Air Quality Assessment Based on a Smart Locally CO 2 Monitoring System with Validation by a Reference Instrument

: This article deals with the development and implementation of a low-cost smart air quality monitoring device. The study aims to develop a simple, efficient, and cheap device for real-time monitoring of Carbon dioxide (CO 2 ) concentration and atmospheric parameters such as temperature (T) and relative humidity (RH). The realized device consists of sensors, a data storage system, a programmable card, and wireless modules for Internet of Things (IoT). It is designed for indoor and outdoor air quality monitoring. Wireless communication between the device and the control PC is ensured by the ZigBee protocol. The MH-Z14A sensor was calibrated according to the manufacturer’s recommendations and then experimentally in the laboratory. A calculation of the CO 2 regression line for the devices using the least squares method gave R 2 = 0.6889. The slope test using the Student method at a significant coefficient of 1% threshold and 19 degrees of freedom gave t(1%; 19) = 2.539. This illustrates that the result obtained by the realized device has less than a 1% chance of being obtained at random. An investigation was carried out to assess air quality because following the proliferation of pollutants, the normal concentration of atmospheric CO 2 (400 ppm) has considerably changed. Thus, the data processed are outdoor concentrations collected in November 2022 in some localities of the Adamawa Region, Cameroon. The average values of CO 2 concentration measured in four subdivisions are 497.30 ± 11.32 ppm in Ngaoundéré 1, 481.83 ± 14.90 ppm in Ngaoundéré 2, 568.63 ± 25.03 ppm in Martap and, 624.14 ± 3.96 ppm in Minim. Martap and Minim are among the most important bauxite deposits in the world. These CO 2 values higher than the normal atmospheric concentration indicate a moderate Air Quality Index (AQI) in these cities during the measurement period. The designed device is inexpensive, reliable, and appropriate for air quality monitoring.


Introduction
Air pollution is the greatest environmental and public health challenge in the world today 1 .One of the main causes of global warming is the emission of CO 2 into the atmosphere 2 .Currently, the decline in air quality in large cities continues to increase due to industrial growth, road traffic, and population growth 3 .Living beings transform oxygen into CO 2 through respiration, while plants use CO 2 for photosynthesis and transform it into organic carbon compounds (sugars, cellulose, etc.) and dioxygen (O 2 ), commonly known as oxygen 4 .The importance attributed to CO 2 comes from the rapid increase in CO 2 concentration in the atmosphere as a result of increasing fossil fuel consumption and a significant reduction in forest cover on a global scale 5 .On a planetary level, CO 2 is part of human-caused pollutant emissions; its emissions have two origins: natural and anthropogenic 6 .The current average rate of CO 2 concentration in the air (in ppm) oscillates around 0.038% (0.1% = 1,000 ppm) with some day-night, seasonal variations, and pollution peaks 7 .CO 2 is a greenhouse gas and contributes greatly to the increases in the earth's temperature by intercepting the infrared rays reflected from the earth's surface 8 .Global warming promotes the atmospheric accumulation of water vapor, the atmosphere becomes humid and traps more and more heat [9][10][11] .From a health perspective, CO 2 has a very high dissolving capacity in the mucous membranes of the human body and causes numerous reactions as soon as its concentration in the inspired air increases.For concentrations close to 0.1%, CO 2 alters the respiratory rate of people who suffer from respiratory insufficiency.Levels above 1,000 ppm in a closed environment can lead to asthma attacks 12,13 .
Due to the danger of CO 2 , researchers developed a low-cost, pocket-sized CO 2 device with flexible applications that yields benefits for students and schools 14,15 .The advantage of this device is that it is low-cost, but does not have an IoT module and is not included in the scope of delivery so that it can be used on-site.Gonçalo Marques et al. realized an IoTbased CO 2 monitoring system to assess indoor air quality assessment 16,17 .They have developed a low-cost device that uses a Wi-Fi module for data transmission to the Cloud.This means that we will need a permanent internet connection and in the long term, this will become expensive.In addition, the device is not in the box and its functionality is only limited to indoor measurement.Dari et al. designed and built a hazardous gas monitoring device (CO, CO 2 , CH 4 ) on cigarette smoke in an enclosed space based on Arduino Uno and GSM SIM 900 A 18 .They have always offered a lowcost device with IoT functionality, but except that they use MQ sensors which are metal oxide semiconductor sensors that are very unstable and difficult to calibrate 18 .Several works on CO 2 are being carried out around the world.Some realize the device and use it to measure CO 2 while others capture CO 2 [19][20][21] .According to the Mauna Loa Observatory (MLO), atmospheric CO 2 is higher today than at any time in the past 200,000 years of human history 22 .In fact, studies of past CO 2 levels have accumulated evidence that CO 2 levels have been below 400 ppm (parts per million) for the past 23 million years 23 .This makes CO 2 records today the highest in human history.Statistics show that we now consider air with 420 ppm of CO 2 and 21 percent of oxygen to be fresh.However, the norm used to be no more than 350 ppm 24,25 .With this rise of almost 100 points in 50 years, it becomes clear there is a need for the constant monitoring of indoor and outdoor levels of CO 2 26-28 .CO 2 monitors in general are decision-making tools regarding air pollution.It is in this context that an intelligent, simple, effective and inexpensive device is produced in this article as an appropriate solution for knowing CO 2 concentrations in real-time.The proposed system is manufactured locally based on the microcontroller card, XBee S2C modules, temperature, relative humidity and CO 2 sensors whose reliability has been confirmed in [29][30][31][32][33][34] .The final device was packaged in the box to facilitate its deployment in the field by using rechargeable batteries.Methods such as the least squares method to determine the correlation coefficient R 2 = 0.6889 and the Student method to test the slope of the regression line of the concentrations of CO 2 were applied during the calibration process of the realized device, in order to note that the result of the realized device has less than 1% chance of being obtained at random.Data were collected in November 2022 in the city of Ngaoundéré and two localities in the bauxite zones of Minim and Martap in the Adamawa region of Cameroon.Real-time data acquisition was supervised and remotely monitored via the IoT functionality of the device.The collected data was stored in a digital card before being analyzed to assess air quality based on the obtained CO 2 levels.These results were also compared to the international standards of the Mauna Loa Observatory (MLO)-USA and the National Agency for Food, Environmental and Occupational Health Safety (ANSES)-French [35][36][37][38] .

