Ordinal time series model for forecasting air quality index for ozone in Southern California

Air quality index (AQI) for ozone is currently divided into six states depending on the level of public health concern. Generalized linear type modeling is a convenient and effective way to handle the AQI state, which can be characterized as non-stationary ordinal-valued time series. Various link functions which include cumulative logit, cumulative probit, and complimentary log-log are considered, and the partial maximum likelihood method is used for estimation. For a comparison purpose, the identity link, which yields a multiple regression model on the cumulative probabilities, is also considered. Random time-varying covariates include past AQI states, various meteorological processes, and periodic components. For model selection and comparison, the partial likelihood ratio tests, AIC and SIC are used. The proposed models are applied to 3 years of daily AQI ozone data from a station in San Bernardino County, CA. An independent year-long data from the same station are used to evaluate the performance of day-ahead forecasts of AQI state. The results show that the logit and probit models remove the non-stationarity in residuals, and both models successfully forecast day-ahead AQI states with almost 90 % of the chance.     

Forecasting Daily Maximum Ozone Concentrations in the Athens Basin

In the work ozone data from the Liossion monitoring station of the Athens/PERPA network are analysed. Data cover the months May to September for the period 1987–93. Four statistical models, three multiple regression and one ARIMA (0,1,2), for the prediction of the daily maximum 1-hour ozone concentrations are developed. All models together, with a persistence forecast, are evaluated and compared with the 1993's data, not used in the models development. Validation statistics were used to assess the relative accuracy of models. Analysis, concerning the models' ability to forecast real ozone episodes, was also carried out. Two of the three regression models provide the most accurate forecasts. The ARIMA model had the worst performance, even lower than the persistence one. The forecast skill of a bivariate wind speed and persistence based regression model for ozone episode days was found to be quite satisfactory, with a detection rate of 73% and 60% for O3 >180 g m-3 and O3 >200 g m-3, respectively.     

Modified grey model for estimating traffic tunnel air quality

This study compared three forecasting models based on the mean absolute percentage errors (MAPE) of their accuracy in forecasting air pollution in a traffic tunnel: the Grey model (GM), the combination model used four sample point and five sample point prediction with GM (1,1)(GM(1,1)^4 + 5), and the modified grey model (MGM). An MGM was combined using the four points of the original sequence using the original grey prediction GM (1,1) for short-term forecasting. The proposed method cannot only enhance the prediction accuracy of the original grey model, but can also solve the jump data forecasting problem something for which the original grey model is inappropriate. The MAPE was applied to the models, and the MGM found the proposed method to be simple and efficient. The MAPE of MGM, calculated over 3 h of forecasts, were as follows: 10.12 (Upwind), 10.07 (Middle) and 7.68 (Downwind) for CO; 10.79 (Upwind), 6.05 (Middle) and 5.98 (Downwind) for NO _x , and 11.67 (Upwind), 7.32 (Middle) and 4.56 (Downwind) for NMHC. The MGM model results reveal that the combined forecasts can significantly decrease the overall forecasting error. Results of this demonstrate that MGM can accurately forecast air pollution in the Kaohsiung Chung–Cheng Tunnel.     

Minimax Statistical Models for Air Pollution Time Series. Application to Ozone Time Series Data Measured in Bordeaux

This paper deals with the application of l(infinity) (or minimax) optimization techniques to statistical modelling of high frequency air pollution data. The method was applied to ground-level ozone time-series data measured in Bordeaux over 4 years from 1998 to 2001. The aim of model building was to develop predictive models in order to provide forecasts of the maximal daily ground-level ozone concentration. Experimental results from this case study indicate that such techniques could be more appropriate than the commonly used l2 setting if only good estimation of high levels is of interest. When the free parameters are fitted by means of l(infinity) optimization techniques, the forecasting errors are more evenly distributed amongst the data points, resulting in a better estimation of high values. The paper compares the quality of forecasts produced by both a linear and a nonlinear model, using l2 and l(infinity) parameter optimization.     

最优化集成方法在城市臭氧数值预报中的应用研究

基于多模式(NAQPMS、CMAQ、CAMx、WRF-Chem)空气质量数值预报业务系统的滚动预报结果,结合站点观测资料,评估了最优化集成方法在城市臭氧数值预报中的可行性和预报效果。一年的评估结果表明:当训练期为15 d时,最优化集成方法能够得到相对较好的结果。总体而言,最优化集成方法对城市臭氧浓度变化趋势和浓度水平的预报效果明显优于单个模式,且在大部分城市优于多模式的最优预报,其预报值和观测的相关系数提高0.11以上,均方根误差降低约10μg/m~3;该方法对城市臭氧污染等级的预报能力也明显优于单个模式,特别是轻、中度污染。此外,在模拟偏差较大的城市,最优化集成方法对预报效果的改进更为显著;在模拟偏差较小的城市,该方法仍可进一步提升预报效果。     

