Modeling and implementation of a low-cost IoT-smart weather monitoring station and air quality assessment based on fuzzy inference model and MQTT protocol

Abstract

The automatic weather system serves to inform farmers, tourists, planners, and others with precise information to help them take the appropriate action. Today, with the advancement of smart technologies, the system has evolved into many sensing methods to gather real-time climate data. This article investigates the modeling and implementation of a low-cost weather station device that also functions to measure air quality. The proposed system based on the Internet of Things (IoT) allows access to real-time climate data for a given area. This system monitors environmental conditions such as ambient temperature, humidity, atmospheric pressure, altitude, and levels of harmful atmospheric gases like CO₂ and NO₂. This real-time telemetry device uses MQ-135, DHT-11 and BMP280 sensors to gather data. The ESP32 board processes the obtained data from all sensors. Additionally, we present a model for a fuzzy inference system (FIS) that performs parameter categorization using a reasoning procedure and incorporates the results into an air quality index (AQI) that describes the levels of pollution for Al Hoceima city. The FIS takes CO₂ and NO₂ values as input and returns the AQI. The AQI for Al Hoceima city is categorized into six levels: Excellent, Good, Regular, Bad, Dangerous, and Very Dangerous. Furthermore, the suggested system's block hardware employs the Message Queuing Telemetry Transport (MQTT) protocol to broadcast collected data to a mobile and web application via the Internet. The suggested IoT-embedded device was tested in real life, and the results were promising.

Introduction

Regarding climate change and the transition in weather state, the determination of weather information is becoming increasingly important. The beneficiaries of this information include organizations in charge of civil security or health (Lee et al. 2020); companies involved in insurance, energy, transport, or agriculture (Ramachandran et al. 2022); and people planning weekends or holidays. For instance, farmers’ work is highly weather-sensitive because their crops depend significantly on weather conditions. Therefore, they are attentive to weather forecasts when planning their daily and seasonal tasks. For example, they carefully monitor fallen and expected rainfall before watering their crops. In this way, agricultural yields can be improved through weather forecasts by providing farmers with timely information. Moreover, attendance at tourist sites is highly dependent on weather conditions (Mahon et al. 2021). Also, attendance varies depending on temperature and sunshine or rainfall. For example, sunny days are preferred for weekends, school holidays, outings, and outdoor activities. In contrast, on days with heavy rain showers, museums, cinemas, and other indoor activities are given preference. In the field of transportation and logistics, accurate weather information is crucial to drivers and pedestrians’ safety (Zhai et al. 2019). Snowfall and freezing rains are hazards that make it difficult to operate roads. To facilitate commuters’ mobility and ensure safety, road managers and departmental councils put winter service plans in place. In addition, airports are commonly affected by weather variations. The airstrip can become frozen in particularly low temperatures, complicating the management of aircraft landings (Midtfjord et al. 2022). Finally, the building trade is heavily impacted by the weather. Certain conditions such as rain or wind can significantly affect tasks in the construction industry (Ballesteros-Pérez et al. 2018).

Meanwhile, environmental concerns are becoming more severe, endangering both nature and human health. The most serious threat is air pollution (Yonar and Yonar 2022). It is increasing due to a variety of factors, including city overcrowding, and industrial processes involving combustion. Air pollution is a combination of particulate matter (PM_x) and gases in the atmosphere. Common air pollutants include NO₂, SO₂, PM_2.5, CO, and CO₂. Some air pollutants can be harmful since ingesting them can cause serious health issues. It may be dangerous, particularly for people who suffer from respiratory illnesses. Each year, outdoor and indoor air pollution kills nearly 7 million people, according to the World Health Organization (WHO) (Yonar and Yonar 2022). Furthermore, the COVID-19 virus pandemic, which emerged in late 2019, highlighted the significance of the interaction between public health and the environment. According to research studies, people exposed to long-term air pollution are more likely to contract and be harmed by viruses such as COVID-19 as a result of emerging chronic disorders (Ghorbani and Zamanifar 2022; Yonar and Yonar 2022). Therefore, it is crucial to continuously monitor the quality of the air in all indoor and outdoor areas, including homes, workplaces, and industrial facilities (Ghorbani and Zamanifar 2022).

However, due to financial constraints, many economically developing countries are unable to invest in professional stations that enable air pollution monitoring and forecasting. Thus, their geographical distribution is very limited, mostly due to the high equipment price and the difficulty and cost of maintenance (Zaric et al. 2021), which leads to limited available climate information. For instance, the national network for monitoring air quality in Morocco consists only of 29 fixed stations to measure, forecast, and notify the public, local authorities, and decision-makers on air quality. Yet, air quality data is strongly needed as a first step toward mitigating the severe health and economic implications of air pollution. However, the cost of air quality monitoring technology devices is still somehow high. It is important to propose low-cost air quality monitors to the market of developing countries to avoid the costly traditional air quality monitoring technology. The affordability of low-cost air sensors can enable air quality assessment in regions where air quality data has never been available previously. Numerous models for the evaluation and surveillance of air pollutants have been proposed in the literature (Megantoro et al. 2021; Messan et al. 2022; Sung and Hsiao 2021; Zaric et al. 2021). However, their inability to address the ambiguity and inconsistency prevalent in this complicated environmental challenge is their most significant flaw.

