Carbon Monoxide (CO) and Ozone (O₃) Concentrations in an Industrial Area: A Dataset at the Neighborhood Level

Abstract

The growth of urban and industrial areas is accompanied by an increase in vehicle traffic, resulting in rising concentrations of various air pollutants. This is a global issue that causes environmental damage and risks to human health. The dataset presented in this research contains records with measurements of the air pollutants ozone (O₃) and carbon monoxide (CO), as well as meteorological parameters such as temperature (T), relative humidity (RH), and barometric pressure (BP). This dataset was collected using a set of low-cost sensors over a four-month study period (March to June) in 2024. The monitoring of air pollutants and meteorological parameters was conducted in a city with high industrial activity, heavy traffic, and close proximity to a petrochemical refinery plant. The data were subjected to a series of statistical analyses for visualization using plots that allow for the identification of their behavior. Finally, the dataset can be utilized for air quality studies, public health research, and the development of prediction models based on mathematical approaches or artificial intelligence algorithms.

Dataset: https://doi.org/10.17632/j4xn7wkhbm

Dataset License: CC BY 4.0

1. Summary

Air pollution poses a significant threat to public health and the environment [1]. In urban and industrial areas, increased concentrations of various air pollutants are associated with vehicular and freight traffic [2], as well as various industrial processes, which are the primary sources of air pollutant emissions [3]. In this regard, the petrochemical industry is an essential source of volatile organic compound (VOC) emissions that contribute as precursors of tropospheric ozone (O₃) [4,5], as well as the emission of air pollutants such as particulate matter (PM_2.5 and PM₁₀), carbon monoxide (CO), and nitrogen oxides (NOx), among others [6], which are released through the production, transportation, and marketing processes of petroleum and its derivatives [7]. Additionally, refineries are characterized by the emission of gases such as benzene, toluene, ethylbenzene, and xylene, collectively known as BTEX, which pose a risk to the health of the population [8] and a threat to the atmosphere and the environment [7]. During periods of restriction due to the COVID-19 pandemic, a significant decrease in the concentrations of pollutants associated with vehicular traffic was reported, with decreasing levels of contaminants such as CO, PM_2.5, PM₁₀, and NOx, which are generated by incomplete combustion in engines [9,10].

On the one hand, CO is generated by incomplete combustion in gasoline and diesel engines, particularly through the exhaust system of vehicles and during travel on urban, rural, and highway roads [11]. It is a highly toxic gas that has adverse effects on the health of the population, affecting the respiratory system, reducing lung capacity, and exacerbating diseases such as asthma [12,13,14,15]. Furthermore, it is one of the main causes of domestic poisoning, resulting in premature deaths [16,17].

On the other hand, O₃ is a pollutant that is formed through photochemical reactions of compounds that are released from vehicles, engine combustion and gasoline vapors, as well as industrial processes [18]; it is an excellent oxidant and participates in the formation of photochemical smog and the intensification of the greenhouse effect, global warming, and therefore climate change [19]. It has been reported that prolonged exposure to this pollutant is associated with damage to the respiratory system [20], cardiovascular diseases [21,22], mental disorders [23], and an increase in medical emergencies due to asthma complications [24].

2. Data Description

The dataset of air pollutants (O₃ and CO) and meteorological parameters (temperature, relative humidity, and barometric pressure) was collected at an air quality monitoring station located northeast of Cadereyta city (Nuevo León, Mexico). This station is situated 5 km from a petrochemical refinery (located southeast of the city) and near one of the main highways in the state of Nuevo León, Mexico, with high traffic flow, including both light and heavy vehicles. The dataset consists of a comma-separated file containing 2928 data records and 14 attributes (columns). These instances were collected using three low-cost IoT sensors from 1 March to 30 June 2024. This period covers the last three weeks of the winter season (March) and fourteen weeks of the spring season (March–June).

Table 1. Description of the columns contained in the dataset.

Column Name	Description	Unit
ID_sample	Unique ID for each observation	Integer
Date	Day of recording of the instance value	dd/mm/aaaa
Hour	Time of recording of the instance value	hh:mm
O3	Mean ozone concentration	ppb
CO	Mean carbon monoxide concentration	ppm
Temperature	Mean air temperature	°C
RH	Mean relative humidity	%
BP	Mean barometric pressure	hPa
O3_linear	Ozone concentration imputed using the linear method	ppb
O3_stine	Ozone concentration imputed using the Stineman method	ppb
O3_spline	Ozone concentration imputed using the spline method	ppb
CO_linear	Carbon monoxide concentration imputed using the linear method	ppm
CO_stine	Carbon monoxide concentration imputed using the Stineman method	ppm
CO_spline	Carbon monoxide concentration imputed using the spline method	ppm