Development of an Ozone Forecasting Model for Non-Attainment Areas in the State of Ohio

Ground level ozone is responsible for the formation ofsmog, and for a variety of adverse effects on bothhuman and plant life. High concentrations of groundlevel ozone occur during the summer months. This paperdescribes the development of a model to forecast themaximum daily concentration of ozone as a function ofthe maximum surface temperature, for ozonenon-attainment regions in Ohio. The model wasdeveloped by statistical analysis of existing data.Site-specific models were developed initially. Theverification and evaluation of the performancecriteria of the model at each site were explored bycomparing the model with an independent datasetcollected from that site. A generalized statewidemodel was developed from the site-specific models. Theperformance criteria of this model were verified andevaluated by employing the same independent datasetsemployed for the site-specific models. An exceedencemodel to predict the occurrence of ozone exceedencesover 100 ppb has also been presented.     

Some Dependence Models for the Time Between Ozone Exceedances in Mexico City

In this paper, we consider several modelling approaches for the mean time between exceedances of a given environmental threshold. The interest here resides in the time between ozone exceedances (also called ozone inter-exceedances times). The proposed models assume two basic density functions for the time between surpassings: the Weibull and the generalised exponential functions. Considering those distributions, a random effect with autoregressive structure is taken into account to determine unexpected changes in the mean of the inter-exceedances density functions. Those unexpected changes could be captured either by their scale parameter or by both their scale and shape parameters. The models are applied to ozone data from the monitoring network of Mexico City. Selection of the model that best explains the data is performed using the deviance information criterion and also the sum of the absolute values of the differences between the estimated and observed means of the inter-exceedances times. An analysis to detect the possible presence of change-points is also presented.     

Estimation of Ozone Trend in Central Greece, Based on Meteorologically Adjusted Time Series

Data referring to an approximately 8-year period (1999–2007) are analyzed in order to estimate the trend of the daily maximum hourly value of ozone concentration at the east coast of central Greece, where the summer background ozone concentration is high. A Kolmogorov–Zurbenko filter is applied to remove the short-term component from the raw time series of ozone and meteorological variables. Regression models are developed in order to produce meteorologically adjusted ozone time series, involving the noise-free temperature, relative humidity, and wind speed as independent variables. The analysis verifies that the meteorological adjustment provides better results on estimating ozone’s trend, which is found to be increasing (α?=?0.001) with an annual rate of 1.34?±?0.07?μg/m³. This trend could mainly be attributed to policy and changes in the emissions of ozone’s precursors. Additionally, the short-term component of ozone concentration is also meteorologically adjusted and its impact on the trend is examined. The analysis shows that its contribution is of minor importance when the ozone trend is adjusted by temperature, relative humidity, and wind speed. Moreover, the sea breeze circulation system that is frequently developed in the area influences the short-term and seasonal ozone variation, and therefore, it should be taken into account when producing meteorologically adjusted time series. The study’s conclusions could be exploited by environmental and agricultural authorities in order to develop their long-term strategies towards the air quality management.     

2016年秋季新乡市空气质量模式预报效果评估

基于新乡市空气质量数值预报平台,采用相关系数（r）、标准化平均偏差（NMB）等统计指标,系统评估2016年秋季新乡市嵌套网格空气质量预报模式（NAQPMS）和通用多尺度空气质量模式（CMAQ）的预报效果,对比分析2套模式不同预报时效和不同水平分辨率的空气质量等级预报准确率。结果显示：2套模式均较好地表征了各主要污染物的浓度变化特征,2套模式的等级预报准确率高于60%,其中CMAQ对中度及重度的预报等级准确率达到70%。对比模式24、48、72 h 3种预报时效效果,24 h预报时效的统计数据最优,说明24 h预报时效模拟结果可作为业务预报重要的支撑。     

臭氧综合指数与臭氧潜势指数的构建与应用

随着社会经济的快速发展,我国臭氧污染日益严重,因此,研发出能定量评估气象条件对臭氧污染影响程度的诊断指数,成为提高和改善气象服务质量的重要任务之一。利用中国大陆地区2018年温度、总云量、风速、风向、相对湿度等气象场数据与臭氧浓度数据,研究臭氧污染敏感气象条件,统计各气象因子分布在不同数值区间时发生臭氧污染事件的相对频率(即分指数),按照分指数最大值和最小值的差值大小进行排序,筛选出10个与臭氧污染密切相关的气象因子,将10个气象因子的分指数进行累加,即得出臭氧综合指数。随后,对各地构建臭氧综合指数时采用的气象要素进行统计,得到出现频率最高的3个气象要素,并参考这些气象要素构建了臭氧潜势指数。分别以臭氧潜势指数和臭氧综合指数对北京市2019年臭氧日最大浓度建立拟合预报模型,结果表明:两类指数的拟合预报值与实测值有着相似的变化趋势;利用臭氧综合指数计算得到的预报值与实测值的相关系数为0.76,优于利用臭氧潜势指数计算得到的预报值与实测值的相关系数(0.64)。     

