999 and 2002. The PC trigger acts as a predictor variable in the RTSE model. It tries to answer the question: will the next day be a possible high O₃ day? To answer this question, three PC trigger rules were developed: (1) Hi-Low Checklist, (2) Discriminant Function Approach I, and (3) Discriminant Function Approach II. Also, a pure RTSE model without including the PC trigger and a persistence approach were tested for comparison. The RTSE model with DFA I successfully forecast the only two 1-hr federal exceedances (124 ppb), one in 1999 and one in 2002. In terms of the O₃ 100-ppb exceedances, 60–80% of the incorrect forecasts occurred with incorrect PC decisions. A few others may have been caused by unexpected O_3- weather relations. When the three approaches used UWM-N data to forecast a 100-ppb exceedance somewhere in the Milwaukee, WI, metropolitan area, their performance was dramatically improved: the false alarm rate was reduced from 0.89 to 0.44, and the probability of detection was increased from 0.71 to 0.88.     

Ozone Formation in California's San Joaquin Valley: A Critical Assessment of Modeling and Data Needs

ABSTRACT

Data from the 1990 San Joaquin Valley Air Quality Study/ Atmospheric Utility Signatures, Predictions, and Experiments (SJVAQS/AUSPEX) field program in California's San Joaquin Valley (SJV) suggest that both urban and rural areas would have difficulty meeting an 8-hr average O₃ standard of 80 ppb. A conceptual model of O₃ formation and accumulation in the SJV is formulated based on the chemical, meteorological, and tracer data from SJVAQS/ AUSPEX. Two major phenomena appear to lead to high O₃ concentrations in the SJV: (1) transport of O₃ and precursors from upwind areas (primarily the San Francisco Bay Area, but also the Sacramento Valley) into the SJV, affecting the northern part of the valley, and (2) emissions of precursors, mixing, transport (including long-range transport), and atmospheric reactions within the SJV responsible for regional and urban-scale (e.g., downwind of Fresno and Bakersfield) distributions of O₃. Using this conceptual model, we then conduct a critical evaluation of the meteorological model and air quality model. Areas of model improvements and data needed to understand and properly simulate O₃ formation in the SJV are highlighted.     

Air Quality Modeling in the South Coast Air Basin of California: What Do the Numbers Really Mean?

Abstract

This study evaluates air quality model sensitivity to input and to model components. Simulations are performed using the California Institute of Technology (CIT) airshed model. Results show the impacts on ozone (O₃) concentration in the South Coast Air Basin (SCAB) of California because of changes in: (1) input data, including meteorological conditions (temperature, UV radiation, mixing height, and wind speed), boundary conditions, and initial conditions (ICs); and (2) model components, including advection solver and chemical mechanism. O₃ concentrations are strongly affected by meteorological conditions and, in particular, by temperature. ICs also affect O₃ concentrations, especially in the first 2 days of simulation. On the other hand, boundary conditions do not significantly affect the absolute peak O₃ concentration, although they do affect concentrations near the inflow boundaries. Moreover, predicted O₃ concentrations are impacted considerably by the chemical mechanism. In addition, dispersion of pollutants is affected by the advection routine used to calculate its transport. Comparison among CIT, California Photochemical Grid Model (CALGRID), and Urban Airshed Model air quality models suggests that differences in O₃ predictions are mainly caused by the different chemical mechanisms used. Additionally, advection solvers contribute to the differences observed among model predictions. Uncertainty in predicted peak O₃ concentration suggests that air quality evaluation should not be based solely on this single value but also on trends predicted by air quality models using a number of chemical mechanisms and with an advection solver that is mass conservative.     

The New England Air Quality Forecasting Pilot Program: Development of an Evaluation Protocol and Performance Benchmark

Abstract

The National Oceanic and Atmospheric Administration recently sponsored the New England Forecasting Pilot Program to serve as a “test bed” for chemical forecasting by providing all of the elements of a National Air Quality Forecasting System, including the development and implementation of an evaluation protocol. This Pilot Program enlisted three regional-scale air quality models, serving as prototypes, to forecast ozone (O₃) concentrations across the northeastern United States during the summer of 2002. A suite of statistical metrics was identified as part of the protocol that facilitated evaluation of both discrete forecasts (observed versus modeled concentrations) and categorical forecasts (observed versus modeled exceedances/nonexceedances) for both the maximum 1-hr (125 ppb) and 8-hr (85 ppb) forecasts produced by each of the models. Implementation of the evaluation protocol took place during a 25-day period (August 5–29), utilizing hourly O₃ concentration data obtained from over 450 monitors from the U.S. Environment Protection Agency’s Air Quality System network.     