Study areas
The CO 2 measurements in this article were carried out in two divisions of the Adamawa region, namely three districts in the Vina division and one district in the Djérem division.Figure 1 shows the location of the Adamawa region and the subdivisions where the measurements were taken.The main objective of this article is to realize a low-cost atmospheric CO 2 rate monitoring device.However, it is important to geolocate the study area to facilitate the identification and visualization of sampling points.Figure 1 shows the state of this study area and the sampling points are represented by the black triangles on the map.For this study, 62 measurement points were carried out, namely 20 points (from N1 to N20) in the district of Ngaoundéré 1, 12 (from K1 to K12) in the district of Ngaoundéré 2, 16 (from M1 to M16) in the district of Martap and 14 (from L1 to L14) in Minim town which belongs to the district of Tibati.Tables 1 and 2 show the geographical coordinates of the different measurement points carried out in the four study cities.

Design of real-time CO 2 monitoring device
The developed real-time CO 2 device is based on an Arduino board (data processing and analysis unit), XBee wireless, temperature, relative humidity, and CO 2 sensors.The final prototype is powered by three rechargeable batteries of 3.7 V each, for autonomous measurements.The main measured molecule is carbon dioxide, which is detected by the MH-Z14A sensor.Data transmission between the device and the remote PC is provided by the XBee radio frequency and the IoT module.A schematic diagram of the CO 2 device and the used monitoring system is shown in Figure 2   The MH-Z14A NDIR infrared gas module is a common small sensor that uses the non-dispersive infrared (NDIR) principle to detect the existence of CO 2 in the air, with good selectivity, non-oxygen dependent and long life.Integrated temperature compensation, digital output and PWM output.

Xbee S2C Transmitter
The AMT1001/AM1001 is a moisture-resistant temperature and humidity sensor, a single wet sensor that sends signals via an analog voltage output.This module features high accuracy, high reliability, consistency, and temperature compensation to ensure long-term stability, ease of use and low price.
The XBee S2C modules use the ZigBee radio communication protocol based on the IEEE 802.15.4 standard AMT1001 sensor (T.RH) with an operating frequency of 2.4 GHz.They facilitate remote data transmission, between an electronic device and a computer through two XBee transmitter and receiver modules and the XCTU application previously installed on the PC.