Forecasting of daily total atmospheric ozone in Isfahan

A neural network combined to an artificial neural network model is used to forecast daily total atmospheric ozone over Isfahan city in Iran. In this work, in order to forecast the total column ozone over Isfahan, we have examined several neural networks algorithms with different meteorological predictors based on the ozone-meteorological relationships with previous day's ozone value. The meteorological predictors consist of temperatures (dry and dew point) and geopotential heights at standard levels of 100, 50, 30, 20 and 10 hPa with their wind speed and direction. These data together with previous day total ozone forms the input matrix of the neural model that is based on the back propagation algorithm (BPA) structure. The output matrix is the daily total atmospheric ozone. The model was build based on daily data from 1997 to 2004 obtained from Isfahan ozonometric station data. After modeling these data we used 3 year (from 2001 to 2003) of daily total ozone for testing the accuracy of model. In this experiment, with the final neural network, the total ozone are fairly well predicted, with an Agreement Index 76%.     

A Long-Term Electricity Dispatch Model with the TIMES Framework

A new Swiss TIMES (The Integrated MARKAL–EFOM System) electricity model with an hourly representation of inter-temporal detail and a century-long model horizon has been developed to explore the TIMES framework’s suitability as a long-term electricity dispatch model. To understand the incremental insights from this hourly model, it is compared to an aggregated model with only two diurnal timeslices like in most MARKAL/TIMES models. Two scenarios have been analysed with both models to answer the following questions: Are there differences in model solutions? What are the benefits of having a high number of timeslices? Are there any computational limitations? The primary objective of this paper is to understand the differences between the solutions of the two models, rather than Swiss policy implication or potential uncertainties in input parameters and assumptions. The analysis reveals that the hourly model offers powerful insights into the electricity generation schedule. Nevertheless, the TIMES framework cannot substitute for a dispatch model because some features cannot be represented; however, the long model time horizon and integrated system approaches of TIMES provide features not available in conventional dispatch models. The methodology of the model development and insights from the model comparison are described.     

Mixed Integer Linear Programming Models for Selecting Ground-Level Ozone Control Strategies

The traditional strategy for ground-level ozone control is to apply emission reductions across the board throughout certain time periods and locations. In this paper, we study various mixed integer linear programming (MILP) models that seek to select targeted control strategies for the Dallas Fort-Worth (DFW) region to reduce emissions, in order to achieve the State Implementation Plan (SIP) requirements with minimum cost. Statistics and optimization methods are used to determine a potential set of cost-effective control strategies for reducing ozone. These targeted control strategies are specified for different types of emission sources in various time periods and locations. Three MILP models, a static model, a sequential model, and a dynamic model, are studied in this research. These different MILP models allow decision makers to study how the targeted control strategies change under different circumstances. Meanwhile, two types of auxiliary variables are considered as supplemental control strategies in the optimization if the current set of control strategies is unable to reduce ozone to comply with the 8-h ozone standard. Results from the different models can provide decision makers with information concerning how the effectiveness of the control strategies varies with daily emission patterns and meteorology.     

江苏省空气质量预报准确度及改进方法分析

介绍了江苏省重污染天气监测预报预警系统以及大气重污染预警会商流程,将2015年13个地级市的模式预报、人工预报结果分别与实际观测值进行比较。结果表明:人工预报更准确,PM_(2.5)日均值、臭氧日最大8 h平均值、AQI 3个指标人工预报和实况的相关性分别比模式预报高出12.8%、0.3%、11.4%,平均标准误差(MNE)分别低20.7%、3.1%、23.1%。依据国家空气质量预报技术指南评分办法,对各市2015年全年空气质量级别为"良"时进行评分。通过开展07∶00预报更新,使2015年上半年空气质量预报级别得分平均提高了0.9分,全年级别得分平均提高了2.6分;通过改进模式预报参数,使PM_(2.5)日均预报值、臭氧日最大8 h平均预报值、AQI预报值和实际观测值的相关性比上年同期分别提高26.0%、5.0%、33.9%,MNE分别降低3.6%、31.3%、7.6%。     