describes the attributes contained in the dataset, along with their corresponding units of measurement. The ID_sample attribute is automatically generated by the software system each time a record is inserted into the database. The O₃ column indicates the concentration of the O₃ pollutant recorded by the sensor, while the CO attribute stores the concentration of the air pollutant carbon monoxide. The second and eighth attributes contain raw values, as measured by the sensor and recorded by the software system in the database, and are presented as hourly averages (see Table 1). These attributes may contain empty values, represented by the value ‘null’. A power outage, transmission failure, wireless communication disruption, or lack of Internet service can cause null values. The ninth to fourteenth attributes contain values calculated for O₃ or CO using different statistical interpolation methods (linear, spline, or Stineman) to impute missing values. All three methods provide excellent results in data imputation, and their choice depends on data characteristics such as gradient changes, gentle jumps, steep slopes, or outliers, as well as whether the data are real or synthetic. Furthermore, the data source must be considered, for example, sensors, experimental data, or time series with or without noise. In the data analysis presented in the next section, instances whose values were imputed using the Stineman method are used because the data have abrupt but natural changes and are real sensor measurements. This method ensures smooth and stable interpolation, following the shape of the data without exaggerated oscillations.

3. Methods

3.1. Study Area

The city of Cadereyta, situated within the Monterrey metropolitan area, Nuevo León, Mexico, has a population exceeding 100,000 inhabitants [25]. Figure 1 shows the coverage area of the city, highlighting the location of the air pollutant monitoring station and the oil refinery, the primary source of air pollution in the area. The main economic activities are focused on industry, commerce, and services [26]. Government records indicate a vehicle fleet of over 50,000 units, comprising private cars, motorcycles, transport vehicles, and cargo trucks [27]. Additionally, a substantial amount of heavy cargo traffic passes through the city daily. The city is characterized by being surrounded by mountainous terrain and has a climate with seasonal variations, with summers with temperatures above 40 °C and winters with temperatures below 0 °C.

3.2. Data Collection

The air quality monitoring station was installed in a residential area between two highways with heavy vehicular traffic, as well as a nearby freeway (less than 600 m away) with high-capacity tractor–trailer trucks. Figure 1 shows the location of the monitoring station between Mex-40 freeway, Metropolitan 16 highway, and Interstate 9. A large-scale oil refining plant in Mexico is located downwind (less than 4.5 km away), between Mex-40 freeway and Interstate NL-161 (see Figure 1). Other industries that emit different types of air pollutants are also located in the area. The software system, powered by Internet of Things (IoT) technology, manages a set of sensors and automatically transmits the data collected in real-time to a private cloud repository hosted on an Internet service. Air pollutants and meteorological parameters were configured to be measured every 3 min, and the system automatically generates an hourly average every clock hour based on the readings. Temperature (°C), relative humidity (%), and barometric pressure (hPa) were measured with a Bosch BME280 sensor. The O₃ (ppb) and CO (ppm) concentration levels were measured using calibrated, low-cost ZE25 and MQ series sensors, respectively. The ZE25 sensor was calibrated through a collocation process with a regulatory-grade reference instrument or equivalent monitor (FRM/FEM) under real-world conditions during an evaluation period specified in the current methodology for calibrating ozone sensors [28]. By the end of the calibration study, the sensor achieved a coefficient of determination (R²) of 0.90 for ozone relative to the reference instrument. The calibration process was like that described in our previous work [29]. The MQ series sensors were calibrated in a laboratory using a vacuum chamber with a controlled internal temperature of 20°C and a relative humidity of 55% (±5%). An electrical resistor with the required capacity for the air pollutant CO was installed on the sensor board, as specified in the data sheet provided by the manufacturer. The sensors that comprise the air quality monitoring station are situated at an approximate height of 9 m above ground level and are free from obstacles in the immediate vicinity.

Figure 1. Identifying the study area, marking the position of the monitoring station, and specifying the primary source of air pollutant emissions in the area.

Figure 1. Identifying the study area, marking the position of the monitoring station, and specifying the primary source of air pollutant emissions in the area.

3.3. Data Analytics

During the statistical data analysis phase, descriptive analyses were performed for each study variable. A Spearman correlation coefficient analysis was also conducted to explore the relationships between meteorological variables and air pollutants. Heat map plots were also created to visualize the temporal variability of temperature throughout the study period. Finally, an analysis of the seasonal behavior of pollutant concentration levels in relation to meteorological parameters was carried out to identify trends and changes over time.