In this context, this current research discusses modeling and implementation of an automatic weather monitoring station device based on IoT and fuzzy inference system (FIS) that provides real-time climate data and indicates air quality for a specific area.

The key novelty of this study is the proposal of a low-cost weather monitoring station based on innovative technologies, including IoT, FIS, and the message queuing telemetry transport protocol (MQTT). This solution will undoubtedly alleviate the barriers that obstruct access to appropriate and accurate weather information.

The rest of the paper is organized as follows: Sect. 2 analyses the relevant work in the literature. Section 3 presents the system model. Section 4 discusses the core elements of our approach as well as materials and methods. We describe the obtained results in Sect. 5 of this article. A discussion is highlighted in Sect. 6. Section 7 concludes the paper.

Related research

The IoT has attracted significant interest in recent years from business leaders, policymakers, regular people, and engineers. The IoT is a network that connects items on the Internet, allowing for extensive data exchange and the use of remote controls to construct decentralized systems. The IoT transforms physical objects into network devices that can share their resources with other participants or even other objects (Brandão et al. 2021). The IoT has been operationalized extensively in many areas, such as healthcare (Yesmin et al. 2022), farming (Patel et al. 2022; Xu et al. 2022), home automation (Almusaylim and Zaman 2019), smart cities (Bellini et al. 2022), and weather monitoring systems. Accurate weather information also benefits various fields, including electricity generation, agriculture, and transportation. Hence, weather monitoring is an important aspect of economic growth.

Various studies on the design of weather monitoring stations have been carried out thus far. For instance, a smart weather station concept based on a peripheral interface microcontroller (PIC) and a cloud platform was proposed by (Djordjevic and Dankovic 2019). The system was built to be simple to set up and scalable for expansion. The system is powered by a potent PIC microcontroller that controls the entire system. It contains embedded sensors to view and measure the environment or a given place when required. Additionally, it contains a general packet radio service (GPRS) module for uploading data to a cloud platform. Furthermore, (Liu 2019) outlined the development of a low-cost, reliable, modular IoT weather station that can be used in third-world countries such as Haiti. These countries lack the necessary weather/climate data to support future agricultural growth or provide climate pattern information for public and infrastructure protection. In (De Oliveira Filho et al. 2022), temperature, relative humidity, pressure, dust, and wind speed are all measured using several sensors in the constructed weather station. Carbon dioxide (CO₂), carbon monoxide (CO), ammonia (NH₃), hydrogen (H₂), and methane (CH₄) are among the gases monitored by the authors' proposed system. The different sensors are shielded from sunlight, rain, and other weather conditions (such as dust storms) by a solar radiation barrier, with the exception of the wind, dust, and light sensors, which must remain open to the outside world. Moreover, an IoT-based climate monitoring system for rural areas was proposed by (Muslim et al. 2021), who chose an Indonesian hamlet as the site for IoT wind behavior, rainfall, and temperature sensor installation because it is subject to floods, landslides, and extremely high winds. The research presented by (Girija 2018), describes a system that includes a microcontroller (ESP8266) that serves as the system's main processing unit, with all sensors and devices attached to it. The sensor is controlled by a microcontroller, which collects, processes, and uploads data to the Internet via a connected Wi-Fi module.

The primary purpose of the work outlined in (Verma et al. 2017) was to design a remote weather monitoring system that is enabled for continuous weather-monitoring parameters. The IoT-based distant weather monitoring station is a complete open-source weather station that can accurately detect temperature, humidity, and light intensity. The values of the recorded parameters were plotted on an open cloud called “ThingSpeak.” The device also included a camera that permitted live broadcasting of the monitored region. Likewise, (Megantoro et al. 2021) proposed the design of a weather station unit that could monitor gas concentrations in the air. The ESP32 board was used to process measurement data in this real-time telemetry system enabled by IoT. Wind speed, wind direction, humidity, ambient air temperature, air pressure, rainfall, and the ultraviolet index were some of the climate parameters examined. Meanwhile, ozone (O₃), H₂, CH₄, NH₃, CO, and CO₂ were the gas concentration factors measured in the air. The ESP32 board processes all sensor readings of ultraviolet (UV) and uploads them to the server. Subsequently, the dataset is received by a client device, processed, displayed on a monitor, and saved as a text file. In addition, the monitor and data are utilized to analyze air quality and weather conditions in the area. Furthermore, Shaker and Walla (2017) proposed a weather station to track a variety of environmental factors in Babylon city, including CO₂, CO, and CH₄ concentrations, humidity, temperature, and light levels. These elements were measured, analyzed, and transferred to a web application, where authorized users worldwide can view them. The system suggested by (Agrawal et al. 2019), demonstrates an IoT-based approach for tracking meteorological conditions in a specific location and making the data available to anyone anywhere globally. The system has a display that shows the readings in real time. It also keeps track of past data on an hourly and daily basis. This data can be shown on a liquid crystal display (LCD) screen, and the information can be sent to a web page and plotted as graphical statistics. The developed system includes a microcontroller (LPC2138) that serves as the system's main processing unit, to which all sensors and devices can be attached. The microprocessor can operate the sensors and retrieve data from them. It runs an analysis of the sensor data before uploading it to the Internet via a GPRS module connected to it. In another study, a low-cost smart weather forecast system based on an artificial neural network and IoT was discussed (Islam et al. 2021). Temperature, humidity, air pressure, rainfall, soil moisture, and other weather data are measured using an IoT platform in the proposed system. An MQTT dashboard connects the various sensors in this approach. Sharma and Prakash (2021) aimed to create a weather monitoring system to continuously monitor live environmental parameters while watching the weather. The weather monitoring system employed a variety of sensors to measure the parameters and used the IoT as well. The authors of this research introduced three different sensors into the system. The sensors were structured into various bundles. The proposed model illustrates the connection between the sensor and the microcontroller unit (MCU). Sensors were linked to the MCU8266 unit in the system. Three sensors were used in the model to measure temperature, pressure, humidity, and precipitation. Similarly, the primary purpose of the work discussed in Molnár et al. (2020) was to describe how to develop and implement a weather station based on the IoT and cloud services to assist in teaching IoT technologies.