Sensors calibration and device operation algorithm
Detection and measurement sensors require calibration and configuration processes before use.The sensors used in this work are experimentally calibrated in the laboratory according to the manufacturer's instructions.For the MH-Z14A sensor, we performed zero point calibration and this module has three methods for its zero point calibration namely manual method, command sending method and self-calibration 39,40 .Its zero point is set at a CO 2 concentration of 400 ppm.The manual method was to connect the HD pin of the sensor to a low level (0 V) for a few seconds (10 seconds) while ensuring that the sensor was stable for half an hour of operation in an ambient environment of 400 ppm.Calibration of the MH-Z14A sensor using the send command method was not performed due to the lack of a serial port (URAT).Self-calibration, on the other hand, was performed by running the module in the laboratory every 24 hours for a week.During these running times, the module intelligently judges the zero point and automatically performs a zero calibration each time the power is turned on.
The final calibration of the MH-Z14A and AMT1001 sensors, whose technical specifications are shown in Table 3, was carried out in the laboratory in the presence of a reference device.The two devices shown and referenced in Figure 4 were used in the same location and under the same atmospheric conditions for three weeks to measure temperature, relative humidity and CO 2 concentration.The temperatures and relative humidities measured by the two devices are similar at some points and show small differences at other points.These results give mean values of 27.52 ± 0.51 °C and 66.37 ± 0.96% for the realized device and 27.17 ± 0.36 °C and 66.32 ± 0.86% for the reference device.These values are almost identical.With the CO 2 sensor, the concentrations given by the realized device remained higher than those of the reference device throughout the first week of measurement (Figure 5 (c)) and from the second and third week of measurement the concentrations delivered by the two devices remained almost identical.The average values obtained are then 742.95 ± 26.67 ppm for the realized device and 739.42 ± 19.80 ppm for the reference device.
An in-depth statistical study of the results obtained is necessary to better assess the reliability of the realized device.Therefore, it is important to calculate the equation of the regression line y = ax + b of the CO 2 concentrations of the two devices by using the least squares method and perform the slope test using the Student method 41,42 .Let x be the CO 2 concentrations from the reference device and y those of the realized device.The values of the slope a and the ordinate at the origin b are obtained from the following equations ( 1) and (2): . We deduce from the cov(x, y) and the standard deviations of x and y, the existence of a very strong positive linear correlation between the two variables: cov( , ) 0.83 Verification of the results of equations ( 3) and (4) obtained by the least squares method was carried out in Microsoft Excel as presented in Figure 6.In both methods, we almost obtained the same regression line y = 1.12x -85.6, with R 2 = 0.6889 ≈ 0.76.To perform the test of the slope using Student's method, it is necessary to calculate the coefficient t so that: (5) 6.22

∑
With r the residue such that r = y i -ŷ = e; the error between the observed value y i and the estimated value ŷ. σ r is the standard deviation of the residuals, σ a the standard deviation of the slope a of the regression line, n the number of observations i.e., the measurements taken and (n -2) the degree of freedom with 2 number of independent variables x and y.It is essential to make assumptions for the significance test of the regression coefficients.Thus, we first pose H 0 : β = 0, when the concentration x of the reference device has no influence on the concentration y of the realized device and secondly H 1 : β ≠ 0, when the concentration x has an influence on the concentration y. , then the hypothesis H 0 is rejected therefore there is a significant linear relationship at the 1% risk between the CO 2 concentrations measured by the two devices.In other words, the result obtained by the produced device has less than 1% chance of being obtained at random.
For the use of sensors in the measurement of air pollutants, computer programming and experimental work were performed on a computer with specifications Intel (R) Core (TM)i7-4600M, 2 Duo CPU 2.9 × 2 GHz, 8 GB RAM with Windows 10-64 bit professional as the operating system.The simulation environment is the Arduino IDE, whose operating algorithm is programmed for the monitoring device.The complete operation algorithm of the realized device is described in detail as Appendix.