Table 2. Descriptive analysis of the air pollutant variables and meteorological parameters considered in the study.

Period	Variable	Mean	SD	Median	IQR	Min	25%	75%	Max
March	CO	1.51	0.32	1.47	0.4	0.63	1.3	1.7	2.78
	O3	13.80	7.88	12.09	7.94	2.06	8.69	16.63	80.42
	T	24.15	6.10	23.1	8.66	10.73	19.65	28.3	41.32
	RH	74.72	21.53	77	36.96	21.03	58.82	95.78	100
	PB	973.33	10.86	973.09	8.86	808.61	969.52	978.38	1012
April	CO	1.32	0.38	1.33	0.46	0.08	1.08	1.54	3.56
	O3	19.00	12.34	16.03	13.04	2.13	11	24.04	94.32
	T	28.39	6.64	27.35	10.08	15.15	23.89	33.96	43.83
	RH	77.93	26.4	92.88	42.71	5.16	57.29	100	100
	PB	973.59	5.59	973.17	7.52	959.14	969.62	977.14	1031
May	CO	1.62	0.34	1.58	0.37	0.23	1.42	1.79	3.16
	O3	17.87	18.03	11.39	15.23	1.44	6.42	21.65	96.01
	T	33.59	6.31	32.05	10.11	21	28.5	38.61	49.88
	RH	84.24	18.97	92.93	28.67	27.45	71.33	100	100
	PB	972.58	13.76	968.81	4.54	878.29	966.6	971.14	1021
June	CO	1.12	0.32	1.14	0.51	0.28	0.85	1.36	2.35
	O3	18.62	15.43	16	18.6	0.71	6.80	25.41	91.85
	T	32.77	6.09	30.81	10.20	23.23	27.78	37.98	47.52
	RH	78.06	24.39	87.58	41.39	22.65	58.61	100	100
	PB	971.75	7.47	971.74	3.97	795.05	969.83	973.8	982.29
Average	CO	1.4	0.39	1.41	0.43	0.08	1.18	1.61	3.56
	O3	17.3	14.09	13.88	13.34	0.71	8.00	21.34	98.01
	T	29.71	7.34	28.88	9.87	10.73	24.80	34.67	49.88
	RH	78.75	23.21	86.56	38.77	5.16	61.23	100	100
	PB	972.81	9.99	971.5	6.52	795.05	968.58	975.1	1031

shows the descriptive analysis of the study variables, presenting the measures of central tendency grouped by month and as a general average for the study period. This analysis identifies the distribution of the data and its variability. The maximum and minimum values allow for the identification of the extreme values around which the records of each variable fluctuated, enabling the detection of outliers during the study period. The last row of Table 2 presents the average of all the records included in the study. The data demonstrate significant variability in meteorological factors. The temperature (T) variable had a mean of 29.71 °C and a standard deviation (SD) of 7.34, indicating moderate variability, along with an interquartile range (IQR) of 9.87 °C, with values ranging from 10.73 to 49.88 °C. The relative humidity (RH) variable exhibited high variability, with a standard deviation of 23.21% and a mean of 78.75%, ranging from an extreme minimum of 5.16% to a maximum of 100%. Barometric pressure (BP) maintained more stable records during the study period, with a median of 971.5 hPa and an interquartile range (IQR) of 6.52 hPa. Regarding air pollutants, CO recorded an average of 1.4 ppm and a SD of 0.39 ppm, indicating low dispersion and greater accumulation. Finally, O₃ showed greater dispersion, with minimum values of 0.71 ppb and maximum values of 98.01 ppb.

Figure 2 displays the correlation coefficients of the study variables, calculated using the Spearman rank correlation technique, which are detailed in a dispersion matrix that includes correlations and densities. On the one hand, the dispersion graphs are displayed, represented by a cloud of points between two variables; each point corresponds to a time record. The colors represent the date variable, grouped by month (March to June). On the other hand, the graphs located on the main diagonal, represented as curve graphs, show the distributions of the univariate density by specific month, which allows the shape and bias of each variable to be detected. Finally, the Spearman correlation coefficient values for bivariate relationships are shown on the right side of Figure 2, considering a p-value of 0.05. In this analysis, the pollutant O₃ in April showed a moderate positive correlation with CO, with a correlation coefficient of 0.36. Similarly, CO exhibits a positive association (moderate to strong) with RH, with a correlation coefficient of 0.49. The highest coefficients are identified in the association between the T and RH variables, with values of -0.85 and −0.87, in May and June, respectively. This behavior is correct for the T and RH, whereas temperature increases, relative humidity decreases in the same proportion, and vice versa.