The prototype’s brain proposed by Kodali and Mandal (2016) was the ESP8266-based Wi-Fi module NodeMCU (12E). The NodeMCU is equipped with four sensors: a temperature and humidity sensor (DHT11), a pressure sensor (BMP180), a raindrop module, and a light-dependent resistor (LDR). Other data parameters, such as dew point, can be calculated using temperature and humidity. In addition to these features, the location's light intensity can also be measured. A room's air pressure can be monitored, and rain quantity can be tracked. In another study (Murugan et al. 2020), an IoT-based weather monitoring system was ideated to predict cyclones in the Cauvery delta area of Tamil Nadu. This study proposed a new architecture for monitoring typical environmental conditions. The design primarily used sensors, such as those that measure temperature, humidity, pressure, light intensity, rainfall, and so on. In addition (Carlos-Mancilla et al. 2020) discussed the design and development of a weather station called MEIoT, which is an IoT device that blends educational mechatronics with IoT. MEIoT connects to the Internet, records eight meteorological factors, and uploads the data to the cloud. Using a mixture of IoT and machine learning, Durrani et al. (2019) proposed a smart weather station that predicts and monitors meteorological data. Additionally, it generates quick warnings for residents of various places to alert them of coming hazards. It is equipped with several sensors that collect meteorological data from the surroundings and send the data to the cloud. At this stage, forecasts are created using various neural network models that have been evaluated for accuracy. Siva Nagendra Reddy et al. (2018) discussed a low-cost, high-performance system that can record minute-by-minute environmental data and transfer it to a web server. If toxic gases, such as hydrocarbons (HC), CO, sulfur dioxide (SO₂), or nitrogen oxides (NOx), are released into the environment, the system sends a warning message to higher authorities in that area and nearby areas. Additionally, the system sounds an alarm. The MQ135 air quality sensor, Node MCU, DHT 11, and buzzer are used in this system.

Kumar and Jasuja (2017) propose the Raspberry Pi as their system's main controller. Sensors detect various environmental characteristics, such as particulate matter (PM), CO, CO₂, temperature, humidity, and air pressure. The sensors are wired to the Arduino board, and the Raspberry Pi is connected to the Arduino Uno through a USB cable. The data collected by the sensors are continuously transferred to the cloud via Raspberry Pi. Finally, the proposed prototype in Kumari et al. (2018) is an IoT-based solution for monitoring temperature, humidity, rainfall, earthquakes, and light intensity. This prototype also displays the current status and history in the console server of the “ThingSpeak” cloud, which can be accessed globally via an Android app.

System model

There is a location L. There are a set of environmental sensors (S) such that {S1, S2, …, Si} ∈ S for the location L. For sensors S_j ∈ S in location L, { R1, R2, …, Rn} ∈ R are the reading taken within a moment t. Some sensors can provide readings for multiple environmental parameters simultaneously. Sensors S measure a variety of parameters, including ambient temperature (T), humidity (H), atmospheric pressure (A_p), altitude (A), and levels of dangerous atmospheric gases such as CO₂ (α) and NO₂ (β). The α and β values are fed into a fuzzy inference model which outputs the air quality index (AQI) for location L. The Fuzzy inference system (FIS) model is described in the a subsection below. The list of symbols employed in the suggested model is shown in Table 1.

Table 1.