Threshold and effects of CO 2 on metabolism
At low concentrations, CO 2 is harmless to human health, but long-term exposure can cause headaches.CO 2 is also a good marker for measuring indoor air renewal.The normal indoor concentration varies between 400 and 1,000 ppm.Above this threshold, the negative effects on human health can increase and can lead to death at concentrations above 40,000 ppm 45,46 .This is described by ANSES 37 describes in Table 4 below:   In the Martap and Minim districts, most CO 2 concentrations are high, around 600 to less than 700 ppm.A low CO 2 concentration of around 200 ppm was obtained in the city of Martap.The measuring device must certainly be influenced by errors at this measuring point.The obtained values in these four localities are higher than the normal concentration of atmospheric CO 2 which is 400 ppm 48,49 .This may be because the measurements were carried out in public places, namely markets, city centers, schools, family homes, and hospitals where there are a large number of people who produce CO 2 through the mechanism of respiration.It was important to carry out the measurements in these environments because the primary objective of the application of this work is the protection of people.Figures 8 and 9 show that temperature and humidity have acceptable values, according to the World Health Organization (WHO) guidelines on housing and health 50 , namely temperatures below 25 °C in the districts of Ngaoundéré 1 and 2 and temperatures range between 25 °C and 30 °C in Martap and Minim.The low temperatures obtained in these localities are one of the peculiarities of the Adamawa region, one of the coldest regions in Cameroon.The measurements were taken in November, the month of the year when it is cold throughout the Adamawa region.This cold climate period generally extends from November to February of the following year.
The second atmospheric parameter, which is the humidity in Figure 9, is very proportional to the temperature.The lower the temperature, the higher the humidity, which explains the high increases in humidity in Ngaoundéré 1 and 2 compared to low and fluctuating values in the cities of Martap and Minim.
Figure 10 presents the distributions of CO 2 concentrations obtained in the four research areas, namely Minim, Martap, Ngaoundéré 1 and 2. Logically, areas with low CO 2 concentrations are in dark green, areas with medium CO 2 concentrations are in light green and yellow, while areas with high CO 2 concentrations are in orange and red, as indicated in the legends of each distribution in Figure 10. Figure 10 presents the distributions of CO 2 concentrations obtained in the four research areas, namely Minim, Martap, Ngaoundéré 1 and 2. Logically, areas with low CO 2 concentrations are in dark green, areas with medium CO 2 concentrations are in light green and yellow, while areas with high CO 2 concentrations are in orange and red, as indicated in the legends of each distribution in Figure 10.
Analogous to the distributions of previous CO 2 concentrations, the log-normal distributions of the number of CO 2 measurements taken at each location are shown in Figure 10.Of the 20 measurement points carried out in Ngaoundéré 1 and shown in Figure 11, 3 points have CO 2 concentrations in the range of 400 to 450 ppm, 10 points between 450-500 ppm, 4 between 500-550 ppm, 1 point in the 550-600 ppm range, and 2 between 600-650 ppm.This explains the approximation of the average CO 2 concentration in Ngaoundéré 1 (497.30± 11.32 ppm) to the highest values.In Ngaoundéré 2 (Figure 11), all CO 2 concentrations are in the range of 450 to 500 ppm and only one value is 600-650 ppm.The average concentration of this locality is 481.83 ± 14.90 ppm.In Martap (Figure 11), only one concentration is in the range of 200-300 ppm and all others are the 500 to 700 ppm range.The same observation is in Minim (Figure 11) where the concentrations range from 600 to 670 ppm.These last two cities are those with high CO 2 concentrations.This can certainly be explained by the low wind circulation in these two cities during the sampling period, as they are located in the middle of forest and bauxite areas of the Adamawa region.A comparative study of the average CO 2 concentrations obtained at the four study sites with reference values such as the normal concentration of atmospheric CO 2 and the average CO 2 measurements from the MLO in November 2022 was performed and presented in Table 5.The concentrations obtained in the Adamawa region are well above the reference value.This is explained by the fact that the measurements were taken in public places.These averages in Table 5 made it possible to display the graphs in Figure 12, which are essential for an overall analysis of CO 2 concentrations and measured atmospheric parameters.These plots visually show the distribution of the numerical data and the asymmetry by displaying the average range of the concentrations obtained.They summarize the data set into three elements, namely the minimum score, the average and the maximum score.