Figure 3 and Figure 4 show a heat matrix for March and June; the days of the month are defined on the x-axis, and the time, represented by the hour (00:00–23:00) of the record, is on the y-axis, where each cell corresponds to an independent value recorded over time. The color scale represents the recorded temperature in degrees Celsius (°C). Purple tones on a scale from medium to dark correspond to temperatures between 15 °C and 25 °C, while orange and yellow tones identify the highest temperatures (up to 45 °C). Figure 3 displays the heat map for March, showing greater temperature variability. In Figure 4, the heat map for June is shown, indicating sustained thermal stability, except for the period from the 19th to the 21st, when temperatures were recorded below 25 °C, corresponding to days with cloudy or precipitation conditions. The heat matrix enables the identification of temperature patterns, extreme heat events, and critical times that are crucial to the photochemistry of the air pollutants considered in this study.

Figure 5 shows a comparison of the seasonal behavior of the pollutants O₃ and CO, considering the transition month between seasons, that is, between winter and spring. Each graph represents the behavior of the pollutants from March to April, respectively. Additionally, the RH percentage is included to identify the behavior of air pollutant concentration in the presence of higher or lower values of this meteorological variable, as shown in Figure 2. The abscissa axis corresponds to the day of the month, and the ordinate axis represents the air pollutant concentration, each in its corresponding measurement unit: parts per billion (ppb) for O₃ and parts per million (ppm) for CO. Additionally, the secondary Y-axis corresponds to the RH variable (%). The solid line represents air pollutants (blue for March; purple for April), and the dashed green line represents RH.

3.4. Discussion

The data analyzed represent the seasonal behavior of two primary pollutants in air pollution and meteorological variables in an industrial city with a large fleet of private cars and heavy-duty transport vehicles, which were evaluated at two different times of the year. Meteorological variables have a direct influence on pollutant concentration levels [30]. In this sense, future studies can benefit from the dataset to identify critical times of high temperatures that favor the formation of air pollutants (O₃) and accumulation due to poor dispersion of the pollutant (CO), as well as increase health risks to people due to prolonged exposure [31]. Furthermore, analysis of the dataset can help determine thresholds for meteorological variables that favor the increase in or dispersion of pollutants, enabling the implementation of efficient air pollution controls [30,32].

According to [32,33], air pollution is closely related to synoptic climates and prevailing meteorological dynamics, which allow for the identification of regional pollution events. Our results demonstrate that temperature was a significant factor in O₃ levels (see Figure 2, Figure 3 and Figure 4), as O₃ is sensitive to high temperatures and UV-B radiation, which favor atmospheric conditions for its formation [34], which is consistent with [35], who found that air pollution levels are higher in the summer due to regional climatic conditions and atmospheric chemical patterns that influence the reactions of these compounds. On the other hand, CO concentrations increased on days with higher RH and unstable ambient temperature (see Figure 5), likewise reported by [11]. Additionally, a positive association was observed between CO and RH (see Figure 2), which favors its accumulation in the atmosphere. Finally, high CO levels were closely related to vehicular traffic on high-flow roads and in the morning.

4. User Notes

The importance of using the presented dataset lies in the fact that it contains values collected from a monitoring system employing low-cost sensors as an alternative for measuring air quality. These sensors are capable of continuous, real-time measurements and are compatible with Internet of Things (IoT) technologies. It is worth noting that these sensors, depending on the model, have a lifespan of between 6 months and 2 years in outdoor monitoring operations. Furthermore, the low-cost sensors were installed in an environmental monitoring station under the parameters established by air quality guidelines, ensuring that the data are representative of the emissions generated by anthropogenic activities in the region. We therefore consider that the presented dataset can be used for the following approaches: (1) spatiotemporal analysis of CO pollution regarding the increase in O₃ in that area, (2) determination of critical temperature schedules in contrast with the levels of atmospheric pollutants, (3) the identification of the pollutant–source relationship to encourage attention to be paid to the control of atmospheric pollution, (4) use by other researchers to compare the measured concentration levels with monitoring results from other study areas with the same geographical conditions, and, finally, (5) as a dataset to exploit training and prediction models using machine learning and deep learning algorithms. In this context, several challenges emerge, such as the nonlinear relationship between the dependent variable and the predictor variables, the spatiotemporal correlations involved in predicting air pollutants, the seasonality in time series, and the long-term forecasting of air pollutant concentrations (for example, the next 8 or 12 h), with high performance and accuracy. Furthermore, predicting the concentration of certain pollutants is heavily influenced by meteorological factors like temperature, relative humidity, wind speed, precipitation, and others.