List of symbols

Symbol	Name
L	Location
S	Set of sensors
R	Readings from sensors
T	Ambient temperature
H	Humidity
Ap	Atmospheric pressure
A	Altitude
α	CO2 level
β	NO2 level

Materials and methods

Figure 1 depicts our proposed low-cost smart weather station model. It includes a microcontroller (ESP32) that serves as the system's main processing unit, and all sensors and devices can be attached to it. We used several sensors for measuring the needed parameters for this study. For weather parameters such as humidity, ambient temperature, we used DHT11 sensor. The atmospheric pressure sensor (BMP280) was used for measuring the atmospheric pressure and altitude. While the levels of harmful gases like CO₂ and NO₂ have been gathered by employing a gas sensor (MQ135). The microcontroller can access the sensors to retrieve data, and it runs the analysis using the sensor data before uploading it to the Internet via an attached Wi-Fi module, using the MQTT protocol. MQTT is a lightweight messaging protocol based on the publish/subscribe concept. It was designed primarily for low-bandwidth IoT applications. With minimal code, MQTT protocol can deliver real-time, credible messaging services to network-connected devices. Furthermore, a fuzzy model is developed to assess air pollution in Al Hoceima city. The fuzzy system has been created by defining an input, a set of fuzzy rules, a fuzzy inference engine, and an output to generate a crisp number (defuzzification). The input variables for the system are the captured levels of NO₂ (β) and CO₂ (α). The unit of the input variables is taken as parts-per-million (ppm). The AQI is the output variable of the fuzzy model. The next subsections describe in detail the main components of the proposed model.

Fig. 1.

Model flowchart

Fuzzy inference model

The concept of fuzzy logic (FL) was proposed in the 1960s by Lotfi Zadeh (1965), an Iranian mathematician and computer scientist, to address the limitations of classical logic (Türkşen 2006). Classical logic represents the degree of membership as either 1 (true) or 0 (false). However, in the case where the membership of certain elements is unclear, classical logic becomes incapable to deal with complex real-life problems that contain some degrees of ambiguity. Therefore, fuzzy logic is used to address fuzziness in real-world problem-solving by assigning the meaning to the values ranging between 0 and 1 (Katushabe et al. 2021).

Mathematically, let's consider, μ_A the membership function of the set A defined over the Universe U. In the classical set theory, the membership function is defined by:

$$\begin{array}{c}\forall x\in U,{\mu }_A(x)=0\text{if}x\notin A\\ \mu {(x)}_A=1\text{if}x\in A\\ \end{array}$$ 1

In the context of fuzzy set theory, a fuzzy set A defined over the Universe U. The membership function is represented as follows:

$$\begin{array}{c}{\mu }_A:U\to [0,1]\\ x\to {\mu }_A(x)\\ \end{array}$$ 2

μ_A assigns a value in the interval [0,1] to each element x of U, representing the degree of belonging of x to A.

Generally, FL can represent and process imprecise or approximate knowledge. In recent years, the number of applications based on FL theory has increased significantly (Belman-Flores et al. 2022; Eriyadi et al. 2021; Wu and Xu 2020). FL is ideal for such applications because it mimics human decision-making by generating precise solutions from ambiguous and imprecise data.

Furthermore, FL facilitates the computation of linguistic variables whose values are counted as words and not numbers. A linguistic variable's main function is to model vague or imprecise knowledge. Formally, a linguistic variable is a triple (V, X, T_V), where:

• V is the name of the linguistic variable, e.g. age, height, temperature, etc.

• X is the set of values (linguistic terms) that can be taken by V, e.g. bad, good, dangerous, etc.

• T_V is a fuzzy partition, where each subset is associated with a value in X.

One of the most famous applications of FL is a technique known as FIS, which analyses the values in the input vector and, depending on a set of rules, provides values to the output. FIS has quickly established itself as one of the most effective and successful control application solutions available today (Mahboob et al. 2022). For that, we have adopted it in the present work for air quality assessment. The FIS is a decision-making system based on “IF…THEN” rules. It comprises four basic components, namely, fuzzy rule base, fuzzification, inference engine, and defuzzification, which are interrelated, as shown in Fig. 2.

Fig. 2.

Fuzzy inference model for the air quality determination

The purpose of the fuzzification step is to transform a crisp input into a linguistic variable. For this process, the fuzzy system designer must create membership functions. A membership function enables the definition of membership’s degree by converting a numerical data item into a linguistic variable. Triangular (Fig. 3a) and trapezoidal (Fig. 3b) membership functions are used in this study because they are computationally effective since utilized to normalize crisp inputs. Mathematical descriptions of these membership functions are given as follows.