Meteorological Factors of Ozone Predictability at Houston,Texas

ABSTRACT

Several ozone modeling approaches were investigated to determine if uncertainties in the meteorological data would be sufficiently large to limit the application of physically realistic ozone (O₃) forecast models. Three diagnostic schemes were evaluated for the period of May through September 1997 for Houston, TX. Correlations between measured daily maximum and model calculated O₃ air concentrations were found to be 0.70 using a linear regression model, 0.65 using a non-advective box model, and 0.49 using a three-dimensional (3-D) transport and dispersion model. Although the regression model had the highest correlation, it showed substantial underestimates of the highest concentrations. The box model results were the most similar to the regression model and did not show as much underestimation. The more complex 3-D modeling approach yielded the worst results, likely resulting from O₃ maxima that were driven by local factors rather than by the transport of pollutants from outside of the Houston domain. The highest O₃ concentrations at Houston were associated with light winds and meandering trajectories. A comparison of the gridded meteorological data used by the 3-D model to the observations showed that the wind direction and speed values at Houston differed most on those days on which the O₃ underestima-tions were the greatest. These periods also tended to correspond with poor precipitation and temperature estimates. It is concluded that better results are not just obtained through additional modeling complexity, but there needs to be a comparable increase in the accuracy of the meteorological data.     

Application of Artificial Neural Networks to Modeling and Prediction of Ambient Ozone Concentrations

ABSTRACT

The deterministic modeling of ambient O₃ concentrations is difficult because of the complexity of the atmospheric system in terms of the number of chemical species; the availability of accurate, time-resolved emissions data; and the required rate constants. However, other complex systems have been successfully approximated using artificial neural networks (ANNs). In this paper, ANNs are used to model and predict ambient O₃ concentrations based on a limited number of measured hydrocarbon species, NO_x compounds, temperature, and radiant energy. In order to examine the utility of these approaches, data from the Coastal Oxidant Assessment for Southeast Texas (COAST) program in Houston, TX, have been used. In this study, 53 hydrocarbon compounds, along with O₃, nitrogen oxides, and meteorological data were continuously measured during summer 1993. Steady-state ANN models were developed to examine the ability of these models to predict current O₃ concentrations from measured VOC and NO concentrations. To predict the future concentrations of O₃, dynamic models were also explored and were used for extraction of chemical information such as reactivity estimations for the VOC species.

The steady-state model produced an approximation of O₃ data and demonstrated the functional relationship between O₃ and VOC-NO_x concentrations. The dynamic models were able to the adequately predict the O₃ concentration and behavior of VOC-NO_x-O₃ system a number of hourly intervals into the future. For 3 hr into the future, O₃ concentration could be predicted with a root-mean squared error (RMSE) of 8.21 ppb. Extending the models further in time led to an RMSE of 11.46 ppb for 5-hr-ahead values. This prediction capability could be useful in determining when control actions are needed to maintain measured concentrations within acceptable value ranges.     

Prediction of the Vertical Profile of Ozone Based on Ground-Level Ozone Observations and Cloud Cover

Abstract

A number of statistical techniques have been used to develop models to predict high-elevation ozone (O₃) concentrations for each discrete hour of day as a function of elevation based on ground-level O₃ observations. The analyses evaluated the effect of exclusion/inclusion of cloud cover as a variable. It was found that a simple model, using the current maximum ground-level O₃ concentration and no effect of cloud cover provided a reasonable prediction of the vertical profile of O₃, based on data analyzed from O₃ sites located in North Carolina and Tennessee. The simple model provided an approach that estimates the concentration of O₃ as a function of elevation (up to 1800 m) based on the statistical results with a ±13.5 ppb prediction error, an R² of 0.56, and an index of agreement, d ₁, of 0.66. The inclusion of cloud cover resulted in a slight improvement in the model over the simple regression model. The developed models, which consist of two matrices of 24 equations (one for each hour of day for clear to partly cloudy conditions and one for cloudy conditions), can be used to estimate the vertical O₃ profile based on the inputs of the current day’s 1-hr maximum ground-level O₃ concentration and the level of cloud cover.     