空气质量预报在世界互联网大会中的应用

为支持世界互联网大会期间大气污染管控工作,利用人工结合数值模式预报的方式在第二届到第五届世界互联网大会期间开展空气质量预报工作。多模式系统中WRF-CMAQ对乌镇及浙北区域大气污染变化的趋势模拟最好,2016—2018年对AQI日均值模拟的平均分数偏差（MFB）和平均分数误差（MFE）分别为-1.3%~1.6%和24.3%~28.3%。会前48 h、72 h和96 h空气质量等级预报准确率分别为37.5%~83.3%、33.3%~90.0%和0~89.9%。会议期间乌镇的AQI日均值48 h预报准确率为33.3%~100%,等级预报准确率为66.7%~100%。与日常空气质量预报不同,会议期间预报还应重点关注大气污染过程,如有污染可能性,需要给出污染过程的起始时间、持续范围和浓度峰值等情况及其关键时间节点,有针对性地提出大气污染管控的措施建议,为会议期间空气质量保障提供技术支撑。     

A Wavelet-based Neural Network Model to Predict Ambient Air Pollutants’ Concentration

The present paper proposes a wavelet based recurrent neural network model to forecast one step ahead hourly, daily mean and daily maximum concentrations of ambient CO, NO₂, NO, O₃, SO₂ and PM_2.5 — the most prevalent air pollutants in urban atmosphere. The time series of each air pollutant has been decomposed into different time-scale components using maximum overlap wavelet transform (MODWT). These time-scale components were made to pass through Elman network. The number of nodes in the network was decided on the basis of the strength (power) of the corresponding input signals. The wavelet network model was then used to obtain one-step ahead forecasts for a period extending from January 2009 to June 2010. The model results for out of sample forecast are reasonably good in terms of model performance parameters such as mean absolute error (MAE), mean absolute percentage error (MAPE), root mean square error (RMSE), normalized mean absolute error (NMSE), index of agreement (IOA) and standard average error (SAE). The MAPE values for daily maximum concentrations of CO, NO₂, NO, O₃, SO₂ and PM2.5 were found to be 9.5%, 17.37%, 21.20%, 13.79%, 17.77% and 11.94%, respectively, at ITO, Delhi, India. Bearing in mind that the forecasts are for daily maximum concentrations tested over a long validation period, the forecast performance of the model may be considered as reasonably good. The model results demonstrate that a judicious selection of wavelet network design may be employed successfully for air quality forecasting.     

An integrated air quality forecast system for a metropolitan area

Air quality forecasting is an important issue in environmental research, due to the effects that air pollutants have on population health. To deal with this topic, in this work an integrated modelling system has been developed to forecast daily maximum eight hours ozone concentrations and daily mean PM10 concentrations, up to two days in advance, over an urban area. The presented approach involves two steps. In the first step, artificial neural networks are identified and applied to get point-wise forecasting. In the second step, the forecasts obtained at the monitoring station locations are spatially interpolated all over the domain using the cokriging technique, which allows to improve the spatial interpolation in the absence of densely sampled data. The integrated modelling system has been then applied to a case study over Northern Italy, performing a validation over space and time for the year 2004 and analyzing if the limit values for the protection of human health set by the European Commission are respected. The presented approach represents a fast and reliable way to provide decision makers and the general public with air quality forecasting, and to support prevention and precautionary measures.     

工业居民混合区降尘含量的灰色GM(1,1)预测

本文采有灰色理论中的预测方法,以９０—９４年度的降尘监测数据为基础,建立了邯郸市工业居民混合区降尘含量的ＧＭ（１,１）预测模型,其数学表达式为：＾Ｘ（１）（ｋ＋１）＝５６０２９ｅ００５６４１２ｋ－５３０８３,经三种方法对模型精度检验,由模式计算的预测值与历年的实测值的平均相对误差仅为－０００８％,关联度系数γ＝０９０５,后验差比值Ｃ＝００３,模型精度为一级,经与９５年度预测值与实测值对比验证,误差２７４％,说明本模型能真实地反映该区降尘的发展变化规律,预测数据可作为制定该区降尘防治规划的依据。     

江苏省空气质量预报与实测结果比对研究

选取2015年1—8月江苏地区NAQPMS、CMAQ、CAMx、WRF-Chem 4个模式预报结果与实测值进行比对分析,结果表明,标准化分数偏差(MFB)为-0.066 5~0.201 1,标准化分数误差(MFE)最大值为0.381 8,均在理想范围内,其中CAMx预报效果相对较好,WRF-Chem有一定误差。4个模式相比,NAQPMS对于PM_(10)的模拟性能较好,各模式对PM_(2.5)模拟性能相近,CMAQ和CAMx对O_3模拟较好,WRF-Chem对CO模拟较好,各模式对SO_2和NO_2的模拟都需进一步优化。