• Triangular membership (Mahapatra et al. 2011):

  $$\text{Tringular}(x;a;b;c)=\left\{\begin{array}{cc}0& x\le a\\ \frac{x-a}{b-a}& a\le x\le b\\ \frac{c-x}{c-b}& b\le x\le c\\ 0& c\le x\end{array}\right)$$ 3

Fig. 3.

a Triangular membership function; b Trapezoidal membership function

Alternatively, we can write Eq. 3 as follows:

$$\text{Triangular}(a;x;y;z)=max\left(min\left(\frac{x-a}{b-a},\frac{c-x}{c-b}\right),0\right)$$ 4

• Trapezoidal membership (Janeela Theresa & Joseph Raj 2013):

  $$\text{Trapezoida}l(x;a;b;c;d)=\left\{\begin{array}{cc}0& x\le a\\ \frac{x-a}{b-a}& a\le x\le b\\ 1& b\le x\le c\\ \frac{d-x}{d-c}& c\le x\le d\\ 0& d\le x\end{array}\right)$$ 5

Eq. 5 can also be written as follows:

$$\text{Trapezoidal}(x;a;b;c;d)=max\left(min\left(\frac{x-a}{b-a},1,\frac{d-x}{d-c}\right),0\right)$$ 6

The random fluctuations that are inherent in air quality measurements might be dangerous for prolonged exposure. It is vital to provide the levels of beneficial or dangerous concentrations to categorize the negative effects of certain parameters. The AQI calculation can use a variety of sources of contaminants (Dionova et al. 2020). However, in this study, two main pollutants were taken into account for calculating the AQI for Al Hoceima city: NO₂ and CO₂. These two pollutants’ levels are crisp inputs for our FIS model, while AQI is the output variable.

The categorization levels of the air quality metrics are shown in Tables 2 and 3, which are mainly based on reported works (Carbajal-Hernández et al. 2012; Dionova et al. 2020; EPA 2014; Katushabe et al. 2021).

Table 2.

Threshold point for air pollution

CO2 (ppm)	Very low	Low	Medium	High	Very high	Hazardous
	0–300	301–600	601–1000	1001–1700	1701–2700	2701–7000
NO2 (ppm)	Very low	low	medium	High	Very high	hazardous
	0–53	54–100	101–360	361–649	650–1249	1250–2049
AQI	Excellent	Good	Regular	Bad	Dangerous	Very dangerous
	0–50	51–100	101–150	151–200	201–300	301–500

Table 3.

The six categories of AQI (EPA 2014)

Air quality index (AQI) values	Levels of health concern	Colors
0 to 50	Good	Green
51 to 100	Moderate	Yellow
101 to 150	Unhealthy for sensitive groups	Orange
151 to 200	Unhealthy	Red
201 to 300	Very unhealthy	Purple
301 to 500	Hazardous	Maroon

In this study, the AQI model for Al Hoceima city is based on pollutant concentration levels determined by the Environmental Protection Agency (EPA) air quality index guidelines (EPA 2014). As shown in Table 3, a higher AQI value indicates that the air quality is poorer; a lower AQI value implies that the air quality is better. The harmful effects of air pollutants on human health may be divided into the following categories in accordance with air quality regulations (Carbajal-Hernández et al. 2012):

• Excellent: Conditions are appropriate for practicing outdoor activities.

• Good: It is possible to participate in outdoor activities.

• Regular: Outdoor activities should be avoided since they may have a negative impact on population health, particularly among young and old persons

• Bad: There are increased negative health impacts on the community as a whole, especially among young and old persons

• Dangerous: There could be a significant negative impact on the general public's health.

• Very Dangerous: Everyone could have severe health problems.

Moreover, the associated fuzzy membership values are described as follows:

• The concentration of NO₂ (β): very low, low, medium, high, very high, and hazardous

• The concentration of CO2 (α): very low, low, medium, high, very high, and hazardous

The AQI is categorized and approximated as the output of the following membership values: excellent, good regular, bad, dangerous, and very dangerous.

The input presented in Table 2 is utilized to determine the corresponding membership function for each pollutant type based on the range of values for each category. The membership functions are shown in Fig. 4, for each input and output. Membership functions of Fig. 4 were obtained using the python matplotlib.pyplot, skfuzzy, mpl_toolkits.mplot3d libraries.

Fig. 4.

The used membership functions: a CO₂ membership function; b NO₂ membership function; c AQI membership function

So far, we have linguistic variables, we can pass them into the inference engine. Here, each rule of the inference engine is written by the fuzzy system designer according to their knowledge. The fuzzy rule base set is a collection of linguistic assertions, specifying how the FIS should decide whether to categorize an input or regulate an output. Fuzzy rules are always expressed in the manner shown here:

$$\text{IF}a\text{is in}{\text{A}}_i,\text{THEN}\text{b}\text{is in}{\text{B}}_i,$$ 7

where A_i and B_i are fuzzy sets that represent linguistic terms of input and output variables.

Table 4 shows the rule base utilized in this study. The probable combination of membership functions influences the number of defined fuzzy rules which can be obtained with the following formula (Cavallaro 2015):

$$\text{Number of possible fuzzy rules}={\text{N}}^{\text{M}}$$ 8

where:

Table 4.