Assessment of a regulatory model's performance relative to large spatial heterogeneity in observed ozone in Houston,Texas

In Houston, some of the highest measured 8-hr ozone (O₃) peaks are characterized by sudden increases in observed concentrations of at least 40 ppb in 1 hr, or 60 ppb in 2 hr. Measurements show that these large hourly changes appear at only a few monitors and span a narrow geographic area, suggesting a spatially heterogeneous field of O₃ concentrations. This study assessed whether a regulatory air quality model (AQM) can simulate this observed behavior. The AQM did not reproduce the magnitude or location of some of the highest observed hourly O₃ changes, and it also failed to capture the limited spatial extent. On days with measured large hourly changes in O₃ concentrations, the AQM predicted high O₃ over large regions of Houston, resulting in overpredictions at several monitors. This analysis shows that the model can make high O₃, but on these days the predicted spatial field suggests that the model had a different cause. Some observed large hourly changes in O₃ concentrations have been linked to random releases of industrial volatile organic compounds (VOCs). In the AQM emission inventory, there are several emission events when an industrial point source increases VOC emissions in excess of 10,000 mol/hr. One instance increased predicted downwind O₃ concentrations up to 25 ppb. These results show that the modeling system is responsive to a large VOC release, but the timing and location of the release, and meteorological conditions, are critical requirements. Attainment of the O₃ standard requires the use of observational data and AQM predictions. If the large observed hourly changes are indicative of a separate cause of high O₃, then the model may not include that cause, which might result in regulators enacting control strategies that could be ineffective.

Implications To show the attainment of the O₃ standard, the U.S. Environmental Protection Agency (EPA) requires the use of observations and model predictions under the assumption that simulations are capable of reproducing observed phenomena. The regulatory model is unable to reproduce observed behavior measured in the observational database. If the large observed hourly changes were indicative of a separate cause of high O₃, then the model would not include that cause. Inaccurate model predictions may prompt air quality regulators to enact control strategies that are effective in the modeling system, but prove ineffective in the real world.     

Effects of 10% biofuel substitution on ground level ozone formation in Bangkok,Thailand

The Thai Government's search for alternatives to imported petroleum led to the consideration of mandating 10% biofuel blends (biodiesel and gasohol) by 2012. Concerns over the effects of biofuel combustion on ground level ozone formation in relation to their conventional counterparts need addressing. Ozone formation in Bangkok is explored using a trajectory box model. The model is compared against O₃, NO, and NO₂ time concentration data from air monitoring stations operated by the Thai Pollution Control Department. Four high ozone days in 2006 were selected for modeling. Both the traditional trajectory approach and a citywide average approach were used. The model performs well with both approaches but slightly better with the citywide average. Highly uncertain and missing data are derived within realistic bounds using a genetic algorithm optimization. It was found that 10% biofuel substitution will lead to as much as a 16 ppb peak O₃ increase on these four days compared to a 48 ppb increase due to the predicted vehicle fleet size increase between 2006 and 2012. The approach also suggests that when detailed meteorological data is not available to run three dimensional airshed models, and if the air is stagnant or predominately remains over an urban area during the day, that a simple low cost trajectory analysis of O₃ formation may be applicable.     

An intercomparison of formaldehyde measurement techniques at ambient concentration

A study was conducted to evaluate five techniques for determining ambient formaldehyde concentrations. One technique used a spectroscopic determination, and the other four techniques used derivatization followed by fluorometric analysis or high-performance liquid chromatography with detection by u.v. absorption. Formaldehyde was generated by two techniques. In the first technique, zero air was bubbled through a solution of aqueous formaldehyde to produce gas-phase formaldehyde. Various compounds serving as possible interferences were added singly or in combination to these air mixtures. In the second technique, formaldehyde was generated as a product from controlled irradiations of hydrocarbons and nitrogen oxides in a smog chamber operated in a dynamic mode. The study was conducted as a blind intercomparison with no knowledge by the participants of the HCHO concentrations or the interferences added.The data from each of the techniques were compared against mean values in each sampling period. For formaldehyde in zero air, average deviations for each of the techniques ranged between 15 and 30%. At a formaldehyde concentration of 10 ppb, each technique showed no evidence for interferences by O₃ (190 ppbv), NO₂ (300 ppbv), SO₂ (20 ppbv), and H₂O₂ (7 ppbv). The agreement for formaldehyde concentrations measured for the photochemical mixtures was similar to that of the mixtures in zero air.Ambient measurements were also performed on three evenings and for one 36-h period. Ambient formaldehyde concentrations ranged from 1 to 10 ppbv. Ambient H₂O₂ measurements were also performed. A strong correlation in the diurnal concentration profile for formaldehyde and H₂O₂ was observed over the 36-h period.     