CO 2 -based outdoor air quality assessment
The aim of this work is to protect people from the harmful effects of air pollution.Therefore, it is then necessary to know the AQI in the four study areas.The following equation ( 9) makes it possible to determine the different indices based on only CO 2 .The results of the obtained AQIs are presented in Table 6.All AQI values obtained are in the range of 51 to 100 and according to Table 6, the corresponding air quality is moderate.This work shows the importance of realizing electronic devices for air quality monitoring, and generally, some devices available on the market, presented in 29,52 , are portable, efficient, and measure few gases.Some use short-range Wi-Fi applications for wireless data transmission.However, the device developed in this article measures CO 2 and atmospheric parameters, helps to determine the AQI, and uses XBee S2C 2.4 GHz band radio communication modules that provide wireless data transmission over long distance (more than 1,200 m) to a remote PC via the XCTU enable application.These low-power XBee S2C modules are suitable for dense and mesh equipment networks [53][54][55] .The special feature of this device is the use of the XBee module, which allows data transfer over long distances without an internet connection.In addition, the device is in the box, autonomous, portable and can be deployed in the field for indoor and outdoor measurements.

Conclusion
This paper presents a real-time CO 2 monitoring device developed and used for air quality sampling.The device uses ZigBee wireless transmission components as an IoT tool.A comparative study of the device carried out with a reference device showed R 2 = 0.6889 and the slope test of the regression line according to the Student method showed t (1%; 19) = 2.539.The measurement campaign carried out in November 2022 in four subdivisions of the Adamawa region of Cameroon made it possible to obtain average levels of CO 2 equal to 497.30 ± 11.32 ppm with a maximum of 618 ppm and a minimum of 443 ppm in Ngaoundéré 1, 481.83 ± 14.90 ppm with a maximum of 640 ppm and a minimum of 453 ppm in Ngaoundéré 2, 568.63 ± 25.03 ppm with a maximum of 663 ppm and a minimum of 223 ppm in Martap and, 624.14 ± 3.96 ppm with a maximum of 661 ppm and a minimum of 608 ppm in Minim town located in Djérem division.These values indicate a moderate AQI in these cities during the measurement period.The realized equipment has shown that low-cost sensors are of great interest for air quality sampling.They are used for air pollution monitoring in developing countries, where air pollution has adverse effects on human health.Improvement and optimization work is currently being carried out on this realized prototype to obtain a final decision support tool for air pollution.

Figure 1 .
Figure 1.Location of CO 2 measurement points

Figure 2 .
Figure 2. Schematic diagram of the realized CO 2 monitoring device

Figure 3 .
Figure 3. Schematic diagram in IoT operation of the CO 2 monitoring system

Figure 4 .Figure 5 .
Figure 4. Calibration process With cov(x, y) = 440.72 the covariance of x and y, σ x = 19.80 and σ y = 26.68 the standard deviations of x and y, x ¯ = 739.42and y ¯ = 742.95 the mean values of x and y.The equation of the regression line of the CO 2 concentrations of the two devices is then obtained:

Figure 6 .
Figure 6.Equation of the regression line and correlation coefficient obtained with Microsoft Excel

Figure 7 .
Figure 7. CO 2 concentrations obtained in the measurement localities

Figure 8 .
Figure 8. Temperature obtained in the measurement localities

Figure 9 .Figure 10 .
Figure 9. Humidity obtained in the measurement localities

Figure 11 .
Figure 11.Log-normal distribution of CO 2 concentrations obtained in the measurement localities

Figure 12 .
Figure 12.Histograms of average concentrations of measured pollutants

Table 1 .
Geographical coordinates of the measurement points taken in Ngaoundéré 1 and 2

Table 2 .
Geographical coordinates of measurement points taken at Martap and Minim

Table 3 .
Technical characteristics of the used electronic sensors

Table 5 .
Comparison of CO 2 concentrations obtained with certain reference values CO 2 Average concentrations during November 2022 around the world

Table 6 .
Air quality index and level of health impact in the study localities[29]