Rule base

Input parameters		Output parameter
CO2	NO2	AQI for Al Hoceima city
Very Low	Very Low	Excellent
Very low	Low	Good
Very low	Medium	Regular
Very low	High	Bad
Very low	Very high	Dangerous
Very low	Hazardous	Very dangerous
Low	Very low	Good
Low	low	Good
Low	Medium	Regular
Low	High	Bad
Low	Very high	Dangerous
Low	Hazardous	Very dangerous
Medium	Very low	Regular
Medium	Low	Regular
Medium	Medium	Regular
Medium	High	Bad
Medium	Very high	Dangerous
Medium	Hazardous	Very dangerous
High	Very low	Dangerous
High	Low	Dangerous
High	Medium	Very dangerous
High	High	Very dangerous
High	Very high	Very dangerous
High	Hazardous	Very dangerous
Very high	Very low	Very dangerous
Very high	Low	Very dangerous
Very high	Medium	Very dangerous
Very high	High	Very dangerous
Very high	Very high	Very dangerous
Very high	Hazardous	Very dangerous
Hazardous	Very low	Very dangerous
Hazardous	Low	Very dangerous
Hazardous	Medium	Very dangerous
Hazardous	High	Very dangerous
Hazardous	Very high	Very dangerous
Hazardous	Hazardous	Very dangerous

M = number of parameters.

N = the number of linguistic terms per input parameter.

In our case, 6² = 36 rules.

Table 4 shows the 36 rules for two inputs (CO₂ and NO₂) with six categories for each input (i.e., very low, low, medium, high, very high, hazardous) and six categories for the AQI output (i.e., excellent, good, regular, bad, dangerous, very dangerous) utilized in this study. For instance, if we pick the first line, the reading of the input is as follows: “IF CO2 is very low AND NO₂ is very low THEN AQI is excellent”.

Finally, the defuzzification step constitutes a decisional phase that makes it possible to transform a fuzzy value of a variable into a real value. The inference engine provides a fuzzy output set. Therefore, the information at the output of the inference block is fuzzy when we need precise information. Hence, it is necessary to transform this fuzzy information into precise information (a real value). Defuzzification is what ensures this transformation.

The mean of maxima approach, center-of-gravity (COG) method, weighted average method, and the max method are some of the defuzzification techniques that have been used in past research (Masoum and Fuchs 2015). The COG approach was employed in the current study since it is considered to be the most frequent and relevant defuzzification technique. The defuzzification procedure was carried out according to the following equation, where z is the final AQI:

$$z=\frac{\int \mu \left(z\right)\ast z\ast \text{d}z}{\int \mu \left(z\right)\text{d}z}$$ 9

The COG returns the center of the area under the curve formed by the output fuzzy function (Carbajal-Hernández et al. 2012).

Message queuing telemetry transport protocol

Message Queuing Telemetry Transport (MQTT) is an open-source messaging protocol that enables non-persistent connections between devices by transferring their messages. It was developed in 1999 by Andy Stanford-Clark, an engineer at IBM, and Arlen Nipper, an engineer at EuroTech, primarily for machine-to-machine (M2M) communication. M2M communication enables the connection of two devices by utilizing various technologies. It is increasingly used to make connected objects communicate; to this end, connected objects collect various information from integrated sensors, and these data are sent via MQTT. The MQTT protocol is compatible with industrial programmable logic controllers and embedded devices, such as Arduino and Raspberry Pi. The MQTT server (broker), the MQTT client (subscriber), and the MQTT publisher are required to initiate MQTT communication. Therefore, as shown in Fig. 5, a broker is the central component of any standard MQTT protocol built on an IoT architecture. All objects and services are connected to the broker as clients. Clients can send messages as publishers and receive messages as subscribers. Published messages contain a topic that describes the content of the message (e.g., the weather in Boston, US). Each subscriber who has subscribed to the subject of the published message receives a copy of the message.

Fig. 5.

How MQTT protocol works

Device specification

Figure 6 represents the wiring of the various components of our model using the Fritzing software.

• ESP32

Fig. 6.

The wiring diagram of the proposed model

The ESP32 shown in Fig. 7a is a microcontroller with built-in Wi-Fi and Bluetooth modules. It is simple to use, lightweight, and has more memory and calculating power than its competitors. This makes it an excellent choice for learning to program and creating connected objects. Furthermore, a dual processor is included in the ESP32. This enables a more fluid and faster process. Communication (Wi-Fi or Bluetooth) is handled by one processor, while the input/output control is handled by the other. In some applications, this makes the microcontroller more efficient than the widely used (and older) ESP8266, for which the processor must occasionally interrupt input/output management to commit itself entirely to communication.

• DHT11

Fig. 7.

Various components used in the model: a ESP32 card; b DHT11 sensor; c BMP280 sensor; d MQ135 sensor

One of the most widely used sensors is the DHT11 sensor, shown in Fig. 7b. It can detect temperature and humidity. The DHT11 is an entirely digital device. There is no need for analog-to-digital conversion, and the sensor is simple to operate. With an accuracy of 1 °C and 1%, the sensor can measure temperatures from 0 °C to 50 °C and humidity from 20 to 90%.