An Assessment of the Emissions Inventory Processing Systems EMS-2001 and SMOKE in Grid-Based Air Quality Models

Abstract

In the United States, emission processing models such as Emissions Modeling System-2001 (EMS-2001), Emissions Preprocessor System-Version 2.5 (EPS2.5), and the Sparse Matrix Operator Kernel Emissions (SMOKE) model are currently being used to generate gridded, hourly, speciated emission inputs for urban and regional-scale photochemical models from aggregated pollutant inventories. In this study, two models, EMS-2001 and SMOKE, were applied with their default internal data sets to process a common inventory database for a high ozone (O₃) episode over the eastern United States using the Carbon Bond IV (CB4) chemical speciation mechanism. A comparison of the emissions processed by these systems shows differences in all three of the major processing steps performed by the two models (i.e., in temporal allocation, spatial allocation, and chemical speciation). Results from a simulation with a photochemical model using these two sets of emissions indicate differences on the order of ±20 ppb in the predicted 1-hr daily maximum O₃ concentrations. It is therefore critical to develop and implement more common and synchronized temporal, spatial, and speciation cross-reference systems such that the processes within each emissions model converge toward reasonably similar results. This would also help to increase confidence in the validity of photochemical grid model results by reducing one aspect of modeling uncertainty.     

A new method to estimate air-quality levels using a synoptic-regression approach. Part I: Present-day O3 and PM10 analysis

In order to make projections for future air-quality levels, a robust methodology is needed that succeeds in reconstructing present-day air-quality levels. At present, climate projections for meteorological variables are available from Atmospheric-Ocean Coupled Global Climate Models (AOGCMs) but the temporal and spatial resolution is insufficient for air-quality assessment. Therefore, a variety of methods are tested in this paper in their ability to hindcast maximum 8 hourly levels of O₃ and daily mean PM₁₀ from observed meteorological data. The methods are based on a multiple linear regression technique combined with the automated Lamb weather classification. Moreover, we studied whether the above-mentioned multiple regression analysis still holds when driven by operational ECMWF (European Center for Medium-Range Weather Forecast) meteorological data. The main results show that a weather type classification prior to the regression analysis is superior to a simple linear regression approach. In contrast to PM₁₀ downscaling, seasonal characteristics should be taken into account during the downscaling of O₃ time series. Apart from a lower explained variance due to intrinsic limitations of the regression approach itself, a lower variability of the meteorological predictors (resolution effect) and model deficiencies, this synoptic-regression-based tool is generally able to reproduce the relevant statistical properties of the observed O₃ distributions important in terms of European air quality Directives and air quality mitigation strategies. For PM₁₀, the situation is different as the approach using only meteorology data was found to be insufficient to explain the observed PM₁₀ variability using the meteorological variables considered in this study.     

Influence of ship emissions on ozone concentrations around coastal areas during summer season

The influence of ship emissions on ozone (O₃) concentrations in a coastal area (CA) including Busan port, Korea was examined based on a numerical modeling approach during a high O₃ episode. The analysis was performed by two sets of simulation scenarios: (1) with ship emissions (e.g., TOTAL case) and (2) without ship emissions (e.g., BASE case). A process analysis (PA) (the integrated processes rate (IPR) and integrated reaction rate (IRR) analyses) was used to evaluate the relative contributions of individual physical and chemical processes in O₃ production in and around the CA (e.g., sites of Dong Sam (DS) and Dae Yeon (DY)). The model study suggested the possibility that pollutant gases emitted from the ships traversing Busan port can exert a direct impact on the O₃ concentration levels in the CA. Largest impacts of ship emissions on the O₃ concentrations were predicted at the coast (up to 15 ppb) and at inland locations (about 5 ppb) due to both the photochemical production of pollutant gases emitted from the ships and meteorological conditions. From the PA, the photochemical production of O₃ (P(O₃)) due to ship emissions in the CA was found to increase by a mean of 1.5 ppb h^?1 (especially by ≥10 ppb h^?1 at the DS site) during the day.