Atmospheric Measurements By Portable Instruments: Data Correction Optimization And Application

Air pollution has an important impact on human health.The existing conventional air quality monitoring stations use traditional rack-mount air pollution monitors,which are large in size and high in cost.The number of stations that can be set up in a city is relatively limited,which results in the inadequacy of spatial coverage and resolution of pollutant monitoring.Particularly,it is difficult to monitor air pollution in urban microenvironments,such as bus stops and street canyons,which poses a challenge to fine-grained air quality management in the new situation.In recent years,portable instruments have gradually emerged with the advantages of small size and low cost,which can obtain a high-spatial-temporal resolution of urban air pollution characteristics,and serve as a complementary to the existing urban air quality monitoring network.However,the application of portable instruments in atmospheric environment observation still faces many challenges.The precision and accuracy of portable instruments are still not as good as traditional air pollution monitors.To improve the accuracy of the portable instrument,this study optimized the data correction algorithms of electrochemical sensors and micro Aethalometers.The real-world application of micro Aethalometer data correction was demonstrated in a vertical measurement campaign at Shenzhen meteorological tower.The work associated with electrochemical sensors was carried out at Jinan University atmospheric supersite in Guangzhou（JNU）.The electrochemical sensor data was obtained by a commercial sensor,Sniffer4D（Shenzhen Soarability Technologies Co.,Ltd.,Shenzhen,China）and the reference data was obtained by a set of regulatory-grade gas analyzers.The observation period was from October 1 2018 to March 312019.This study includes three gas pollutants CO,NO₂ and O₃,with a time resolution of 1 minute.The readings of the working electrode of the electrochemical sensor during the observation period covered three orders of magnitude（10⁰ to 10² n A）,indicating that the data obtained in this study are representative.Due to the weak electrode signal（n A magnitude）,it is vulnerable to the drift of electrical signal caused by environmental factors（such as temperature and humidity）,which is the main source of observation error of electrochemical sensors.A variety of electrochemical sensor data correction algorithms have been proposed in previous studies,but the performance of these algorithms has not been compared and evaluated.In this study,eight data correction algorithms were compared,among which six algorithms were from the literature（AS＿SA,AS＿AA,NZ＿BA,NZ＿OPT,PXB,CL）.Two algorithms（LY＿A and LY＿B）are newly developed in this study.To achieve a quantitative algorithm comparison,a new evaluation scheme is proposed in this study,which covers the following statistical parameters:Mean Bias（MB）,Mean Gross Error（MGE）,Root Mean Squared Error（RMSE）,correlation coefficient（Pearson’s r）precision and coefficient of variations in the mean absolute error（Cv MAE）,linear regression slope（DRS）and intercept（DRI）of reference data,Therefore,the correction performance of each algorithm can be evaluated comprehensively.The major findings of the evaluation are as follows:PXB performs best in CO correction,followed by CL.However,the scoring difference between CL and PXB is less than 5%,so both algorithms are suitable for CO correction.For the correction of NO₂,NZ＿BA has the highest score,followed by CL,which is only5%lower than NZ＿BA.Therefore,both of these algorithms can be used for the correction of NO₂.For the correction of O₃,LY＿B,the new algorithm proposed in this study,has the best result,followed by AS＿SA,whose score is 10%lower than that of LY＿B.Therefore,LY＿B and AS＿SA are recommended in this study for O₃ data correction.In addition,this study conducts an in-depth analysis of the dependence of each algorithm on different factors,such as temperature,humidity,pollutant concentration and cross-sensitivity were investigated.The results confirm that how well the algorithm performs strongly relies on how well it reduces its dependence on these factors.The work of the data correction algorithm development and application of the micro Aethalometers was carried out in the JNU supersite and the Shenzhen meteorological tower,respectively.The observation period at JNU site was from April13,2018 to April 22,2018,with a temporal resolution of 1 minute.Two models of micro Aethalometers（MA200 and AE51）and one reference instrument（AE33）were used.Previous studies had proposed the ONA algorithm to reduce the noise of electrical signals,and the Virkkula algorithm to correct the load effect caused by the nonlinear response between black carbon concentration and optical attenuation signals,respectively.However,these two algorithms only consider a specific aspect of data correction,and the full correction cannot be achieved by using them alone.In this study,a three-step method is proposed for the correction.The two algorithms are systematically combined to achieve the best correction results.In the first step,the ONA algorithm is used to reduce the signal noise.In the second step,the Virkkula algorithm is used to correct the nonlinear response.In the third step,a reference instrument is used to correct the slope of the signal response of micro Aethalometers.The results show that each step of the three-step method can improve the correlation coefficient with the reference instrument,and the three-step method can get better results than using either ONA or Virkkula algorithm alone.This correction application of black carbon data was carried out in Shenzhen Meteorological Tower from August 17 to September 1,2018,with a time resolution of 5 minutes.Measurement was performed at four heights,including 2m,100m,200m and 350m of the meteorological tower,respectively.The data were corrected using the three-step method mentioned above.The observation results were summarized as follows:The average e BC concentration at each height was2.5±1.4,3.1±1.9,2.6±1.6,2.8±2.2μg·m^-3,respectively,indicating that the black carbon in the boundary layer was not uniformly mixed in the vertical direction,but had a certain gradient distribution.Moreover,the gradient distribution has considerable diurnal variations.The gradient was weak in the evening and early morning.The gradient was more pronounced in the afternoon,with an increasing trend with height.Existing studies on the vertical distribution of black carbon in the boundary layer remain limit.The observational data accumulated in this study will be useful for future studies on the climate effects of black carbon.