• BMP280

The BMP280 pressure sensor (Fig. 7c) is the best low-cost sensing solution for measuring atmospheric pressure. Because pressure changes with altitude, it can also be used as an altimeter. Using the BMP 280 sensor is easy. If we use an ESP32, we simply connect the VIN pin to the 5 V voltage pin, GND to ground, SCL to analog pin 5, and SDA to analog pin 4. Afterward, we declare the library in Arduino IDE BMP180/ BMP280, and an example code for air pressure calculation is then added automatically in the program.

• MQ135

The MQ-135 gas sensor (Fig. 7d) can detect gases like NH₃, sulfur (S), benzene (C₆H₆), CO₂, and other harmful gases and fumes. Similar to other MQ series gas sensors, this sensor has a digital and analog output pin. When the level of these gases exceeds a threshold limit in the air, the digital pin rises. This threshold value can be adjusted using the onboard potentiometer. The analog output pin releases an analog voltage that can be used to approximate the level of these gases in the atmosphere. The MQ135 air quality sensor module operates at 5 V and draws around 150 mA. It must warm up before it can give accurate results.

Experimental setup and results

Figure 8 shows the realized embedded weather station prototype with its components for reading meteorological parameters and assessing air quality. A power source feeds each system component. The sensor block that is employed for meteorological measurements and observations is managed by the ESP32, which serves as the device's brain.

Fig. 8.

Photograph of the proposed embedded weather station

The suggested smart meteorological station based on FIS was tested and validated via an experimental research study. A set of data collected by different sensors was examined to conduct this investigation.

The meteorological parameters were measured using the suggested smart weather station prototype throughout 24 h (07 April 2022). Measurements were performed in the Faculty of Sciences and Technologies (FSTH) in Al Hoceima city in Morocco. Al Hoceima is geographically located in the northern center of Morocco on the Mediterranean coast, with a surface area of 3550 km². Its province has approximately 300,000 inhabitants in the urban community. Al Hoceima has a Mediterranean climate with dry and moderately hot summers and rainy, snowy, and cool winters. The system location is presented in Fig. 9.

Fig. 9.

Embedded weather station location using Google Maps

In our model, the DHT-11 sensor is employed to measure humidity and temperature. After placing the attached DHT-11 sensor in the weather station, in the described earlier location, we noticed that the model starts to record various humidity values. Figure 10a depicts the obtained values. The sensor's readout ranges from 69 to 89%. The outcome of this experiment is regarded as a demonstration that confirms the sensor's response to humidity levels. As a result, it demonstrates that our system works efficiently for detecting different humidity levels for a given location. We also conducted a further simple test to evaluate the temperature values using the same DHT-11sensor. The retrieved temperature measurement is represented in Fig. 10c. The result shows that the temperature range is between 10.8 and 21 °C. This change demonstrates the system's capability to detect various temperature values. Moreover, the BMP-280 sensor is used to test our proposed model’s ability to measure various atmospheric pressure values. The output of the atmospheric pressure sensor is shown in Fig. 10b. The air pressure ranges from 1023.5 to 1025.2 hPa. To provide the altitude, the BMP-280 sensor was also used as an altimeter in this experiment.

Fig. 10.

Measured meteorological parameters: a the recorded relative humidity in [%]; b the recorded ambient atmospheric pressure in [hPa]; c the recorded temperature in [°C]

Besides, the suggested low-cost weather device uses a fuzzy logic model to monitor air quality. To evaluate the efficiency of the model, the MQ-135 sensor was used to measure CO₂ and NO₂ levels in the air. The two values were taken as inputs for our model, while the output is the AQI. Two types of membership functions were used for input and output parameters (i.e., trapezoidal and triangular). Based on the range of values for each category, the boundary values presented in Table 2 were used to determine the corresponding membership function for each pollutant type as well as for the AQI. In our experiment, we evaluated the state of the air quality in Al Hoceima city. As indicated in a single measured point, values provided by the MQ-135 sensor are 20 ppm and 11 ppm for CO₂ and NO₂, respectively. According to Table 2, the particular membership functions with a specific range of (0.0 ≤ 20.5 ≤ 300.0) and (0 ≤ 11 ≤ 53) are selected. In other words, the membership function “Very low” is assigned for the two inputs in this case. Consequently, the appropriate fuzzy rule among the list in Table 4 is chosen to be applied. Afterwards, the fuzzy inference model aggregates all the input data to provide the fuzzy AQI value in the output. Finally, a defuzzification procedure was performed using Eq. 9. The AQI equals 35.00000000000002 after this final phase. Figure 11 depicts the AQI membership function highlighted with a black line corresponding to the calculated AQI. In this case, that means the air quality is excellent.

Fig. 11.

Example of the defuzzification process using the COG method

Moreover, as shown in Fig. 12, the surface view demonstrates how the output AQI is affected by CO₂ and NO₂ levels in the air. The three-dimensional surface view is used to see how one value of the outputs is dependent on one or two values of the input data. The X and Y-axes correspond to the input values of air pollutants, while the Z-axis indicates the AQI output value. The plot indicates that when one of the air pollutant input concentrations rises, corresponds to an increase in the AQI.

Fig. 12.

3D surface view showing the relationship between AQI and the concentration of CO₂ and NO₂

The block hardware system of the proposed system uses the MQTT protocol to broadcast via the Internet sensor data to a mobile/web application designed by Node-RED (Klawon et al. 2017). Node-RED is a “low code”, simple, fundamental, and free programming tool with an intuitive graphical user interface that makes it easy to link physical components, the application programming interface (API), and web services. It was developed in JavaScript and built on Node.js to run on the network using low-cost hardware, such as Raspberry Pi.

In particular, Node-RED employs a graphical programming approach based on flows that allow programming almost entirely without code. Rather, it uses predefined blocks of code called “nodes” that must be linked together to form a program. Therefore, a developer or non-developer can easily design useful applications utilizing simple flows. Figure 13a represents the web application dashboard layout for users, and Fig. 13b shows the mobile app layout. The web application provides the possibility to consult the values of meteorological characteristics in real time remotly. Besides, the mobile application was developed to complement the website's work and to provide a more convenient and accessible version to users on medium and small devices. It communicates to the user an appropriate interface for consulting data and manipulating the system while remaining remote from the system. In addition to the AQI, every label on the dashboard displays a distinct sensor value. Each reading panel is featured with a label and a level indicator.

Fig. 13.

The end-users dashboard: a web dashboard version; b mobile dashboard version

Discussion

The climate change impacts and the transition in weather states, as well as the various effects of air pollution, led researchers to get interested in monitoring environmental parameters and air quality to minimize and handle its intensity. Very few of the proposed weather stations in the literature had achieved the guarantee of a quick and easy weather update, due to the lack of real-time data acquisition. Furthermore, the majority of the designs are simulation-based and did not provide all the necessary information under real conditions (Dionova et al 2020; Ghorbani and Zamanifar 2022; Katushabe et al. 2021).

Table 5 establishes a comparison of some reported works based on the adopted methodologies and used technologies in these research works (Hasanh et al. 2021; Dayananda et al. 2021; Saini et al. 2022; Si Tayeb et al. 2022), with our proposed model. In contrast, although the authors (Dayananda et al. 2021; Si Tayeb et al. 2022) were successful in obtaining values for some dangerous gases such as CO₂, CO, NH₃, and NO₂, but they did not use this information to assess air quality. Furthermore, authors (Hasanh et al. 2021; Dayananda et al. 2021; Saini et al. 2022) provided no solution for visualizing the meteorological data collected remotely. Hence, in the present work we did assess the air quality using the collected values of some dangerous gases.

Table 5.

Methodology comparison

Reference	Air quality assessment	Fuzzy logic	Used communication technologies	Web app	Mobile app	Used sensors	Measured parameters
(Hasanh et al. 2021)	–	–	–	–	–	MQ4, MQ7	CO, CO2
(Dayananda et al. 2021)	–	–	SIM900A GSM module	–	–	DHT11, BMP180, FC37, MG811, Hall-effect, nfra-Red	Temperature, humidity, Pressure, rain, light intensity, CO2 level, wind speed, wind direction
(Saini et al. 2022)	–	–	-	–	–	CCS811, SDS011, MP503, DHT11	CO2, tVOC, PM10, PM2.5, CO, NO2, Temperature, Humidity
(Si Tayeb et al. 2022)	–	–	HTTP REST-API, JSON,	–	–	DHT22, MH-Z19, CJMCU-30, CJMCU-6814, CJMCU-680v2	CO, CO2, NH3, NO2,temperature humidity
Our proposed model	–	–	MQTT, Wi-Fi	–	–	DHT11, BMP280, MQ135	temperature, humidity, atmospheric pressure, altitude, levels of CO2 and NO2

Conclusion and future work

Due to the traditional weather monitoring stations' limited data access, large size, high cost, and inability to expand, many researchers have recently focused on other alternative solutions that take advantage of advanced sensing technologies such as IoT. Thus, this study was carried out in this context. The purpose of the proposed system is to create a low-cost, robust, modular weather station based on IoT and a fuzzy inference model. The proposed model monitors various weather parameters including ambient temperature, relevant humidity, atmospheric pressure, altitude, and CO₂ and NO₂ levels. A model based on a fuzzy inference system has been also presented to determine the state of air quality for a given location. The model was tested in real-time to evaluate the air quality in al Hoceima city based mainly on air pollutant data concentrations (CO₂ and NO₂). It is concluded that the fuzzy logic model is capable of efficiently determining the air quality index. The final data are broadcasted wirelessly using the MQTT protocol to the end-users dashboard. The suggested system can be improved in various ways in the future. For example, increasing the number of sensors would boost the precision of the measured environmental parameters. Additionally, the air quality could be calculated by considering additional pertinent pollutants, such as sulfur dioxide (SO₂), carbon monoxide (CO), O₃, and PM_x.

Author contributions

All the authors have contributed equally to this study.

Funding

The authors declare that there is no funding involved in this research.

Availability of data and materials

The authors declare that there is no external data or materials involved in this study.

Declarations

Conflict of interest

The author declares no conflicts of interest.